Vous êtes sur la page 1sur 396

Symposium Proceedings:

Transformer Reliability:
Management of Static Electrification
in Power Transformers

TR-113741

Proceedings, October 1999

EPRI Project Manager


S. Lindgren

EPRI 3412 Hillview Avenue, Palo Alto, California 94304 PO Box 10412, Palo Alto, California 94303 USA
800.313.3774 650.855.2121 askepri@epri.com www.epri.com
DISCLAIMER OF WARRANTIES AND LIMITATION OF LIABILITIES
THIS DOCUMENT WAS PREPARED BY THE ORGANIZATION(S) NAMED BELOW AS AN
ACCOUNT OF WORK SPONSORED OR COSPONSORED BY THE ELECTRIC POWER RESEARCH
INSTITUTE, INC. (EPRI). NEITHER EPRI, ANY MEMBER OF EPRI, ANY COSPONSOR, THE
ORGANIZATION(S) BELOW, NOR ANY PERSON ACTING ON BEHALF OF ANY OF THEM:

(A) MAKES ANY WARRANTY OR REPRESENTATION WHATSOEVER, EXPRESS OR IMPLIED, (I)


WITH RESPECT TO THE USE OF ANY INFORMATION, APPARATUS, METHOD, PROCESS, OR
SIMILAR ITEM DISCLOSED IN THIS DOCUMENT, INCLUDING MERCHANTABILITY AND FITNESS
FOR A PARTICULAR PURPOSE, OR (II) THAT SUCH USE DOES NOT INFRINGE ON OR
INTERFERE WITH PRIVATELY OWNED RIGHTS, INCLUDING ANY PARTY'S INTELLECTUAL
PROPERTY, OR (III) THAT THIS DOCUMENT IS SUITABLE TO ANY PARTICULAR USER'S
CIRCUMSTANCE; OR

(B) ASSUMES RESPONSIBILITY FOR ANY DAMAGES OR OTHER LIABILITY WHATSOEVER


(INCLUDING ANY CONSEQUENTIAL DAMAGES, EVEN IF EPRI OR ANY EPRI REPRESENTATIVE
HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES) RESULTING FROM YOUR
SELECTION OR USE OF THIS DOCUMENT OR ANY INFORMATION, APPARATUS, METHOD,
PROCESS, OR SIMILAR ITEM DISCLOSED IN THIS DOCUMENT.

ORGANIZATION(S) THAT PREPARED THIS DOCUMENT

Stig Nilsson, Consultant

ORDERING INFORMATION
Requests for copies of this report should be directed to the EPRI Distribution Center, 207 Coggins
Drive, P.O. Box 23205, Pleasant Hill, CA 94523, (800) 313-3774.

Electric Power Research Institute and EPRI are registered service marks of the Electric Power
Research Institute, Inc. EPRI. POWERING PROGRESS is a service mark of the Electric Power
Research Institute, Inc.

Copyright 1999 Electric Power Research Institute, Inc. All rights reserved.
CITATIONS

This report was prepared by

Stig Nilsson, Consultant


20715 Brush Road
Los Gatos, California 95033-9138

This report describes research sponsored by EPRI.

The report is a corporate document that should be cited in the literature in the following manner:

Symposium Proceedings: Transformer Reliability: Management of Static Electrification in


Power Transformers, EPRI, Palo Alto, CA: 1999. TR-113741.

iii
REPORT SUMMARY

Certain mitigation procedures could have prevented most of the known static electrification
failures of very large forced-oil-cooled power transformers. This conclusion was confirmed at an
EPRI-sponsored symposium where participants shared worldwide research results and field
experiences spanning 15 years.

Background
In January 1979, a prototype high-voltage DC valve (RP213) developed pinhole leaks in cooling-
loop insulators containing flowing refrigerant 113. A subsequent study (EPRI reports EL-4501
and EL-6138) investigated static electrification as a possible explanation for these leaks and
developed a number of models. Several 1979 reports from Japan also indicated that abnormally
high-velocity oil flow cold cause static electrification problems in power transformers. Although
U.S. transformers operate with lower flow rates, over a dozen U.S. failures have been associated
with the problem since 1982. EPRI has sponsored ongoing research to develop possible
solutions, and sponsored workshops in 1986, 1989, and 1994 to assess results from related EPRI
projects (EPRI reports EL-6081, EL-6880, EL-6918, TR-101216, TR-102112, TR-104973,
TR-105019, TR-111386, TR-113381, TR-113441, TR-113461, and TR-113535).

Objective
To review the latest worldwide research on the management of static electrification in power
transformers with emphasis on results from EPRI-sponsored field experiments (cosponsored by
Con Edison) on a 336 MVA 525/345 kV autotransformer and laboratory experiments on a
1/4 scale flow model of this transformer.

Approach
EPRI invited researchers, transformer manufacturers, component suppliers, utility engineers, and
other experts to participate in a symposium held in Monterey, California, May 19-21, 1999. Over
50 attendees participated.

Results
The main conclusion from the symposium is that static electrification failures of existing
transformers are preventable. The technology and understanding exists to identify potentially
high risk operating conditions; monitoring equipment is available or can be readily designed
to detect incipient failures.
Design guidelines for new transformers can help minimize the risk of static electrification
failures for future large forced-oil-cooled power transformers.

v
EPRI Perspective
Despite the culmination of 15 years of EPRI-sponsored effort, much about static electrification
(SE) remains a mystery with the potential for committing the perfect crime in large forced-oil-
cooled shell-form or core-form power transformers. Build-up of DC stress, caused by flowing
oil, can contribute to a dielectric breakdown without leaving the slightest trace of evidence.
Similar mysterious problems occur in many insulated moving processes, such as printing,
spinning, weaving, and flying through the air, and not just in transformers. Even lightning is not
understood as well as it could be. SE will always be present to some extent in very large
directed-flow power transformers when pumps are running. A certain set of conditions can occur
to cause a catastrophic failure, especially when it is believed there is no problem because no
problem has been recognized before. Proper use of the SE knowledge developed in EPRI studies
will help control electrification-related transformer problems and lead to improvements in high-
voltage transformer reliability.

TR-113741
Keywords
Electrification
Transformers
Insulation
Failures
Reliability

vi
ABSTRACT

This report contains the papers presented at EPRIs Transformer Reliability: Management of
Static Electrification in Power Transformers Symposium held May 19th through 21st, 1999 in
Monterey, California. Also included is a record of the panel session and other discussions taking
place at the event.

The symposium is the culmination of 25 years of research on static electrification in large power
transformers. The EPRI sponsored research began in 1984 but builds on work that goes back to
the early to mid 1970s when the first signs of static electrification in large power transformers
began to be seen in the industry. Several workshops on the subject have been sponsored by EPRI
in the course of leading this research. EPRI has therefore been in position to obtain research
results produced in other countries for the benefit of the entire industry. This Symposium is
culmination of all of this work. The presented papers contain information about tests run on a
transformer owned by Consolidated Edison of New York, Inc. These tests were conducted in the
field at Con Edisons Ramapo substation. The report also contains the latest thinking on how to
detect static electrification, manage transformers with a propensity for static electrification and
information about how to design new transformers with lower risk for static electrification
failure. One case study presented at the symposium discusses how a utility detected a critical
static electrification related operating emergency and how the operation of the transformer was
changed to bring the transformer back out of the critical operating region to avoid failure.

vii
ACKNOWLEDGMENTS

There are many that contributed to making the symposium a success. Static electrification is a
rather special subject that is not well understood. The research has made progress in a jerky
fashion. At times with no apparent progress and then, upon some breakthrough, we have seen
accelerated learning. Such a breakthrough was the realization by Stan Lindgren that moisture
dynamics appeared to play a significant role in static electrification failures. Very dry oil seemed
to be a factor that increased the risk of failure. This hypothesis has been substantiated through
the tests of a Consolidated Edison owned transformer. These tests began in 1995 and ended with
a teardown and inspection of the unit in March of 1997. The support of Con Edison in helping to
fund the tests is gratefully acknowledged because the research has brought about a much better
understanding of the static electrification phenomena in large transformers. A number of people
from Con Edison contributed to this. Leo Savio, Noah Tai, Don Chu, Mike Lebow were
instrumental in obtaining funding for the work. The data from these tests, if applied correctly,
should lead to no more static electrification failures in existing transformers. Of course, the
efforts by Harold Moore in technically guiding the research are also recognized. Many papers in
the symposium had their origin in this work.
However, the contributions by Japanese researchers are also recognized. The symposium has
many excellent contributions written by a distinguished group of researchers. From Japan the
papers contain information developed by a Working Group led by Professor Dr. Hitoshi Okubo.
Other members of this WG that contributed to the success of the symposium are: Motoo Tsuchie,
Noboru Hosokawa, Susumu Isaka, Hiroshi Miyao, Toshitsugu Ishii, Shunichi Kobayashi,
Takayuki Kobayashi, and Takahiro Ono. Their effort is greatly appreciated.
Static electrification phenomena would however, not be as well understood as they are without
contributions from academia. This work focused on fundamental material properties, the physics
of static electrification, modeling of electric field stresses and elekrokinetic aspects of the oil
flows. Also, understanding the mechanism behind partial discharges that appear as a result of the
electrification, monitoring of transformers etc. The late Jim Melcher, who worked at MIT,
Markus Zahn at MIT and Keith Nelson, RPI have produced a number of graduate students in the
course of doing research on static electrification. Professor Praxl at Graz participated in the
symposium representing academia from outside the US. The contributions by academia to this
symposium fill in the gaps between imperical results and theory.
The assistance from Bruce Gavioli, Con Edison; Jean Poittevin, Alstom, France; Don Rose,
TU Electric; Mike Lebow, Con Edison; and Dan Crofts, retired from TU Electric in chairing the
paper and panel sessions is also gratefully acknowledged.

Finally, last but not least, the assistance from Paige Polishook in handling the arrangements for
the symposium and Michelle Samoulides for her help at the hotel are acknowledged. Also, a
thank you to Laine Gates, who did a marvelous job in transcribing the tapes for the discussion
sessions.

ix
CONTENTS

1 PANEL SESSION FOR PRESENTATION OF RAMAPO TEST RESULTS ......................... 1-1


1.1 Application of In Service Transformer Monitoring Systems during Tests at
Consolidated Edisons Ramapo Station............................................................................. 1-3
Abstract .............................................................................................................................. 1-3
Introduction......................................................................................................................... 1-3
Outline of Tests.............................................................................................................. 1-3
Summary of Tests Reported in 1996 Paper ........................................................................ 1-4
1996 Test Series - Series III and IV .................................................................................... 1-4
Series III: September 1996............................................................................................ 1-5
Series IV: October 1996................................................................................................ 1-5
Test Setup for Series III and IV ...................................................................................... 1-5
Summary of Series III and IV Tests .................................................................................... 1-6
Series III......................................................................................................................... 1-6
Type A....................................................................................................................... 1-6
Type B, Second Day.................................................................................................. 1-7
Type C, Third Day ..................................................................................................... 1-7
Summary Series III......................................................................................................... 1-8
Series IV ........................................................................................................................ 1-8
Type A....................................................................................................................... 1-8
Type C, Second Day ................................................................................................. 1-8
Type B on Day 3........................................................................................................ 1-9
Type C on Day 3 ....................................................................................................... 1-9
Type C on Day 4 ....................................................................................................... 1-9
Overview of Program Results ........................................................................................... 1-10
Static Electrification...................................................................................................... 1-10
Performance of Sensors and Monitors.............................................................................. 1-11
Partial Discharge Detection.......................................................................................... 1-11
Water and Gas in Oil Monitors ..................................................................................... 1-12

xi
Diagnostic Systems .......................................................................................................... 1-12
Summary and Recommendations..................................................................................... 1-14
Acknowledgement ............................................................................................................ 1-14
Reference......................................................................................................................... 1-14
1.2 Acoustic Partial Discharge and Leakage Current Measurement Results of
Ramapo Tests .................................................................................................................. 1-16
Abstract ............................................................................................................................ 1-16
Acoustic Discharge Detection Scheme ............................................................................. 1-16
Overview of Acoustic Partial Discharge Detection System ........................................... 1-16
Performance of Acoustic Partial Discharge System ..................................................... 1-19
Winding Leakage Current System .................................................................................... 1-24
System Description ...................................................................................................... 1-24
Performance of Leakage Current Measurement System .............................................. 1-25
Relative Humidity and General Oil Tests .......................................................................... 1-27
Humidity Sensor........................................................................................................... 1-27
Oil Tests....................................................................................................................... 1-28
Summary and Conclusions ............................................................................................... 1-29
1.3 Ramapo Static Electrification Project Test Results from J. W. Harley Inc.................... 1-30
Abstract ............................................................................................................................ 1-30
Introduction....................................................................................................................... 1-30
Experimental Instrumentation ........................................................................................... 1-30
Non-Energized Test Results ............................................................................................. 1-31
Energized Test Results..................................................................................................... 1-35
Conclusion........................................................................................................................ 1-39
Acknowledgements........................................................................................................... 1-40
1.4 Ramapo Tests: Results from Streaming Current Monitoring Using a Nilsson Plate
and an Absolute Charge Sensor ....................................................................................... 1-41
Abstract ............................................................................................................................ 1-41
Keywords.......................................................................................................................... 1-41
I. Introduction.................................................................................................................... 1-41
II. Monitoring System ........................................................................................................ 1-42
III. Tests with No Applied Voltage ..................................................................................... 1-43
A. March 1995 Test Series........................................................................................... 1-43
B. Tests on July 10, 1995............................................................................................. 1-44
C. Tests September 1996 ............................................................................................ 1-46

xii
D. Tests during October 15, 1996 ................................................................................ 1-47
IV. Tests on Energized Transformer ................................................................................. 1-48
A. July 11 and 12, 1995 ............................................................................................... 1-48
B. October 16, 1996 Data ............................................................................................ 1-50
V. Other Test Observations .............................................................................................. 1-51
VI. Conclusions and Observations .................................................................................... 1-51
Acknowledgements........................................................................................................... 1-52
VII. References................................................................................................................. 1-52
VIII. Biographies ............................................................................................................... 1-53
1.5 Dismantling of Ramapo Transformer .......................................................................... 1-54
Abstract ............................................................................................................................ 1-54
Introduction....................................................................................................................... 1-54
Location of Discharges ..................................................................................................... 1-55
Discharges at the Top End ............................................................................................... 1-55
Lead and Lead Support Structure at Top End .............................................................. 1-55
Discharges on Walls Adjacent to HV Bushing Shield ................................................... 1-56
Discharges on Lower HV Bushing Shield ..................................................................... 1-56
Relationship of Discharges to Partial Discharge Locations at Top during Tests ........... 1-57
Significance of Change in Discharge Pattern when Flowing Hot Oil into Tank.............. 1-57
Location of Discharges at Bottom End .............................................................................. 1-57
Discharges in the V Region ....................................................................................... 1-58
Discharges at the 45 Degree Region ........................................................................... 1-58
Progression of Failure from the 45 Degree Region into the Windings .......................... 1-60
a. Low Voltage Winding........................................................................................... 1-60
b. High Voltage Winding .......................................................................................... 1-60
Summary and Conclusions ............................................................................................... 1-60

2 MODELING AND PREDICTION OF STATIC ELECTRIFICATION ...................................... 2-1


2.1 Static Electrification at Solid/Liquid Interfaces of Power Transformers .......................... 2-3
Abstract .............................................................................................................................. 2-3
Charge Generation ............................................................................................................. 2-3
Adhesion and Cohesion...................................................................................................... 2-7
Charge Transfer ............................................................................................................... 2-10
Conclusions ...................................................................................................................... 2-12
6 References .................................................................................................................... 2-15

xiii
2.2 Study on Increase of Electrostatic Charging Tendency of Insulating Oil and the
Diagnosis on Streaming Electrification.............................................................................. 2-18
Abstract ............................................................................................................................ 2-18
Introduction....................................................................................................................... 2-19
Study on the Increase of ECT of Insulating Oil.................................................................. 2-19
Causes of Increase in ECT and Verification Method .................................................... 2-19
Deterioration Due to Aging Caused by a Slight Amount of Oxygen Contained in the
Transformer Oil ............................................................................................................ 2-20
Deterioration of Filled Oil as it got Mixed with the Impregnated Oil, which was
Deteriorated Factory Oil Used for Drying at the Time of Production ............................. 2-20
Air Exposure Test of the Transformer Oil ..................................................................... 2-21
Mixing of Aged Transformer Oil with New Oil at the Time of Exchanging the
Transformer Oil ............................................................................................................ 2-21
Results of Verification Test........................................................................................... 2-21
Deterioration Due to Aging Caused by a Slight Amount of Oxygen Contained in
the Transformer Oil.................................................................................................. 2-21
Deterioration of Filled Oil as it gets Mixed with the Impregnated Oil Produced by
the Impregnation of Factory Oil, Used for Drying at the Time of Production into
the Insulator ............................................................................................................ 2-22
Air Exposure Test of Transformer Oil ...................................................................... 2-23
Mixing of Aged Transformer Oil with New Insulating Oil at the Time of
Transformer Oil Exchange....................................................................................... 2-25
Study on Suppression of the Increase of ECT in Insulating Oil ......................................... 2-26
Verification Method of ECT Suppression...................................................................... 2-26
Result of Verification Test ............................................................................................ 2-26
Diagnosis of Streaming Electrification............................................................................... 2-27
Mechanism and Diagnosis of Streaming Electrification ................................................ 2-27
Increase in ECT of Insulating Oil and Diagnosis........................................................... 2-28
Criteria for Electrification Properties ............................................................................. 2-30
Conclusion........................................................................................................................ 2-33
References ....................................................................................................................... 2-33
2.3 The Electrostatic Charging Tendency of In-Service Oils and The Evaluation
Method of Leakage Current from Transformer Windings................................................... 2-34
Abstract ............................................................................................................................ 2-34
Introduction....................................................................................................................... 2-35
ECT of In-Service Oils ...................................................................................................... 2-35
Study Method of In-Service Oils ................................................................................... 2-35

xiv
Investigation Results of ECT ........................................................................................ 2-35
Evaluation Method of the Leakage Current ....................................................................... 2-36
Flow Electrification Diagnosis and Leakage Current from Transformer Windings ......... 2-36
Estimation Method of the Leakage Current .................................................................. 2-37
Conclusion........................................................................................................................ 2-38
References ....................................................................................................................... 2-38
2.4 Flow Electrification Measurements of Transformer Insulation ..................................... 2-48
Abstract ............................................................................................................................ 2-48
1 Background ................................................................................................................... 2-48
1.1 The Flow Electrification Problem ............................................................................ 2-48
1.2 Flow Electrification Scenario .................................................................................. 2-49
2 Stages of Electrification ................................................................................................. 2-49
2.1 Charge Generation - The Electrical Double Layer .................................................. 2-49
2.2 Transport................................................................................................................ 2-50
2.3 Charge Storage...................................................................................................... 2-50
2.4 Charge Leakage..................................................................................................... 2-50
3 Measurement Methodologies......................................................................................... 2-51
3.1 Absolute Charge Sensor ........................................................................................ 2-51
3.2 Couette Charger..................................................................................................... 2-53
3.2.1 Operating Principles ....................................................................................... 2-53
3.2.2 Open-Circuit Voltage and Short-Circuit Current Measurements...................... 2-54
4 Representative Couette Facility Results ........................................................................ 2-54
4.1 Rotation Rate Dependence .................................................................................... 2-55
4.2 Temperature Dependence ..................................................................................... 2-55
4.3 Moisture Dependence ............................................................................................ 2-59
4.4 Applied DC Voltage Dependence........................................................................... 2-59
4.5 Conditioning Transients.......................................................................................... 2-60
4.6 Effects of Benzotriazole (BTA) ............................................................................... 2-62
4.6.1 BTA Measurements in Oil ............................................................................... 2-62
4.6.2 BTA Measurements in Pressboard ................................................................. 2-63
4.6.3 Electrification Measurements.......................................................................... 2-63
5 Summary and Conclusions ............................................................................................ 2-66
Acknowledgment .............................................................................................................. 2-67
References ....................................................................................................................... 2-67

xv
2.5 Measurement of Oil Charge Densities in a Transformer Model ................................... 2-69
Abstract ............................................................................................................................ 2-69
Background ...................................................................................................................... 2-69
Description of -scale Model Design................................................................................ 2-70
Experimental Procedure Description................................................................................. 2-73
Discussion of Results ....................................................................................................... 2-74
Experiments with Model in Base Case Position............................................................ 2-75
Experiments Involving Raised Model ........................................................................... 2-77
Mitigation Experiment................................................................................................... 2-79
General Oil Tests ......................................................................................................... 2-80
Summary and Conclusions ............................................................................................... 2-81
References ....................................................................................................................... 2-81
2.6 Streaming Electrification Dynamics - Duct Modeling and Analysis .............................. 2-83
Abstract ............................................................................................................................ 2-83
Introduction....................................................................................................................... 2-83
Charge Transport ............................................................................................................. 2-84
Implementation ................................................................................................................. 2-84
Initial Charge Density Distribution ................................................................................ 2-85
Electric Potential Distribution........................................................................................ 2-86
Flow Velocity Distribution ............................................................................................. 2-87
Dynamic Charge Distribution and Currents .................................................................. 2-88
Model Performance .......................................................................................................... 2-88
Base Case ................................................................................................................... 2-88
Operating Conditions ................................................................................................... 2-91
Velocity.................................................................................................................... 2-91
Applied Field............................................................................................................ 2-93
Geometry ..................................................................................................................... 2-93
Channel Half Height ................................................................................................ 2-94
Leading Edge Aspect Ratio ..................................................................................... 2-94
Channel Length ....................................................................................................... 2-95
Chemical Characteristics ............................................................................................. 2-96
Conductivity............................................................................................................. 2-96
Wall Charge Density................................................................................................ 2-97
Coefficient of Diffusion............................................................................................. 2-98

xvi
Conclusions ...................................................................................................................... 2-98
Acknowledgements........................................................................................................... 2-99
References ....................................................................................................................... 2-99
2.7 Investigation for Standardization of Electrostatic Charging Tendency
Measurement of Transformer Oil in Japan...................................................................... 2-101
Abstract .......................................................................................................................... 2-101
Introduction..................................................................................................................... 2-101
Investigation on Mini-Static Tester .................................................................................. 2-102
Mini-Static Tester ....................................................................................................... 2-102
Current Waveform Reading........................................................................................ 2-103
Algebraic Calculation ................................................................................................. 2-104
Reproducibility Assessment by Plural Facilities on the Same Sample Oils................. 2-105
Correlation between Mini-Static Tester and Conventional Methods............................ 2-105
Characteristics of Mini-Static Tester........................................................................... 2-106
Temperature Dependence ......................................................................................... 2-106
Oil Flow Rate Dependence ........................................................................................ 2-107
Verification of Oil Flow Dependence on ECT Measurement due to Heterogeneous
Pores .............................................................................................................................. 2-110
Conclusion...................................................................................................................... 2-111
References ..................................................................................................................... 2-111

3 OPERATING EXPERIENCE OF TRANSFORMERS ........................................................... 3-1


3.1 A Static Electrification Transformer Survey ................................................................... 3-3
Abstract .............................................................................................................................. 3-3
Rationale ............................................................................................................................ 3-3
Survey Methodology ........................................................................................................... 3-4
Sources of Existing Information...................................................................................... 3-4
Survey Questionnaire..................................................................................................... 3-4
Database Refinement .................................................................................................... 3-5
Derived Measures .............................................................................................................. 3-5
Equipment Characteristics ............................................................................................. 3-5
Oil Data.......................................................................................................................... 3-7
Operational Parameters ................................................................................................. 3-8
Other Factors ............................................................................................................... 3-10
Appraisal .......................................................................................................................... 3-11

xvii
Speculation on the Mechanism of Failure..................................................................... 3-11
Countermeasures ........................................................................................................ 3-12
Conclusions ...................................................................................................................... 3-13
Acknowledgements........................................................................................................... 3-13
References ....................................................................................................................... 3-13
3.2 A Static Electrification Failure at the Navajo Generating Station ................................. 3-16
Abstract ............................................................................................................................ 3-16
Pre-fault Information ......................................................................................................... 3-16
The Failure ....................................................................................................................... 3-18
Dismantling Investigation .................................................................................................. 3-20
Coil Packet Removal.................................................................................................... 3-21
Static Electrification Evidence ...................................................................................... 3-21
Coil Failure at the Neutral............................................................................................. 3-22
Low Voltage Conductor Heating................................................................................... 3-22
Life Evaluation ............................................................................................................. 3-22
Degree of Polymerization Tests............................................................................... 3-22
2-Furaldehyde (2FAL) Analysis ............................................................................... 3-24
Conclusions ...................................................................................................................... 3-24
Action Plan................................................................................................................... 3-24
Perplexity ..................................................................................................................... 3-25
Recommendation......................................................................................................... 3-25
Summary .......................................................................................................................... 3-26
Reference......................................................................................................................... 3-26
3.3 Whitpain #1C Transformer Failure .............................................................................. 3-32
Abstract ............................................................................................................................ 3-32
Investigation ..................................................................................................................... 3-32
3.4 Static Electrification Failure of MPT at Powerton Generating Station .......................... 3-36
Abstract ............................................................................................................................ 3-36
Background ...................................................................................................................... 3-36
Incident Description .......................................................................................................... 3-36
Operations........................................................................................................................ 3-37
Teardown Inspection ........................................................................................................ 3-37
Analysis ............................................................................................................................ 3-38
Conclusions ...................................................................................................................... 3-38

xviii
References ....................................................................................................................... 3-38
3.5 Investigation of a Failed Westinghouse 500/345 kV 500 MVA Single Phase
Transformer ...................................................................................................................... 3-41
Abstract ............................................................................................................................ 3-41
History .............................................................................................................................. 3-41
Incident............................................................................................................................. 3-42
Preliminary Failure Investigation ....................................................................................... 3-42
Transformer Teardown ..................................................................................................... 3-44
Online Monitoring ............................................................................................................. 3-46
Conclusion........................................................................................................................ 3-46

4 MONITORING AND MANAGEMENT METHODS ................................................................ 4-1


4.1 Further Analysis of Streaming Current as Recorded from the Ramapo
Transformer ........................................................................................................................ 4-3
Abstract .............................................................................................................................. 4-3
Introduction......................................................................................................................... 4-3
Installation of Nilsson Plates ............................................................................................... 4-3
Circuit Analysis ................................................................................................................... 4-4
Review of Test Records without Applied Voltage................................................................ 4-5
July 10, 1995..................................................................................................................4-5
March 27, 1995 .............................................................................................................. 4-7
September 17, 1996....................................................................................................... 4-7
October 15, 1996 ........................................................................................................... 4-8
Review of Test Records with Applied Voltage..................................................................... 4-9
July 11, 1995..................................................................................................................4-9
July 12, 1995................................................................................................................ 4-10
July 13, 1995................................................................................................................ 4-11
October 16, 1996 ......................................................................................................... 4-12
Review of ACS Data ......................................................................................................... 4-14
Summary and Conclusions ............................................................................................... 4-16
References ....................................................................................................................... 4-17
Appendix A. Equations Used for the Analysis of the Equivalent Circuit ............................. 4-17
Appendix B. Equations Used for the Analysis of the Nilsson Plate Currents ..................... 4-19
4.2 Static, The Ramapo Experience and What can be Done ............................................ 4-21
Abstract ............................................................................................................................ 4-21

xix
Introduction....................................................................................................................... 4-21
Electrical Detection of Static Events ................................................................................. 4-23
Measurement Configuration, Ramapo Test Site ............................................................... 4-24
The Distinctive Electrical PD Signal, The Cascade - Event ............................................... 4-30
Hazardous, Loud Audible Bang Discharge Events .......................................................... 4-30
Discharge Track Marks After Cascade-Type Events......................................................... 4-31
Conclusions ...................................................................................................................... 4-31
Acknowledgement ............................................................................................................ 4-35
Reference......................................................................................................................... 4-35
4.3 Implementing a Static Electrification Diagnostic Model ............................................... 4-36
Abstract ............................................................................................................................ 4-36
History .............................................................................................................................. 4-36
1. Uninsulated T-beams ............................................................................................... 4-36
2. Open Winding Failures............................................................................................. 4-37
Early Detection ................................................................................................................. 4-38
Correlation with Oil Temperature.................................................................................. 4-40
Cooling Modifications ................................................................................................... 4-40
Diagnostic Model .............................................................................................................. 4-41
Acknowledgements........................................................................................................... 4-44
4.4 Mitigation of Streaming Electrification by Intelligent Control of Oil Circulation ............ 4-45
Abstract ............................................................................................................................ 4-45
Introduction....................................................................................................................... 4-45
Design Approach .............................................................................................................. 4-46
Controller Design.............................................................................................................. 4-47
Input Preprocessing ..................................................................................................... 4-48
Main Processor ............................................................................................................ 4-49
Pump Operator ............................................................................................................ 4-49
Training Considerations.................................................................................................... 4-50
Example Case .................................................................................................................. 4-51
Results ............................................................................................................................. 4-52
Conclusion........................................................................................................................ 4-54
Acknowledgements........................................................................................................... 4-54
References ....................................................................................................................... 4-55
4.5 Transformer Reliability: Management of Static Electrification in Power
Transformers .................................................................................................................... 4-56

xx
Abstract ............................................................................................................................ 4-56
Background ...................................................................................................................... 4-56
Static Electrification in Power Transformers...................................................................... 4-58
Oil ................................................................................................................................ 4-59
Insulation ..................................................................................................................... 4-59
ECT Free Transformer Design and Construction ......................................................... 4-60
Solid Insulation............................................................................................................. 4-62
External Circuit............................................................................................................. 4-62
Summary and Conclusions ............................................................................................... 4-63
Recommendations............................................................................................................ 4-64
Acknowledgements........................................................................................................... 4-64
References ....................................................................................................................... 4-65

5 PANEL SESSION: ARE WE THERE YET? ......................................................................... 5-1


5.1 Mitigation of Static Electrification Discharges in Power Transformers ........................... 5-3
Abstract .............................................................................................................................. 5-3
Introduction......................................................................................................................... 5-3
Enhanced Understanding from the Ramapo Project ........................................................... 5-3
Impact of Temperature................................................................................................... 5-3
Effect of Oil Flow Volume and Velocity........................................................................... 5-4
Effect of Temperature Cycling........................................................................................ 5-5
Effect of Running Pumps before Applying Voltage ......................................................... 5-5
Recommended Mitigation Procedures ................................................................................ 5-5
Control of Oil Flow.......................................................................................................... 5-5
Control of Pumps as a Function of Temperature ............................................................ 5-6
Special Precautions after Drying, Oil Processing or Long Term Idle Periods.................. 5-7
Change of Cooling ......................................................................................................... 5-7
Design of New Transformers .............................................................................................. 5-7
Summary ............................................................................................................................ 5-8
Reference........................................................................................................................... 5-9
5.2 Record of Discussion from TRANSFORMER RELIABILITY: Management of Static
Electrification in Power Transformers Symposium ............................................................ 5-10
Session: Panel Session for Presentation of Ramapo Test Results ................................... 5-10
Paper: Application of In Service Transformer Monitoring Systems during Tests at
Consolidated Edisons Ramapo Station, by Harold Moore and Donald Chu................ 5-10

xxi
Paper: Acoustic Partial Discharge and Leakage Current Measurement Results of
Ramapo Tests, by George K. Frimpong, Michael J. Walden and Stan Lindgren ......... 5-11
Paper: Ramapo Static Electrification Project Test Results from J. W. Harley Inc.
by J. C. Harley ............................................................................................................. 5-12
Paper: Ramapo Tests: Results from Streaming Current Monitoring Using a
Nilsson Plate and an Absolute Charge Sensor, by Stig Nilsson, Markus Zahn and
Stan Lindgren............................................................................................................... 5-18
Paper: Dismantling of Ramapo Transformer by Harold Moore ................................... 5-18
Session: Modeling and Prediction of Static Electrification ................................................. 5-21
Paper: Static Electrification at Solid/Liquid Interfaces of Power Transformers,
G. Praxl and K. Neuner ................................................................................................ 5-21
Paper: Study on Increase of Electrostatic Charging Tendency of Insulating Oil and
the Diagnosis on Streaming Electrification, by Motoo Tsuchie, Noboru Hosokawa,
Susumu Isaka, Hiroshi Miyao, Toshitsugu Ishii, Shunichi Kobayashi, Takayuki
Kobayashi, Takahiro Ono, Hitoshi Okubo..................................................................... 5-22
Paper: The Electrostatic Charging Tendency of In-Service Oils and the Evaluation
Method of the Leakage Current from the Transformer Winding, by Hiroshi Miyao,
E. Mori, S. Isaka, M. Tsuchie, K. Takamoro, S. Kobayashi, T. Koybayashi, T. Ono
and H. Okubo............................................................................................................... 5-23
Paper: Flow Electrification Measurements of Transformer Insulation, by Markus
Zahn, Andrew P. Washabaugh and Darrell E. Schlicker............................................... 5-24
Paper: Measurement of Oil Charge Densities in a Transformer Model, by George
K. Frimpong, Michael J. Walden and Stan Lindgren..................................................... 5-26
Spontaneous Contribution by Professor Dr. Hitoshi Okubo .......................................... 5-28
Paper: Streaming Electrification Dynamics - Duct Modeling and Analysis, by
J. Keith Nelson and J. A. Palmer.................................................................................. 5-33
Paper: Investigation for Standardization of Electrostatic Charging Tendency
Measurement of Transformer Oil in Japan, by Susumu Isaka, Masama Ikeda,
Hiroshi Miyao, Motoo Tsuchie, Toshitsugu Ishii, Shunichi Kobayashi, Takayuki
Kobayashi, Takahiro Ono and Hitoshi Okubo............................................................... 5-34
Session: Operating Experience of Transformers............................................................... 5-35
Paper: A Static Electrification Transformer Survey, J. Keith Nelson........................... 5-35
Paper: A Static Electrification Failure at the Navajo Generating Station, by
Thomas Lundquist ....................................................................................................... 5-36
Paper: Whitpain #1C Transformer Failure, by David F. Goodwin ............................... 5-37
Paper: Static Electrification Failure of MPT at Powerton Generating Station, by
Raymond F. Cameron and Christopher P. Stefanski .................................................... 5-37
Paper: Investigation of a Failed Westinghouse 500/345 kV 500 MVA Single Phase
Transformer by Don Angell and Kent Venosdel .......................................................... 5-38
Session: Monitoring and Management Methods ............................................................... 5-39

xxii
Paper: Static, the Ramapo Experience and What Can Be Done, by Chatham
Cooke .......................................................................................................................... 5-39
Paper: Mitigation of Streaming Electrification by Intelligent Control of Oil
Circulation, by J. Keith Nelson and J. A. Palmer ......................................................... 5-42
Panel Session: Are We there Yet?.................................................................................... 5-43
Paper: Mitigation of Static Electrification Discharges in Power Transformers,
Harold Moore ............................................................................................................... 5-43
Panel Presentation by Ray Cameron ........................................................................... 5-49
Panel Presentation by Professor Dr. Hitoshi Okubo ..................................................... 5-51
Panel Presentation by Tom Prevost ............................................................................. 5-52
Panel Presentation by Jean Poittevin........................................................................... 5-52
Panel Discussion ......................................................................................................... 5-54
Closing Remarks by Stig Nilsson............................................................................. 5-61
Chairmans Remarks ............................................................................................... 5-62
Closing Remarks: Stan Lindgren, EPRI........................................................................ 5-62

A FINAL PROGRAM FOR EPRI SYMPOSIUM ON STATIC ELECTRIFICATION..................A-1

B LIST OF REGISTRANTS BY NAME ...................................................................................B-1

C LISTS OF REGISTRANTS BY COMPANY .........................................................................C-1

xxiii
1
PANEL SESSION FOR PRESENTATION OF RAMAPO
TEST RESULTS

1-1
Panel Session for Presentation of Ramapo Test Results

1.1 Application of In Service Transformer Monitoring Systems during


Tests at Consolidated Edisons Ramapo Station
Harold Moore
Harold Moore & Associates
1062 Lakeway Drive
Niceville, FL 32578
Donald Chu
Consolidated Edison Company of New York
4 Irving Place
New York, NY 10003

Abstract
A paper presented at the EPRI February 1996 SUBSTATION EQUIPMENT DIAGNOSTICS
CONFERENCE IV titled Static Electrification Tests at Ramapo outlined results from static
electrification and monitoring tests made at Consolidated Edisons Ramapo Substation. These
tests included Series I in which only the pumps were run a short time after extensive field drying
was performed on the transformer. Series II tests were made several months after the field
drying. The results of this initial work indicated that much additional learning could be
accomplished by making additional tests. EPRI with assistance from Consolidated Edison then
sponsored two additional series of tests in 1996, the results of which are outlined in this paper.
Series III tests were conducted in September 1996 about 13 months after the Series II tests were
completed in 1995, and the final Series IV tests were made in October 1996.

Introduction
The 1996 program1 was successful in that static electrification discharges were produced under
different conditions, which enhanced the understanding of the electrification process. Detection
of the partial discharges was improved in some of the monitoring systems, which resulted in
information for improved detection of partial discharges and other information in operating
transformers.
The results of this program contain much valuable data, all of which cannot be reported in this
paper. In order to present an overview of this tremendous amount of data, the paper is organized
such that a presentation of the overall test results is made. A general summary of the learning
that resulted appears at the end of the paper.

Outline of Tests
For clarification, the conditions used during the tests are as follows:
Type A: Pumps only with no voltage at ambient temperatures.

Type A-1: Pumps only with no voltage with increasing temperature.

Type B: Pumps with voltage at ambient temperature.

Type C: Pumps with voltage and increasing temperature.

1-3
Panel Session for Presentation of Ramapo Test Results

Summary of Tests Reported in 1996 Paper

During the Series I tests using Type A conditions, (pumps and no voltage) made immediately
after the drying operation, the ambient temperatures were in the order of 8 to 15C. Some
discharges were detected, and the magnitude of the discharges increased as the ambient
temperature increased to around 15C. The charging effects in the oil as detected by the Nilsson
plate in the ERM Smart Manhole system were recorded, and some of the indications were
quite high at the low temperatures with pumps only.

The transformer set for 4 months after the Series I tests prior to the start of Series II. The
following tests were performed in Series II. Type A (pumps only), Type B (pumps and voltage at
ambient temperature), and Type C (pumps and voltage with increasing temperature). The
ambient were in the order of 16 - 23C during this series. Rather high level discharges, some of
which were audible, were measured when running all four pumps with no voltage at ambient
temperature.

Subsequently, with four pumps in operation with rated voltage on the transformer windings at
ambient temperature (Type B), the discharges were enhanced by an order of magnitude when
voltage was applied. Rather high level discharges occurred after only a few minutes. After about
45 minutes with pumps and voltage, a series of loud discharge type noises inside the transformer
were heard.

When tests with three pumps and voltage and increasing temperature were conducted (Type C),
high discharges were measured and loud discharge type noises were heard about one hour after
starting the test with the temperature at 25C. As expected, the discharge level decreased as the
temperature increased and was almost non-existent at 50C and higher.

The discharges were quite low when the Type C tests were repeated after the transformer system
had been up to 50C on the previous day. In fact, the discharge level remained low for all
subsequent tests after the transformer had been through the temperature cycle.

All of the partial discharge monitoring systems were effective with the ESEERCO sponsored
MIT electrical system being the most sensitive. However, the internal and external acoustic
systems did detect the discharges well before the magnitude became high or the discharges
became audible.

1996 Test Series - Series III and IV


The objectives of the 1996 program were as follows:

A. To further expand the understanding of the static electrification process.

1. Determine the effects of time and drying on the process.

2. Install an additional Nilsson plate at the bottom to enhance the understanding of the
charging process.

3. Evaluate the impact of heating cycles on the electrification process.

1-4
Panel Session for Presentation of Ramapo Test Results

B. Further evaluate different partial discharge detection systems, gas in oil monitors, and water
in oil sensors to enhance transformer monitoring and diagnostics.

1. Further evaluation of the MIT (ESEERCO) electrical partial discharge system connected
to bushings.

2. Evaluate different acoustic and other partial discharge detection systems.

3. Evaluate gas in oil and water in oil monitors as the system goes through the test program.

In order to meet these objectives, two sets of tests were conducted.

Series III: September 1996

This series was conducted to study the effects of long time setting on moisture equilibrium and
electrification processes. The transformer was allowed to set for 14 months after the Series II
tests were completed in July 1995. The transformer was not entered nor was any processing
performed during this period.

Series IV: October 1996

The transformer was drained, inspected, and another field drying operation was performed to
determine if a drying cycle would change the electrification processes. The tests started one
week after the field drying process was completed.

Test Setup for Series III and IV

The test setup was essentially the same as in Series I and II1. The 500/345 kVA, 333 MVA shell
form transformer was installed on a special pad at the Ramapo station. A 4.16 - 46 kV mobile
transformer was installed to energize the autotransformer to rated voltage. A 46 kV test bushing
was installed to energize the transformer through the induced test tap, which had been built into
the common winding of the transformer. Schematics of the setup are shown in Figures 1 and 2.

To provide the reactive excitation power for the transformer, shunt capacitor banks were
installed on the 4.16 kV side of the mobile transformer. The necessary disconnects and other
control devices were installed on the power supply.

Since many monitoring devices were installed in and on the transformer, provisions had to be
made for supplying power to the devices and for running all of the wiring to the instrumentation
trailer that was provided.

The electrification and relaxation processes that determine the static voltages, which result in
discharges, is a function of several variables including temperature and oil flow. External gas
heaters had been provided and installed under blankets to conserve the heat for Series II and
III tests. Consolidated Edison arranged for an in-line 200 kW, 50-gpm oil heater to be installed
for the Series IV tests for faster heating of the oil. It should be noted that the use of this heater
might have resulted in new learning about the electrification process.

1-5
Panel Session for Presentation of Ramapo Test Results

The sensors that were used for the Series III and IV 1996 tests were as follows:
Internal acoustic partial discharge waveguides installed at the top and bottom of the
transformer.
ABB TTI and J. W. Harley, Inc.
Increased number of external acoustic partial discharge sensors.
ABB TTI and J. W. Harley, Inc.
Electrical partial discharge sensors connected to the H, X, HoXo, and tap bushings.
MIT (ESEERCO)
EPRI smart manhole with Nilsson plate and MIT ACS sensors in a wall near the
top of the tank and in the bottom of the tank for Series IV.
ERM and EPRI
Leakage Current measurements
ABB TTI
Pump oil flow.
ABB TTI
Oil temperature
Various sensors including Luxtron fiber optic sensors in the insulation and oil flow
paths.
Photodiode and photomultiplier partial discharge sensors.
J. W. Harley, Inc.
Micromonitors and Syprotec gas in oil sensors.
Consolidated Edison
Water in oil sensor.
J. W. Harley, Inc.
On site water in oil analysis.
ABB TTI

Summary of Series III and IV Tests

Series III

Type A

This series started with an average oil temperature of 16C. The MIT electrical system started to
pick up signals 5 minutes after starting the pumps. The internal acoustic waveguides started to
detect discharges a few minutes later. Some rather high level discharges were detected on the
MIT system, and the internal top acoustic waveguides detected high signals. Rain occurred
during this series. The MIT system and the internal acoustic waveguides did not respond to the
rain, but the external acoustic sensors indicated signals, which were caused by the rain striking
the transformer tank.

1-6
Panel Session for Presentation of Ramapo Test Results

It is interesting to note that most of the discharges detected during this test were at the top of the
tank. The external heaters were turned on, and the magnitude of the discharges did increase as
the temperature increased from around 17C to 20C.

Type B, Second Day

Again, the discharges were enhanced by an order of magnitude when the voltage was applied to
the windings compared to pumps only. High level discharges were detected on the MIT electrical
system almost as soon as the voltage was applied. Very high level discharges were detected later
by the MIT system and the top waveguides.

Approximately 50 minutes after applying voltage, almost continuous discharges were detected
on the MIT system, the top internal acoustic, and some top external acoustic sensors.

Approximately one hour after applying the voltage with the temperature at 20C, a loud
discharge noise originating inside the tank, was heard. Another loud discharge noise was heard a
few minutes later. The pumps were turned off, and the audible discharges and high level
discharges detected by the sensors stopped.

Over the next 12 minutes three pumps were put back in operation. Almost immediately, the MIT
system and the internal and external acoustic sensors at the top indicated high level discharges.
Almost exactly one hour after the initial loud noises with the temperature at 24C, another loud
discharge originating from inside the tank, was heard. The sensor indication just prior to the
noises indicated the discharges at the top were much higher than at the bottom. The pumps were
turned off, and the discharge activity decreased to a low level.

Type C, Third Day

Voltage was applied with the pumps running. The MIT system indicated discharges almost
immediately after applying the voltage. After a few minutes, the top acoustic sensors again
indicated rather high level signals. One hour and 10 minutes after the voltage was applied, the
MIT system indicated high level signals similar to those prior to the loud noises on the previous
day. One hour and 13 minutes after applying voltage, one loud discharge noise was heard from
inside the tank. Two pumps were turned off for 5 minutes and then a third pump was started. Ten
minutes after the first loud noise, another loud discharge sound was heard. All three pumps were
left in service, and rather high level discharges were detected. However, there were no
indications of the high level discharges that had occurred just prior to the loud sounds.

All four pumps were put into operation when the temperature was 26C. The MIT system
indicated high level discharges almost immediately, and the top acoustic sensors had high level
signals. The MIT system was indicating a discharge pattern similar to the previous loud noises,
and the top acoustic sensors readings were building up in magnitude. Seven minutes after
starting the fourth pump, another loud discharge noise was heard. All pumps were left in
operation, and another loud noise was heard 35 minutes later. It was not as loud as the previous
loud noises on this date. The temperature was 28C at this time.

1-7
Panel Session for Presentation of Ramapo Test Results

The external gas heaters were started with one pump in operation and the discharges decreased
as the temperature increased. All four pumps were put into operation when the temperature
increased to 40C. There were a number of high level discharges, but they were not as high at
this temperature as on the previous day.

Summary Series III

The behavior during test Series III was very similar to that during Series II in July 1995. The
only real change was that the discharges after going through a temperature cycle were much
higher than in Series II. However, it should be noted that the temperature only reached 45C
when the Type C test was made during Series III; but it reached over 50C and was held at
temperatures over 50C for over one hour during Series II. This may account for the different
behavior during the Series III tests.

Series IV

The transformer was drained, inspected, dried and refilled with oil before the Series IV tests
were started. The objective was to determine the effect of drying on the electrification process.
It had been anticipated that partial discharge indication would be found on the insulation since
loud discharge sounds had been heard inside the tank, but no partial discharges tracks or trees
were found.

Type A

These tests were started with the oil temperature at 10C. The behavior was very similar to the
original Series I tests in March 1995 with pumps only at low temperatures. The Nilsson plate
readings were almost identical, and the MIT system indications were very similar.

Type A-1 tests were then conducted with some external heat, and the discharge activity increased
with the temperature. There were high level discharges on the MIT system when the temperature
reached 26C. The top acoustic sensors also indicated high level signals. The ERM smart
manhole partial discharge sensor indicated discharge activity as the temperature increased.

Type C, Second Day

The gas heaters and the external oil heater were used to increase the oil temperature. Shortly
after starting the flow of warm oil from the external oil heater during this test, the discharge
activity on the top acoustic sensors decreased noticeably. The flow from the heaters was stopped,
and the discharge activity as measured by the top acoustic sensors increased to the original
levels. After the system stabilized, the warm oil was again pumped into the top of the
transformer. The discharges as measured by the top acoustic sensors again deceased in
magnitude. This appeared to be an important finding. It was decided that this was best studied in
the scale model that will be used in another EPRI project and that the original plan for the work
at Ramapo should not be disrupted. The use of the external oil heater was discontinued.

1-8
Panel Session for Presentation of Ramapo Test Results

This series was conducted by starting the pumps one at a time until all four were in operation.
As expected, the discharge activity was a function of the number of pumps.

When all four pumps were in operation, the high level discharges started almost immediately.
The MIT system showed high level discharges and cascades, which indicated that very high
discharges would occur. After 42 minutes with all four pumps in operation, two loud discharge
type noises were heard. Seven minutes later, there were two loud discharges two minutes apart.
The noises were not as loud as previous high level audible discharges. Four minutes later, there
was another loud but lower level internal discharge noise.

Eleven minutes later, a very loud discharge type noise was heard. A number of sensors and
monitoring devices were made inoperative by the transients that occurred during these loud
noises. These included the Micromonitors and Syprotec gas in oil monitors.

For the next 1 hour and 5 minutes, there were a number of the loud but lower level discharges.
After a short time, the magnitude of the discharges were so high on the MIT system that it was
decided to shut off the pumps because there was concern that the transformer might fail or all of
the equipment would be disabled if even higher level discharges were to occur.

The tests were continued later with reduced numbers of pumps and added heating of the oil.
There were some discharges as the oil was heated to higher temperatures, but there were no high
magnitudes. The oil heating continued until the oil had been above 40C for about 30 minutes.
The pumps, voltage and external heaters were turned off to let the transformer cool.

Type B on Day 3

The Type B tests with pumps and voltage were repeated for 2 hours to determine the effects of a
heating cycle on the electrification process. The MIT system and other sensors indicated some
discharges, but the magnitudes were not high for the first hour. However, the MIT system started
to indicate the type discharges that resulted in the loud noises. The oil temperature was in the
range of 20-21.5C during this test.

Type C on Day 3

The external gas heaters were activated 2 hours after starting the Type B test. Eleven minutes
later, there was a loud noise which disabled the bottom smart manhole and a number of the
acoustic sensors and recording devices. The pumps were left in operation, and there were a
number of high level events on the MIT system, but there were no additional loud noises.

Type C on Day 4

This test was repeated after the oil had cooled.

The procedure was changed slightly in that all four pumps were run for one hour and 20 minutes
before applying the voltage. The behavior was similar to the previous day. One loud but lower
level noise was heard 17 minutes after applying the voltage. No other loud noises were heard.

1-9
Panel Session for Presentation of Ramapo Test Results

Overview of Program Results

The objectives of the program were achieved. A summary of the results is as follows:

Static Electrification

A. The behavior patterns of the transformer system during the static electrification tests were
almost identical in July 1995 and September 1996. In both of these series, the transformer
had set for several months. The total water in the system would have been the same since no
processing or other changes occurred during the set periods from the first processing and
between the Series II tests in July 1995 and the Series III tests in September 1996. In both
series of tests, high level discharges with audible loud bangs occurred.

B. Processing the transformer changed the static electrification process. The behavior of the
charge generation and relaxation were almost identical during the initial tests in March 1995
which were conducted immediately after drying and in October 1996 which were conducted
one week after drying. The results confirm that changing the water equilibrium in the system
changes the electrification process.

C. Flowing warm oil over the top of the transformer seemed to change the electrification
process particularly at the top of the transformer.

D. The time constant for producing enough static voltage to cause loud audible discharges
appears to be in the order of 50 to 60 minutes in this transformer.

E. The maximum discharges occurred in a temperature range of 20 to 30C although some loud
discharge type noises were heard up to temperatures approaching 35C. However, it should
be noted that the temperature range where maximum discharges occur is believed to be a
function of the oil, the water equilibrium and other conditions of the transformer insulation,
and other variables. Therefore, a different electrification pattern as a function of temperature
will probably exist for other transformers and system conditions.

F. The dismantling report shows that a failure occurred between the shield on the high voltage
bushing to the neutral lead by creepage over lead support structures. This is an indication that
concentrations of charge around the high voltage bushing resulted in very high potentials,
which initiated the failure path found.

THIS IS THE IDENTICAL FAILURE PATTERN, WHICH OCCURRED IN A NUMBER


OF UNEXPLAINED SHELL FORM FAILURES. One result of this work is that it
explains some of the past failures.

G. The questions arise as to why only EHV transformers are involved in the failures and why
does it involve some of the high voltage bushings but not others. An analysis of the Ramapo
data indicates that most of the high level audible discharges were at the top of the transformer
and location tests showed that they were near the bottom shield on the high voltage bushing.
The answers to both questions appear to be as follows:

1-10
Panel Session for Presentation of Ramapo Test Results

1. The total charge at the top of the transformer is a function of the charges separated in the
oil as it passes through the insulation structure. This in turn is a function of the number of
insulation sheets in the high to low space between windings which the oil passes as it
flows from the bottom to the top end. The number of insulation sheets in a 345 kV
transformer is approximately 50 % greater than in a 230 kV shell design. Therefore, it
appears that the number of sheets combined with the oil flow through the sheets in this
shell design is critical for producing static electrification.

2. The shield at the lower end of the high voltage bushing is usually located over the end of
the sheets between the high and low voltage windings so that it is in the region where the
maximum charges in the oil occur.

Performance of Sensors and Monitors

Partial Discharge Detection

The results demonstrated that the partial discharge detection systems detected the discharges in
the incipient stages well before the discharge level became great or damaging. The relative
sensitivity and other factors are discussed below.

A. MIT electrical discharge system was the most sensitive. It also gave early warning of
major discharges that occurred. This system connected to the bushings made it possible to
determine the nature of the discharges and the trends as the tests progressed and in most
cases increased in intensity. Since the sensors are connected to different winding terminals,
it should be possible to determine the location of the discharge source. This system is non-
obstructive and can be installed without disturbing the internal parts of transformers.

B. The internal acoustic waveguides were the second most sensitive. They are much less
responsive to external sources such as rain or sleet than external acoustic sensors. The
acoustic rods can be installed in existing transformers if valves are located at appropriate
places on the cover using glands to seal around the rod. The rod is an insulating material, but
it is not as good as transformer insulation so that location of the rods in existing transformers
must be done by transformer engineers.

C. The external acoustic sensors were not as sensitive as the internal acoustic sensors, but the
results demonstrated that they had the sensitivity to detect discharges in the incipient stages
well before they become damaging.

The use of multiple sensors on the tank increased the sensitivity of the system and made it
possible to locate the source of the discharges. They are susceptible to external noises.
However, the advantage is that they can be installed on transformers in service without
disturbing the internal parts. In fact, they can be installed on energized transformers if they
are located on the tank away from high voltage areas. They are of interest for making tests on
transformers in service.

1-11
Panel Session for Presentation of Ramapo Test Results

D. The EPRI Smart Manhole has some interesting possibilities for monitoring static
electrification. The partial discharge feature also indicated possibilities for detecting
discharges in general.

E. The other sensors such as the photo diodes and the photo multipliers showed promise for
further development as partial discharge sensors.

F. The test program showed the need for further field hardening of some sensors and
monitors. A number were made inoperative by the transients that occurred during the
loud bangs.

Water and Gas in Oil Monitors

The water in oil monitor appeared to result in good data. There were some differences compared
to laboratory measurements. However, the results did produce useful data.

A Micromonitors gas in oil monitor and a Syprotec Hydran Hydrogen sensor were installed for
the Series IV tests because it was desirable to monitor the hydrogen as the static electrification
discharges occurred. No meaningful data was obtained during the first sets of tests with pumps
only when low-level discharges occurred. Unfortunately, both sensors became inoperative after
one of the loud bangs in the transformer so that it was not possible to obtain the desired data as
this series of tests was conducted. Additional protection for substation faults and flashovers is
needed.

Diagnostic Systems

The results of this program demonstrated that diagnostic capability exits using these sensors for
analysis of static electrification discharges in operating transformers and discharges that occur as
the result of other weakness or phenomena in the transformer system.

If discharges occur under the following conditions, the probability is very high that static
electrification is occurring.
The temperatures of the transformer insulation system are in the order of 15 to 40C.
All pumps or most of the pumps are in operation with voltage on the transformer.
The discharges occur at intervals rather than on a continuing basis.

A simple model could be developed using partial discharge sensors, oil temperature sensors that
are already available, and pump oil-flow sensors that are already on the transformer.

1-12
Panel Session for Presentation of Ramapo Test Results

This work has demonstrated that the partial discharge sensors have the capability for making
diagnostic tests on transformers in service. This can be illustrated as follows:
The MIT Electrical System has demonstrated the sensitivity for detecting discharges
in operating transformers. It could be used to make tests on transformers in service to
determine the existence and magnitude of discharges. Discharge site location
capability also exists. If needed, the devices can be installed on transformers in
service to determine trends.
Internal waveguide acoustic rods installed into the oil have good sensitivity, and these
recent tests at Ramapo demonstrated that they can be made to be largely insensitive to
external noises such as rain. It is not practical to install these devices in operating
transformers unless valves exist at the needed locations, but they can be installed in
new transformers or in transformers on which there is reason to install on-line
devices.
External acoustic sensors that were used at Ramapo had good sensitivity. They can be
installed on operating transformers since they are non-obtrusive.
A number of sensors can be located around the transformer to increase the sensitivity
and for location purposes. This work demonstrated the value of this approach for
making partial discharge diagnostic tests on transformers that have hydrogen gassing
or other reasons to suspect partial discharges. They can be used for one time
diagnostic tests or as on-line sensors.

Unfortunately, it was not possible to relate the gases evolved from the oil on an ongoing basis,
due to the failure of the gas in oil sensors. However, the use of the partial discharge sensors in
conjunction with gas in oil monitors results in more powerful diagnostic systems. For example,
if a transformer is generating hydrogen and one or more of the partial discharge sensors detects
partial discharges, it is clear that a potentially dangerous situation exists in the transformer.
Further diagnostic tests such as partial discharge location tests can then be made. The electrical
and acoustic sensors demonstrated the ability to determine trends in the partial discharge
behavior, which is essential for diagnostic systems.

Partial discharge monitoring and diagnostic systems must have the following essential features.

1. The sensors must be reliable and furnish meaningful data.

2. The system must not respond to events that are not partial discharge initiated.

3. The system must give early warning to make operator intervention possible before permanent
damage or failure occurs.

4. The system must be simple and protected for system transient conditions.

In summary, this work has demonstrated that technology and devices exist for making diagnostic
tests on transformers in service. The devices can be installed as on-line monitors or used to make
diagnostic tests on transformers that are suspected of having a partial discharge problem.

1-13
Panel Session for Presentation of Ramapo Test Results

Summary and Recommendations

From monitoring and diagnostic viewpoints, it is appropriate for utilities to use the techniques
used in the Ramapo program for making diagnostic tests on operating transformers when
potential problems are suspected. It is necessary to utilize persons in diagnostic applications that
not only understand the monitoring systems but also have an appreciation for the behavior of
partial discharges in transformer insulation systems.

It was concluded that additional work on static electrification using this full size transformer
could not be justified. However, the input of these results into the planned scaled model tests will
be very beneficial to further understanding of this complex process. This work again confirmed
the need to take reasonable precautions such as using temperature controls to activate pumps on
transformers in service to minimize the probability of failure from static electrification.

In spite of this and past work, some utilities continue to operate pumps when the load is low
and/or oil temperatures are in the range where static electrification is at maximum.

Acknowledgement

The authors wish to thank the many persons from Consolidated Edison, who assisted with the
test set up and conducting the tests. We appreciate the efforts of ABB during the processing and
the use of equipment during the tests. The use of information from tests made by TU Electric for
planning and execution of these tests is also greatly appreciated.

This paper contains material from a paper of the same name authored by Harold Moore, Harold
Moore & Associates, Donald Chu, Consolidated Edison; Dr. Chatham Cooke, MIT; William
Gish, Electric Research & Management; Dr. Michael Brubaker, then of ABB TTI; Stig Nilsson,
Consultant; George Frimpong, ABB TTI; and John Harley, J. W. Harley, Inc. at the EPRI
Substation Equipment Diagnostics Conference V.

Reference

1. EPRI Report TR-111282, September 1998, Proceedings: Substation Equipment Diagnostics


Conference V, pages II-10 through II-46.

1-14
Panel Session for Presentation of Ramapo Test Results

500 kV

4/33 kV Mobile
345 kV
2.5 ohm
Test Tap

100 kVAR

Figure 1
Ramapo Static Project

Supply from Station

Temporary
Bushing

Supply, Mobile to
Transformer

Instrumentation
Trailer
Figure 2

1-15
Panel Session for Presentation of Ramapo Test Results

1.2 Acoustic Partial Discharge and Leakage Current Measurement Results


of Ramapo Tests
George K. Frimpong and J. Michael Walden
ABB Electric Systems Technology Institute
Raleigh, North Carolina 27606

Stan Lindgren
EPRI-PDG
Palo Alto, CA 94303

Abstract
This paper summarizes the results of the data collected by ABB Electric systems Technology
Institute during the large-scale static electrification tests performed at Con Edisons Ramapo
substation. During these tests, ABB monitored acoustic partial discharge sensors, a leakage
current meter, and temperature sensors, pump flow meters, a relative humidity meter and a set of
video cameras looking into the transformer. The results collected from these sensors during a
representative set of measurements are described in this paper. The performance of the pertinent
instruments for the prediction of static electrification in transformers is presented. In general, it
was observed that the acoustic partial discharge sensors were effective in detecting the
occurrence of electrostatic discharges. The leakage current sensor was useful in understanding
the charge accumulation in combination with other sensors, however, it may not be effective in
indicating the dangerous conditions that can result from static electrification in transformers.
Both instruments indicated that the most significant activity occurred with the combination of all
four pumps, low temperature and AC energization and significantly less activity when any fewer
pumps are used. For completeness, a description of the sensors and their setup are presented.

Acoustic Discharge Detection Scheme

Overview of Acoustic Partial Discharge Detection System


The ABB acoustic detection system was comprised of a Physical Acoustics R15I transducer with
built-in integral pre-amplification and a bandwidth extending from 70-200kHz, a peak and hold
amplifier and event counting circuits. A schematic diagram is shown in the Figure 1. The
acoustic sensors were either attached to the external tank wall with a magnetic mount or they
were attached to an internally mounted waveguide. The signal from the transducer was first
filtered to remove a DC offset. The signal was then amplified and connected in parallel to the
peak hold and the counting circuits. The peak hold is designed to capture the largest single
acoustic signal and hold the level until the recorder samples it. It has a fast response and a slow
decay to allow for sampling the peak value accurately. The counting circuitry is designed to tally
the number of events above a set threshold during each one-minute sampling interval. The count
was then converted to a proportional voltage level, which simply represents the amount of
activity during the sampling interval. Both of these measurements are important in making the
correct analysis of the discharge activity occurring inside the transformer. Figure 2 shows the
layout of the test site, pumps locations on the transformer and orientations used for identifying
the acoustic sensor locations as shown in Figure 3, Figure 4 and Figure 5.

1-16
Panel Session for Presentation of Ramapo Test Results

Power Supply/ 32 dB Peak Detector


Amplifier
Filter Recorder
Transformer Acoustic
Sensor
Tank Wall Counter Circuit

Serial port

Oscilloscope PC
(Optional)

Figure 1
Acoustic Partial Discharge Detection System

2 3

EAST Transformer
WEST

1 4

NEAR (NORTH)

TRAILER

Figure 2
Top View of Test Site

1-17
Panel Session for Presentation of Ramapo Test Results

104 75
Near Top Ext.
33
Far Top Ext.
FAR 35 NEAR

31 31
69

4.5 6.5

Figure 3
External Acoustic Sensor Layout (East Side)

West East

33.5 35.5

4.5

(This waveguide was recessed in a wet tap valve)

Figure 4
Far Bottom Acoustic Waveguide Placement

1-18
Panel Session for Presentation of Ramapo Test Results

East West

38 31

4.5

(Extends 8 Pass The Tank Wall, but behind boxing


Insulation)

Figure 5
Near Bottom Acoustic Waveguide Placement

Performance of Acoustic Partial Discharge System

The acoustic system was effective in detecting discharge activity in the transformer. The internal
rods were usually more sensitive than the external sensors for detecting discharge activity inside
the unit. The peak signals of the external sensors were more sensitive to external disturbances
(e.g. rain drops) than the internal rods. However, the frequency of raindrop impacts is in general
much lower than discharge activity, so the count measurements were not affected to any great
extent by external disturbances. An effective acoustic detection scheme must therefore
incorporate both the peak and count signals of the acoustic signal.

More activity was detected with the upper plenum acoustic sensors. It may be that the sensors
were located close to a critical area or that there was less of an acoustic barrier between the
sensor and the source of the activity. Discharges occurring deep within the winding may not be
detectable with acoustic sensors.

The first graph in Figure 6 shows the parameters of a test sequence for sample date in
October 1996. This graph shows at each minute of the day the number of the pumps operating,
both top and bottom oil temperatures and the AC energization of the transformer. The graph of
the test parameters is included with each set of data for reference. Figure 6 and Figure 7
respectively show the peaks and counts for each of the acoustic sensors located at the top part of
the transformer and Figure 8 and Figure 9 show similar data for the sensors located toward the
bottom part of the transformer.

1-19
Panel Session for Presentation of Ramapo Test Results

60 5
Top RTD
50

Temperature (C)
4 Bot. RTD

# of Pumps
40
3 Pumps
30 AC
2
20
10 1
Energized
0 0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Peak Voltage Level

3
Near Top Ext

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Peak Voltage Level

3
Far Top Ext

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Peak Voltage Level

3
Top South Ext

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Peak Voltage Level

3
Top W.G.

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

Figure 6
Top Acoustic Sensor Peak Signals for 10/16/96

1-20
Panel Session for Presentation of Ramapo Test Results

60 5
Top RTD
50

Temperature (C)
4 Bot. RTD

# of Pumps
40
3 Pumps
30 AC
2
20
10 1
Energized
0 0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Proportional Count

3
Near Top Ext

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Proportional Count

3
Far Top Ext

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Proportional Count

3
Top South Ext

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Proportional Count

3
Top W.G.

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

Figure 7
Top Acoustic Sensor Counts for 10/16/96

1-21
Panel Session for Presentation of Ramapo Test Results

60 5
Top RTD
50

Temperature (C)
4 Bot. RTD

# of Pumps
40
3 Pumps
30 AC
2
20
10 1
Energized
0 0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Peak Voltage Level

3
Bottom Near Ext

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Peak Voltage Level

3
Near Bottom W.G.

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
4
Peak Voltage Level

3
Far Bottom W.G.

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

Figure 8
Bottom Acoustic Sensor Peak Signals for 10/16/96

1-22
Panel Session for Presentation of Ramapo Test Results

60 5
Top RTD
50

Temperature (C)
4 Bot. RTD

# of Pumps
40
3 Pumps
30 AC
2
20
10 1
Energized
0 0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
0.4
Proportional Count

0.3
Bottom Near Ext

0.2

0.1

0.0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
0.4
Proportional Count

0.3
Near Bottom W.G.

0.2

0.1

0.0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
0.4
Proportional Count

0.3
Far Bottom W.G.

0.2

0.1

0.0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

Figure 9
Bottom Acoustics Sensor Counts for 10/16/96

1-23
Panel Session for Presentation of Ramapo Test Results

From the data presented above and those collected for other days of testing, the following
general observations can be made:
1. High peak values with corresponding low counts were not dangerous. This type of activity
results from gradual static charge accumulation which periodically establishes a critical field
for local discharge.
2. High peak values with virtually no count could be attributed to either external noise or
discharge activity similar to that described in 1.
3. Increasingly high peaks with corresponding increasingly high counts eventually led to loud
dangerous discharges. This behavior was shown to be very consistent when approaching
failure.
4. Very good correlation between the acoustic sensors and the MIT electrical partial discharge
detection system was observed, particularly with respect to the upper plenum waveguides.
5. The most significant activity occurred with the combination of four pumps running, low
temperature and AC energization.
6. More discharge activity was detected with the upper plenum acoustic sensors.
7. The waveguides were generally more sensitive than external clamp-on sensors.
8. The same level of activity for both peak and count, was measured both during the heat-up
and cool-down cycles of the transformer.

Winding Leakage Current System

System Description
The winding of the auto-transformer was used as a collector electrode for DC leakage current
measurements. An IFI low pass filter and electrometer system was obtained from ALF
Engineering in Switzerland for implementation in the test circuit as shown in Figure 10.

HV

Tank

SUPPLY N

Con Ed. Spark Gap 400uF

METER

IFI

Figure 10
Leakage Current Measurement System

1-24
Panel Session for Presentation of Ramapo Test Results

Note that the entire energization circuit was isolated and grounded through the leakage current
system to ensure that no stray leakage paths were available outside the transformer.

Performance of Leakage Current Measurement System

The leakage current is an integral measurement of charge generation and relaxation within the
transformer winding oil ducts. Charges will be separated locally due to agitated flow of oil
against the cellulose surfaces. Typically, the oil will become positively charged, while the
cellulose becomes negatively charged. The positive charges in the oil relax to the windings and
tank surfaces. The polarity of the winding leakage current will depend on the relative amounts of
negative charges leaking through the cellulose and the positive charges relaxing from the oil to
the winding. Hence, the measurement on the winding, which acts as a large integrating electrode,
can be viewed as the difference between relaxation and generation of charge to the winding.

The leakage current responds to flow-rate, applied voltage and temperature. When the pumps are
activated, the current exhibits an abrupt change followed by an exponential decay to a steady
state value. The effect of temperature on leakage current results from the temperature
dependence of the transformer oil/cellulose insulation conductivity. The relaxation time constant
of the oil is inversely proportional to the thermally enhanced conductivity and hence decreases
with temperature. Consequently, positive charges in the oil relax to the winding and the tank wall
at a faster rate, so the leakage current tends toward an increasing positive value as the
temperature rises. Although the effect is not as pronounced, AC energization tends to increase
the leakage current. This could be attributed to the fact that an applied field enhances charge
separation at the oil/cellulose interface.

These observations are evident in Figure 11. When all the pumps are started, the leakage current
experiences a sharp spike and decay over 20 to 30 minutes. When the transformer is energized
with all four pumps, the leakage current is enhanced, but AC alone has no effect on the leakage
current. The occasional spikes on the leakage current were caused by large discharges in the
transformer that coincided with the sampling of the leakage current.

1-25
Panel Session for Presentation of Ramapo Test Results

35 10
Top RTD
30

Temperature (C)
8 Bot. RTD
25

# of Pumps
6 Pumps
20
15 AC
4
10
2
5
Energized
0 0
6:00 8:00 10:00 12:00 14:00 16:00 18:00
4
2
Leakage Current (A)

0
IFI leakage
-2
-4
-6
-8
-10
-12
6:00 8:00 10:00 12:00 14:00 16:00 18:00

Figure 11
IFI Leakage Current Measurement for 7/11/95 Tests

60 5
Top RTD
50
Temperature (C)

4 Bot. RTD
40 # of Pumps Pumps
3
30 AC
2
20
10 1
Energized
0 0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
14
12
Leakage Current (A)

10
IFI leakage
8
6
4
2
0
-2
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

Figure 12
IFI Leakage Current Measurement for 10/16/96

Typically, the winding leakage current tends to mirror the behavior exhibited by the upper
plenum Nilsson plate. This was very consistent throughout the 1995 test series where the
winding current was negative and the plate current positive (Figure 11). However, during the
1996 tests (Figure 12), the winding current exhibited an relatively large positive offset such that
the currents were normally positive while showing similar transient behavior to the 1995 testing.

1-26
Panel Session for Presentation of Ramapo Test Results

This bias may be related to modified leakage paths due to moisture equilibration, a shift in the
balance between generation and relaxation, or possibly due to a problem with the
instrumentation. In any case, the transient behavior is similar to the 1995 testing so a few
observations can still be made regarding the leakage current:

1. When the pumps are switched on, the current jumps to a negative peak value and then decays
exponentially back to a steady state value.

2. The maximum leakage current is observed with all pumps operating and is significantly less
when any fewer pumps are used.

3. The leakage current is enhanced by energization, but this is a second order effect relative to
oil flow.

4. The temperature peak observed at 35 to 40C for partial discharge activity is apparent in the
behavior of the winding current.

5. The winding current typically mirrors the behavior of the top Nilsson plate current.

Relative Humidity and General Oil Tests

Humidity Sensor

During the testing, ABB monitored a Harley Moisture Sensor that was installed in a pipe at one
of the radiators. The transformer was effectively dry for all stages of testing. Figure 13 shows the
humidity measurements from 10/16/96. The moisture content in the oil changes over the course
of a test because the moisture migrates between the oil and the cellulose insulation. The relative
humidity was generally higher in the early morning and reached equilibrium soon after the
pumps were started. The relative humidity of the oil stayed nearly the same for the rest of the day
as the temperature rose. However, the total moisture content is actually higher at the higher
temperatures with the same relative humidity, because the saturation of the oil rises with
temperature. External gas heaters were used to heat the radiators during the morning of this test.

The transformer was only heated for about 7 hours in this test, which is not enough time to drive
a significant amount of moisture out of the insulation.

1-27
Panel Session for Presentation of Ramapo Test Results

60 5
Top RTD
50

Temperature (C)
4 Bot. RTD

# of Pumps
40
3 Pumps
30 AC
2
20
10 1
Energized
0 0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00
12
Relative Humidity (%RH)

10

8 Harley Moisture

0
8:00 10:00 12:00 14:00 16:00 18:00 20:00 22:00

Figure 13
Relative Humidity Measurements

Oil Tests

In addition to the relative humidity sensor, samples were taken once or twice a day during the
experiments to be tested for water content using an on-site Karl Fischer titration system. At the
same time, oil samples were taken for general oil quality tests.

The oil tests for 10/16/96 are shown in Table 1. A sample was taken at the beginning of the day
with oil temperature at 6.5C and late in the day with oil temperature at 48C. The oil is shown
to be in good condition and to have a relative low electrostatic charging tendency.
Table 1
Ramapo Oil Quality Tests

Date/Time Collected 10/16/96, 08:20 10/16/96, 19:45


Bottom oil temperature at time of sample (C) 6.5 48
On-site Moisture Measurement (ppm) 3.34 7.41
Laboratory results
General Condition Clean Clean
Dielectric Strength (kV) 52.8 48.2
Power Factor (%) @ 60 Hz/25C 0.009 0.019
Neutralization No. (mg. KOH/g) 0.012 0.012
Moisture Content (PPM) 6.58 7.47
Conductivity (pico-Siemens/meter) 0.97 0.97
Electrostatic Charging Tendency (C/m3) 104 143

1-28
Panel Session for Presentation of Ramapo Test Results

Summary and Conclusions

The overall conclusion on the acoustic discharge sensing is that it can be effectively used to
predict failures due to static electrification. However, two components of the acoustic signal, the
frequency of occurrence (counts) and the peak of the signal, must be analyzed concurrently for
proper diagnosis to be made. A trend that shows increasing high peaks with associated increasing
counts may point to impending danger to the transformer. The next level of concern will be high
peaks with associated low counts, which may point to single isolated high level discharges that
are easily dissipated. It is important to note that the same levels of acoustic discharge activity can
be present during cool-down of the transformer as during heat-up. Therefore, any operational
procedures that are instituted to suppress static electrification discharges when a transformers oil
temperature is increasing must also be present when the temperature is decreasing below the
critical temperature.

Although leakage current measurements can also indicate static electrification in a transformer,
its use for predicting impending failure is not quite clear. The difficulty may lie in the fact that is
an integrative representation of the charge generation and relaxation processes, which are in
competition with each other under all conditions.

1-29
Panel Session for Presentation of Ramapo Test Results

1.3 Ramapo Static Electrification Project Test Results from J. W. Harley Inc.
J. C. Harley, Ph. D.
J. W. Harley Inc.
9177 Dutton Dr.
Twinsburg, OH 44087

Abstract
J. W. Harley Inc. participated in the Ramapo Static Electrification project, conducted at
Consolidated Edisons Ramapo substation from 3/27/95 to 10/18/96. The project was conducted
to experimentally observe static electrification processes occurring in a full scale power
transformer under conditions that closely approximated actual field conditions. Measurements of
acoustic energy from 11 different acoustic partial discharge sensors, a Harley Moisture Sensor,
two photomultiplier tubes, and several other sensors was recorded using a Harley TPAS system.
Experimental observations and test results are presented here. Results indicate that there is a
dependence of static electrification on number of pumps in operation, energization of the unit,
position within the transformer, oil temperature, and oil moisture level. Because of the
impossibility of controlling independent variables separately under field conditions, only general
trends in the data are presented here.

Introduction
The static electrification process has long been suspected of being one type of failure mechanism
of large power transformers. Units that have failed under circumstances that lead investigators to
suspect static electrification processes show severe carbon tracking of the paper insulation
system, indicating breakdown of the insulation system over a large area. Fortunately, despite the
severity of this failure mechanism, it appears to be isolated to only certain combinations of
operational parameters, and only in certain types of transformers. However, opportunities to
experimentally measure and observe the static electrification process as it occurs under field
conditions are quite rare. Consolidated Edison experienced failure of two Westinghouse
345/500 KV 336 MVA single phase autotransformers at the Ramapo substation. Since the third
unit was going to be replaced anyway, it was decided that this presented an ideal opportunity to
perform tests on a large power transfomer that was susceptible to static electrification. The test
unit was taken off-line and placed on a special temporary pad near the edge of the substation.
Provision was made to re-energize the transformer by re-connecting the test tap to an external
bushing and applying 33 KV to the test lead. This allowed energization to the full 345/500 KV.
Numerous test ports were also cut in the tank wall to allow for instrumentation. Three days prior
to the first set of tests, the oil (Shell Diala-A) was vacuum dried to Consolidated Edison
specifications by an ABB crew.

Experimental Instrumentation

The transformer tested was a 1969 Westinghouse 345/525 KV, 336 MVA single phase
autotransformer. The transformer was heavily instrumented with sensors from four primary
investigative teams: ABB, ERM, J. W. Harley Inc, and MIT. The Harley instrumentation
consisted of the array of sensors shown in Table 1, connected into a Harley TPAS system.

1-30
Panel Session for Presentation of Ramapo Test Results

Table 1
Sensor types and locations

Sensor Designation Sensor Type Location


PD-1 Rod Mount Acoustic (PA R15I) Top
PD-2 Rod Mount Acoustic (PA R15I) Top
PD-3 External Acoustic (PA R15I) Bottom
PD-4 External Acoustic (PA R15I) Bottom
PD-5 Internal Acoustic (Harley Fisheye) Bottom
PD-6 External Acoustic (Harley) Bottom
PD-7 External Acoustic (Harley) Bottom
PD-8 Internal Acoustic (Harley Fisheye) Bottom
PD-9 External Acoustic (Harley) Top
PD-10 External Acoustic (Harley) Top
PD-Ambient External Acoustic (PA R15I) Not on the tank
PMT I Photomultiplier Tube Bottom
PMT II Photomultiplier Tube Top
PDS Photodiode Sensor Bottom
HMS Harley Moisture Sensor Loop
TopOil RTD Top
BotOil RTD Bottom
LoopOil RTD Loop
Ambient RTD Not on the tank

In addition to the data automatically recorded by the TPAS system, a large set of data in the form
of signal waveforms was obtained using an oscilloscope and a high speed data acquisition card.
The response of the PMTs and PDS were also interrogated using the oscilloscope. In order to
reduce the overall amount of data to a reasonable level, only combined results for all top and all
bottom sensors grouped together are shown here. For more detailed information, the reader is
1
referred to .

Non-Energized Test Results

Testing was performed from 3/27/95 to 3/29/95 and on 7/10/95 with the unit completely
de-energized. The only mechanism for creating static electrification within the transformer was
the action of the pumps. Under these conditions, very energetic events were observed with all
types of internal and external acoustic sensors, as well as with photo-multiplier light sensors
mounted inside the tank. By plotting the level of PD activity as a function of the number of
pumps that are on, the effect of the pump on/off state can be seen.

1
J. C. Harley, Harley Static Electrification Test Report for tests conducted at Consolidated Edisons Ramapo
Substation from 3/27/95 to 10/18/96, 55 pp., 1/16/97, J. W. Harley Inc, Twinsburg, Ohio.

1-31
Panel Session for Presentation of Ramapo Test Results

1800
1600 27-March
1400 28-March
1200 29-March

PD Counts
1000 10-July
800
600
400
200
0
0 1 2 3 4
# of Pumps On

Figure 1
Bottom Acoustic Activity vs. # of Pumps On

600
27-March
500 28-March
400 29-March
PD Counts

10-July
300
200
100
0
0 1 2 3 4
# of Pumps On

Figure 2
Top Acoustic Activity vs. # of Pumps On

From Figure 1 and Figure 2, it is evident that very energetic events were detected only when 3 or
4 pumps were being run. Very little activity was noted with only 1 or 2 pumps running. Also
note that the most energetic events were concentrated in the bottom of the unit. Temperature also
plays an important role in the static electrification process, as shown below.

1-32
Panel Session for Presentation of Ramapo Test Results

1800
1600 27-March
1400 28-March
1200

PD Counts
29-March
1000 10-Jul
800
600
400
200
0
0 10 20 30
Temp, Deg. C

Figure 3
PD Activity vs. Top Oil Temp

Figure 3 indicates that for all acoustic events recorded in the transformer, both top and bottom,
there is a correlation with temperature. The most energetic events occur at lower temperatures.
While the scatter in the temperature data is large, the peak activity appears to be between 8 and
12 degrees C. Unfortunately, the non-energized data set is too small to establish the true nature
of the correlation here, as most of the non-energized tests were run at low temperatures, so the
results may be skewed by the limited data at higher temperatures. However, during the energized
tests, short periods when the unit was not energized substantiate the fact that for temperatures
above 40 degrees, there is very little static activity.

1800
27-March
1600
1400 28-March
1200 29-March
PD Counts

1000 10-July: no data


800
600
400
200
0
0 2 4 6
% Relative Humidity

Figure 4
PD Activity vs. % Relative Humidity

Figure 4 demonstrates that the relationship between static electrification and relative humidity is
significant. Again, the data set is too limited to determine the exact nature of the relationship.
The Ramapo transformer was extremely dry and hence, at low temperatures especially, there is a
very small range of values of relative humidity that were available. No moisture data was
available on the 10th of July due to problems with the sensor.

1-33
Panel Session for Presentation of Ramapo Test Results

In addition to the automatically collected TPAS recording of acoustic counts for each sensor,
a great deal of data was taken using a digital oscilloscope to examine the response of various
sensors. This soft of data is shown for two different static events that occurred on 3/28/95 and
3/29/95 in Figure 5 and Figure 6.

Figure 5
Static Electrification Event. Time Stamp: 3/28/95 12:44. Record 44. Ch1=PD-9, ch2=PD-8,
ch3=PD-5, ch4=PMT, M1=FFT of ch1, M2=FFT of ch2, M3=FFT of ch3

Figure 6
Static Electrification Event. Time Stamp: 3/29/95 11:31. Record 55. Ch1=PD-1, ch2=PD-8,
ch3=PD-3, ch4=PMT, M1=FFT of ch1, M2=FFT of ch2, M3=FFT of ch3

1-34
Panel Session for Presentation of Ramapo Test Results

Figure 5 shows the response of several of the sensors to a large static event. This event was heard
as an audible click by both Harold Moore and Donald Chu. In addition to the acoustic detection
of the event, the photomultiplier tube (PMT-I) also recorded this event, as shown by the small
spike-like signal trace that precedes the acoustic signals. Events like this were energetic enough
to totally saturate the acoustic input circuitry used for the recording of counts. Curiously, the
event shown in Figure 5 only produced a maximal acoustic count of 398 on sensor PD-9, here
shown fully saturated. This discrepancy was not noted until after the first set of experiments was
completed. When it was noted, it lead to an improvement in the data acquisition system that was
implemented in the data starting in 1996 and that lead to a 7 to 8 fold improvement in the
acoustic sensitivity of the TPAS system. Figure 6 also shows an event for which the
photomultiplier tube obtained a signal. Numerous events like this, similarly energetic, were
recorded using the oscilloscope during the non-energized portion of the testing. It is not clear if
the photomultiplier tube was actually responding to the presence of light within the transformer,
or simply to an electromagnetic pulse created by the static event.

Energized Test Results

Tests were conducted with the unit energized to full rated voltage from 7/11/95 to 7/14/95, from
9/17/96 to 9/19/96, and from 10/15/96 to 10/18/96. These results are presented in the following
section. It is important to note that during testing on 7/11/95, an extremely energetic cascade of
events was observed at 13:02, shown in Figure 7.

Figure 7
Static Electrification Event. Time Stamp: 7/11/95 12:59. Record 102. Ch1=PMT-1, ch2=PD-3,
ch3=PD-8, ch4=PD-7

This cascade of events severely damaged equipment from several of the other test groups, and it
was believed to have been close to failing the unit. After this event, numerous events were
recorded on the top of the tank that were synchronized with the line frequency, as shown in
Figure 8.

1-35
Panel Session for Presentation of Ramapo Test Results

Figure 8
Static Electrification Event. Time Stamp: 7/13/95 16:11. Record 168. ch2=PD-2, ch3=AC
Waveform

(Note that the anomalous shape of the AC waveform is believed to be an artifact of the
measurement technique). It is unclear as to whether this synchronization of static events with the
AC waveform was the result of damage to the unit, but it seems likely. It is important to note that
this activity was not solely a function of energization. For instance, at 17:00 of 3/11/95, the
pumps were shut off, and the AC synched signal on top of the tank abruptly disappeared, even
though the unit remained energized until 17:08. Clearly, these particular events could only occur
in the presence of static charging of the oil and insulation system by the motion of the oil in
conjunction with the energizing voltage.

Data from the energized tests indicate that the locus of the main activity shifted to a location
higher in the transformer. Figure 9 and Figure 10 show that the highest concentration of PD
events was recorded by the sensors in the top of the tank. Please note that this is an extremely
crude localization of events, based solely on the unweighted response of several different types
of sensors. Nonetheless, there does appear to be a shift in the concentration of data from the
bottom of the transformer for non-energized results, as shown in Figure 1 and Figure 2, to a
location higher in the transformer.

1-36
Panel Session for Presentation of Ramapo Test Results

10000
9000
8000
7000

PD Counts
6000
5000
4000
3000
2000
1000
0
0 1 2 3 4
Number of Pumps On

Figure 9
Max Bottom PD Activity vs. # of Pumps On

10000
9000
8000
7000
PD Counts

6000
5000
4000
3000
2000
1000
0
0 1 2 3 4
Number of Pumps On

Figure 10
Max Top PD Activity vs. # of Pumps On

In general, energizing the unit increased the quantity and intensity of static events by close to an
order of magnitude. The dependence on temperature remains strong for events on the bottom of
the tank, as shown in Figure 11.

1-37
Panel Session for Presentation of Ramapo Test Results

10000
9000
8000
7000

PD Counts
6000
5000
4000
3000
2000
1000
0
0.00 10.00 20.00 30.00 40.00 50.00 60.00
Temp, Deg. C

Figure 11
Max Bottom PD Activity vs. Top Oil Temp

It appears that the majority of the events recorded in the bottom of the tank for energized tests
were at top oil temperatures below 40 degrees C. Figure 12 indicates that the largest
concentration of events occurring on top of the tank were at temperatures near 20 degrees, but
they can also occur at temperatures up to 50 degrees C.

10000
9000
8000
7000
PD Counts

6000
5000
4000
3000
2000
1000
0
0.00 10.00 20.00 30.00 40.00 50.00 60.00
Temp, Deg. C

Figure 12
Max Top PD Activity vs. Top Oil Temp

It is speculated here that the different spread of events on the top and bottom of the unit may be
associated with different mechanisms of static charging, as the AC coupled events were very
clearly localized in the upper portion of the transformer, and they appeared to be somewhat
independent of the activity occurring in the bottom of the transformer.

Data shown in Figure 13 demonstrates a correlation between activity on the top of the tank and
relative humidity level.

1-38
Panel Session for Presentation of Ramapo Test Results

10000
9000
8000
7000

PD Counts
6000
5000
4000
3000
2000
1000
0
3.00 4.00 5.00 6.00 7.00 8.00
% Relative Humidity

Figure 13
Max Top PD Activity vs. % Relative Humidity

However, the correlation between events on the bottom of the tank and the relative humidity is
relatively weak, as shown in Figure 14.

10000
9000
8000
7000
PD Counts

6000
5000
4000
3000
2000
1000
0
3.00 4.00 5.00 6.00 7.00 8.00
% Relative Humidity

Figure 14
Max Bottom PD Activity vs. % Relative Humidity

Even for the data shown in Figure 13, there is insufficient data to draw a trend line through the
data with any reasonable level of confidence. Nonetheless, it appears that moisture level
expressed as % relative humidity may be an important factor in the static electrification process.

Conclusion

The results of the non-energized and energized static electrification tests indicate that very high
levels of static electrification were occurring in the Ramapo test transformer. These static events
were observed by four independent test groups with a vast array of different sensing
technologies. The Harley effort consisted of a fully instrumented TPAS I system with seven to

1-39
Panel Session for Presentation of Ramapo Test Results

eleven (depending on test date) acoustic channels. A total of eleven different acoustic sensors
were installed in the transformer, as well as three light sensors, a moisture sensor, and four
RTDs. Harley was able to obtain strong static electrification signals on all acoustic sensors, as
well as what may have been a response to light with one of the PMTs.

The events were often quite energetic, completely saturating the electronics in many cases,
sometimes creating banging noises audible to the naked ear from more than a hundred meters
distant from the unit. These massive static events also proved to be a good test bed for transient
protection systems, as much of the monitoring gear used by all of the groups involved was
damaged due to transients associated with the very energetic static electrification events. While
the majority of activity for non-energized tests appears to have occurred on the bottom of the
unit, the primary locus of activity appeared to shift to the top of the unit during energized tests.
This may have been the result of damage that may have been done to the unit during a
particularly energetic cascade of events that occurred on July 11, 1995. Subsequent to this event,
numerous events were observed that were synched with the energization AC waveform. Static
electrification was clearly a function of the number of pumps being run, with very little static
evident during the non-energized tests with less than all 4 pumps on. Highly energetic static
events were almost an order of magnitude more frequent when the unit was energized than when
it was non-energized. Also, the static events are far more common at lower temperatures, and
lower relative humidity levels. The results of the tests clearly indicate that almost any of the
sensing technologies employed had merit as a potential way of alarming on static electrification
problems.

Acknowledgements

The effort to perform the Ramapo static electrification testing was enormous, and so is the cast
of people who made it possible. I would like to personally acknowledge the efforts of Donald
Chu, John Sandstrom, and the entire Consolidated Edison crew that made this test possible.
Thanks Harold Moore for his essential insight and guidance throughout the testing project. From
J. W. Harley, the efforts of Richard Berris, Mike Hostetler, and Jeff Parmelee were noteworthy,
and the work of and many others was essential as well.

1-40
Panel Session for Presentation of Ramapo Test Results

1.4 Ramapo Tests: Results from Streaming Current Monitoring Using a


Nilsson Plate and an Absolute Charge Sensor

Stig Nilsson, Fellow, IEEE


Consultant
Los Gatos, CA 95033 USA

Markus Zahn, Fellow, IEEE


MIT
Cambridge, MA 02139 USA

Stan Lindgren, Sr. Life Member, IEEE


EPRI
Palo Alto, CA 94303 USA

Abstract
Static electrification of transformers was identified in the late 1970s as a potential cause of
failure of large forced cooled transformers. R&D spanning more than 20 years has led to an
understanding of the conditions under which transformers may have the highest risk of failure.
Tests conducted by EPRI and Con Edison at ConEds Ramapo substation over a three-year
period has demonstrated that static electrification is detectable and conditions under which static
electrification arises are repeatable. This paper presents static electrification data from a plate
sensor and an absolute charge sensor.

Keywords
Transformers, static electrification, On-line diagnostic, charge sensing, electrokinetic activity.

I. Introduction
Static electrification was discovered as a potential failure cause of large power transformers in
the 1970s. Since that time, research has been conducted all over the world [1] with the objective
of understanding how to detect critical electrification and if possible to prevent failure.
Laboratory research has been sponsored by EPRI in an effort to understand the fundamental
aspects of the electrification process. This included development of sensors for measurement of
charge density and the so called streaming current.

The moisture content of the transformers dielectric system has been proposed as one significant
factor to consider. The possible transformer flow electrification failure mode that is dependent on
moisture dynamics often arises when a transformer that has been out of service is being re-
commissioned. A plausible scenario is that as the transformer heats up, moisture is driven out of

1999 IEEE. Reprinted, with permission, from IEEE Transactions on Power Delivery. Paper to be presented at the
Transmission & Distribution Conference & Exposition, New Orleans, LA, April 10-17, 1999, paper number
99TD351.

1-41
Panel Session for Presentation of Ramapo Test Results

the pressboard insulation into the oil. The moisture first comes from near the pressboard
interface, leaving a very dry interface that is highly insulating, which allows surface charge to
accumulate until tracking discharges occur. These discharges cause gas formation, which can rise
into the high electric field region causing a high-energy discharge that causes transformer failure.
If failure does not occur during the early times of the process, the moisture deeper in the
pressboard diffuses to the surface reaching equilibrium with the oil moisture. The moisture
diffusion time can at low temperatures easily be a few weeks. However, once there is no longer
an interfacial dry zone, there is a leakage path for interfacial surface charge so that the charge
density cannot rise to cause electric field strengths beyond the breakdown strength of oil and
pressboard. Thus, once the transformer is in moisture equilibrium, this flow electrification hazard
is minimized.

When Con Edison experienced transformer failures believed to be caused by static


electrification, it allowed one of the transformers to be used for experiments. These tests were
conducted over about three years so that moisture equilibrium could be reached between tests.
Although, changing environmental conditions did not allow complete reproducibility, tests were
run in the fall and spring under similar conditions. Also, early and late summer conditions
enabled close to repeatable conditions. The tests were conducted at the Ramapo substation. Parts
of the data from these tests are reported below.

II. Monitoring System

The monitoring system consists of an MIT developed absolute charge sensor (ACS). The ACS
[2] measures the charge carried in the oil by drawing in a small volume of oil into an expanding
Faraday cage bellows volume followed by ejection of the same sample. A Wilson Plate,
converted for use under oil inside the transformer tank, is also included. This system, built for
EPRI by Electric Research and Management, Inc. is hereafter referred to as the Nilsson Plate
also abbreviated NP [3, 4].

Figure 1 shows very schematically the coupling between the Nilsson Plates and the
electrification process. The oil flow strips charges from the interfaces with solid materials;
primarily the paper and pressboard materials used for insulation in the transformer. This is
illustrated by the current source. The current is returned to the source through leakage (resistive
paths) but the flowing oil also carries the charges. The space charge also results in image charges
in the tank walls, windings and other conducting structures inside the transformer. Consequently,
displacement currents caused by changes in the spatial distribution and amount of charge are also
seen by the plates. The electrification process is chaotic and can not be described using a linear
model, but the simple model may still be useful for understanding the data and the electrification
process in general.

1-42
Panel Session for Presentation of Ramapo Test Results

Top Plate

Plate Current

Leakage
Tank
Wall

Current Source

Bottom Plate

Plate Current

Figure 1
Schematic Representation of the Coupling between the Electro-Kinetic Charge Generator
Inside the Transformer and the Plate Sensors

The top NP was mounted on the side of the transformer below the gas space. (A low oil flow
region at the top of the tank would have been the preferred location but this was not possible
2
because of the nitrogen blanket.) This plate has an area of 0.147 m . The area of the bottom plate
is a 0.118 m2.

III. Tests with No Applied Voltage

A. March 1995 Test Series

The first test run using the Ramapo transformer actually took place between March 27th and
29th, 1995. These tests were run without energizing the transformer after processing the
transformer in the fall of 1994. The ACS data is shown in Figure 2 and the corresponding plate
currents are shown in Figure 3. The plate current is sampled about once every one to two minutes
through a long time-constant filter. Pump switching causes the steps in the readings. The
temperature during this run was about 7 to 8C.

Ramapo ACS Charge


03/27/95

80
70
ACS Charge uC/m3

60
50
40
30
20
10
0
11:00 12:12 13:24 14:36 15:48 17:00
Time

Figure 2
ACS Data from the March 27, 1995 Test Series

1-43
Panel Session for Presentation of Ramapo Test Results

Ramapo Nilsson Plate Current


03/27/95

7000

Nilsson Current pA
6000
5000
4000
3000
2000
1000
0
-1000
11:00 12:12 13:24 14:36 15:48 17:00
Time

Figure 3
Plate current during the March 27, 1995 tests

B. Tests on July 10, 1995

A significant event was the first day of testing on July 10, 1995. During this test, the transformer
was not energized. The plate current is shown in Figure 4. A delta current quantity, computed as
the difference between the samples plate current data and a filtered (smoothed) current is also
shown. The smoothing function is a simple, first-order digital filter of the type:
ak = ak k + (1 k ) ak 1 (1)

where for this analysis, k has been set to 0.30.

Ramapo Tests 7/10

11000 30
Top Oil
NP Current Curve Fit
Plate current (pA)

9000 Temperature 29

7000 28
deg-C

5000 27

3000 26

1000 Delta - NP 25

-1000 24
14.00

14.50

15.40

16.30

17.20

18.10

19.02

Time: Hours.Min

Figure 4
Ramapo Tests during July 10, 1995. Four Pumps running from 16.30 to about 18.30 Hours

By comparison to other PD instruments used during the test, it was found that a positive delta
was a result of a static electrification type discharge at the bottom of the transformer and a
negative delta arose from a discharge closer to the top of the unit. This can be intuitively
understood by considering that a discharge at the top of the unit relaxes positive charge and
results in a reduced electric field closer to the top plate. The opposite happens as a result of a
relaxation at the bottom of the unit. Thus, the plate can be used to sense significant static
electrification related discharge activity within the tank.

1-44
Panel Session for Presentation of Ramapo Test Results

When pumps were turned off, a current from a trapped charge (Figure 5) became visible in the
plate measurements. Since the relaxation time for oil-paper dielectric is much longer than for the
oil dielectric, the discharge time should relate to the relaxation time of the oil-paper system.

The time constant of oil and oil impregnated pressboard as a function of temperature is shown in
Figure 6. These time constants have been computed based on available information about the
resistivity of the oil-paper dielectric [5]. The conductivity of the oil has been estimated from
Figure 4 of Ref. 5 and that of pressboard by using the 26V/mil stress level shown in Figure 3 of
Ref. 5. The dielectric constant for oil has been set to 2.2 and that of pressboard to 3.2. The
relaxation time for the oil-paper dielectric system using the recorded data as shown in Figure 5,
is about 2700s. The calculated relaxation time for a temperature of about 35C is about 2000s.
Because material properties can be expected to vary in a wide range and are stress dependent, the
comparison is surprisingly good. Note that with a 70 to 100 s sample interval for the plate current
measurements, the time constant of the oil will be negligible first at temperatures above
40 degrees.

Ramapo Tests 7/10

800 36
700 Top Oil Temperature 34
Plate Current (pA)

600 32
500 30

deg C
400 28
300 26
200 24
Top Plate Current
100 22
0 20
17.35

18.14

18.54

19.34

20.14

20.54

21.34

22.14

22.54

23.34
Time: Hours.Min

Figure 5
Upper Plate Current after switching off the Pumps at the End of the July 11 Test Series

100000

10000
Relaxation Time (s)

PRESSBOARD
1000

OIL
100

10
0 10 20 30 40 50 60
Temperature (Celsius)

Figure 6
Relaxation Time as a Function of Temperature for Transformer Oil (solid) and Oil
Impregnated Pressboard. (Dashed line)

1-45
Panel Session for Presentation of Ramapo Test Results

The behavior of the system during the July 10 test can be reasonably well described using an
exponentially decaying plate current combined with an ohmic or a charge transfer flow from the
moving oil. A curve fitted using the calculated relaxation time of 4000-s (26C) has been added
in Figure 4 to show this. The data shown in Figure 4 might be viewed as follows:
Assuming a relatively constant electrification current generated inside the transformer, the
plate current has an exponential decay to a steady state level with a time constant close to
that of the paper.
The decaying component could be a displacement current flowing through the paper
dielectric to the plate.

If this is correct, the streaming current might be estimated from the plate current. The voltage at
the output of the simple Norton equivalent shown in Figure 1 could then be estimated from the
plate current using an estimate for the impedance, through which the current has to return to its
source. This might be valid at least for a specific transformer design if the charge distribution is
reasonably spatially stable. It could at least be used for alarm purposes to alert the operators
about an immediate problem.

C. Tests September 1996

The September 1996 test series was run just prior to opening of the transformer tank for
inspections and repair as needed and to enable modifications of the instrumentation systems. The
first run is shown in Figure 7 below.

4000 25
NP Current
3500
Top Oil Temperature
20
3000
Current (pA) or Count

2500
Temperature

15
2000
PD Count
1500
10
1000
Delta -NP Current
500 5

0
11 12 13 14 15 16 17
-500 0
Time: Hours

Figure 7
Initial Test Run, September 17, 1996

The temperature at the beginning of this test is about 16 degrees but as for all of the tests, the
time temperature history is not known. Since September can still be a relatively hot month, the
temperature could have been higher for many days during the previous past several weeks.
However, the data is not unlike that from the March 95 test.

The plate current behaves as expected. The initial transient can be seen. At this temperature, the
relaxation time for the oil should be on the order of 4 to 5 minutes. After about one and a half
hour, severe high frequency content is picked up on the plates PD sensor, which prior to this test
had been modified to enable continuous PD monitoring. A similar significant increase in the

1-46
Panel Session for Presentation of Ramapo Test Results

delta-NP current readings is not seen. The reason for this could be that the discharge activity
might be too rapid to generate any changes in the slow plate current sensor. (In this case, the
filter constant for the delta-NP current calculation has been set to 0.2 instead of to 0.3 as was
used for the July 95 tests.) Therefore, it is clear that a PD detector is needed to supplement a
delta-NP current sensor to detect static electrification activity. The temperature dependence of
the plate current/electrification process is also seen again.

D. Tests during October 15, 1996

A second NP was added at the bottom of the transformer prior to the October 1996 tests. The
system was also re-scaled to avoid saturation of the amplifiers as had happened above 10 nA in
some of the earlier tests. The modifications also enabled continuous PD recording as well as
running PD detection and the ACS at the same time.

Figure 8 shows the plate currents during the start of the tests. The ramp seen in the bottom plate
current can only be explained if positive charge is injected into the bottom of the unit. The
sampling rate for this record is three times per minute and therefore, the relaxation time of the
oil, which is about 400 s or 20 samples, can not be ignored.

5000 12.5

Top Oil
4000 12

Temperature (deg-C)
3000 11.5
Plate Current (pA)

Top Plate Current


2000 11

1000 10.5

0 10
11:41

12:00

12:20

12:40

13:00

13:20

13:40

14:00

14:20

-1000 9.5
Bottom Plate Current
-2000 9
Time: Hours.Min

Figure 8
The First Test from which Data Was Obtained from Both the Top and Bottom Nilsson
Plates

Furthermore, at the flow rate of the pumps (four 800 gal/min pumps), it takes at least 5 minutes
for the oil to be circulated once through the system. Thus, the transients seen during the first
10 min. would have to be carefully analyzed before any conclusions could be reached as to its
origin. Beyond this time, the time constant of the paper dielectric system, which for a
temperature of about 10C is about 5 to 7 hours, would be dominant. The bottom plate amplifier
failed about one hour after the pumps were turned on.

It can be seen that, in contrast to the July 10 tests when the temperature of the system was
significantly higher, the streaming current does not exhibit a decay until about 75 minutes after
the beginning of the test. The ACS data plotted in Figure 9, shows that the charge density is
actually growing for this time period. Thus, the response of the plate is actually the response to a
ramp instead of a step function.

1-47
Panel Session for Presentation of Ramapo Test Results

A close examination of Figure 10 shows a slight delay between the plate current and the ACS
data. The plate will respond instantaneously to a charge placed anywhere within the tank but the
ACS will see charges first when the oil flow has brought the charges to the nozzle of the ACS.
Thus, since it will take minutes for oil circulation to be established, a delay can be expected.

4000 20

ACS Charge (uC/m ) or Oil Temp. (Deg-


18
3500
N P C u rr e n t 16
3000
14

Plate Current (pA)


2500
12

C)
2000 10

3
8
1500
T o p O il T e m p e r a tu r e
C h a r g e D e n s ity (A C S ) 6
1000
4
500
2

0 0
1 1 .0 0 1 2 .0 0 1 3 .0 0 1 4 .0 0 1 5 .0 0
T im e (h o u rs )

Figure 9
ACS Readings on October 15, 1996

16 00 0 70

Charge Density and Temperature


14 00 0 60
A C S D a ta
12 00 0
50
Plate Cuurent

10 00 0
N P C u rren t 40
80 00
T o p O il T e m p e ra tu re 30
60 00
20
40 00

20 00 10

0 0
18 19 20
T im e : H o u rs )

Figure 10
Continued Tests on October 15, 1996

IV. Tests on Energized Transformer

A. July 11 and 12, 1995


The July 11 tests are shown in Figure 11. During these tests the transformer was energized. The
energization by itself did not result in any plate current, which means that the 60-Hz filters
worked as expected. As soon as the pumps were turned on, the plate current increased until the
sensor saturated at about 12 nA. Although peak currents from the two sequential tests can not be
directly compared it is clear that energization causes a substantial increase in the plate current
and presumably in the electro-kinetic activity, too.
During the initial operation with four pumps, an increase in the PD activity was seen. This
preceded a heavy discharge in the transformer at which time the pumps were shut off. The test
was later restarted with only two pumps running. Later, a third pump was added. Note that after
this test, the transformer was conditioned for moisture equilibrium at about 30C.

1-48
Panel Session for Presentation of Ramapo Test Results

Pumps 0 4 2 0 2 3 0

Plate current in pA or PD
14000 40
Top Oil Temperature
12000 35
NP Current
10000 30

deg-C
8000 25

count
NP Current
6000 20
4000 PD Count 15
Delta-NP Current
2000 10
0 5
-2000 0

.13

.38

.03

.28

.53

.18

.43

.08

.33

.58

.23

.48

.13

.38

.03
11

11

12

12

12

13

13

14

14

14

15

15

16

16

17
Time: Hours.Min

Figure 11
July 11, 1995. Test. Transformer Energized and Four Pumps Operating at 12.21, Reduced
Pumps at 12.50 and Switched off at 13.03. Restarted at 13.54 with Two Pumps with a Third
Added at 15.20 Hours

The tests on July 12, Figure 12, began with the transformer energized, three pumps in operation
and an oil temperature down to about 24C. During the first five hours, the plate current stayed
relatively constant. At the same time, the conductivity of the oil was increasing and therefore,
any ohmic current flow to the plate should increase if the charge is constant. However, if the
electro-kinetic activity is increasing at the same rate as the capacitive displacement current is
diminishing, the plate current will also stay constant. This may very well be the case. A clear
reduction of the plate current is seen after 3 p.m. indicating that the electrification activity
probably began to go down. When the temperature was approaching 50C, the fourth pump was
added. This immediately increased the plate current but some of this could be expected to be a
displacement current from an increased charge inside the system. The relaxation time for oil-
paper dielectric at about 45C is about 900 s. Thus, any capacitive current coupled to the plate
should diminish in 15 to 30 minutes.

9000 60
Plate Current (pA)

8000 Top Oil Temperature 50


7000
6000 40
deg-C

5000 NP Current 30
4000
3000 20
2000
10
1000
0 0
10.29

12.09

13.49

15.31

17.11

18.51

20.31

22.11
7.10

8.49

Time: Hours.Min

Figure 12
July 12, Test. Transformer Energized, Three Pumps in Operation and an Oil Temperature
Down to about 24C

Shortly after the fourth pump was added, the additional oil heaters were turned off and the
transformer allowed cooling down as fast as possible. An immediate increase in the plate current
is seen. This can only mean that the electrification activity is increasing because any resistive
current flow to the plate should go down as soon as the temperature is reduced if the current is
driven from a constant voltage source inside the transformer. Alternatively, if the current comes
from a constant current source inside the transformer, the current would stay unchanged but the

1-49
Panel Session for Presentation of Ramapo Test Results

electric field will go up assuming a constant charge in the system. Instead the current is going up.
Thus, the conclusion must be that the electrification is increasing. Since there was increasing PD
activity in the transformer, the assumption must be that the stresses are increasing. However,
shortly after 8 p.m. something changed and the plate current began to diminish again and the PD
activity began to go down. The transformer seemed to have been conditioned and no longer
exhibited any real high electrification activity. Indeed, after this test, the rest of the test series did
not result in any serious discharge activity. The test demonstrated that rapid cooling of the oil
will increase the electrification activity and can lead to a higher risk of failure from static
discharges. It also demonstrated that a transformer, after a few days of conditioning at elevated
temperatures, might no longer be at a high risk of failure from static electrification.

B. October 16, 1996 Data

The tests on October 16, 1996, which are shown on Figure 13 below, added information about
the relationship between the ACS and plate readings. Also shown on the figure is a filtered ACS
output. The same first order digital algorithm given in Eq. (1) above was used with a filtering
constant equal to 0.2. Also, some manual manipulation of the data was applied to improve the
response of the filter to large perturbations. Both the filtered and unfiltered responses are shown
in Figure 13.
Pumps 0 4 2 4 1 4 0
18000 120

16000 NP Current
100

Oil. Temp. (deg-C) or Charge


14000
NP Current (pA)

12000 80
ACS Charge
10000 Density (uC/m 3)
Top Oil
Temp. 60
8000

6000 Smoothed 40
ACS
Data
4000
20
2000

0 0
11 13 15 17 19 21
Time: Hours

Figure 13
Ramapo Tests on October 16, 1996 with Different Pump Sets in Operation

The plate current differs from what was seen in e.g. Figure 11 because the sensor is not reaching
saturation. There are also some noticeable differences between the plate current and the ACS
data. It can be seen that the plate current increases when the temperature increases. This has been
taken as an indication of an increasing electrification activity. However, the ACS readings are
going down, or are reasonably stable, indicating a stable or slightly lower charge density close to
the plate. A couple of points, shown in the table below, have been selected to illustrate the
relationships.

1-50
Panel Session for Presentation of Ramapo Test Results

Time ACS Plate Curr. Temp Relaxation


3
(hr) (C/m ) (pA) (C) time (s)
13.32 27.1 3520 25.5 4400
14.87 16.6 5280 31.3 2400
Delta -10.5 1760 5.8 -2000
Delta - -
(%) 39 50 23 45

The change in relaxation time and the charge density appears to be correlated (as expected). This
indicates that less charge is surviving all the way to the wall before it is neutralized as a result of
the increased conductivity of the oil. The plate current reading could therefore indicate an
increased electrification activity in the system. At the 40-degree temperature, the relaxation time
is down to about 1200 s, down 73% from the relaxation time at 13.32 hours. As the temperature
begins to go down during the last part of the test, the opposite happens. ACS readings come back
up and the plate current goes down. That is, even with a seemingly lower electrification activity,
more charges survive and reach the ACS at the wall. The data just shows how complex the
electrification process really is. Further analysis is needed to understand the data.

V. Other Test Observations


The relative and absolute humidity of the oil was monitored both with on-line and off line (oil
sample tests) methods. The hypothesis is that the electro-kinetic activity is strongly related to the
relative humidity of the interface between the pressboard and the oil. There appeared to be a
correlation between the relative humidity of the oil and the plate current but the data to show the
relationship is not a part of the plate current records. It was noted that only a small change (a few
tens of percent) of the relative humidity showed a significant increase in electrification activity.
However, this is left for others to explore.

VI. Conclusions and Observations

The understanding of the electrification process in a full-scale transformer has been advanced
immensely through these tests. The following can be seen from the data:

1. The initial tests in March of 1995 and the same test performed on October 15 resulted in a
plate current of about 3 nA. Both of these tests were run with the same pump set in operation
and the same test protocol on a transformer, which had been reprocessed in the field
indicating that the state of the transformer was about the same for the two tests.
2. The tests run on July 11 and on September 18 (not shown) shows a similar plate current
(about 8 nA) when operating at about 45 to 50C with four pumps and the system energized.
3. All tests showed a remarkably reduced electrification activity after three days of testing and
after the transformer had been operated at above 40 to 50 degrees for several hours.
4. Any major static electrification related discharge occurred first after about 40 minutes to an
hour of operation.

1-51
Panel Session for Presentation of Ramapo Test Results

5. The question about the usefulness of the system for monitoring of transformers may not be as
easy to answer. One conclusion is however, indisputable. That is that static electrification
related discharge activity was visible as changes in the plate current as well as were recorded
on the high frequency PD detector portion of the plate.
6. Although more work is needed before the ACS and the NP can be used as a static
electrification monitors, both responded well to static electrification activity
7. The relative humidity of the transformer dielectric system, although not a part of this paper,
appeared to be well correlated with the electrification activity.

Acknowledgements
The late Professor Jim Melcher, MIT, gave the Nilsson Plate name to the Wilson plate for this
application. The name has been kept here in honor of Dr. Melchers accomplishments and his
contributions to the static electrification research. Without ConEds support and the use of the
Ramapo transformer the progress in static electrification research and transformer monitoring
would not have been possible. A number of people have contributed but especially Leo Savios,
Don Chus and Mike Lebows efforts are acknowledged. Harold Moore, consultant to EPRI and
Con Edison for this project, has also made major contributions to the project.

VII. References

1. Proceedings: Static Electrification in Power Transformers, EPRI Report TR-105019,


May 1995.

2. A. J. Morin II, M. Zahn, J. R. Melcher, and D. M. Otten: An Absolute Charge Sensor for
Fluid Electrification Measurements, IEEE Transactions on Electrical Insulation, Vol. 29,
No. 2, April 1991, pp. 181-199.

3. William B. Gish and Stig Nilsson: Static Electrification Monitoring on Union Electric
Mason Substation Transformer, Proceedings: Static Electrification in Power Transformers,
Report EPRI TR-105019, May 1995, Pages 4-1-1 through 4-1-11.

4. W. B. Gish: Static Electrification Monitor Acceptance Testing, Proceedings: Static


Electrification in Transformers, Report EPRI TR-102480, June 1993,
Pages 1-7-1 through-7-12.

5. William Kennedy: Resistivity of Oil and Pressboard Insulation and their effect on
Transformer Design, Proceedings: Static Electrification in Transformers, Report EPRI
TR-102480, June 1993, Pages 1-4-1 through 1-4-13.

1-52
Panel Session for Presentation of Ramapo Test Results

VIII. Biographies

Stig Nilsson, (F 1988) was born in Hr, Sweden on March 4, 1938. He has an EE degree from
HTL, Malm, Sweden in 1960 and an MBA from Santa Clara University, 1985. He left EPRI in
1994 as the Executive Engineer of the Electrical Systems Division after close to 20 years at the
Institute. He left a position as the Executive VP of the Silicon Power Corporation in 1997 and is
now a Consultant specializing in power quality and technology marketing. He is Chair of the
IEEE-PES T&D Committee.

Markus Zahn (F 1993) was born in Bergen-Belsen, Germany on Dec. 3, 1946. He received his
BSEE and MSEE degrees in 1968 and ScD-degree in 1970, all from MIT, where he is now
Professor of EE working in the Lab. for Electromagnetic and Electronic Systems and the High
Voltage Research Laboratory. He is also the director of the EECS VI-A internship program.
Dr. Zahn is Chair of the IEEE Liquid Dielectrics Committee and associate editor of the IEEE
Transactions on Dielectrics and Electrical Insulation.

Stanley R. Lindgren (Sr. Life Member), born on February 19, 1928, obtained his BSEE degree
from Kansas State University in 1950. He is Manager, Power Transformers, in the Substation
Target of the Energy Delivery and Utilization Division of EPRI in Palo Alto, California. Before
joining EPRI in 1986, Mr. Lindgren was with the Paragon Electric Co., Inc., Wisconsin. At RTE
Corporation, Waukesha, Wisconsin, between 1971 and 1978, and at Power Transformer Division
of Allis Chalmers, 1952-1971.

1-53
Panel Session for Presentation of Ramapo Test Results

1.5 Dismantling of Ramapo Transformer

Harold R. Moore
Harold Moore & Associates
1062 Lakeway Drive
Niceville, FL 32578

Abstract

The Ramapo transformer was dismantled and inspected in an effort to try to relate the discharge
events to specific damage to the insulation system. This was done under the direction of
Consolidated Edison. The windings and lead structure had be removed intact so a procedure was
developed for supporting the phase during core removal, lifting the phase out of the tank, and
laying it in a horizontal position. During the inspection, discharges were found at the top of the
transformer, which correlated with partial discharge measurements, and observations of audible
noise locations during the tests at Ramapo. Discharges were also found at the bottom in the
insulation between the high and low voltage windings and in the insulation between the high
voltage winding and the core. The paper discusses the findings of the tear down and conclusions
that might be possible to draw from this inspection for past failures and also discusses
implications for new designs.

Introduction

The phase was removed from the transformer by G & S technologies under the direction of
Consolidated Edison. Since it was very important that the windings and lead structure be
removed intact; a procedure was developed for supporting the phase during core removal, lifting
the phase out of the tank, and laying it in a horizontal position.

Due to the efforts of John Sandstrom, the windings and lead structure arrived at the
G & S Technologies plant in good condition with very little distortion to any parts including the
leads or lead supports.

Discharges were found at the top of the transformer, which correlated with partial discharge
measurements, and observations of audible noise locations during the tests at Ramapo. The
discharge pattern at the top was very similar to failure patterns seen in a number of failed
Westinghouse shell form transformers. The results of this inspection probably explain a number
of shell form failures where the cause was undetermined.

Discharges were also found at the bottom in the insulation between the high and low voltage
windings and in the insulation between the high voltage winding and the core. Some were
located at the end of the barriers which ended at the 45 degree region, but a number also existed
in the V where the corner pieces intersect the straight pieces. A number of discharges were
found on the washers [sheet barriers] between the high and low voltage windings. The location
of these discharges compares well with the partial discharge measurements. The various sensors
indicated high level discharges at the bottom of the windings, and many of the loud noises heard
from the interior of the tank appeared to be at the bottom.

1-54
Panel Session for Presentation of Ramapo Test Results

Therefore, the location of the discharges correlate well with the partial discharge locations made
during the various tests. This is for the long discharges at the top and the discharges into the high
and low voltage winding coils at the bottom which are in the regions where high acoustic signals
were measured and loud audible noises were heard. Where possible, the results of the inspection
will be related to partial discharge measurements and other reactions during the static
electrification tests.

Location of Discharges

Discharges were found at the top end of the transformer which involved the barriers around the
lower end of the high voltage bushing, the shield on the high voltage bushing, and creepage
along a pressboard support wall to the neutral lead.

Discharges were also found at the bottom end of the windings. The origin of these discharges
was in the main high voltage to low voltage insulation spaces or near the 45-degree region.
Significant discharges were found on the bottom end of pressboard barriers between the
windings and in the enclosing insulation pieces between the high voltage winding and the core.

The location and nature of the discharges will be described in the following sections of the
report.

Discharges at the Top End

Lead and Lead Support Structure at Top End

The lead and lead support structure is similar to the cutaway shown in Figure 1. In Westinghouse
shell form transformers, walls made from laminated pressboard were used to support the leads
and components such as de-energized tap changers. In this design, the walls provide the
following functions:

a. Leads and connections are clamped to the walls to maintain electrical clearance and provide
short circuit support.

b. Support for the de-energized tap changer board.

c. Support for the load tap changer leads.

d. The walls provided a box electrical barrier between the lower shield on the high voltage
bushing, and ground or other leads.

The winding arrangement for this transformer is shown in Figure 2. It is the typical two High-
Low arrangement used for EHV shell autotransformers. Figure 3 shows the winding arrangement
with respect to the main lead support walls and the high voltage bushing.

1-55
Panel Session for Presentation of Ramapo Test Results

Discharges on Walls Adjacent to HV Bushing Shield

There were many discharges on the main wall labeled A and B on Figure 3. These pressboard
walls were located around the spun copper shield which has the shape shown in Figure 4. The
discharges on the end wall A are illustrated on Figure 5. The discharges originated in the
region of the HV bushing shield.

They then progressed along the surface of the wall to a pressboard support bolted to the wall.
This support was 2 x 2 x and about 15 long. There was one large black area between the
support and the wall. The discharges then went to the top edge of the wall. There was
considerable delamination of pressboard and the edge of the wall was black.

The discharges went along the edge of the wall to the neutral lead. There were two small
puncture holes through the tape on the neutral lead. The discharges went into the conductor.
There was no disruption of the metal in the lead, which indicates a low-energy failure.

There were several discharge trees on wall B. The discharge pattern is shown in Figure 6. There
was rather deep erosion of the pressboard surface. No indication of a failure path from this wall
to any other part was found. However, there did appear to be a connection from some of these
discharges to those on wall A in the corner where the walls are joined.

Discharges were found on the top end of washers 22 and 23 below the bottom end of the HV
bushing. These trees extended down from the top edge for about 5-6, but they did not extend
into the winding. There was no indication of discharges on any other washers.

Discharges on Lower HV Bushing Shield

This spun copper shield has a thin coating of epoxy resin. Crepe paper tape approximately
1/8 thick had been applied over the epoxy. The shield was carefully inspected for evidence of
failures before the crepe paper was removed. There were multiple locations where there was
evidence of carbon, but there was no real evidence of holes. As the tape was removed, two
locations were found where the tape was punctured. The holes were very small in diameter. The
epoxy had holes approximately 1/8 diameter in line with the two small holes in the tape. The
copper surface had a minimum amount of disruption at these two locations.

When all of the crepe tape was removed, 12 additional punctures of the epoxy coating were
found. The epoxy flaked off at these spots and there was a small amount of carbon around the
holes.

There was no way to determine the location of the punctures in the epoxy with respect to the
discharges on the walls. However, they extended about 120 degrees around the circumference.
It is believed that the region with these punctures on the shield was in the corner where the
discharges on the walls appeared to originate.

1-56
Panel Session for Presentation of Ramapo Test Results

Relationship of Discharges to Partial Discharge Locations at Top during Tests

There were numerous occasions during the tests when the partial discharge sensors indicated that
high level discharges were occurring at the top. There were occasions when the loud audible
discharges from the interior of the tank sounded as if they were at the top.
The acoustic sensors indicated that the discharges at the top were higher than at the bottom after
the drying operation was performed in 1996. During one set of tests, the Harley Company
performed triangulation calculations using the output from acoustic sensors located on the upper
part of the tank and the lower end of the HV bushing was indicated.
Therefore, the partial discharge measurements and the discharge pattern found at the top are
consistent.

Significance of Change in Discharge Pattern when Flowing Hot Oil into Tank

Consolidated Edison arranged for ABB to furnish an oil heater for increasing the rate of oil
temperature rise. The output from the oil heater was connected to a valve on the upper tank wall.
A check of the drawings showed that this valve is located such that the warm oil flowed through
the cutout in main wall B to the bottom end of the HV bushing. The behavior when the heater
was on and off was as follows:
a. Significant discharges were being measured at the top before the hot oil flow was started.
b. Discharge activity decreased to low levels within 1 or 2 minutes after the flow of hot oil.
(about 60C) was started. The flow rate was only 200 gpm.
c. The discharge activity at the top increased to about the same level within 5 minutes when the
hot oil flow was stopped.
d. The discharges decreased significantly when the hot oil flow was started for a second time.
e. The activity increased to its original level when the hot oil flow was again stopped.
The use of the hot oil was discontinued since it changed the discharge pattern, and we did not
have time to further evaluate the phenomena during this series. It is believed that this behavior
was significant, and it seemed to modify the charges in the vicinity of the walls and bushing
shield.

Location of Discharges at Bottom End


The windings were arranged so that coil 30 was on top, and the dismantling started by removing
coils 30 to 25 one at a time to determine if failures or discharges existed in this group. No
indication of discharges was found.

The main high to low space between coils 24 and 25 was then dismantled along with the
enclosing insulation around the HV winding. The enclosing insulation is interleaved with the
high-low washers.

1-57
Panel Session for Presentation of Ramapo Test Results

Discharges were found at the following locations:

A. At the intersection of the straight items that come all the way to the end of the assembly and
the corner pieces that goes around the coil radius. See Figure 7. The turbulence in the V
region caused discharges, which extended several inches from the intersection. Details on
these discharges are listed below.

B. Discharges occurred at the 45-degree region where the barriers around the high voltage
winding terminate. See Figure 7. Details on individual insulation pieces where this occurred
are outlined below.

Discharges in the V Region

a. The first indication was on straight angle. The discharges were about 6 from the end where
it overlaps the corner item, and went down over the long leg. See Figures 8 and 9.

b. A corner angle, which overlaps a straight item, had severe discharges starting at the
V region where the overlap occurs. The discharges extended down from the V for about
15 on the leg of the angle.

c. A straight item had discharges starting at the overlap with the corner angles. The discharge
trees extended down the long leg of the angle.

Discharges at the 45 Degree Region

Refer to Figure 10.

a. The first full-length washer in the high-low space had a small tree at the 45-degree line where
the large angles around the winding end. Two spacer blocks above this point had carbon on
the surface. The opposite side of this washer had two trees about 3-4 long near 45-degree
region. These discharges may have been connected to the larger ones further down in the
high-low space. They may have been extensions of other discharges going toward the low
voltage winding.

b. The next washer had a tree 3-4 long at the bottom overlap on the centerline.

c. The last full length washer adjacent to the first radiused end washes had long trees, (18-20)
which originated at the 45 degree region on the LV side only.

d. The first radiused end washer had long discharges (16-18) starting at the 45. This washer
had some small trees 3-4 long on the opposite side which appeared to start near the inside
corner.

e. The next radiused washer had some small trees starting at the corner of spacer blocks. This
occurred about 12 from the outside corner along a line from the 45 point. The opposite side
of this washer, which had no spacer blocks, had long trees, 18-24, starting at the 45-degree
region. This washer also had some small trees, which appeared to start at the inside corner on
the HV side.

1-58
Panel Session for Presentation of Ramapo Test Results

The remaining radiused washers will be described in sequence.

f. The washer had a discharge, which appeared to start at the inside radius on the HV side. This
arc delaminated the surface of the pressboard.
g. The next washer had a long discharge, which started at the 45-degree region and progressed
up along the 45-degree line to the inside corner.
h. The adjacent washer had some small discharges on the surface of spacer blocks on the
LV side near the 45-degree region. The plain side had small trees 5-6 long starting at the
45-degree region on the LV side.
i. Washer had some minor trees starting at the inside corner on the LV side.
j. Washer had long trees starting at the 45 on the LV block side. The side with blocks had some
long trees along the 45-degree line with the greatest erosion being toward the inside.
However, there were some large discharges at the outer 45-degree region.
k. The next washer had no discharges.
l. Very small discharges on the block side, LV only.
m. The next washer had small trees 2-3 long near the 45 on the LV side. The HV side had some
small trees 2-3 long on the non-block side near the 45.
n. Washer had trees 3-4 long along the 45-degree line near the middle of the washer on
LV side. There were some small trees near the inside radius on the HV side.
o. Washer had discharge 8-10 long on LV side starting at 45-degree region.
p. Washer had some small trees 1-3 long at corners of spacers near the 45 on LV side. Had
very small trees near the inside corner on the HV side.
q. Washer had rather heavy discharge 12-15 long on LV side starting at 45.
r. Washer had discharges on surface of 6 spacers starting at corners near the 45 on LV side.
s. No discharges found on the next three washers, which are close to the high voltage end of the
winding. However, there were some small discharges at the outer radius near the 45-degree
region of washer, which is adjacent to the static plate on high voltage line coil 24. These
discharges progressed over the edge of the static plate, and some trees were found on the
surface of the plate on the HV side of the plate where the 60-Hertz stress is very low. These
discharges were on the LV side.

It is interesting that the most significant discharges at the bottom were on the low voltage side
starting at or near the 45-degree region where the large enclosing angles around the winding
terminate. The discharges on the high voltage side were not as intensive as on the low voltage
side. In addition, the discharges on the HV side originated primarily near the inside radius rather
than at the outer radius. Discharge patterns found on other shell form transformers were also at
the 45-degree region, but the discharges were usually on both sides of the assembly. That is, on
both the high voltage and low voltage sides.

1-59
Panel Session for Presentation of Ramapo Test Results

For example, the other two Ramapo transformers that failed from static electrification had major
discharges on both sides.

The HV side of the high to low space is under the region at the top where the discharge pattern
occurred around the bottom end of the HV bushing. Therefore, the major discharges at the
bottom were at the other side of the transformer away from the region where the major
discharges occurred at the top.

The reason for the discharge pattern being so different on the HV side is not understood at this
time. The location of the pump inlets with respect to both sides is the same so that the difference
cannot be explained by flow patterns.

Progression of Failure from the 45 Degree Region into the Windings

a. Low Voltage Winding

The discharges progressed from the 45-degree regions on the LV side into the low voltage
winding. The progression was as follows:

The discharges went from the 45-degree region where the discharges were found on the last full-
length washers, over the edge of an angle. It then went over the surface of a corner channel item
to the middle of turn 2, which was at the edge of a cut out in the channel. There was a tiny hole
in the conductor insulation, which penetrated the layers of paper tape to the conductor.

b. High Voltage Winding

The discharges seen at the 45 region on one of the washers and the static plate progressed over
the edge of the angles enclosing HV line coil 24 to the space between coils 23-24.

The discharge punctured one of the washers between coils 23 and 24 near the 45-degree region,
LV side. There were discharges on the surface of the washer from the point where it was
punctured down to the middle of the 2nd turn in coil 23. This again was a very small hole where
it penetrated the turn insulation on the conductor. Since it did not start in the turn to turn space
where the stress is highest and the hole was very small, the indications are that this was the end
of a discharge rather than the starting point.

Summary and Conclusions

The discharge and failure pattern at the top of the transformer appears to be the same as
experienced on most of the shell form transformer failures, which have been attributed to static
electrification. It is also the same as experienced on several Westinghouse shell form failures in
which the failure causes were not determined when they were dismantled for inspection.

1-60
Panel Session for Presentation of Ramapo Test Results

It is possible that some of the hypotheses proposed for the failure mechanisms in the past may
not have been correct. For example, the Palo Verde failure mechanism was believed to be
migration of gas and carbon up through the coils and insulation, which caused the flashover at
the top. However, the discharge evidence at the bottom of the Palo Verde transformer was
actually less than at the bottom of the Ramapo transformer.

An examination of the records on all shell form static electrification failures or discharge
situations showed that there is only one failure where clear evidence exists that discharges at the
bottom caused a flashover at the top. This was the second Ramapo failure where there was a
definite carbon path from the discharges at the bottom to a start-start connection, which then
caused a flashover to the low voltage winding.

Therefore, it appears that there can be two mechanisms that cause failure from static
electrification in shell form transformers.

A. Discharges at the bottom caused by turbulence which results in gas bubbles and carbon
flowing to the top end where they are trapped by insulation. Ionization then occurs at the top
resulting in flashovers.

B. A cloud of charge results at the top end around the bottom end of the high voltage bushing
which is usually located over part or all of the high to low voltage winding space. Potential
gradients exist which result in discharges on the walls or barriers around the bottom end of
the bushing. Gradients from the bushing to the cloud result in discharges from the lower
shield on the bushing. The presence of these static discharges in the high AC electrostatic
field causes the discharge path to be extended resulting in the eventual creepage failure to a
different potential part such as a lead.

It is possible that the high DC field caused by the cloud of charge distorts the AC field, which
then results in high gradients. The high gradients cause the discharges and eventual failure. The
discharge mechanism around the bottom of the bushing should be investigated from the more
theoretical viewpoint.

The records are not available for some shell form failures that could not explained at the time of
inspection. A number of them were at the top end from the high voltage bushing to the low
voltage winding or leads or from the high voltage winding to the low voltage winding or leads.
At least some of them had discharges on the cylindrical barriers or walls around the bottom end
of the high voltage bushing, and some had multiple failures on the lower HV bushing shield. At
least some had discharges at the bottom which at the time were attributed to carbon, etc. being
pumped through the insulation since pumps were usually in operation at the time of failure.

After the Ramapo experience, it appears plausible that most of the known static electrification
failures have resulted from the charge cloud at the top. The moisture conditions, flow velocity,
and oil characteristics which are involved in generation of the cloud at the top also results in the
discharges at the bottom in high turbulent regions. Stripping of charges at the bottom could
contribute to the net charge at the top end.

1-61
Panel Session for Presentation of Ramapo Test Results

The position of the bottom end of the high voltage bushing with respect to the high - low space
may be a contributing factor in some of the failures from the lower bushing region. For the
construction with a large high to low space as used for the Ramapo design, the lead normally
comes up in the plane of the high voltage line coil so that a part or all of the shield is over the
high - low space. The high - low space and the enclosing insulation around the windings are
where the major oil leakage occurs.

This high - low scenario would not explain the discharges that have been found in the
hourglass configuration because the high voltage bushing and lead can be at the center where
there is no large high low space. However, the bushing would have been offset to one side in
many instances to get clearance to other leads and parts so that the lower end of the HV bushing
would have been near the enclosing insulation where maximum leakage would be expected.

On puzzling question has been the lack of documented static electrification problems in
McGraw-Edison (Cooper) shell transformers. One reason is that a different type of heat
exchanger was used so that large capacity oil pumps were not used. However, the total flow in
some of them would have been similar to some of the Westinghouse that has experienced the
problem. After considering the known design differences, the following McGraw - Edison
features may have been important.

1. They did not bring the large angles around to the 45 degree region. All of their large angles
between the high voltage winding and the core went all the way to the end of the washers.
Most of their high - low washers were full length.

2. They brought one or more of the coil to coil washers 3-4 beyond the top and bottom edges
of the coils. This would tend to break up the flow pattern, which might reduce local
turbulence in the bottom.

3. Perhaps the most important difference is their literature states that directed oil flow was not
used through the windings. They did take some actions to reduce the oil flow through the
high to low washers and the enclosing pieces around the high voltage winding. However, the
degree to which they went to prevent leakage through the insulation not involved in the
cooling spaces adjacent to the coils was much less than Westinghouse. Perhaps this greater
leakage in all spaces did not result in high velocity leakage which results in more stripping of
the charges from the oil such as occurred in the Westinghouse designs.

1-62
Panel Session for Presentation of Ramapo Test Results

Figure 1
Transformer Cutaway

Figure 2
Transformer Winding Arrangement

1-63
Panel Session for Presentation of Ramapo Test Results

Figure 3
Transformer Winding Arrangement Physical

1-64
Panel Session for Presentation of Ramapo Test Results

Figure 4
End Wall A Discharge Pattern

1-65
Panel Session for Presentation of Ramapo Test Results

Figure 5
End Wall B Discharge Pattern

Figure 6
HV Bushing Shield Shape

1-66
Panel Session for Presentation of Ramapo Test Results

Figure 7
High-Low Washer Insulation Pattern

Figure 8
Straight Angle Discharge Pattern

1-67
Panel Session for Presentation of Ramapo Test Results

Figure 9
V Region Discharge Area

Figure 10
High-Low Insulation Arrangement

1-68
Panel Session for Presentation of Ramapo Test Results

Figure 11
Low Voltage Insulation System

1-69
2
MODELING AND PREDICTION OF STATIC
ELECTRIFICATION

2-1
Modeling and Prediction of Static Electrification

2.1 Static Electrification at Solid/Liquid Interfaces of Power Transformers

G. Praxl
K. Neuner
Institute on High Voltage Engineering
Graz University of Technology
Inffeldgasse 18
A-8010 Graz, Austria

Abstract

Static electrification of forced cooled power transformers is caused by two different mechanisms:
charge generation at oil/paper or oil/board interface and charge transfer induced by energy of oil
flow. Due to Coulomb, Keesom, London, Debye and Born forces, oil ions are generated
predominantly by intermolecular impulse processes. This results in an electron transfer from the
fluid to the solid component and positively charged oil ions at interface of the cooling ducts.
Probability of these processes depends on energy of oil flow, temperature, intensity of adhesion
and cohesion forces (material property of both insulating components). Electrostatic phenomena
arise, if the oil ions are separated from the interface and moved by the streaming oil (i.e. Stokes
forces of oil flow are higher than adhesion and electrical forces on the ions). Separation process
depends on energy of oil flow, oil viscosity, oil temperature and oil conductivity (reverse
current). This paper deals with intermolecular processes of static electrification. Theoretical
modelling and other technical aspects are discussed to enable more reliable transformer designs
and operation as well as for better assessment of material properties.

Charge Generation

Any charge generation at the solid/liquid interfaces of power transformer is related to diffusion
and intermolecular impulse processes. Diffusion processes take place when the energy of
molecules at surface is different from that of those inside the volume (interaction of cohesion and
adhesion forces of solid and liquid phase, Figure 4). Therefore a higher charge density is
obtained at the surface of the components (compared with charge density inside the volume).
This results in a short-time potential difference at the interface and a resulting diffusion current,
if surfaces of different materials (phase A: oil, phase B: paper or board) will be very close
together (intermolecular distances). Charge generation by diffusion processes is based on charge
transfer through the solid/liquid interface charging the molecules of fluid at interface. Static
electrification occurs, if these oil ions are separated from the interface by the energy of oil flow.
This results in opposite polarity charging of the board interface and in generation of a charged oil
(ions) moved by the oil flow. The intensity of static electrification caused by diffusion relates to
temperature and local oil velocity at the interface (convective diffusion). High oil velocities
forces separation of oil flow into layers of high charge density (Helmholtz layer) just as high
temperatures enhances charge transfer through the interface according Butler-Volmer-Equation.
Diffusion processes do not generate additional charges, only charge separation at interface.
Furthermore charge density by diffusion could not exceed 3 to 5 times the charge density inside
the oil volume. Sensitive electrochemical methods (modified DC polarografic methods) have
30
proved , that measured charge density at interface is lower than leakage current density

2-3
Modeling and Prediction of Static Electrification

measured from duct models (up to pA/cm2)2, 8, 10. Additionally no time characteristics have been
observed, which are significant at diffusion processes (characteristic time constant). From these
results may be concluded, that static electrification of power transformer could not be caused by
diffusion processes predominantly. On the other hand diffusion processes must not be ignored in
the generation of adsorption layers (adhesion forces of oil additives, impurities, particles,
moisture).

Figure 1
Charge Generation at Oil/Board Interfaces of Power Transformer Caused by Intermolecular
Processes (Simplified Scheme)

Charge generation in the cooling ducts relates to multi-stage intermolecular impulse processes
that are present when the flowing oil interacts with the board components. All these processes
are based on intermolecular forces within their different interaction spheres (Coulomb: 1/r2,
Keesom & London: 1/r6, Debye: 1/r6 or in specific cases 1/r3, Born: 1/r9 to 1/r12). Without oil
flow, an adhesion layer will be generated at the interface. The layer contains uncharged oil
molecules fixed by Debye, London and Keesom forces. These forces are in equilibrium with
repulsive Borne forces. There is no interaction relating to Coulomb forces (i.e. interaction of
charged particles). Due to charge generation by intermolecular impulses, any charge generation
at the interface relates to a multi-stage energy transfer from the streaming oil to the unmoved oil
and board molecules. Oil molecules in the adhesion layer cause an energy transfer due to impulse
processes (elastic impulses M1-M2, plastic impulses M2-M3 acc. Figure 1). In general there is
mostly an electron transfer from oil to board resulting in a layer of positively charged oil ions at
the interface. Energy of impulse processes is based on energy of oil flow and temperature. The
probability function for these processes depends on the thickness of the adhesion layer and the
energy of the oil flow (local flow velocity and turbulence at interface). The thickness of the
adhesion layer is a function of adhesive and cohesive forces. Cohesive forces depend on oil
viscosity and temperature of insulating components. Adhesive forces are a function of
temperature and additionally of oil additives, impurities, particles and moisture. Charge
generation from plastic intermolecular impulses is a function of the insulating components
(direction of impulse, material property i.e. structure of molecule). In regards to oil velocity there
are two mechanisms. On the one hand high oil velocity increases the impulse energy of the oil
flow. On the other hand, high oil velocity decreases the probability for charge generation. The
conductivity of the oil has no effect on the charge generation process (other than on the resulting
leakage current, the so-called reverse current as shown in Figure 5). Charge generation is due to

2-4
Modeling and Prediction of Static Electrification

cohesion and adhesion forces (interfacial tension) and not to oil conductivity. From theoretical
considerations, increasing static electrification will be expected when adding surface-reactive
additives to insulating oil. Extensive experimental investigations confirm this theory (oil
additives based of non-ionic tensides).

Figure 2
Interfacial Tension versus Tenside Concentration Log C (CMC Critical Micelle
Concentration)

Figure 3
Interfacial Tension versus Portion of Oil Sludge in Power Transformers (Insulating Oil
31
against Water According ISO 6295)

There are a lot of different surface-reactive tensides (hydrophobic, hydrophilic, anionic, cationic,
ampholytic, non-ionic). The characteristic relationship between tenside concentration and
interfacial tension is shown in Figure 2. In general the interfacial tension of a fluid (resulting
intermolecular forces at surface) decreases strongly with increasing tenside concentration (c).
Below a critical concentration (cmc) so-called micelles are generated and the interfacial tension
stops decreasing. Three typical areas may be observed.

2-5
Modeling and Prediction of Static Electrification

1. The adhesion of the tensides at interfaces is caused by free surface forces (Langmuir
adsorption theory). When adding tensides to transformer oil the resulting surface forces will
decrease and the molecules of the tenside will compensate some of the free surface forces.

2. At critical micelle concentration there is a single molecular tenside layer and most of surface
forces have been compensated.

3. Increasing the tenside molecule concentration causes the tenside to move into the oil volume
and micelle generation occur (this mechanism is not relevant at static electrification).

Additives may be used over a long time to mitigate static electrification of high resistivity fluids
(pumping of crude oil, fuel etc.)12. Most additives decrease the specific resistance of fluid
(< 1010 Ohm.m). Therefore mitigation of static electrification is based exclusively on creating a
reverse current flow (high conductivity of fluid). Such a change of conductivity can not
successfully be applied to transformer oils. Although some results are reported on mitigation of
15
static electrification in power transformers by additives based on Benzotriazol (BTA) , there are
some uncertainty about the long-term ageing and impact on the breakdown voltage by BTA.
Table 1
Impact of a Hydrophobic and Non-Ionic Tenside (Dehyphon G 2084) on Electrostatic
Charging Tendency (ECT) of Transformer Oil

Percentage of tenside at 20C % 0,00 0,01 0,10 1,00

Interfacial tension 1 mN/m 43,0 0,35 0,10 0,03

Specific resistance 2 Ohm.m 1,2 .1013 1,2 .1013 0,75 .1013 0,95 .1013

Electrostatic charging tendency3 C/m


3
17,9 26,3 40,3 601
1
Interfacial tension of oil against water according ISO 6295
2
Specific resistance according IEC 247
3 1, 28, 29
Electrostatic charging tendency (ECT) measured with Ministatic test at 20C

Extensive lab tests from adding surface-reactive additives to transformer oil confirm theoretical
models for charge generation at solid/liquid interface of insulating components. The main
emphasis of our own tests have been to investigate the effect of a non-ionic and hydrophybic
tenside (Dehypon G 2084)30. Its chemical structure belongs to long-chained alkyl- and
alkamylgroups (substituted polyglykolethers responsible for the hydrophobic properties). The
concentration of additive has been varied systematically (Table 1). There is no impact of the
additive on conductivity and moisture content of the tested transformer oil (non-ionic and
hydrophobic tenside). A tenside concentration of 0.1% causes twice ECT and 1% leads to
34 times ECT. These results correspond with expected interaction of surface-reactive additives
on static electrification (compensation of surface forces by the intermolecular forces of tenside).

Similar interaction of surface tension may be observed with sludgy oil (interfacial tension against
water measured acc. ISO 6295, Figure 3). Each decrease of the interfacial force results in
mitigation of adhesion layer. Then the oil flow penetrates the laminar sublayer and the
probability of charge generation will increase. This mechanism explains why any additional
particle (impurities, moisture, ageing compounds, additives relating to crude oil origin or oil
refining procedure) strongly changes the electrification of power transformers (comparable with

2-6
Modeling and Prediction of Static Electrification

doping of semiconductors). From this point of view, storage of insulating oil and handling of
refurbished transformer windings under a non-reactive nitrogen atmosphere is suggested
(generation of adsorption layer due to oxygen). Any increase of adhesion layer thickness at
oil/board or oil/paper interface will mitigate static electrification. Thermal or laser treatment of
board or paper interfaces could change surface forces of the solid components. At present these
methods have been tested with simple hydrocarbons, but never on board or kraftpaper. Another
method for mitigation of static electrification is to use oil additives reacting opposite to tensides
(hydrophobic and non-ionic additives enhancing surface forces of oil and have effects just a glue,
not just a tenside). An additional method to mitigate electrification is to decrease the local oil
viscosity at the interface (change of viscosity-temperature or viscosity-velocity-characteristic).
From the point of transformer design and operation there are restrictions on changing the
viscosity characteristic of the transformer oil (cooling).

Adhesion and Cohesion

The properties of the adhesion layer are most important for understanding the static
electrification process of power transformers. Therefore we would like to discuss the adhesion
layer in more details. Macroscopic forces (adhesion forces of solid and fluid) may be calculated
from surface and intermolecular forces. Any force between similar molecules is called adhesion
and any force between different molecules cohesion (Figure 4). Adhesion and cohesion forces
affect generation of adhesion layer, the number of adhesive fluid particles as well as streaming
condition at the interface (local oil velocity, turbulence). A small adhesion layer and a high oil
velocity increases the probability of charge generation and charge transfer. Therefore these
forces dominate the static electrification process. Adhesion and cohesion forces or the thickness
30
of the adhesion layer may be calculated from the theory developed by Hamaker and Lifshitz . In
general, numerical evaluation of adhesion forces will fail at the oil/board or oil/paper interface;
the required intermolecular coefficients could be measured only at simple molecular structures.
At present there are some results from hydrocarbon use in petrochemical technologies but there
are no results available for naphtenic or paraffinic based transformer oils, board or kraftpaper.

Lifshitz theory is based on interactions of macroscopic materials (fluctuating electro-magnetic


fields). These fluctuations are a result of mutual intermolecular distortion and thermal movement
of the molecules (London forces). Hamakers theory is restricted to a single oscillating
frequency. The theoretical considerations of Lifshitz assume an infinite frequency spectrum. The
evaluation of the resulting adhesion energy includes all types of intermolecular forces (Debye,
Keesom, London). Molecular structures with hydrogen bonds results in an additional energy
term for intermolecular interactions. Molecules at the interface react with neighbouring
molecules as well as with distant molecules. Neighbouring molecules involve predominantly
adhesion energy caused by strongly decreasing interaction forces enlarging distances between
molecules. Therefore, any evaluation of adhesion energy at solid/liquid surfaces has to start with
these complex interactions. Surface energy VA (phase A) of any materials having polarisation
properties share with Debey, Keesom and London forces (transformer oils with long-chained
molecule structure).
VA = VD + VK + VL (1)

2-7
Modeling and Prediction of Static Electrification

Resulting adhesion energy of polar fluids is higher than for non-polar fluids because of
additional energy from Debey and Keesom forces. Therefore, high interfacial tension will be
measured at polar fluids (transformer oil: > 42 mN/m at 20C). For materials with high
polarisation properties, there is some uncertainty in the calculated values (unsymmetrical
structure of transformer oil, measurement problems of London coefficient). Surface tension of
solids will be observed just as those of fluids. However, in solids, surface molecules can not be
moved by interfacial forces (no generation of drops). So far, all considerations relate to forces at
the interface but there is an increasing impact of interaction forces exceeding critical density of
neighbouring volume phases. Interfacial tension tends to go to zero if molecules of both phases
compensate the free surface forces because then no interfacial phenomena or significant phases
will be observed. General at interface of different phases (board, oil) we observe (Figure 4):
interactions of molecules of same kind (cohesion forces of board or oil)
interactions of molecules of different kind (adhesion forces board/oil)

Figure 4
Interaction of Intermolecular Forces at Oil/Board Interface (Simplified Scheme),
1 Cohesion Forces (Board), 2 Adhesion Forces (Board/Oil), 3 Cohesion Forces (Oil)

Index A refers to molecules of board or paper (phase A), index B to oil molecules (phase B).
Resulting energy of interaction VR will be calculated as the sum of interaction energies VAA
and VBB minus interaction energy VAB:

VR = VAA + VBB 2 VAB (2)

Interaction energy VAB may be evaluated approximately from dissipation (index D). Dissipation
energy VABD for the interaction between phase A and phase B leads to the following result:

VAB VABD = (VAAD . VBBD )1/2 (3)

2-8
Modeling and Prediction of Static Electrification

Adhesion forces of macroscopic components may be obtained from resulting interaction


energy VR. Resulting interface forces and interfacial tension may be calculated by
differentiation of Equation 2:
AB A + B - 2 ( AD BD )
1/ 2
(4)

From Equation 1 4 we may obtain approximately the dissipation energy from measured
interfacial tensions AB , A and B (Antow rule). Interfacial tension according Equation 4 will
be valid for any solid/liquid interface:
SOLID / LIQUID = SOLID + LIQUID - 2 (SOLID LIQUID )
1/ 2
DISSIPATION DISSIPATION (5)

Evaluation of adhesion layer thickness as a function of local flow velocity may be obtained from
vector addition of the intermolecular adhesion forces and the macroscopic forces of oil flow
(Stokes forces). The intensity of the static electrification may be estimated from the thickness of
the adhesion layer. The objective of these intermolecular considerations is to clarify the complex
phenomena at solid/liquid interfaces. From this modelling may be concluded that adhesion forces
at the interface are proportional to the difference of the sum of both surface tensions and
dispersion of both surface tensions. From this modelling we may derive how to evaluate
adhesion forces and adhesion thickness from measured surface and interfacial tensions. These
molecular-physical considerations clarify, why any change of surface tension of solid or liquid
components will strongly affect static electrification (surface-reactive additives, impurities,
particles, and moisture). This behaviour should be considered as a way to force growth of the
thickness of adsorption layer (storage of transformer oil, board and paper under nitrogen
atmosphere to mitigate impact of moisture and oxygen on static electrification) because a few
adsorbed molecules may change the static electrification properties dramatically. Surface forces
of solid components relate to the molecule structure (intermolecular angle of impulses).
Mitigation of the static electrification propensity at the solid/liquid interface of the insulating
components may be achieved by surface-reactive additives (liquid phase) and/or by thermal or
laser treatment of board or kraftpaper (solid phase). Disadvantage of additives relates to the
large oil volume (reaction only at small regions of interface) and to the unknown impact on
ageing and breakdown voltage of transformer oil.

2-9
Modeling and Prediction of Static Electrification

Charge Transfer

Static electrification will occur, when oil ions at the oil-solid interface are separated and moved
with the oil flow. Separation is caused by energy transfer from streaming and uncharged oil
molecules to stationary oil ions bound to the interface. Energy is required for three different
functions. Energy is needed to:

1. overcome adhesion forces at interface (adhesion energy),

2. accelerate the separated oil ions up to flow velocity, and

3. to generate the electric field of the shifted oil ions (reverse current acc. Figure 5).

Static electrification would occur, if the flow energy were larger than sum of these three
elements. Energy transfer may be characterised by the effect of Stokes forces. From the
understanding of the separation process we have identified additional methods for mitigation of
electrostatic charging. An important factor relates to the viscosity characteristics of transformer
oil. Basically static electrification will decrease at low Stokes forces. Simultaneously static
electrification will increase at a smaller adhesion layer thickness caused by decreasing oil
viscosity. However, there is an impact of temperature and velocity on the viscosity of
transformer oil. A temperature increase from 20 to 80C reduces oil viscosity approximately one
decade. Additionally, viscosity of macromolecular fluids will increase non-uniformly at
increasing flow velocity (non-Newton fluids i.e. silicon oils at pour point). In general mitigation
of static electrification based on increasing temperature and oil velocity will fail based on
decreasing thickness of adhesion layer. Additionally increasing electrification will occur as a
result of Stokes forces. There is an increasing electrification summarising all effects. Reverse
current mitigates leakage current. Therefore any increase of conductivity will result in mitigation
of static electrification. In general leakage current will increase on rising temperatures
(decreasing thickness of adhesion layer).

Figure 5
Charge Separation and Transfer Induced by the Oil Flow (Simplified Scheme) V OIL Flow
Velocity, IA Leakage Current, IR Reverse Current, IA IR Resulting Leakage Current (Static
Electrification)

2-10
Modeling and Prediction of Static Electrification

An important factor for the static electrification is the transient characteristic of leakage current
versus transient characteristic of oil velocity (Figure 6). At the interface, a very low
concentration (charge density) of streaming oil ions is present immediately after start of
circulating pumps. The resulting electrical field in the direction towards regions of charge
separation is low because of a low charge density. Therefore the reverse current is low. The
impact of uncharged oil molecules decreases with increasing charge density of streaming oil
(Figure 5). From transient characteristics we know that static electrification of power transformer
could not reach the highest level of electrification immediately after start of the circulating
pumps. From these discussed mechanisms we can not derive the transition of leakage current
from a transient to a steady-state level. The steady state leakage current is a result of having a
low probability for any charge generation at the interface, when the charge density of oil flow
increases. According to the intermolecular impulse processes, each charge generation at the
interface will alter the local potential at the interface. Therefore at constant oil velocity
(i.e. constant impulse energy) the probability that local electron transfer will occur from
uncharged oil molecules to negatively charged paper molecule will decrease. From this
mechanism we could conclude that any simultaneous start of the circulating pumps will result
in high static electrification activity.

Figure 6
2
Transient Characteristic of Leakage Current Density (Measured at Duct Models) ,
P Pressure at Oil Inlet after Start of Circulating Pumps, A, B, C, D Leakage Current Density
at Different Distance from Inlet

2-11
Modeling and Prediction of Static Electrification

Conclusions

Any ranking of relevant parameters and mechanisms affecting static electrification of power
transformers is controversial. Most interactions could be taken from Figure 7. The main
emphasis is given to causal connections and their interactions. Based on this overview
quantitative assessments and ranking of the key parameters will be given (Table 2). Both figures
should consider our present technology base and restrictions of transformer design and
transformer operation for mitigation of static electrification. In general main attention for low
charging systems should be given to:
high cohesion forces (large thickness of adhesion layer by low oil viscosity
or changed viscosity-temperature or viscosity-velocity characteristic of oil)
high adhesion forces (large thickness of adhesion layer by surface-reactive additives,
changed impulse dynamics by thermal or laser treated paper or board surface)
high impact of moisture and adhesion layers at interface (change of surface or interfacial
tension caused by temperature/moisture transients or after refurbishing of transformers)
low oil velocity at interface of cooling ducts (not average of flow velocity)
low turbulence of oil flow (design of spacer and cooling ducts)
not too low an oil velocity (energy of oil flow less resulting energy for charge separation, ion
acceleration and electrical field may result in discharging and generation of wormholes)
start of circulating pumps step by step could mitigate static electrification
(transient of the leakage current)
low relaxation constant (permittivity or conductivity of insulating components enforce
reverse current)
oil conductivity no impact on charge generation (only on reverse current)
high electrification after refurbishing (extremely low reverse current by drying process,
non-uniform distribution of moisture, simultaneous start of circulating pumps, additional
adhesion layers of oxygen e.a.)

2-12
Modeling and Prediction of Static Electrification

Table 2
Impact of Parameters on Static Electrification of Power Transformers (Ranking)

2 high positively correlation - 2 high negatively correlation x unknown correlation


1 low positively correlation - 1 high negatively correlation pos. or neg. correlation
main correlation additional
parameter parameters
impulse energy 2 velocity, turbulence, temperature
oil turbulence 2 velocity, viscosity
oil velocity 2 temperature, viscosity, transformer design
Stokes friction 2 viscosity, temperature, velocity, type of oil
temperature 2 load, transformer design and operation
transformer design 2 Power, load, voltage, type, operation
transformer operation 2 load, cooling design
transformer refurbishing 2 moisture, particles, impurities
transient pump operation 2 load, temperature, cooling design
convective diffusion 1 type of oil and board
relaxation length 1 oil conductivity and permitivity, temperature
board surface treatment laser or thermal treatment
board type transformer design, temperature
impulse dynamics oil additives, impurities, surface treatment
interfacial tension additives, impurities, ageing, type of oil, temperature
oil additives conductivity, ageing, reliability, dissipation factor
oil type transformer operation and design, temperature
particles & impurities oil processing, refurbishing
board conductivity -1 type of board, temperature, moisture, impurities
board moisture -1 temperature transients, oil moisture, refurbishing
oil moisture -1 temperature transients, board moisture, refurbishing
adhesion forces -2 additives, impurities, moisture, temperature
adhesion layer -2 velocity, adhesion and cohesion forces, temperature
cohesion forces -2 viscosity, temperature, type of oil
oil conductivity -2 type of oil, temperature, moisture, impurities
oil viscosity -2 type of oil, temperature, velocity

2-13
Modeling and Prediction of Static Electrification

Figure 7
Impact of Different Parameters on Static Electrification of Power Transformers

2-14
Modeling and Prediction of Static Electrification

This paper deals with the physical background of static electrification in forced cooled power
transformers. Emphasis is given to intermolecular processes at the interface of oil/paper or
oil/board. Complex interactions of relevant characteristics and parameters have been discussed in
detail. The main goal is further the understanding of fundamental mechanism and interactions.
At present with our physical knowledge and limited experimental facilities, we could not
measure all the necessary intermolecular coefficients of macromolecules such as sulfatic
cellulose (board, paper) or transformer oil. Therefore numerical evaluations have not been made.
From theoretical considerations and experimental results we may derive, that intermolecular
impulse processes are the most important mechanism for static electrification. The discussed
model could be a key to mitigate static electrification of large power transformers based on more
reliable assessment of insulating components as well as on transformer design and operation.

6 References

1. T. V. Oomen: Static electrification control in power transformers. Proc. EPRI-Workshop


on Static Electrification in Power Transformers, Monterey CA (1986).

2. W. Schobel: Electrostatic charging of streaming insulating oils. Thesis (in German),


Graz University of Technology: Institute of High-Voltage Engineering (1990).

3. P. Sinz: Impact of moisture and particles on the electrical breakdown of insulating oil.
Thesis, Graz University of Technology: Institute of High-Voltage Engineering (1990).

4. G. Praxl: Electrostatic charging of solid/liquid components and methods characterising the


electrostatic charging tendency (ECT). Proc. CIGRE Joint Session 15/12.13, Paris (1992).

5. G. Praxl: Impact of particles on the electrostatic charging of large power transformers.


Proc. CIGRE Joint Session 15/12.13, Paris (1992).

6. G. Praxl: Impact of the electrostatic charging on the breakdown mechanism of solid/liquid


components. Proc. CIGRE Joint Session 15/12.13, Paris (1992).

7. G. Praxl: Impact of the oil flow on the electrostatic charging of large power transformers.
Proc. CIGRE Joint Session 15/12.13, Paris (1992).

8. G. Praxl, R. Stonitsch, G. Spandonis, C. Krause, H. P. Moser: Electrostatic charging in large


size models of power transformer cooling ducts. WG-Paper CIGRE Joint Working Group
12/15.13, Paris (1992).

9. D. W. Crofts: Addition of BTA to transformer oil for control of static electrification. EPRI
TR 102480, Project 1499-99, Proc. June 1993, Workshop on Static Electrification in Power
Transformers, San Jose CA (1992).

10. G. Praxl, R. Stonitsch, G. Spandonis, H. P. Moser, C. Krause, H. Brechna: Influence of


Kraft- and Nomexboard on the electrification of power transformers. EPRI TR 102480,
Project 1499-99, Proc. June 1993, Workshop on Static Electrification in Power Transformers,
San Jose CA (1992).

2-15
Modeling and Prediction of Static Electrification

11. G. Praxl, G. Lemesch: Electrification of power transformers screening of materials.


WG-Paper CIGRE Joint Working Group 12/15.13 TF-01, Palo Alto CA (1992).

12. W. Umdasch: Electrostatic flow measurement of fuel. Thesis (in German), Graz University
of Technology: Institute of High-Voltage Engineering (1993).

13. G. Praxl, G. Lemesch: Measurement of electrostatic charging tendency (ECT) with the
Cigre-test-cell, Status Report. WG-Paper CIGRE Joint Working Group 12/15.13,
Madrid (1993).

14. G. Praxl, G. Lemesch, W. Umdasch, K. Zikulnig: Impact of oil resistivity on the efficiency
of the ministatic testmethod. WG-Paper CIGRE Working Group 15.01, Gotenba,
Japan (1993).

15. Ch. Krause, E. Knoll, R. Stonitsch, J-J. Alf: Measurements showing the enhancement of
electrostatic charging by AC-voltage in a duct model and a full-scale power transformer.
EPRI TR 105019, Project 3334-78, Proc. May 1995, Workshop on Static Electrification in
Power Transformers, Milwaukee, Wisconsin (1994).

16. A.P. Wahabaugh, M. Zahn: A study of the effect of BTA on flow electrification using
rotating cylindrical electrodes. EPRI TR 105019, Project 3334-78, Proc. May 1995,
Workshop on Static Electrification in Power Transformers, Milwaukee, Wisconsin (1994).

17. G. Praxl: Progress Report of Task Force 1, Cigre Joint Working Group 12/15.13, Wrap-Up
Meeting, Milwaukee, Wisconsin (1994).

18. G. Praxl, G. Lemesch, R. Stonitsch, Ch. Krause: Measurement of the electrostatic charging
tendency. WG-Paper CIGRE Joint Working Group 12/15.13, Paris and Wrap-Up Meeting
Milwaukee (USA), 1994.

19. G. Praxl: Progress Report of Task Force 1, CIGRE Joint Working Group 12/15.13, Wrap-Up
Meeting, Milwaukee Wisconsin (1994).

20. K. Stonitsch: Impact of flow condition on the electrostatic charging of HV-power


transformers. Thesis (in German), Graz University of Technology: Institute of High-Voltage
Engineering (1995).

21. Ch. Krause, E. Knoll, J-J Alf, R. Stonitsch: Impact of AC-fields on electrostatic charging in
a full-scale power transformer. Proc. 9th International Symposium on High Voltage
Engineering, Graz (1995).

22. G. Lemesch, G. Praxl: Measurement of the electrostatic electrification tendency of


solid/liquid insulating materials. Proc. 9th International Symposium on High Voltage
Engineering, Vol. 1, No. 1006, Graz (1995).

23. G. Lemesch: Assessment of the electrostatic charging tendency of solid/liquid insulating


systems. Thesis (in German), Graz University of Technology: Institute of High-Voltage
Engineering (1996).

2-16
Modeling and Prediction of Static Electrification

24. G. Praxl, G. Lemesch, R. Stonitsch: Electrostatic charging in high voltage power


transformers - methods and devices. 5th Hflers Days of High Voltage Technique,
Potoroz/Slovenia (1996).

25. G. Praxl, G. Lemesch: Static electrification in power transformers, screening of materials,


Preliminary Draft 08/1996, CIGRE JWG 12/15.13-TF01, Paris (1996).

26. G. Praxl, K. Neuner: Electrostatic charging at liquid/solid dielectrics - interfacial


phenomena. Cigre JWG 12/15.13, Paris (1996).

27. G. Praxl, Lemesch: Static electrification in power transformers, screening of materials.


Preliminary Draft, CIGRE JWG 12/15.13-TF01, Graz (1997).

28. G. Praxl, G. Lemesch: Static electrification in power transformers, screening of materials,


Final Report, CIGRE JWG 12/15.13-TF01, Sydney (1997).

29. G. Praxl, G. Lemesch: Methods and devices to evaluate electrostatic charging tendency of
solid/liquid insulating components (in German). Proc. 43. Internationales
Wissenschaftliches Kolloquium der Technischen Universitt Ilmenau (1998).

30. K. Neuner: Electrostatic charging at liquid/solid dielectrics - interfacial phenomena. Thesis


(in German), Graz University of Technology: Institute of High-Voltage Engineering (1998).

31. G. Newesely, H. Halbwirth, A. Gefaell: Technol transformer oils. Seminar (in German),
Vienna (1999).

2-17
Modeling and Prediction of Static Electrification

2.2 Study on Increase of Electrostatic Charging Tendency of Insulating Oil


and the Diagnosis on Streaming Electrification

M. Tsuchie and N. Hosokawa S. Isaka


Mitsubishi Electric Corporation Toshiba Corporation
651, Tenwa, 2-1, Ukishima-cho, Kawasaki-ku,
Ako, 678-0256 Japan Kawasaki, 210-0862 Japan

H. Miyao T. Ishii
Hitachi, Ltd. Yuka Industries Co., Ltd.
1-1, Kokubo-cho 1-chome, 7-15 Hutaba, 1-chome, Shinagawa-ku,
Hitachi, 316-8501 Japan Tokyo, 142-0043 Japan

S. Kobayashi, T. Kobayashi, T. Ono H. Okubo


Tokyo Electric Power Company Nagoya University
1-3, Uchisaiwai-cho 1-chome, Chiyoda-ku, Furo-cho, Chikusa-ku
Tokyo, 100-0011 Japan Nagoya, 464-8603 Japan

Abstract

The investigation of electrostatic charging tendency (ECT) of new oil has been studied since
1970s, but that of aged oil has never been studied. e studied following three factors on ECT and
method to prevent the increase of ECT. (1) Effect of the increasing of ECT of new oil by aging.
(2) The influence on new oil mixed with oxidized oil or aged oil. (3) The influence on aged oil
by exposure to air.

It was found that the ECT of new oil increased in the case of containing small amounts of
oxygen and that the ECT of new oil mixed with aged oil increased. The ECT of aged oil exposed
to air changed variously depending on the condition of oil.

It is confirmed that benzotriazole (BTA) is effective to prevent the increasing of ECT of aged oil.

Streaming electrification of the transformers in service can be diagnosed by four measurement


indices of such as (1) ECT, (2) leakage current from transformer windings, (3) partial discharge,
and (4) supporting properties relating to ECT such as dissolved copper in oil.

Measurement of dissolved copper is supplemental method to estimate the deteriorating state of


oil. Measurement of partial discharge is effective to detect the partial discharge by streaming
electrification, but it is not suitable as preventive measurement of streaming electrification.
Leakage current from transformer windings seems to be the most reliable method to prevent
streaming electrification among above four methods, but it is necessary to stop transformer
operation for the measurement. ECT is the most practical diagnosis method among above four
methods.

2-18
Modeling and Prediction of Static Electrification

Introduction

It is a known fact that the streaming electrification takes place between solid insulator and
insulating oil flowing inside a transformer, and various researches have been made to this regard
since the latter half of the 1970s1-4. Studies on the streaming electrification have so far been made
mostly on the assumption of the initial stage after the transformer is produced, with due attention
paid on the flow properties of insulating oil and the transformer insulation structure.

The insulating oil, used over a long time period inside a transformer, is considered to go through
changes as time elapses. It is thought that the following four factors cause changes in the
characteristics of the insulating oil.

1. Deterioration due to aging caused by a slight amount of oxygen contained in the transformer
oil

2. Deterioration of filled oil as it got mixed with the impregnated oil, which was factory oil,
used for drying at the time of production of the transformer

3. Exposure of transformer oil to air during transformer oil treatment

4. Mixing of aged oil with new insulating oil at the time of transformer oil exchange

Since no study has so far been made on the cause of increase in electrostatic charging tendency
(hereafter ECT) of the insulating oil used over a long time, we took up this matter as the theme
of our study.

Benzotriazole (BTA) has so far been widely used as an inhibitor for streaming electrification in
5
Japan . As for the new insulating oil, investigation has been made on electrification
characteristics, etc., but no investigation has so far been made on the aged transformer oil.

Further, different organizations have been carrying out diagnosis of streaming electrification
differently in Japan, and there was no standard method of diagnosis. In recent years, however,
investigation was made by Electric Technology Research Association on the electrification
characteristics of field transformer oil, and on the basis of the results thus obtained, the diagnosis
items for insulating oil electrification characteristics and the criteria were determined.

Study on the Increase of ECT of Insulating Oil

Causes of Increase in ECT and Verification Method

The ECT of insulating oil is considered to increase due to the four factors shown in Figure 1.
Verification tests were conducted on each factor in the following manner.

2-19
Modeling and Prediction of Static Electrification

Deterioration Due to Aging Caused by a Slight Amount of Oxygen Contained in


the Transformer Oil

Four liters of insulating oil was heated together with 2700 cm2 copper wire at 120C for
450 hours. Three kinds of insulating oil (A, B, C) refined in Japan were tested after degassing.
However, one sample was tested with a slight amount of oxygen left. The ECT (miniature
electrostatic tester method) and dissipation factor were measured at the initial condition, and then
taking samplings 100 hours after heating, 250 hours after heating and 450 hours after heating.
When calculated using half-reduction rule of 7C and supposing the transformer oil of 60C as
the base, the heating condition: 120C 450 hours, is an acceleration test equivalent to
approximately 20 years.

Increase of ECT by a slight amount of oxygen, heat and copper catalysis etc.

New Oil Deterioration due to aging for long term Increase of ECT

Increase of ECT by deterioration of filled new oil mixed with impregnated oxidized factory oil

Residual deteriorated Deteriorating of filled new Increase of ECT by


factory oil oil mixed with changing in quality
impregnated oxidized oil of insulating oil
for the long term

Increase of ECT by Oil Treatment such as Overhaul

Aged Oil Deterioration due to


exposure to air

Residual aged oil Deterioration of filled new


oil mixed with
impregnated oxidized aged
oil

Figure 1
Flow of Insulating Oil before ECT Increase

Deterioration of Filled Oil as it got Mixed with the Impregnated Oil, which was
Deteriorated Factory Oil Used for Drying at the Time of Production

Here, the deteriorated factory oil left inside a transformer is assumed to have got mixed with the
new insulating oil. The new insulating oil suffered by forced deterioration is regarded as the
deteriorated factory oil. A sample is made by mixing 10% deteriorated oil with 90% new
insulating oil of the same brand. The initial characteristic values are confirmed immediately after
mixing, before carrying out the heating test as mentioned above.

2-20
Modeling and Prediction of Static Electrification

Air Exposure Test of the Transformer Oil

Here, the transformer oil is assumed to get temporarily exposed to air at the time of oil treatment.
The oil is exposed to air for 72 hours at the room temperature, while the ECT and dissipation
factor are measured every 24 hours, before carrying out the heating test as mentioned above
(heating condition: 120C 450 hours). Three kinds of sample oil (D, E, F) with different
brands, collected from the field transformer, are used as the aged oil.

Mixing of Aged Transformer Oil with New Oil at the Time of Exchanging the
Transformer Oil

Here, the aged transformer oil left inside a transformer is assumed to get mixed with the new
insulating oil. A sample is made by mixing 10% aged transformer oil with 90% new insulating
oil of the same brand. The characteristic values are confirmed immediately after mixing, before
carrying out the heating test as mentioned above (heating condition: 120C 450 hours).

Results of Verification Test

Deterioration Due to Aging Caused by a Slight Amount of Oxygen Contained in the


Transformer Oil

Figure 2 shows the aging properties of the ECT of the three kinds of sample oil. Samples A, B
and C-1 are tested after degassing, and C-2 with such a slight amount of oxygen contained in oil,
as 2000~3000 ppm (8000~12000 ppm when including their above the oil surface), with C-1 and
C-2 being of the same brand but of different lot.

Figure 2
Aging Properties of ECT

2-21
Modeling and Prediction of Static Electrification

The ECT stands at low level in case of sample A and sample B, with the sample A showing a
peak value of ECT when measured after 100 hours heating, but the value going lower than the
initial value when measured after 250 hours or 450 hours heating. Sample B also shows a peak
value of ECT, though not as conspicuous as in sample A after 100-hour heating, with the value
going lower than the initial value after 250 hours or 450 hours heating. The ECT does not show a
conspicuous increase in the case of sample C-1. However, in the case of sample C-2, which
contains a slight amount of oxygen, the ECT is confirmed to increase as the heating time
increases, the value being particularly large after the first 100 hours of heating. It can, therefore,
be concluded that the existence of oxygen has some influence on the aging properties of the
ECT.

Deterioration of Filled Oil as it gets Mixed with the Impregnated Oil Produced by the
Impregnation of Factory Oil, Used for Drying at the Time of Production into the Insulator

Figure 3 shows the heating test results of the deteriorated oil mixed with the new insulating oil.
When two kinds of insulating oil are mixed, it is normally estimated that the characteristic value
is the value obtained through proportional distribution of the value before mixing at the rate of
the mixing ratio. The actually measured values of new insulating oil and mixed oil, the
proportional distribution values due to the mixing with the oil subjected to forced deterioration of
the three kinds of insulating oil A, B and C-2 are given in Tables 1- 3. The ECT of the oil
subjected to forced deterioration is 139 pC/ml for oil A, 193 pC/ml for oil B, and 267 pC/ml for
C-1. The substantial increase of ECT due to mixing is seen in oil C-2 at 0 hour and 100 hours of
heating, the value being 500% up the proportional distribution value. The test of oil C-2 has been
carried out under the condition that a slight amount of oxygen is contained in the oil as shown in
Figure 2.

New oil A+ Oxidized oil from A


New oil B+ Oxidized oil from B
New oil C-2 + Oxidized oil from C-2
10000

1000
ECT (pC/ml)

100

10

1
0 100 200 300 400 500

Aging time (hr)

Figure 3
Heating Test Results of Mixed Oxidized Insulating Oil with New Insulating Oil

2-22
Modeling and Prediction of Static Electrification

Air Exposure Test of Transformer Oil

Figure 4 shows the test results of the aged oil after exposed to air. The maximum air exposure
time here is 72 hours, but the substantial trend of change is seen after the first 24 hours of
exposure, with the change being small during the next 48 hours.

Table 1
Trend of Increase in ECT of Oil A

Heating Actually Actually Proportional ECT increase


time measured value measured value distribution value factor
of new insulating of mixed oil (pC/ml) (%)
oil(pC/ml) (pC/ml) a b a/b 100

0 hour 4.9 45.6 18.3 249

100 hours 42.7 12.2 52.3 23

250 hours 2.5 21.7 16.2 134

450 hours 0.5 2.9 14.4 20

Table 2
Trend of Increase in ECT of Oil B

Heating Actually Actually Proportional ECT increase


time measured value measured value distribution value factor
of new insulating of mixed oil (pC/ml) (%)
oil(pC/ml) (pC/ml) a b a/b 100

0 hour 13.1 59.2 31.1 190

100 hours 15.9 115 33.6 342

250 hours 9.6 32.9 27.9 118

450 hours 3.7 24.6 22.6 109

Table 3
Trend of Increase in ECT of Oil C-2

Heating Actually Actually Proportional ECT increase


time measured value measured value distribution value factor
of new insulating of mixed oil (pC/ml) (%)
oil(pC/ml) (pC/ml) a b a/b 100

0 hour 3.9 166.9 30.2 552

100 hours 218.1 1247.3 223.0 559

250 hours 384.7 893.5 372.9 240

450 hours 667.7 520.2 627.6 83

2-23
Modeling and Prediction of Static Electrification

Viewed kind-wise, the ECT tends to decrease in the case of oil D when exposed to air, while
little change is seen in the case of oil E. In the case of oil F the value tends to increase, indicating
that the effect is different according to the kind of oil.

Furthermore, the result of heating test after exposure to air for 72 hours is given in Figure 5, and
the test result of the aged oil when heated without being exposed to air is given in Figure 6 for
comparison. Compared with the test result of aged oil only as shown in Figure 6, the heating test
result of the oil exposed to air shows ECT value not at so high a level. It is, therefore, considered
that the effect is large at the early stage of exposure as shown in Figure 4.

10000

before explosure in air


after 24hr
after 48hr
E CT (pC/ml)

after 72hr

1000

100
A ged oil A ged oil A ged oil
Oil S amples

Figure 4
ECT after Exposure to Air for 72 Hours

10000

1000
ECT(pC/ml)

100

10

1
Aged oil c
Aged oil d
0.1 Aged oil e
0 100 200 300 400 500
Aging time(hr)

Figure 5
Heating Test Results after Exposure to Air for 72 Hours

2-24
Modeling and Prediction of Static Electrification

Mixing of Aged Transformer Oil with New Insulating Oil at the Time of Transformer Oil
Exchange

Figure 7 shows the test result when aged transformer oil is mixed with new insulating oil.
Making evaluation of the mixing of aged transformer oil with new insulating oil in the same
manner as in the case of mixing test of deteriorated oil and new insulating oil refers to comparing
proportionate distribution values of the heating test result of new insulating oil only as shown in
Figure 1 and the heating test result of aged transformer oil only as shown in Figure 6 with the
values in Figure 7. The values obtained through the mixing test of (oil B + oil E) at 250 hours
heating and 450-hour heating are equivalent to 600~900% of the proportionate distribution
values, tending to approach closer to the characteristics of aged oil E, which is amounting to only
10% of the total quantity of oil.

10000

1000

100

10

1 A ged oil c
A ged oil d
A ged oil e
0. 1
0 100 200 300 400 500

Ag i n g t i me ( h r )

Figure 6
Heating Test Results of Aged Transformer Oil without Exposure to Air

2-25
Modeling and Prediction of Static Electrification

10000

1000
ECT (pC/ml)

100
New oil A+ Aged oil D
New oil B+ Aged oil E
10
New oil C-2 + Aged oil F

0.1
0 100 200 300 400 500
Aging time (hr)

Figure 7
Heating Test Results of Mixed Aged Transformer Oil with New Insulating Oil

Study on Suppression of the Increase of ECT in Insulating Oil

Verification Method of ECT Suppression

The use of BTA started in 1980s in Japan to restrain the streaming electrification 5, with the
addition of BTA approved by JIS in 1988. However, before that most of the transformer oil did
not have BTA added to it. A test was, therefore, conducted to prove that BTA added to the
transformer oil in operation for long term has the effect to suppress the streaming electrification.
Three kinds of insulating oil samples taken from field transformers are added with 10 ppm BTA,
before being subjected to heating test as in the case of verification test of the increase of ECT.

Result of Verification Test

The test result after adding BTA to aged transformer oil is shown in Figure 8. Compared with the
values of the aged transformer oil only in Figure 6, the ECT values in all cases are found to be at
lower levels. This confirms that BTA does have an ECT suppression effect as has been reported
for the new insulating oil added with BTA.

2-26
Modeling and Prediction of Static Electrification

10000

1000
ECT (pC/ml)

100

10

Aged oil D + BTA 10ppm

1 Aged oil E + BTA 10ppm


Aged oil F +BTA 10ppm

0.1
0 100 200 300 400 500
Aging time (hr)

Figure 8
Heating Test Results of Transformer Oil after Adding BTA

Diagnosis of Streaming Electrification

Mechanism and Diagnosis of Streaming Electrification

It is known that streaming electrification inside a transformer undergoes three processes: charge
transfer, charge separation and charge relaxation, while ECT, dissipation factor and specific
resistivity are counted as the characteristics of insulating oil. The relation between streaming
electrification processes and insulating oil characteristics is shown in Figure 9. In the case of the
first two streaming electrification processes i.e. charge transfer and charge separation, the static
electricity is generated, which can be evaluated through ECT measurement. As for the charge
relaxation that takes place when the excess charge gets neutralized for its balance, it is
considered to have mutual relation with the specific resistivity of insulating oil. It has so far been
taken for granted that there is a mutual relation between ECT and dissipation factor. The
dissipation factor is also thought mutually related to the charge transfer. It is, therefore, the
measurement of leakage current from high voltage (HV) windings of a transformer, which flows
through the neutral bushing grounded that enables overall evaluation inside a transformer
including all three processes mentioned above. However, there is a restrictive condition that the
leakage current from HV windings can be measured only when the transformer is not energized.
Hence, there is no other way but to use ECT and dissipation factor for diagnosis of streaming
electrification at steady state.

2-27
Modeling and Prediction of Static Electrification

Phenomena in insulating Process for streaming Phenomenon in


electrification transformer

Charge transfer
ECT
Leakage current
Charge separation
from windings
Dissipation factor
Charge relaxation
Specific resistivity

Figure 9
Relation between Streaming Electrification Processes, and Phenomena in Insulating Oil
and the Transformer

Increase in ECT of Insulating Oil and Diagnosis

In view of the phenomenon of the increase in ECT of insulating oil and the design-related factors
about streaming electrification, which have been so far studied since 1970, the development flow
of streaming electrification and the diagnosis method for each aspect are as shown in Figure 10.
It is possible to detect the design defects such as excessively high flow rate of oil, etc. by
measuring the leakage current from windings at the time of factory test.

The defect mode from increase of ECT in insulating oil to electrostatic discharge, on the other
hand, cannot be detected at the time of factory test, calling for periodical measurement at field.
The safe operation of a transformer after the increase of ECT in insulating oil is confirmed by
means of the leakage current from windings. As the ECT increase develops, electrostatic
discharge takes place, calling for dissolved gas analysis, and discharge pulse measurement or
acoustic measurement to check for the discharge. However, it is desirable from the standpoint of
preventive maintenance to carry out diagnosis before discharge occurs.

Further, the oxidizing properties of insulating oil such as dissolved copper, etc. do give a clue to
the condition of insulating oil, but the dissolved copper is not necessarily in good mutual relation
with ECT as shown in Figure 11. However, a high level of ECT value is seen when the dissolved
copper is excessively large in quantity, so that it can be used as an auxiliary means.

2-28
Modeling and Prediction of Static Electrification

Figure 10
Progress Flow of Streaming Electrification

2-29
Modeling and Prediction of Static Electrification

1000

100
ECT ( C/m3)

10

1
0 0.2 0.4 0.6 0.8 1
Dissolved copper in oil (ppm)

Figure 11
Relation between Dissolved Copper Quantity in Oil and ECT

Criteria for Electrification Properties

The criteria for leakage current from windings largely depends on the insulation structure of
transformer. Table 4 shows the reference values obtained from the discharge values of lower
limit leakage current from windings, etc. provided by various transformer makers through careful
study and review.

Table 4
Reference Values of Transformer Leakage Current from Windings

Type Core Type Shell Type


Item Transformers *1 Transformers *2

Leakage Current Below Below


from Windings -0.3 to -1.5A/leg -1.75A/group
*1 : 8 Companies
*2 : 1 Company

The criteria for electrification properties of insulating oil are given in Table 5. From the relation
between leakage current from windings and ECT as shown in Figure 12, 500 pC/ml is considered
as the criteria for ECT of insulating oil. Since Figure 12 shows the data of several transformer
makers, the X-axis represents the value of leakage current when the reference criteria of each
maker is supposed to be 100%. The mark in the figure indicates the leakage current from
windings when a transformer, without any countermeasures against streaming electrification,

2-30
Modeling and Prediction of Static Electrification

causes discharge, mark the leakage current from windings when a transformer, without any
countermeasures against streaming electrification, is in normal state, and mark the leakage
current from windings when a transformer has the countermeasures against streaming
electrification duly taken. The criteria for ECT is set to 500 pC/ml when the regression line of ,
more likely to cause electrostatic discharge, and cross the 100% leakage current from
windings.

Table 5
Criteria for Electrification Properties of Insulating Oil

Item Criteria
ECT at Installation Below 200 pC/ml
Dissipation Factor Below 0.3%
at Installation
ECT in Service Below 500 pC/ml
Dissipation Factor Below 0.5%
in Service

Figure 12
Relation between Leakage Current from Windings and ECT of Insulating Oil

The criteria for dissipation factor is set to 0.5% on the basis of the relation between ECT and
dissipation factor obtained through the measurements made on 75 units of field transformer as
shown in Figure 13.

The criteria at installation is set to 200 pC/ml obtained by adding the double standard deviation
to the value of regression line extrapolated to 0 year based on the relation between ECT and
number of service years of a transformer shown in Figure 14. The dissipation factor at insulation
is set to 0.3% which is equivalent to 200 pC/ml in Figure 13.

2-31
Modeling and Prediction of Static Electrification

Figure 13
Relation between Dissipation Factor and ECT

Figure 14
Relation between Electrification Properties of Field Transformer Oil and Number of Service
Years of Transformer

2-32
Modeling and Prediction of Static Electrification

Conclusion

1. The heating test using new insulating oil reveals that the ECT tends to increase with the
elapse of time when a slight amount of oxygen is contained in the oil as compared with the
test with no oxygen included.

2. In the case of mixing test, when new insulating oil is mixed with aged transformer oil, the
ECT may get increased depending on the kind of oil.

3. The ECT may get changed when the aged transformer oil is exposed to air; it may increase or
decrease according to the kind of oil.

4. Addition of BTA to aged transformer oil shows suppression effect against the increase of
ECT.

5. The ECT and dissipation factor are the items of periodical diagnosis of streaming
electrification for maintenance control of an operating transformer, with the criteria for ECT
set to 500 pC/ml and the criteria for dissipation factor set to 0.5%.

References

1. T. Takagi, T. Ishii, T. Okada, K. Kurita, R. Tamura, H. Murata, Reliability Improvement of


500 kV Large Capacity Power Transformer, CIGRE Paper, 12-02, 1978.

2. R. Tamura, Y. Miura, T. Watanabe, T. Ishii, N. Yamada, T. Nitta, Static Electrification by


Forced Oil Flow in Large Power Transformer, IEEE PES Meeting F 79 640-4 (1979).

3. S. Shimizu, H. Murata, M. Honda, Electrostatics in Power Transformers, IEEE Trans.


PAS-98, No 4 pp. 1244-1250, (1979).

4. M. Higaki, Y. Kako, M. Moriyama, M. Hirano, K. Hiraishi, K. Kurita, Static Electrification


and Partial Discharges Caused by Oil Flow in Cooled Core Type Transformers IEEE PAS
Meeting F 79, 231-2 (1979).

5. M. Yasuda, K. Goto, H. Okubo, T. Ishii, E. Mori, M. Masunaga, Suppression of Static


Electrification of Insulating Oil for Large Power Transformers IEEE PES Winter Meeting
WN 82 197-2 (1982).

2-33
Modeling and Prediction of Static Electrification

2.3 The Electrostatic Charging Tendency of In-Service Oils and The


Evaluation Method of Leakage Current from Transformer Windings

H. Miyao
E. Mori
Hitachi, Ltd.
1-1, Kokubu-cho 1-chome, Hitachi, 316-8501 JAPAN

S. Isaka
Toshiba Corporation
2-1, Ukishima-cho, Kawasaki-ku, Kawasaki, 210-0862 JAPAN

M. Tsuchie
Mitsubishi Electric Corporation
651 Tenwa, Ako, 678-0256 JAPAN

K. Takamoto
Kansai Tech Corporation
3-1-176, Fukuzaki-cho, Minato-ku, Osaka, 552-0013 JAPAN

S. Kobayashi
T. Kobayashi
T. Ono
Tokyo Electric Power Company
1-3, Uchisaiwai-cho 1-chome, Chiyoda-ku, Tokyo, 100-0011 JAPAN

H. Okubo
Nagoya University
Furo-cho, Chikusa-ku, Nagoya, 464-8603 JAPAN

Abstract

The electrostatic charging tendency (ECT) was measured for oils from 75 in-service power
transformers (voltage classes, 275 kV and 500 kV) and the influences of operation years,
operation conditions, addition of BTA (1,2,3-benzotriazol) and repair work history in the field
were studied. It was found that the ECT increased with operation years. For BTA-added oils the
ECT also showed an increasing tendency with operation years except for low ECT at the
beginning of transformer operation. The ECT change with operation years had no significant
relationship with oil types, transformer types and operation conditions. It was also found that the
ECT had good correlation with power factor and volume resistivity of oil, but Cu content in oil
had no positive correlation, except for high ECT oils with high Cu content. An estimation
method of the leakage current from transformer windings by using ECT of oil was investigated.
It was found that the correlation between the leakage current and the value of (ECT) (flow rate
of windings) was good. This method was judged to be useful as an accurate estimation method
for the leakage current.

2-34
Modeling and Prediction of Static Electrification

Introduction

In the mid 1970s, problems with power transformers caused by flow electrification were
reported in Japan. Since then many studies on flow electrification of transformers have been
performed [1]-[6] and it has become clear that the flow electrification phenomenon of the power
transformer is influenced by various factors such as flow velocity, flow passage structure,
insulation treatment and characteristics of insulating oil, especially the ECT of oil. Recently,
there has been some concern that the increase of ECT by the aging of insulating oil may possibly
cause flow electrification trouble in transformers. In 1990 EPRI reported the study of ECT of oils
from 124 in-service transformers [7]. But in Japan there have been few reports about the ECT of
oils from in-service power transformers to date. The purpose of the present study was to get
information about the increasing ECT of oils for in-service power transformers by investigating
the relationships between ECT and other properties of insulating oil and operation years of
75 in-service transformers. These data were obtained by the expert committee of Electric
Technology Research Association.

It was considered that the flow electrification diagnosis of power transformers should be judged
by the leakage current from the transformer winding to get good diagnosis accuracy. Leakage
current was chosen because it gives information about the flow electrification phenomenon in the
whole transformer. However it is not possible to measure leakage current during transformer
operation. Then, the estimation method of leakage current from the ECT of transformer oil,
which could be done without suspension of the transformer operation, was discussed.

ECT of In-Service Oils

Study Method of In-Service Oils

75 transformers of 275 kV and 500 kV voltage rating were selected based on getting many
combinations of manufacturers, ages, transformer types, operation conditions, addition of BTA
and varieties of oil. For testing the ECT of the service-aged oils, samples were obtained by
avoiding contact with air as much as possible and they were stored in metal containers under a
nitrogen atmosphere to maintain their integrity. The tests were performed within a week after
sampling. The test method to determine the ECT of the samples used the modified mini-static
tester seen in Figure 1; this was the same method as reported in the EPRI report [7]. The
method used pressurized nitrogen to force oil to flow through a special cellulose filter
(thickness: 0.16 mm, pore size: 22 m) (which was held perpendicular to the flow). The resultant
charge generated was measured as a current using an electrometer and was recorded on a strip
chart.

Investigation Results of ECT

Figures 2, 3 and 4 show the relationship between the ECT and operation years. Judging from all
the data, the ECT had a tendency to increase with operation years and shows large scattering.
The ECT of BTA-added oil had a tendency to show low ECT at an early operation stage
(before 10 years) and after that it tended to increase gradually with operation years.

2-35
Modeling and Prediction of Static Electrification

The increasing tendency of ECT with aging did not depend on varieties of oil, but the ECT of
some transformer manufacturers showed high values after 15 to 25 operation years. In this
investigation the ECT of transformers which had a repair work history in the field did not
provide high ECT values compared with other transformers.

Figure 5 compares the cumulative ECT distribution of samples taken from generator and
transmission transformers. In general, the generator transformers tended to be heavily loaded
while the transmission transformers were light to moderately loaded. For both types of
transformers, 90% of the oils had an ECT under 300 pC/ml and there was almost no difference in
ECT values between the two. Figure 6 provides a comparison of the cumulative ECT distribution
of samples taken from different age transformers. Fifty percent of the transformers built before
1970 had an ECT of over 200 pC/ml and the proportion of ECT over 200 pC/ml became smaller
for newer transformers.

Figure 7 compares the cumulative ECT distribution of samples taken from different
manufacturers. The ECT values of 90% of samples were less than 300 pC/ml, but some
manufacturers tended to have a greater number in the 201-300 pC/ml range than other
manufacturers did.

Figure 8 compares the cumulative ECT distribution of varieties of oils. No appreciable difference
was apparent.

Figure 9 shows a graph of ECT against power factor at 80C. It was found that the ECT had a
good correlation with power factor and the ECT had a tendency to increase with power factor.

Figures 10 and 11 show the ECT plotted against the volume resistivity at 80C ( 1: 1 minute
value) and room temperature ( o: 0 minute value) respectively. Both graphs appeared to have
some correlation as a log function but o had a little better correlation with the ECT than 1 did.

Figure 12 shows the relationship between Cu content in oil and the ECT. It was found that
Cu content had no positive correlation, except for high ECT oils with high Cu content.

Evaluation Method of the Leakage Current

Flow Electrification Diagnosis and Leakage Current from Transformer Windings

It was considered that the flow electrification diagnosis of power transformers should be judged
by leakage current from transformer windings to get good diagnosis accuracy. The leakage
current was chosen because flow electrification phenomenon is determined by the total amount
of static charge generation and relaxation, and the leakage current from transformer windings is
the best measurable parameter which includes both charge generation and relaxation in the
transformer. Table 1 summarizes controlled values and their background data of leakage current
for Japanese transformer manufacturers. Basically each controlled value was determined from
the leakage current at the lower discharge limit. The limit was the value compensated by
temperature and AC voltage application with an added margin. In Japan each transformer
manufacturer judges whether the transformer condition is dangerous or not regarding flow
electrification discharge.

2-36
Modeling and Prediction of Static Electrification

It is not possible to measure the leakage current while in-service because of superimposition of
the terrestrial magnetism DC current, so that it is necessary to have an estimation method for
leakage current.

Estimation Method of the Leakage Current

As an estimation method of the leakage current, the ECT of transformer oil, which could be
obtained at the in-service transformer, was investigated. Figure 13 shows an example of
relationship between the ECT and the leakage current. In this figure the ECT data were obtained
by the mini-static tester mentioned earlier for core type transformers. The correlation between
ECT and the leakage current was not so good, so that the relationship between (ECT flow rate:
per winding leg for a core type transformer or per winding group for a shell type transformer)
and the leakage current was investigated to improve the accuracy of the leakage current
estimation from the ECT. The parameter (ECT flow rate) is thought to have better accuracy
than the ECT for the following two reasons.

1. The dimensions of (ECT flow rate) are A (C/s), the same as the leakage current.

2. Generally flow rate is almost proportion to the size of the windings (surface area of
windings), so that the variation of the leakage current with the transformer size can be taken
into account by multiplying the flow rate by ECT.

Figure 14 shows the relationship between (ECT flow rate/leg) and the leakage current for one
core type transformer manufacturer. The correlation of the leakage current with (ECT flow
rate/leg) was better than the one with ECT. In this way it was found that the estimation accuracy
of the leakage current from ECT could be improved by taking account of (ECT flow rate)
which reflected the transformer size.

Figure 15 also shows the relationship between (ECT flow rate) and the leakage current for
plural core type transformer manufacturers and one shell type transformer manufacturer. For
core type transformers the correlation of the leakage current with (ECT flow rate/leg) becomes
fairly good. Also for shell type transformers there was a correlation of the leakage current with
(ECT flow rate/group).

There is another way to improve the accuracy of the leakage current estimation, that is
(ECT o). This parameter is considered to be able to take into account charge relaxation effect
of the leakage current. Figure 16 shows the relationship between (ECT o) and the leakage
current for shell type transformers of one manufacturer. It was found that the correlation between
(ECT o) and the leakage current was improved compared to (ECT flow rate) and the
accuracy to estimate the leakage current seemed to be better. It was decided that this method was
useful for shell type transformers. However there were not enough o data for core type
transformers, so more data should be obtained.

2-37
Modeling and Prediction of Static Electrification

Conclusion

The relationships between ECT and other properties of insulating oil and operation years of
75 in-service transformers in Japan were investigated. An estimation method of leakage current
from ECT of oil for flow electrification diagnosis of power transformers was considered. The
results can be summarized as follows.

1. It was found that the ECT increased with operation years and BTA added oils showed low
ECT at the beginning of transformer operation, but the ECT of oil alone immediately showed
increasing tendency with operation years.

2. It was found that the ECT change with operation years had no significant relationship with
oils, transformer types and operation conditions.

3. It was found that the ECT had good correlation with power factor and volume resistivity of
oil, but Cu content in oil had no positive correlation, except for high ECT oils with high
Cu content.

4. It was found that the correlation between the leakage current and the value of (ECT) (flow
rate of windings) was good. The proposed method could be used to estimate leakage current
with good accuracy.

References

1. M. Higaki, Y. Kako, M. Moriyama, et al., Static Electrification and Partial Discharges


Caused by Oil Flow in Forced Oil Cooled Core Type Transformers, IEEE Trans. PAS-98,
pp. 1259-1267, 1979.

2. S. Shimizu, H. Murata and M. Honda, Electrostatics in Power Transformers, IEEE Trans.


PAS-98, pp. 1244-1250, 1979.

3. R. Tamura, Y. Miura, et al., Static Electrification by Forced Oil Flow in Large Power
Transformer, IEEE PES Summer Meeting, F79640-4, 1979.

4. R. Tamura, K. Kurita and H. Murata, Flow Electrification in Large Power Transformer,


JIEE, 99, pp. 913-920, 1979.

5. H. Okubo, M. Ikeda, M. Honda, T. Yanari, Charging Tendency Measurement for


Transformer Oil, IEEE PES Winter Meeting, A79 151-4, 1979.

6. M. Higaki, H. Miyao, K. Endou, H. Ohtani, Calculation of Potential Distribution Caused by


Static Electrification Owing to Oil Flow in an Oil-Paper Insulation System and Its
Application to Partial Discharge Phenomena in Oil, IEEE PES Winter Meeting,
F79 230-4, 1979.

7. P. J. Griffin, The Electrostatic Charging Tendency of In-service Oils, Proceedings of the


Second EPRI Workshop Static Electrification in Power Transformers, 3-5, Princeton,
NJ, 1990.

2-38
Modeling and Prediction of Static Electrification

Figure 1
Ministatic Tester

Figure 2
Relationship between ECT and Operation Years (Oil Varieties)

2-39
Modeling and Prediction of Static Electrification

Figure 3
Relationship between ECT and Operation Years (Manufacturers)

Figure 4
Relationship between ECT and Operation Years (Influence of Repair Work at Site)

2-40
Modeling and Prediction of Static Electrification

Figure 5
Distribution of ECT in Way of Uses

Figure 6
Distribution of ECT in Operation Years

2-41
Modeling and Prediction of Static Electrification

Figure 7
Distribution of ECT by Manufacturers

Figure 8
Distribution of ECT in Varieties of Oil

2-42
Modeling and Prediction of Static Electrification

Figure 9
Relationship between ECT and tan (80C)

Figure 10
Relationship between ECT and tan 1 (80C)

2-43
Modeling and Prediction of Static Electrification

Figure 11
Relationship between ECT and o (R.T.)

Figure 12
Relationship between ECT and Cu Content in Oil

2-44
Modeling and Prediction of Static Electrification

Table 1
Examples of Controlled Values of the Leakage Current from Windings and their Determination Back Ground

2-45
Modeling and Prediction of Static Electrification

Figure 13
Relationship between Leakage Current and ECT (Core Type)

Figure 14
Relationship between Leakage Current and (ECT Flow Rate) (Core Type)

2-46
Modeling and Prediction of Static Electrification

Figure 15
Relationship between Leakage Current and (ECT Flow Rate)

Figure 16
Relationship between Leakage Current and (ECT o) (Shell Type)

2-47
Modeling and Prediction of Static Electrification

2.4 Flow Electrification Measurements of Transformer Insulation

Andrew P. Washabaugh and Darrell E. Schlicker


JENTEK Sensors, Inc.
Watertown, MA 02172
and
Markus Zahn
Massachusetts Institute of Technology
Department of Electrical Engineering and Computer Science
Laboratory for Electromagnetic and Electronic Systems
Cambridge, MA 02139

Abstract

Static electrification due to liquid dielectric flow has been found to be the cause of about two
dozen field failures worldwide of large forced-oil-cooled power transformers. Problems stem
from the entrainment of diffuse double layer charge at interfaces into circulating liquid. The
accumulation of this charge in the volume, or the deposition of this charge on insulating or
isolated surfaces, causes the electric potential to rise in the same fashion as voltage build-up in a
Van de Graaff generator. The potential builds up until the rate of charge accumulation equals the
rate of charge leakage, or until spark discharges occur. Electrification effects are initiated by
fluid flow, whether or not the equipment is energized. Energization generally augments the
electrification effect. This paper focuses on electrification measurements made on transformer oil
and pressboard materials including the effects of energization using the Absolute Charge Sensor
in the MIT Couette Flow Facility with and without antistatic additive benzotriazole.

1 Background

1.1 The Flow Electrification Problem

Liquid dielectrics are commonly used in electric power apparatus to both transfer heat and
withstand high electric stress. Charge separation at interfaces between moving fluid and
boundaries with the accumulation of charge on insulators or isolated conductors can lead to high
field strengths and electrical discharges. The potential builds up until the rate of charge
accumulation equals the rate of charge leakage, or until spark discharges occur. Electrification
effects are initiated by fluid flow, whether or not the equipment is energized. Energization
generally augments the electrification effect. Understanding of electrification requires the
coupling of the laws of electromagnetism, fluid mechanics, heat, and electro-chemistry to
describe the generation, transport, accumulation, and leakage of charge and to relate how these
relate to the observed factors of temperature; moisture; flow rate and turbulence; contaminants,
additives, and surface active agents; wall surface condition; energization; and flow configuration.
What makes understanding difficult is that there is a lack of common denominators in failures.
Identical side-by-side transformers have different failure experiences. Trace differences in
material properties and impurities that are not easily controlled apparently have a strong effect on
flow electrification.

2-48
Modeling and Prediction of Static Electrification

Electrification has been a past problem in the flow of petroleum liquids in pipes and through
filters and in charge accumulation in storage tanks [1]. Electrostatic spark discharges in the vapor
of these combustible fluids has resulted in explosions and fires. To avoid these hazards all
metallic surfaces are grounded to prevent the electric potential from rising above the breakdown
strength of the surrounding medium, the flow rate is reduced to decrease the rate of charging, and
anti-static additives are added to the fluid to raise its conductivity so that any accumulation of net
charge relaxes with time constant given by the dielectric relaxation time. These measures have
essentially eliminated flow electrification problems in the transport, filtering, filling, and storage
operations of petroleum products.

The fixes used in the petroleum industry cannot always be applied to electric power equipment.
Grounded conductors are not possible in high field regions, while a decrease in the flow speed of
coolant may be a major limitation in the power rating. Raising the conductivity of the insulating
fluid with anti-static additives may also increase the losses, making the device less efficient.

1.2 Flow Electrification Scenario

The flow electrification problem has often arisen when a transformer that has been out of service
is being recommissioned. One scenario for this behavior links the electrification problem with
the moisture dynamics. As the transformer heats up, moisture is driven out of the pressboard
insulation into the oil. This moisture first comes from near the pressboard interface, leaving a
very dry interface that is also highly insulating. Any accumulating surface charge has no leakage
and thus the surface charge density can increase until spark tracking discharges occur. These
sparks cause gas formation, which can rise into the high electric field region causing a high
energy discharge that causes transformer failure. If failure does not occur during the early times
of the recommissioning process, then moisture deeper in the pressboard diffuses to the surface
reaching equilibrium with the oil moisture. The moisture diffusion time can easily be a few
weeks, but once there is no longer an interfacial dry zone, there is a leakage path for interfacial
surface charge so that the surface charge density cannot give rise to electric field strengths
beyond the breakdown strength of oil and pressboard. Thus, if this scenario is correct, once the
transformer is in moisture equilibrium the flow electrification hazard is minimized.

2 Stages of Electrification

Charge is separated at interfaces between dissimilar materials by the equilibrium between


migration and diffusion. Moving liquid entrains the separated charge on the liquid side and
transports it to insulating walls or isolated conductors where the charge can build up until the
leakage currents equal the rates of charge accumulation or until the insulation fails and spark
discharges occur due to the high generated electric field.

2.1 Charge Generation - The Electrical Double Layer

With slight fluid ionization, either by normal dissociation, trace impurities, or by use of
additives, insulating liquids carry positive and negative ions. These ions try to neutralize each
other in the bulk, but at boundaries there is a preferential adsorption of one species with the
opposite carrier diffusely distributed over a thin boundary region called the electrical double

2-49
Modeling and Prediction of Static Electrification

layer. The degree of net charge and the depth to which it penetrates into the liquid volume are
related by the balance of ion diffusion, migration, and convection. In stationary equilibrium,
diffusion due to concentration gradients is balanced by the electric field induced by the separated
charges.

2.2 Transport

At low flow velocities, the velocity distribution is determined by the fluid viscosity . As the
mean velocity U is increased, inertial effects become important. The non-dimensional parameter
that describes the relative importance of inertial and viscous effects is the hydrodynamic
Reynolds number R = mUL/ , where m is the fluid mass density and L is a typical system
length. For small Reynolds numbers the flow is smooth and steady (laminar) while at moderate
and large Reynolds numbers the flow becomes agitated with fluctuations (turbulent). Turbulent
or non-fully developed flow enhances the charge transfer at the interface.

2.3 Charge Storage

For conducting pipes, the electric fields due to the separated charge are generally much smaller
than that required for spark breakdown. The fields can only build up to breakdown strength
values if there is an isolated conductor or insulating surface which allows charge to accumulate
at a rate that exceeds the leakage rate. For example, an isolated conducting sphere of radius Rs
has capacitance with respect to infinity of C = 4 Rs , being ~ 2.5pf for a 1 cm radius sphere in
transformer oil. For a constant charging streaming current I the voltage build-up is V = It/C. For
a I = 2.5 nA streaming current the voltage builds up at the rate of 1 kV/s. If the sphere is
connected to ground via a leakage resistance RL , the steady state voltage if breakdown does not
occur first is Vss = IRL . With I = 2.5 nA, this voltage can exceed 100 kV if RL exceeds 4x10
13

ohms. If RL is just due to leakage through oil, then RLC equals the dielectric relaxation time
e = / . To limit the steady state voltage rise to less than 100 kV, e = RLC must be less than
100 seconds, for this example.

Insulating pipes allow charge to be distributed on the inside surface along the pipe length. If the
insulating pipe is surrounded by a conductor or has a grounded conductor nearby, the field in the
insulating pipe wall could become very high and break down, often creating a hole allowing fluid
to leak out of the pipe.

2.4 Charge Leakage

For insulating surfaces or isolated conductors, charge leakage through surfaces and volume
determines the steady state potential distribution. If the leakage resistance is too high, the
potentials can exceed breakdown values and there is a discharge. To increase the leakage in
applications where electrical insulation is not required, anti-static additives that raise the
dielectric conductivity are often used.

2-50
Modeling and Prediction of Static Electrification

3 Measurement Methodologies

A conducting probe placed into the flow results in an open-circuit voltage or short-circuit current
but does not provide an unambiguous value of charge density as one may never be certain of the
extent to which the probe response is due to impacting charge from the flow, or from electrical
double layer charge removed from the probe by the flow by the same charge separation process
that is being studied. There is no way to separate the effects of entering charge due to upstream
flow electrification from charge separation at probe surfaces.

3.1 Absolute Charge Sensor

The Absolute Charge Sensor (ACS) shown in Figure 1 avoids this ambiguity and measures the
charge density independent of the fluids electrical properties, the velocity of the fluid, and any
electrification process within the instrument [2]. Here fluid is pumped in and out of a metal
bellows Faraday cage driven by a linear actuator so that the bellows volume periodically changes
with time. The metal bellows is connected to an electrometer so that the measured short-circuit
current or open-circuit voltage is directly proportional to the fluid charge density. The small
entrance pipe to the bellows volume is at the end of a shielded pipe, so that streaming currents
generated from electrification on the outer surface of the probe are not measured.

If the fluid is to be returned to the system at a down-stream location or to a relaxation region, a


pair of check valves is used so that the exiting fluid is not in the proximity of the sampling probe.
The charge density is calculated from the known filling flow rate and either the current or
voltage, respectively measured by low or high impedance electrometers connected to the
electrically isolated Faraday cage. For the fluid within the Faraday cage, streaming currents
generated by electrification on the inner surface of the probe, associated check valve, and
bellows are not reflected in the measurement as long as no charged fluid is leaving the Faraday
cage while charged fluid is entering. The measured current or voltage is proportional to the net
charge density in the sampled fluid regardless of whether that charge is still entrained in the fluid
or has been conducted to the walls of the Faraday cage.

2-51
Modeling and Prediction of Static Electrification

Figure 1
Electromechanical Schematic of the Absolute Charge Sensor (ACS)

Figure 2
Couette Charger with Electrical Terminals Connecting Inner and Outer Cylinders for
Measuring Open-Circuit Voltage or Short-Circuit Current or for Applying a Voltage
between the Cylinders. The ACS Measures the Charge Density in the Rotating Fluid

2-52
Modeling and Prediction of Static Electrification

A charged body introduced into a Faraday cage induces an opposite polarity charge on the
interior wall of the conducting volume. Thus, without actually being transported to the wall, the
net charge induced on the conductive inner wall would be equal to but of opposite polarity to that
in the fluid. Because the rate of accumulation of the charge so induced on the inner wall of the
container comprises a capacitive electrical current that can be measured externally, it is possible
to measure this net charge with an electrometer. In addition to the image charge, there is a
conduction current associated with the relaxation of the charged fluid and a current due to flow
induced electrification on the internal surfaces of the ACS volume. However, because the total
charge within the Faraday cage volume whether due to imaging, charge relaxation, or
electrification remains constant, with no fluid leaving the Faraday cage during the measurement,
the measured current i depends only on the inlet convection current. For inlet flow rate Q the
fluid charge density is = i / Q .

3.2 Couette Charger

3.2.1 Operating Principles

Pipe flow configurations are often used in modeling electrification phenomena in electric power
apparatus because of their similarity to cooling ducts in the inter-winding region of transformers
and in coolant tubes of power semiconductors. The Couette charging apparatus of coaxial
cylinders in Figure 2, where the inner cylinder can rotate at angular speed , offers many
advantages over pipe flow measurements [3-8]. The fluid under test fills the annulus between
cylinders that are covered with the solid material to be tested such as oil-impregnated paper
backed by metal. The inner cylinder speed is varied to give controlled laminar or turbulent flows.
This compact apparatus allows for flexibility in testing liquids, trace impurities, and solid
insulation without involving large amounts of material and allows materials to be easily changed.
By virtue of the re-entrant flow, it acts like a pipe of infinite length and thus provides for the
measurement of basic electrification parameters even in highly insulating systems, where the
electrical development length Lelec V / for fluids with dielectric permittivity , conductivity
and average flow velocity V in a pipe flow system is likely to be longer than the practical
length of a test section. By sampling the charge entrained in the azimuthal flow at a sufficiently
slow rate to not disturb the azimuthal flow, equilibrium electrification parameters can be
deduced. The flow Reynolds number, and thus the diffusion sub-layer thickness, is set by the
shaft rotational velocity of the inner cylinder, independent of the flow rate used to sample the
charge carried by the liquid. Effects of energization can also be studied with the brush contact
allowing a high voltage to be imposed. The brush contacts also allow measurement of open-
circuit terminal voltage and short-circuit current due to inner cylinder rotation, which can also be
related to the volume charge density measured by the ACS. The charge trap in Figure 2 is a large
volume where the fluid residence time is much larger than the fluid dielectric relaxation time. It
is used so that charged fluid exiting the ACS and then entering the Couette Charger is uncharged.
The Couette Charger also models the mixing of pumps and plenums, as well as secondary flows
due to spacer blocks, protuberances, and bends.

2-53
Modeling and Prediction of Static Electrification

3.2.2 Open-Circuit Voltage and Short-Circuit Current Measurements

With the terminals of the Couette Charger shown in Figure 2 connected to a low impedance
electrometer, the short-circuit current was measured as shown in Figure 3. For aluminum
cylinders a negative current was measured which increased in magnitude with increasing speed.
Figure 3 also shows the transient response of the terminal voltage when switched from an initial
short-circuit to an open-circuit at t = 0. For each value of rotational speed, the speed was held
constant for at least 5 minutes before the terminals were opened so that the core charge density
was at its steady state value. For each speed, the negative voltage increased in magnitude
exponentially with a measured time constant of 43 seconds. This time should be the dielectric
relaxation time of the oil, which was estimated to be e 30 seconds. Also shown in Figure 3 is
the response of the open circuit voltage decay after rotation of the inner cylinder was stopped.
Because the core charge density was also relaxing, the voltage decayed at a faster rate. For the
terminal measurements in Figure 3, the core charge density was positive and changed linearly
with rotational speed being about 11 C / m at 1800 rpm as measured by the ACS.
3

Figure 3
Measured Terminal Short-Circuit Current (Left) and Open-Circuit Voltage (Right) as a
Function of Time in the Couette Charger for Various Inner Cylinder Rotational Speeds with
Bare Aluminum Cylindrical Electrodes

Although there is no mean flow in the direction of the generated radial electric field, the terminal
current can deliver a small amount of electrical power to a load connected between the cylinders.
Turbulence introduces fluctuating components of velocity in the radial direction of the electric
field and fluctuating components of electric field in the azimuthal direction of the mean flow.
These terminal voltage and current measurements thus illustrate how turbulent diffusion can do
work and hence be the mechanism for mechanical to electrical energy conversion.

4 Representative Couette Facility Results

Several types of representative measurements are presented in this section [5-8]. The general
approach was to measure the baseline response for the system with bare stainless steel cylinders
and with transformer pressboard covering the inner cylinder. Then the effects of adding copper,
which is present in large quantities in transformers, moisture and the non-ionizable antistatic
additive BTA were explored. In each case, the controllable parameters (inner cylinder rotation

2-54
Modeling and Prediction of Static Electrification

rate, temperature, and applied dc or ac voltage) were varied as the measurable quantities of
charge density, terminal current or voltage, oil moisture content, oil conductivity, and pressboard
conductivity were monitored. When possible, both steady state and transient conditions were
explored.

4.1 Rotation Rate Dependence

The primary variable parameter that is critical to flow electrification is the fluid flow rate. In the
Couette Charger, the flow rate was controlled by the rotation rate of the inner cylinder.
Physically, as the rotation rate increased, the flow rate also increased but the diffusion sublayer
thickness decreased. The net result was that the core charge density tended to increase with
rotation rate as the sublayer thickness decreased and more charge was transported by convection
out of the interfacial electrical double layers into the core region of the flow.

Sample sets of steady state electrification data are shown in Figure 4 for identical bare stainless
steel cylinders and in Figure 5 for pressboard insulation covering a copper layer on the inner
cylinder. For the bare stainless steel cylinders in Figure 4, the charge density had the same
polarity as the terminal current and voltage at high rotation rates, but slight opposite polarities at
very low rotation rates. This is opposite to the measurements of Figure 3 with aluminum
cylinders, which had opposite polarity terminal voltage and current (both negative) to the
measured positive volume charge density. As shown in Figure 2, the terminal voltage and current
polarity in the external circuit is defined and measured from the inner cylinder to the grounded
outer cylinder. When pressboard insulation covered the inner cylinder in Figure 5, the charge
density had a polarity opposite to the terminal current and voltage for all rotation rates. Also
significant about this data is that the measured terminal quantities tended to increase in
magnitude as the rotation rate increased and did not reach a limiting or saturation value. This
indicates that the interfacial charge transfer process was probably rate limited by the transport of
ionic species to the interface, not the interfacial reaction rate.

The polarity reversal of the terminal quantities as the rotation rate increased is shown more
clearly in Figure 6. As the rotation rate was varied, either the short-circuit current or the open-
circuit voltage was monitored. The current was measured during both the startup and shutdown
transients for the rotation of the inner cylinder. For the voltage measurements, the voltage was
initially set to zero (short-circuited) until the inner cylinder reached a steady state speed and then
the terminals were open-circuited as the voltage transient was measured.

4.2 Temperature Dependence

The system temperature is readily controlled but the temperature dependent material properties
(viscosity, conductivity, relative moisture content between the oil and the pressboard, ion
mobility and diffusivity, and the interfacial zeta potentials creating the electrical double layers)
can lead to competing effects. For example, the viscosity of the oil decreases as the temperature
increases. This causes the diffusion sublayer thickness to decrease and more charge to be carried
out of the electrical double layers into the bulk of the fluid. In contrast, the conductivity increases
with temperature, which results in an increase in charge relaxation and a decrease in the volume
charge density.

2-55
Modeling and Prediction of Static Electrification

Sample sets of steady state electrification data showing the variations with temperature are also
given in Figures 4 and 5. In both sets of data, the magnitudes of charge density and short-circuit
current increased with temperature at each rotational speed, to within the accuracy of the
measurement. In contrast, the open-circuit voltage was largest at an intermediate temperature
(near 50C) for the bare metal cylinders but essentially independent of the temperature when
pressboard covered the inner cylinder. Although these pressboard measurements had a copper
sleeve over the inner cylinder, measurements without the copper sleeve present gave similar
results. 3
charge density (C/m )

30

20

10

0
160
o
C pS/mppm %rS
70 16.7 8.7 2.5
120 50 7.9 5.7 3.2
Voc (v)

35 4.7 4.1 4.0


80 15 1.7 3.1 7.0

40

6
Isc (nA)

0
0 200 400 600 800 1000 1200

Rotation Rate (rpm)

Figure 4
A Set of Electrification Measurements for Bare Stainless Steel Cylinders and Shell Diala
A Transformer Oil. The Charge Density, Open-Circuit Voltage and Short-Circuit Current
Usually had the Same Polarity and Increased with Rotation Rate. At Each Temperature and
Rotation Rate, the Open-Circuit Voltage and Short-Circuit Charge Densities are Plotted and
are Essentially the Same. The Maximum Voltage for each Rotation Rate Appeared at an
Intermediate Temperature (50C)

2-56
Modeling and Prediction of Static Electrification

charge density (C/m )


50

3
40

30

20

10

0
0

-100
Voc (v)

o
C pS/mppm %rS
70 9.0 4.40 1.3
50 4.6 2.12 1.2
-200
35 2.5 1.26 1.2
15 0.9 0.54 1.2

-300
0

-2
Isc (nA)

-4

-6

-8
0 200 400 600 800 1000 1200 1400

Rotation Rate (rpm)

Figure 5
A Set of Electrification Measurements for EHV-Weidmann HiVal Pressboard Insulation
Covering a Copper Sleeve on the Inner Stainless Steel Cylinder, a Bare Stainless Steel
Outer Cylinder and Shell Diala A Transformer Oil. The Charge Density (Open and
Short-Circuit) had a Polarity Opposite that of the Terminal Voltage and Current at each
Temperature

During temperature transients, the short-circuit current, charge density, oil moisture, and oil
conductivity usually increased in magnitude with the temperature, as shown in Figure 7 for a step
increase in temperature. The charge density and current also showed long time scale transients
after the temperature stabilized.

2-57
Modeling and Prediction of Static Electrification

2000 o
70 C, bare metal cylinders

Isc (nA)
200 rpm 400
-2000
600

-4000 800
1000

-6000
0 100 200 300 400 500
5

0
200 rpm 400
V oc (v)

-5
600
-10

-15 800

1000
-20
0 100 200 300 400 500 600 700

Time (s)

Figure 6
Time Transient Terminal Measurements with Bare Stainless Steel Cylinders for Various
Inner Cylinder Rotational Speeds. The Current and Voltage were Set to Zero (by Open or
Short-Circuiting the Electrometer) as the Rotation Rate (in rpm) was Set. Note the Polarity
Reversal between Low and High Rotation Rates. The Charge Density was Positive for all
Rotation Rates

Oil Moisture (ppm)


60
Oil Temp. ( C)

4
o

40 3
2
20
1
0 0
0 10 20 30 40 0 10 20 30 40
charge density (C/m )
3

8 3
0rpm=-0.13C/m 1.6
6
1.2
rS (%)

4 0.8
2 0.4
0 0.0
0 10 20 30 40 0 10 20 30 40
Conductivity (pS/m)

I0rpm=62.77pA
25
0.0 20
Isc (nA)

-1.0 15
10
-2.0
5
-3.0 0
0 10 20 30 40 0 10 20 30 40
Time (hours)

Figure 7
Temperature Driven Transient Measurements for EHV-Weidmann HiVal Pressboard
Insulation Covering the Inner Cylinder, a Bare Stainless Steel Outer Cylinder, Shell Diala
A Transformer Oil and a Rotation Rate of 400 rpm. The Oil Moisture Content and
Conductivity Followed the Temperature. This Indicated a Relatively Rapid Transfer of
Moisture from the Pressboard into the Oil at Elevated Temperatures. The Current and
Charge Density Became Larger as the Temperature Increased

2-58
Modeling and Prediction of Static Electrification

4.3 Moisture Dependence

One ubiquitous impurity present both in the solid and liquid insulation is water. Standard
procedures for reducing the moisture content of the insulation include heating the system and
vacuum processing, but some water is always present. In the steady state, the moisture
equilibrium curves show that the concentration of water in the hydrophilic pressboard is much
higher than that of the hydrophobic oil. Changing the temperature of the system induced
moisture transients, so that the equilibrium pressboard and oil moisture contents also change.
The change in oil moisture levels in ppm as well as relative saturation (rS%) as the temperature
is changed between 15 and 70C are also shown in Figure 7.

4.4 Applied DC Voltage Dependence

An applied dc field enhances the migration of charge across the diffusion sublayers, which can
lead to an increase in the net charge density in the well-mixed core region of the flow. In
addition, application of dc voltages causes net charge to develop on the interfaces in the system,
similar to the separation and accumulation of charge inherent in the flow electrification process.
With small dc applied voltages the charge density increases linearly with the applied voltage.

Related measurements show that the charge density has a slight dependence on the terminal
constraints. By toggling the electrometer across the Couette Charger electrical terminals between
voltage (open-circuit) and current (short-circuit) modes, perturbations in the volume charge
density were observed. At low rotation rates, when the terminal voltage and current were small,
the charge density was unaffected by the terminal impedance. At the higher rotation rates, when
the terminal current became appreciable, the short-circuit charge density was found to be slightly
smaller than the open-circuit charge density. These results are consistent with the applied voltage
measurements showing the largest change in the charge density with applied voltage at the
lowest rotation rates. When the applied voltage is large compared to the open-circuit voltage, at
low rotation rates, the perturbation in the charge density is larger than that observed by changing
the load impedance, but when the applied voltage is small compared to the open-circuit voltage,
at high rotation rates, the charge density perturbation is also smaller.

When the rotation of the inner cylinder is started, some of the positive volume charge is
transported by the fluid into the core region. When a positive voltage is applied, the volume
charge near the interface is decreased because negative volume charge accumulates to shield the
applied field from the bulk of the fluid. This leads to a decrease in the core charge density as
well. Because some of this negative charge gets adsorbed onto the interface to form a surface
charge, the volume charge densities return to near their steady state levels. After removing the
voltage source and keeping the terminals open-circuited, the accumulated surface charge leads to
a terminal voltage that decays with the relaxation time of the pressboard. The charge density
rapidly returns to its original level because the oil relaxation time is small compared to the
charge density measurement time. In contrast, after removing the voltage source and keeping the
terminals short-circuited, the accumulated negative surface charge is imaged by a positive
volume charge on the liquid side of the interface and the core charge density then increases.
When the surface charge decays, the core charge density also decays. The charging and
discharging transients have similar time constants, consistent with the accumulation and decay of
interfacial free surface charge for ohmic dielectrics between stationary cylinders. The opposite
polarity effect occurs when a negative voltage is applied.

2-59
Modeling and Prediction of Static Electrification

A representative plot of the time transients in the charge density as the terminal voltage was
varied for two rotation rates is given in Figure 8. By increasing the rotation rate, the diffusion
sublayer thickness was decreased. Since the measured time constants associated with the
transient decays in the charge densities were roughly the same at both rotation rates, the charge
dynamics were probably reaction limited. In contrast, the charge density increased with rotation
rate when there was no dc excitation, suggesting that the charge dynamics were at least partially
transport limited. From these observations it would appear that the transport rate and the reaction
rate for charge transfer at the interface are comparable for these semi-insulating materials.
o
15 C, 1.2 pS/m oil, 1.4 ppm oil mois ture
300
400 rpm
10
200
charge dens ity
o (C/m )
3

Voc (v)
100

0 0

-100
-5 voltage
-200
0 5 10 15 20 25
20 300
800 rpm
15 200
charge dens ity
o (C/m )
3

Voc (v)
10 100

5 0

0 -100
voltage
-5 -200
0 5 10 15 20 25 30 35

Time (hours)

Figure 8
Effects of Applied dc Voltages (100 Volts) on the Core Charge Density with
EHV-Weidmann HiVal Pressboard Insulation Covering the Inner Cylinder and a Bare
Stainless Steel Outer Cylinder. A Positive Voltage Tends to Decrease the Volume Charge
while a Negative Voltage Tends to Increase the Volume Charge

Previous work [4] has shown that AC energization significantly increases the volume charge
density entrained in the fluid flow when 2bEo/()>1, where b is the charge mobility, Eo is the
peak amplitude of the AC electric field at radian frequency , and is the hydrodynamic
diffusion sub-layer thickness. The applied AC electric field drives electrical double layer charge
over one half of the AC cycle to cross the diffusion sub-layer into the turbulent flow.

4.5 Conditioning Transients

While most of the measurements discussed thus far have focused on experiments that had a
relatively short duration, long time transient variations were also observed in the experiments. In
general terms, electrification (charge density, terminal current and voltage) measurements
showed long time variations that were not correlated to measurements of the moisture content or
conductivity. Since the electrification data is sensitive to the interfacial properties, it is concluded
that the interfacial dynamics (adsorption and desorption of ions) are significant over these long
transients.

2-60
Modeling and Prediction of Static Electrification

Although relatively rare, the oil volume charge density was occasionally negative. Usually this
resulted from long time transient changes in the system, but on one occasion, the charge density
was negative immediately after the rotation was started. As shown in Figure 9, during the initial
15C equilibration period, the moisture in the oil migrated into the pressboard and the oil
conductivity remained essentially constant. On the other hand, the charge density was negative
and the terminal current was positive and relatively large. During this period, sharp
discontinuities in the charge density and current were present which appeared to be consistent
with discharges in the system. Subsequent measurements showed the charge density and current
repeatedly reversing polarity. Some of these polarity reversals were driven by changes in the
temperature but others occurred at a constant temperature. In these extended measurements, the
moisture content and conductivity increased slowly with time, indicating that the system was
probably slightly exposed to the atmosphere.

16 1.2

Oil Moisture (ppm)


Oil Temp. ( C)

12
o

0.8
8
0.4
4
0 0.0
0 5 10 15 20 25 0 5 10 15 20 25
charge density (C/m )
3

3.0
0
2.0

rS (%)
-2
-4 1.0
-6
0.0
0 5 10 15 20 25 0 5 10 15 20 25
1.0

Conductivity (pS/m)
1.2 0.8
Isc (nA)

0.8 0.6
0.4
0.4
0.2
0.0 0.0
0 5 10 15 20 25 0 5 10 15 20 25
Time (hours)

Figure 9
Initial 15C Equilibration Period for HiVal Pressboard Covering the Inner Cylinder, a Bare
Stainless Steel Outer Cylinder, Shell Diala A Transformer Oil and a Rotation Rate of
400 rpm. The Oil Moisture Content Slowly Decreased as the Pressboard Dried the Oil. The
Sharp Discontinuities in the Charge Density and Terminal Current are Consistent with
Discharges in the System

One possible mechanism leading to the discharges is that surface charge was accumulating at the
oil/pressboard interface as the fluid flow redistributed the charge in the system. Once enough
surface charge accumulated, the associated electric field exceeded the local breakdown field,
which, in turn, caused all of the accumulated charge to discharge. The discharge appeared to be
global rather than local because the terminal current, which gives the net current flowing across
the entire inner and outer cylinder surfaces, was discontinuous. The total charge involved in the
discharge can be estimated from the discontinuity in the charge density and the volume of the
annular gap. Using a volume of 6500 cm3 and a charge density decrease of 4 C / m 3 , 26 nC of
charge were discharged. With a rotational speed of 400 rpm corresponding to an average oil
velocity of 190 cm/s (based on a radius of 9 cm), this quantity of charge is consistent with
published magnitudes for electrostatic discharges created by flowing oils [9].

2-61
Modeling and Prediction of Static Electrification

These results were somewhat surprising because discharges were not seen in previous or
subsequent measurements. Although the oil moisture content was very low in this measurement,
indicating that the pressboard insulation was also very dry, other experiments did not exhibit
these discharges. Furthermore, the change in the volume charge density as a result of the
discharge was relatively small; other measurements showed larger volume charge densities at the
same temperature, but no discharges. After disassembling the system and looking at the
pressboard, the only surface blemish that was found was near the top of the inner cylinder. The
blemish was slightly darker than the surrounding area and could have been due to the discharges.
Apparently, once the local area of the discharge had been damaged, the discharges stopped.

4.6 Effects of Benzotriazole (BTA)

Japanese research has shown that concentrations of the additive BTA of order 10 ppm in
transformer oil eliminates the flow electrification problem in transformers [9]. To better
understand the role of the BTA, the effects of the BTA were measured in the Couette Charger
during the transient period in which BTA was added to oil and also during relatively short
duration measurements in which the charge density, voltage, and current were measured as the
rotation rate and temperature were varied. In addition to the electrification data, the oil moisture
content and conductivity were monitored and the pressboard properties were inferred from
transient measurements of the open-circuit voltage and short-circuit current decays, just after
disconnecting a dc voltage source or abruptly stopping the rotation of the inner cylinder. In the
experiment discussed here, concentrated solutions of BTA were periodically added to the oil,
with the same oil and pressboard used throughout the experiment.

4.6.1 BTA Measurements in Oil

Before considering the effects of BTA on electrification, first consider the mass transfer
dynamics of BTA in the oil. After adding BTA to the oil in the Couette Facility, the BTA
concentration slowly decreased. This decrease in BTA concentration is similar to that observed
in actual transformers [10].

Some of the concentration decrease can be attributed to the BTA absorption into the paper. Like
any constituent present in trace amounts, the relative concentrations of the BTA in the oil and
pressboard are determined by partition coefficients at each temperature. Although the details of
the experiment were not provided, the equilibrium absorption content of BTA in paper
insulation has been given as 4.2 ppm of BTA ( g BTA/g pressboard) at 15C and 21 ppm of
BTA at 70C [10]. Similar results have been obtained by adding BTA to a beaker containing
known masses of oil and paper insulation. By measuring the decrease in the oil BTA
concentration and assuming that the concentration decrease was due to BTA migrating into the
paper, the BTA concentration in the paper was calculated. Since the BTA content of the oil was
not reported, this would appear to be the solubility of BTA in the paper, but the measurements
given in the next section contradict this conclusion. For comparison, the oil volume in the
Couette Charger was ~ 7500 cm3 and the pressboard covering the inner cylinder only had a
volume of ~ 200 cm 3 so that only a small fraction of the BTA was expected to be transferred
from the oil into the pressboard.

2-62
Modeling and Prediction of Static Electrification

4.6.2 BTA Measurements in Pressboard

Direct measurements of BTA in the pressboard showed that significant amounts of BTA were
present in the pressboard [5-7]. When the pressboard was diced, peeled, and mixed thoroughly
with the extract solutions, the pressboard BTA concentration was estimated to be 130 ppm
( g BTA/g pressboard). The concentration was reasonably uniform over the height of the
cylinder, being roughly 123 ppm at the top of the inner cylinder and 139 ppm at the bottom.

Additional measurements were performed to determine if the BTA was in the bulk of the
pressboard or at the surface. In this case the pressboard samples were diced, but not peeled, and
were mixed thoroughly with the extract solutions to yield BTA concentrations of 67 ppm at the
top of the inner cylinder, 62 ppm in the middle, and 59 ppm at the bottom. In the first set of
measurements, the peeling of the pressboard increased the surface area exposed to the extract
solution, which, with the extra mixing, apparently allowed more of the BTA to be extracted from
the bulk of the pressboard. These results indicate that a substantial portion of the BTA is found in
the bulk of the pressboard and is not simply at the surface. Otherwise, the second set of
measurements would have had BTA concentrations comparable to those of the first
measurement. Based on these measurements, the solubility of the BTA in the pressboard is much
higher than that given by the previously mentioned equilibrium levels in [10].

4.6.3 Electrification Measurements

Consider next the transient when the concentrated solution of BTA was added to the oil in the
Couette Charger, as shown in Figure 10. The measurements show that BTA had both short term
and long term effects. In the short term, immediately after adding BTA, the charge density
decreased and the current became positive and large. After this initial transient period, the charge
density tended to be slightly enhanced and the current tended to become more negative. In the
long term, the charge density and current were initially reduced when the BTA concentration
was increased to ~10 ppm, but the magnitudes of the charge density and current were increased
when the BTA concentration was increased further.

2-63
Modeling and Prediction of Static Electrification

1st: t=0 2nd: t=164 3rd: t=392 4th: t=782 5th: t=159 6th: t=149

[BTA] (ppm)
60

Charge (C/m )
12

3
8 40

4 20

0 0
3
Current (nA)

2
1
0
-1
-2
8
Moist. (g/g)

6
4
2
0
Conduct. (pS/m)

25
20
15
10
5
0
0 10 20 160 170 180 560 570 580 1340 1350 1360 1490 1500 1510 1640 1650 1660
Time (hr)

Figure 10
Time Transient Measurements Associated with the Addition of BTA to Oil in the Couette
Charger. The System was at 70C with HiVal Pressboard Covering the Copper Sleeve on the
Inner Cylinder Rotating at 400 rpm. Concentrated Solutions of BTA in Oil, also at 70C, were
Added as Indicated. The Oil was Periodically Sampled and the BTA Concentration, Illustrated
by the Circles in the Top Row, was Measured using Ultraviolet Spectrophotometry. The Time
t in Hours between Additions of the BTA is Indicated

After the BTA concentrated oil was added, the oil was mixed for ~30 minutes before samples
were withdrawn for BTA concentration measurements. The initial decay of the BTA was
probably caused by the BTA being absorbed into the pressboard, but it could also have been an
experimental artifact because the BTA had not yet been mixed uniformly into the oil of the
Couette Charger or because the same T port was used for both adding the BTA filled oil and
withdrawing oil samples. The much slower decay of the BTA probably resulted from further
absorption into the pressboard, adsorption onto the interfaces, or degradation of the BTA
molecule itself, which is known to be light sensitive. These experiments were done with a
minimal exposure to light, so that the degradations due to light should have been minimal.

In these measurements, the oil moisture content shown in Figure 10 remained essentially
constant and only increased when the relatively wet BTA concentrated oil (which contained
~15 ppm of water) was added to the system. The fact that the moisture content remained nearly
constant indicates that the system was reasonably well sealed.

The BTA appeared to have a strong effect on reducing the oil conductivity as shown in
Figure 10. This was a surprising result because BTA has been found to contribute little to the oil
conductivity [9] or to increase it [11], but those measurements showing an increased conductivity
were performed with paraffinic oils rather than the naphthenic oil used here. One possible
mechanism by which BTA can reduce the conductivity of oil is that the BTA molecules cluster
around the ionic species similar to the hydration of ions in water, thereby increasing the effective

2-64
Modeling and Prediction of Static Electrification

size and reducing the mobility of the ions. If the mobility of the ions is reduced enough, the
clustering of the BTA around the ions could effectively reduce the number density of ions
contributing to the conduction process. During the first addition, shown in Figure 10, the
conductivity increased probably as a result of the BTA concentrated oil having a high
conductivity; in subsequent additions, the kettle containing the BTA concentrated oil, which was
open to atmosphere, was raised to 70C for much shorter periods of time so that oil oxidation
would be minimized. The significant increases in conductivity before the second and third
additions of BTA may have been caused by the BTA concentrated oil not being degassed so that
the oil could be oxidized more readily. The pressboard conductivity appeared to be unaffected by
the BTA, but once the oil became very insulating, the pressboard appeared to become more
insulating as well.
To help explain the role that BTA played in reducing, and then enhancing, the electrification
charge density, data were also obtained at various rotation rates and temperatures for
experiments having a short duration compared to the decay in oil BTA concentration. The results
are shown in Figure 11. During each measurement, the BTA concentration was essentially
constant at the indicated levels. At each BTA concentration, the charge density usually increased
with both rotational speed and temperature, consistent with previous measurements. Similarly,
the open-circuit voltage and short-circuit current tended to become larger in magnitude as the
rotational speed and the temperature increased. Exceptions to this behavior occurred when the
BTA concentration was near 5 to 10 ppm; the voltage then decreased as the rotation rate
increased, in some cases, and the current occasionally reversed polarity.
0(0) ppm; t=0 3(7) ppm; t=30 4(10) ppm; t=247 8(16) ppm; t=232 19(22) ppm; t=23 33(42) ppm; t=16
charge density (C/m )
3

30 o
C pS/m
o
C pS/m o o
C pS/m o o
C pS/m
C pS/m C pS/m
15 1.0 15 1.1 35 3.4 15 0.50 15 0.27 15 0.15
35 2.2 35 3.0 50 6.4 35 1.5 35 0.80 35 0.35
20 50 4.2 50 5.6 70 11.4 50 2.7 50 1.6 50 0.90
70 9.5 70 10.0 70 6.2 70 3.6 70 2.5

10

0
0

-100
Voc (v)

-200

-300
2
0
Isc (nA)

-2
-4
-6
-8
0 4 8 12 0 4 8 12 0 4 8 12 0 4 8 12 0 4 8 12 0 4 8 12
Rotation rate (100 rpm)

Figure 11
The Steady State Electrification Data as the Rotation Rate of the Inner Cylinder, the
Temperature, and the BTA Concentration are Varied. The System had HiVal Pressboard
Covering the Copper Sleeve on the Inner Cylinder. The Oil BTA Concentration during the
Reported Measurements is given, with the Number in Parentheses Denoting the Initial
measured Concentration. The time t (Hours) between the Current Addition of BTA and the
Start of the Set of Electrification Measurements is Indicated. The Symbols Denote the
Temperature as Circles: 15C, Triangles: 35C, Squares: 50C, and Stars: 70C. The Oil
Conductivity (pS/m) at each Temperature is Indicated

2-65
Modeling and Prediction of Static Electrification

The primary goal of this particular experiment was to determine the effect of BTA on the
electrification process. These measurements showed that the addition of small concentrations of
BTA, 10 ppm, did not affect the electrification data significantly in charge density, voltage,
and current. Once the oil contained 10 ppm of BTA, the magnitudes of the charge density,
voltage, and current were reduced. The electrification data remained at these reduced magnitudes
even as the oil BTA concentration decayed. As more BTA was added, the charge density,
voltage, and current increased with BTA concentration. Subsequent measurements taken after
the BTA had been in the system an additional 515 h showed that the electrification data again
became smaller in magnitude. These measurements confirm general conclusions that an oil BTA
concentration of about 10 ppm is critical for reducing the electrification.

It appears that BTA had two separate but competing effects on electrification. As an illustration,
represent the volume charge density on the liquid side of the solid/liquid interfaces with a wall
charge density w . This wall charge density is assumed to be constant at a given temperature that
is independent of rotational speed but dependent upon the BTA that has been added to the
system. The first effect of the BTA was to reduce w , which reduces the charge in the interfacial
electrical double layer and the magnitude of the core charge density, voltage, and current. The
second effect of the BTA was to reduce the oil conductivity, which tends to cause an increase in
the core charge density. An exception to these effects is the data showing a decrease in the
charge density as the oil conductivity increased; this increase in conductivity does not appear to
be associated with BTA and was probably the result of oil oxidation. In essence, the
measurements are a superposition of the effects of BTA reducing the wall charge density and the
variations in the oil conductivity.

This physical model can also be used to describe qualitatively the time transient data when BTA
is added into the Couette Charger. On the short time scale, as BTA is added to the oil, some of
the BTA is immediately absorbed into the pressboard, or, at least, onto the surface so that it can
diffuse into the pressboard. The BTA disrupts the diffuse charge distribution in the double layer,
possibly by shielding the adsorbed interfacial surface charge. In the limit that all of the surface
charge is shielded, then the wall charge density at the pressboard interface would become zero.
The net effect is that the core charge density would also be reduced since charge would only be
stripped away from the outer cylinder interface. Furthermore, the terminal current would become
positive, assuming positive charge is diffusely distributed in the outer cylinder layer, as positive
charge is entrained in the fluid flow and transported away from the outer cylinder interface. As
the BTA diffuses into the bulk of the pressboard, less is present to shield the surface charge at
the pressboard interface and the charge density and current return to their original levels. On the
long time scale, as the BTA gets adsorbed onto the interface, the diffuse charge in the double
layer is again reduced. Although somewhat speculative, this physical picture appears to be
consistent with the essential features of the data.

5 Summary and Conclusions

The physical phenomenon of charge accumulation due to flow that leads to spark discharges has
been well studied. Japanese transformer manufacturers believe that the problem has been solved
by using the antistatic additive benzotriazole, but even today the long-term effects of BTA as it
leaves the oil volume onto pressboard surfaces and tank walls is not well understood. Even
without additive, the electrification problem can be minimized by not turning on cooling pumps

2-66
Modeling and Prediction of Static Electrification

when the transformer is lightly loaded. The most dangerous period seems to be when the
transformer has been out of service and is then being recommissioned. The best method for
recommissioning to avoid electrification induced discharges is to increase load slowly so that the
transformer can rise in temperature slow enough to allow the oil and pressboard to be near
moisture equilibrium, thus avoiding the formation of highly insulating interfacial dry zones
where charge can accumulate to high enough values to cause spark discharges. This paper briefly
described several measurement methodologies, instrumentation, and simulation systems,
together with representative results, which help to characterize the electrification processes.

Acknowledgment

This research has been supported by EPRI, managed by Mr. Stan Lindgren.

References

1. A. Klinkenberg and J. L. Van der Minne, Electrostatics in the Petroleum Industry.


Amsterdam: Elsevier, (1958).

2. A. J. Morin II, M. Zahn, J. R. Melcher, and D. M. Otten, An Absolute Charge Sensor for
Fluid Electrification Measurements, IEEE Trans. on Elec. Insul., V. 26, No. 2, pp. 181-199,
(April 1991).

3. D. J. Lyon, J. R. Melcher, and M. Zahn, Couette Charger for Measurement of Equilibrium


and Energization Flow Electrification Parameters: Application to Transformer Insulation.
IEEE Trans. on Elect. Insul., V.23, No. 1, pp.159-176, (February 1988).

4. A. J. Morin II, M. Zahn, and J. R. Melcher, Fluid Electrification Measurements of


Transformer Pressboard/Oil Insulation in a Couette Charger, IEEE Trans. on Elect. Insul.,
V. 26, No. 5, pp. 870-901, (October 1991).

5. A. P. Washabaugh, Flow Induced Electrification of Liquid Insulated Systems, ScD Thesis,


MIT, December 1994.

6. A. P. Washabaugh and M. Zahn, Flow Induced Electrification of Liquid Insulated Electrical


Equipment, EPRI TR-104973, Project RP 3334-01 Final Report, May 1995.

7. A. P. Washabaugh and M. Zahn, Flow Electrification Measurements of Transformer


Insulation Using a Couette Flow Facility, IEEE Trans. Diel. and Elect. Insul., V. 3, No. 2,
pp. 161-181, (April 1996).

8. Darrell E. Schlicker, Flow Electrification in Aged Transformer Oils, MS Thesis, MIT,


September 1996.

9. M. Ieda, T. Yanari, T. Miyamoto, and M. Higaki, Investigation of Static Electrification In


Large Power Transformers In Japan, in Proceedings: Static Electrification in Power
Transformers, EPRI TR-102480, Project 1499-99, pp. 2-3-1 to 2-3-31, (June 1993).

2-67
Modeling and Prediction of Static Electrification

10. T. Miyamoto, A. Kishi, and Y. Muria, Effect of Insulating Oils on Streaming Electrification
in Transformer, EPRI Workshop Proceedings; Static Electrification in Power Transformers,
pp. 3-2-1 to 3-2-12, EPRI EL-6918, (July 1990).

11. T. S. R. Murphy, Meenakshi, P. G. Agashe, and K.M. Kamath, Studies of Electrostatic


Charge Suppressants for Paraffinic Transformer Oils, 7th BEAMA International Electrical
Insulation Conference, Brighton, England, (May 23-26, 1994).

2-68
Modeling and Prediction of Static Electrification

2.5 Measurement of Oil Charge Densities in a Transformer Model

George K. Frimpong and J. Michael Walden


ABB Electric systems Technology Institute
Raleigh, North Carolina 27606

Stan Lindgren
EPRI-PDG
Palo Alto, CA 94303

Abstract

The magnitude and evolution of the charge density of the oil in the upper plenum of a power
transformer plays a very important role in occurrence of discharge activity in the transformer. It
is therefore useful to record such a measurement during the operation of a transformer.
Experiments are performed on a quarter scale model of a shell-type auto-transformer in which
such parameters as the flow rate and temperature are varied to see their effect on the generation
of static charges and discharges in the transformer model. In addition to several other
measurements, the charge densities of the oil in both the upper and lower plenums are measured
using Absolute Charge Sensors (ACS). This paper presents the results of the ACS measurements
and experimentally verified operational procedures that could potentially minimize the
occurrence of static electrification failures in transformers.

Background

This project originated as a follow-up to the testing of the shell-form autotransformer at Ramapo
[1], but in a controlled environment. The goal was to study the evolution of streaming
electrification under various operating conditions and devising mitigation strategies for
minimizing the effect of streaming electrification. It is a well-known fact that as oil is forced
through the ducts of a transformer, charges are separated at the interface of the oil and the paper.
Whereas the negative charges get attracted to the surface of the cellulose insulation, the positive
charges get swept to the upper plenum of the transformer in the oil stream. Depending on the
condition of the oil and really the overall insulation system, these charges will be neutralized by
recombining with negative charges on the insulation, relax to grounded surfaces in the tank or
build up a charge cloud in the upper plenum of the tank.

An extremely high density of charges in the oil will create an enormous potential between the
cloud and surfaces of opposite polarity. Beyond a certain threshold, the breakdown strength of
the insulation between the cloud and the surfaces will be exceeded and flashover will occur,
which can be potentially harmful to the insulation system and transformer in general. It is
therefore beneficial to monitor the evolution of the density of charges in the oil space of the
upper plenum of the transformer. This paper describes results of measurements of charge density
in a laboratory scale transformer model under various operating conditions with a view to
predicting insulation failure due to streaming electrification.

2-69
Modeling and Prediction of Static Electrification

Description of -scale Model Design

Figure 1 shows the winding layout of the transformer tested at Ramapo. There are two high-low
insulation spaces filled with pressboard barriers and spacers arranged in a way to allow oil to
flow across the surface of the windings to remove heat from the winding surfaces and into the
upper plenum of the tank. Each winding package shown in the figure is made up of several
pancake windings connected in series. There are oil channels around each pancake to allow for
its cooling.

Neutral HV Bushing LV Bushing


(550kV) (345kV)

LV1 HV
HV LV2

Tank

Figure 1
Illustration of Full-Scale Rampo Transformer Winding Configuration

In an effort to preserve as much as possible the features of the Ramapo transformer, its insulation
design served as the template for the design of the model studied in this report. Several features
deemed desirable in obtaining results comparable to the Ramapo tests were considered and are
listed below:

1. Maintain relative oil residence times in the model structure where possible.

2. Incorporate the main high-low space, which will provide an well-insulated region for charge
accumulation.

3. Include the capability to apply voltage across the high-low space such that the stresses in the
oil ducts are comparable to the actual design.

4. Attempt to preserve local oil velocities in critical generation regions and incorporate multiple
ducts for more generation than the original design.

5. Allow for different combinations and orientations of cooling loop flows in the lower plenum.

2-70
Modeling and Prediction of Static Electrification

6. The model must have all of the required DC ground planes including the core, tank and
actual winding conductors.

7. Incorporate the capability for temperature control of the system such that frictional heating
can be controlled. This should also allow operation below ambient temperature to simulate
cold start-up.

Using these guidelines, and assuming a simple control volume model, the final design arrived at
is as shown in Figure 2. All linear dimensions are reduced by , meaning the volume spaces are
reduced by 1 64 .

Aluminum Coil Assembly - Top View


Frame
These Ducts are blocked Lifting
in the center of the leg Hook

LV LV HV HV HV HV HV HV HV HV HV HV
2 1 10 9 8 7 6 5 4 3 2 1

Core

.26 .51

5 1/4 6mm Hi-Val 6mm Hi-Val 6mm

3mm 3mm
3mm
Ducts Duct
Ducts

15.7

Figure 2
Top View of -Scale Model Winding Structure

The model is installed in a tank (Figure 3) fitted with 6 pumping loops and outfitted with
instrumentation ports. Figure 4 shows the position of the ACS meters relative to the winding
packages.

2-71
Modeling and Prediction of Static Electrification

Figure 3
1/4 Scale Model Test Tank

Model Dimensions (side view)


Normal Position Raised Position

Top Nilsson
8 dia.

Top Nilsson
Top
8 dia. 9 above ACS
boxing
Aluminum
Top Frame
6 above
boxing ACS

72
HV LV

HV LV
48

10

Bottom 6 1 Bottom 13 8
ACS ACS

Figure 4
Sensor Locations

2-72
Modeling and Prediction of Static Electrification

Experimental Procedure Description

Streaming electrification has been studied extensively since the late 1970s by researchers
around the world [2, 3, 4, 5]. While a great deal of interesting information has been obtained, the
problem remains to be satisfactorily resolved from the utility industry viewpoint. Furthermore,
the phenomenon has been demonstrated to be sufficiently complex that a complete physical
understanding may not be possible. The objective of this project is thus to develop engineering
guidelines for mitigating the streaming, electrification hazard based on realistic empirical
studies.

Starting with geometry based on a transformer design, which has had electrification problems
(i.e. Ramapo) a parametric study has been performed. Modifications to the design and operating
procedure have been implemented for comparison. Key parameters will be tuned to minimize the
static electrification threat.

Accepting therefore, that streaming electrification will always occur to some degree in forced-oil
cooled power transformers, an optimization must be developed to maintain dielectric integrity.
This study will pursue such an optimization based on geometry, flow rate, temperature, oil
selection and operating procedure.

The project consists of several phases conducted in sequence as outlined below:

Phase 1 (Base Case): Begin with a model geometry based on a transformer design, which has
demonstrated static electrification problems and a typical insulating oil. Such a design is
hopefully, the -scale Ramapo model. Measurements are made of the streaming currents, oil
charge density and partial discharge over a range of temperature, flow rate and pump
operation sequences. This establishes the baseline for optimization and can be correlated with
the Ramapo test results. An attempt will be made to identify the worst case scenario for
streaming electrification in the base case design. In addition, a safe operating temperature
range will be established. This will be defined as the temperature range above which the
streaming electrification measurements are the critical streaming electrification parameters
are at reduced levels.
Phase 2 (High Charging Oil): It has been reported in the literature that the charging tendency
of the insulating oil plays a very important role in establishing a transformers tendency to
produce streaming electrification. As a matter of fact, efforts have been made by some
Japanese manufacturers to reduce the charging tendency of the oil by the addition of charging
reducing additives, for example, BTA [6]. In this phase, goal is to determine if for a given
trouble geometry, a low charging oil could be substituted for the high charging oil and
thereby reduce the possibility for streaming electrification. Since the charging tendency of
the mineral oil we ordered (Shell Diala A) was quite low (~40-80C/m3) out of the drum, a
high charging oil was made by dissolving PPM levels of polyvinyl butyral (or Butvar)
polymers in the oil [7]. With this additive it was possible to increase the charging tendency
without significant change in the conductivity of the oil.
Phase 3 (Raised Model): In this phase, consideration is given to geometrical changes to the
lower plenum with a view to reducing the charge generation in this part of the transformer. It
is obvious that given the height of the lower plenum, there could be considerable charge

2-73
Modeling and Prediction of Static Electrification

separation from the tips of the insulation washers extending down into this region as the oil is
forced into the lower plenum. This will be even more pronounced if the flow is directed
directly at the tips of the washers. Moving the tips of the washers higher in the plenum will
reduce the likelihood of charge shearing from the washer tips. In this phase, a determination
will be made as to if a simple geometry change will make a significant difference in
electrification behavior.
Phase 4 (Mitigation): Given the results of the phases 1-3, a mitigation strategy will be
devised for reducing the likelihood of occurrence of streaming electrification through
operational procedures. This simplify involves experiments where the model is operated with
the fewest number of pumps till the oil temperature rises above a safe operating temperature.
Beyond this point all pumps are turned on and the experiment proceeds. The goal here is to
make a determination if such a simple operational change could reduce the likelihood of
occurrence of streaming electrification in the model.

For each phase, tests were performed with 2, 4 and 6 pumps to represent the variation of static
electrification with flow rate of oil through the ducts. Also, the temperature was allowed to rise
from approximately 10C to 60C or till the electrification phenomenon has transitioned through
it most active state. This represents the heat-up phase of the experimental run. At this point the
system was allowed to cool back down to 10C.

Discussion of Results

For reference purposes the percent relative humidity in the oil during a typical run of
experiments is shown in Figure 5, which shows a typical moisture evolution profile for oil/paper
systems. Typically the oil is moister during cool down than heat up. This obviously plays a role
in the generation and relaxation of charges in the system.

High Charging Oil


15
AC, Raised Lower Plenum, Heat Up
AC, Raised Lower Plenum, Cool Down
AC, Regular Lower Plenum, Heat Up
AC, Regular Lower Plenum, Cool Down
Relative Humidity (%)

10

0
0 10 20 30 40 50 60
Upper Temperature (0C)

Figure 5
Relative Humidity During High Charging Oil Measurements

2-74
Modeling and Prediction of Static Electrification

Experiments with Model in Base Case Position

In this section the experimental results involving the model in its base case position relative to
the top and bottom of the tank are presented. Figure 6 and Figure 7 show respectively the charge
density in both the upper and lower plenum oil spaces for the case of high charging oil. Each
graph contains the charge density for three different oil flow rates obtained by running 2, 4 and
6 pumps. On the other hand, Figure 8 and Figure 9 show similar graphs but for regular Shell
Diala A oil.

250
Top ACS (6 Pumps)
Top ACS (4 Pumps)
Top ACS (2 Pumps)
200
Charge Density (uC/m3)

150

100

50

0
10 20 30 40 50 60
o
Upper Temperature ( C)

Figure 6
Charge Density Measurements in Upper Plenum for High Charging Oil

60

Bottom ACS (6 Pumps)


Bottom ACS (4 Pumps)
50 Bottom ACS (2 Pumps)
Charge Density (uC/m )
3

40

30

20

10
10 20 30 40 50 60
o
Upper Temperature ( C)

Figure 7
Charge Density Measurements in Lower Plenum for High Charging Oil

2-75
Modeling and Prediction of Static Electrification

250
Top ACS (6 Pumps)
Top ACS (4 Pumps)
Top ACS (2 Pumps)
200

Charge Density (uC/m )


3
150

100

50

0
0 10 20 30 40 50 60
o
Upper Temperature ( C)

Figure 8
Charge Density Measurements in Upper Plenum for Regular Shell Diala A Oil

50
Bottom ACS (6 Pumps)
Bottom ACS (4 Pumps)
Bottom ACS (2 Pumps)
40
Charge Density (uC/m3)

30

20

10

0
0 10 20 30 40 50 60
Upper Temperature (oC)

Figure 9
Charge Density Measurements in Lower Plenum for Regular Shell Diala A Oil

The immediate observations are that there is less charge concentration in the bottom of the tank
than at the top. The charges in the bottom of the tank are predominately generated at the inlets to
the ducts and also from the edges of the pressboard barriers that extend into the lower plenum in
the H-L space area. The dependence of charging on the flow rate is quite evident and the
temperature dependence of the charge density is also quite evident. The charge density profiles
depict to general states of electrification in the tank based on the overall temperature of the
insulation system. Below a certain threshold temperature, the electrification process is dominated
by charge generation (or separation) processes. Fewer of these charges either recombine with
charges of the opposite polarity in the oil or relax to grounded surfaces.

2-76
Modeling and Prediction of Static Electrification

As the temperature increases and the conductivity of both the oil and the oil-impregnated
pressboard insulation increase, the relaxation time for charges in these media decrease.
Associated with this is an increase in the relaxation and recombination activity, which become
the dominant electrification processes. The result is a net decrease in the charge density in the as
shown in the figures. Eventually an equilibrium state is approached where generation equal
relaxation and the charge density reach a steady state.

Experiments Involving Raised Model

In the lower plenum of the transformer, the height of the bottom of the H-L washers, the boxing
insulation and the inlet to the ducts to the inlets of the pumping loops may significantly affect
charge generation in the lower plenum. With an increased height local eddies in the bottom of
the tank may be reduced thereby reducing charge generation from the tips of the washers and the
slot inlets to the ducts. In order to study the effect of the lower plenum height on static
electrification, the winding assembly was moved up linearly by 7 inches, which represents at
least a tripling of the distance from the H-L washer to the bottom of the tank (see Figure 4). The
complete set of experiments involving 2, 4 and 6 pumps is repeated for this new configuration.

Figure 10 and Figure 11 shows the results of measurements using high charging oil and
Figure 12 and Figure 13 show similar results for regular Shell Diala A oil. For the tests involving
regular Shell Diala A oil, the raised model gave consistently less charge density for all flow rates
over the entire test temperature range. The high charging oil gave similar results at high
temperatures (>27C). However, at low temperatures, the raised model exhibited high charge
densities in both the upper and lower plenums than the base case model. It is not quite clear why
the charges are higher for the raised model at the lower temperatures during the heat-up cycle.
On the cool down cycle, the results are similar to the Shell Diala A oil results.

350

300 Top ACS (6 Pumps)


Top ACS (4 Pumps)
Top ACS (2 Pumps)
Charge Density (uC/m )
3

250

200

150

100

50

0
10 15 20 25 30 35 40 45 50
o
Upper Temperature ( C)

Figure 10
Charge Density Measurements in Upper Plenum for High Charging Oil

2-77
Modeling and Prediction of Static Electrification

60
Bottom ACS (6 Pumps)
Bottom ACS (4 Pumps)
50 Bottom ACS (2 Pumps)

Charge Density (uC/m )


3
40

30

20

10

0
10 15 20 25 30 35 40 45 50
o
Upper Temperature ( C)

Figure 11
Charge Density Measurements in Lower Plenum for High Charging Oil

50

Top ACS (6 Pumps)


Top ACS (4 Pumps)
40
Top ACS (2 Pumps)
Charge Density (uC/m )
3

30

20

10

0
10 15 20 25 30 35 40 45 50
Upper Temperature ( oC)

Figure 12
Charge Density Measurements in Upper Plenum for Regular Shell Diala A Oil

30

Bottom ACS (6 Pumps)


Bottom ACS (4 Pumps)
25
Bottom ACS (2 Pumps)
Charge Density (uC/m )
3

20

15

10

0
10 15 20 25 30 35 40 45 50
o
Upper Temperature ( C)

Figure 13
Charge Density Measurements in Lower Plenum for Regular Shell Diala A Oil

2-78
Modeling and Prediction of Static Electrification

Mitigation Experiment

Upon analyzing the charge density results for the experiments, it is evident that there is a
characteristic temperature when charge relaxation becomes the dominant process in the
insulation system. Beyond this temperature, charges cannot accumulate to dangerous levels. It is
also evident that charge densities are quite low over the entire temperature range when the flow
rate is low (2 pumps running). It can be hypothesized that the system can be operated under safe
conditions by running the fewest possible pumps till the oil temperature gets beyond the critical
temperature. At this point the full flow capability of the system can be utilized.

This section describes results of a test as described above. The system is operated with
two pumps until the top oil temperature reaches 45C. All six pumps are then turned on for the
duration of the experiment. The reverse procedure is implemented on the cool-down phase of the
measurement. Figure 14 shows the results of the mitigation measurements superimposed on the
complete measurements using both two pumps and six pumps.

It is obvious from the results that the mitigation experiment track rather well the results of the
two and six pump runs for the ranges of temperature they overlap. It is also evident that for the
mitigation experiment, the charge density in the upper plenum of the tank never goes through
any high peak transitions. It is assumed that procedure such as describe above will minimize the
risk of static electrification failures during operation of a transformer.

300
Top ACS (6 Pumps)
Bottom ACS (6 Pumps)
250 Top ACS (2 Pumps)
Bottom ACS (2 Pumps)
Top ACS (Mitigation Test)
Bottom ACS (Mitigation Test)
200
Charge (C/m )
3

150

100

50

10 20 30 40 50 60
o
Upper Temperature ( C)

Figure 14
ACS Measurements Under Operational Mitigation

2-79
Modeling and Prediction of Static Electrification

General Oil Tests

Oil samples were taken at the start and finish of each series of tests and analyzed for oil
condition. The complete set of results are shown on Table 1. Of particular interest in these results
are the oil tests from the sample taken at the end of the run with high charging oil and the core
and coil assembly in its base case position. The amount of hydrogen in the oil increased from
trace amounts to 85 PPM, signifying some degree of discharging. The acid neutralization number
which gives an indication of the level of deterioration of oil increased five times, from 0.004 to
0.021 mgKOH/gm. Notice that in conjunction with these changes in the oil, the ECT actually
decreases from 580 to 330 C/m3. This will indicate that over the course of the experiments, the
high charging impurities get absorbed/adsorbed into the cellulose fibers.

Table 1
General Oil Test Results

Oil Tests Shell Diala A, Shell Diala A, High-charging oil, High-charging oil,
Base Case Raised Model Raised Model Base Case

Start End Start End Start End Start End

Date 27-Jul 4-Sep 10-Sep 18-Sep 21-Sep 29-Sep 7-Oct 16/26-Oct

Moisture 7.39 6.1 3 3.76 7.92 3.88 2.24

Conductivity 0.5 0.39 0.4 0.5 0.71 0.59 0.55 0.47

Gas-in-Oil 6.80% Low 4.24% 5.51% 5.29% 9.04% 3.24%

N2 63321 33304 39267 38046 65005 28746

O2 9666 8814 12658 9003 17007 5261

CO2 137 16 135 121 142 66

CO 66 16 28 38 27 30
H2 trace 0 trace 0 trace 85
CH4 Methane 1 trace 1 1 1 1

C2H4 Ethylene 2 64 1 1 1 trace


C2H6 Ethane trace 1 1 0 trace

C2H2 Acetylene 1 trace trace 1 1 trace

General Oil

Dielectric Strength 52 57.8 55.5 54.5 48.7 57.2 55.4 53.5

Power Factor 0.008 0.004 0.004 0.0047 0.011 0.0095 0.0063 0.008
Interfacial Tension 26.3 31.1 34.1 34.8 34.6 37.32 37.06
Neutralization No. 0.03 0.008 0.005 0.002 0.0063 0.003 0.004 0.021
ECT 46 28 56 50 655 560 587 329

2-80
Modeling and Prediction of Static Electrification

Summary and Conclusions

The results of the experiments presented in this paper clearly show the evolution of charges in
the oil spaces of a transformer during operation and as a function of the oil temperature. It is
clear, as has always been known that the charge density in the oil of a transformer is directly
related to the electrostatic charging tendency (ECT) of the oil. The higher the ECT, the easier it
is for charges to separate in the oil when forced through the transformer insulation system. The
dependence of electrostatic charging on the flow rate of the oil through the ducts has been
demonstrated, with the charge generation increasing with oil flow rate. It is also clear that as the
oil temperature increases, charge relaxation and recombination become the dominating processes
as opposed to charge generation in the insulation system.

Following this realization, perhaps the most important confirmation is that the potential dangers
posed by static electrification can be minimized by modifying the operational procedure of a
transformer. The mitigation experiment demonstrated that if the transformer is operated with the
fewest possible pumps at temperatures below a critical temperature (approximately 50C), then
the remaining pumps can be added at high temperature without increase in charge accumulation
in the oil spaces of the transformer. It should be noted that the same procedure should be
enforced during cool-down of the transformer since similar generation/relaxation processes exist
during both cycles.

Another important result of these experiments deal with minor design changes to the geometry of
the lower plenum of the transformer. The results show that an increase in the lower plenum
height reduces charge generation both in the lower plenum and the ducts. The core and coil
assembly of a transformer exhibiting symptoms of static electrification can possibly be
rearranged in such a way that the effective height of the lower plenum is increased.

References

1. Moore H., et al., Ramapo Streaming Electrification Report.

2. Shimizu, S. Electrostatics in Power Transformers, IEEE Transactions on Power Apparatus


and Systems, Vol. PAS-98, July/August 1979.

3. Tamura, R., Static Electrification by Forced Oil Flow in Large Power Transformers,
IEEE Transactions on Power Apparatus and Systems, Vol. PAS-99, No. 1. IEEE PES
Summer Meeting, Vancouver, British Columbia, July 1980.

4. Crofts, D. W., Static Electrification in Power Transformers, Electrical Insulating Oils,


STP 998, H. G. Erdman, Ed., American Society for Testing and Materials, Philadelphia,
1988, pp. 136-151.

5. Roach, J. F. and Templeton, J. B., An Engineering Model for Streaming Electrification in


Power Transformers, Electrical Insulating Oils, STP 998, H. G. Erdman, Ed.,
American Society for Testing and Materials, Philadelphia, 1988, pp. 119-135.

2-81
Modeling and Prediction of Static Electrification

6. Okubo, H., Charging Tendency Measurements for Transformer Oil, Transaction A


79 051-4, IEEE Winter Power Meeting, New York, 1979.

7. Oommen, T. V., Frimpong, G. K. and Walden, J. M., Certain High Charging Polymeric
Materials in Transformer Oil, Conference Proceedings of International Conference on Solid
Dielectric (ICSD), Vasteras, Sweden, 1998.

2-82
Modeling and Prediction of Static Electrification

2.6 Streaming Electrification Dynamics - Duct Modeling and Analysis

J. A. Palmer
Division of Engineering
Colorado School of Mines
Golden, CO 80401

J. K. Nelson
Electric Power Engineering Department
Rensselaer Polytechnic Institute
Troy, NY 12180

Abstract

The problem of streaming electrification in transformers has most frequently manifested itself
within a short time after the initiation of flow or energization of the unit. Various experimental
studies have indicated that temporal and spacial transient phenomena exist in the distribution and
accumulation of charge. In the interest of better understanding the nature of those dynamics, a
computational model has been developed of streaming electrification within a duct. Through the
use of this model, details of the charge redistribution resulting from the initiation of flow are
made observable, and new understanding of the dynamic nature of the phenomenon is revealed.
This paper briefly describes the nature of the model including the charge transport, flow, and
boundary conditions used in a finite difference representation. The details of the results for a
number of simulations are provided for parametric assessment in flow velocity, applied electric
field, geometry, and chemical characteristics of the fluid. The spacial and temporal data from the
model is analyzed to extract the practical implications that may be drawn from the study,
including recommendations on physical design of flow ducts, flow rate, and system energization.

Introduction

The phenomenon of streaming electrification has been implicated in the failure of dozens of
large power transformers worldwide. Extensive research has brought to light many inherent
characteristics, but many aspects of the phenomenon are still not very well understood,
particularly issues such as the spacial and temporal development of electrostatic charge. These
are especially important features, as evidenced by the nature of most of the relevant failures. In
most cases, the teardown of those transformers revealed extensive tracking and carbonization of
insulation in the entry region to the ducts, but very little towards the exit, highlighting the
importance of spacial development, while the frequency of failures occurring within a few hours
of startup is suggestive of the significance of charging time constants.

An extensive study [1] has been conducted to better understand these transient aspects of the
phenomenon. Temporal transients were observed on two scales, a short-term transient having a
duration of seconds to tens of seconds, and a long-term transient lasting several hours. To
provide more insights into the short-term transient and its relationship to the spacial development
of charge, a simplified system was simulated using a finite-difference implementation of the
charge transport equations. While absolute precision is unattainable, much information about
trending and relative performance of various systems and operating conditions may be obtained.

2-83
Modeling and Prediction of Static Electrification

A computational model may provide such information without the expense and effort required
for the implementation of a physical model. Furthermore, some parameters of the physical
model, such as the spacial development of the charge profile, may not be measurable to the
desired resolution without significantly disturbing the system. A computational model can
readily incorporate both temporal and spacial transients in two dimensions [2].

Having completed the mathematical representation, a number of important parametric behaviors


are observed by exercising the model. Many of these characteristics have been previously
observed in the laboratory, but the present work utilizes the details available in the model, to
provide a more complete explanation for the observed behavior [3]. The objective of this model
is to simulate the development of charge density profiles and currents in space and time at the
leading edge of a channel at the initiation of flow. The paper also explores the practical
implications of the parametric investigation.

Charge Transport

In modeling the charge redistribution process, it is assumed that transport of charge occurs by
three mechanisms. The conduction mechanism is the displacement due to the force of the electric
field on the ions present, diffusion driven charge displacement occurs due to the effects of
Brownian motion, and convective charge transport
& results from the flow of the fluid in which the
charge is contained. The net current density j is the sum of the three components as follows:

j = j cond + j diff + j conv = E Dq + v q (1)

where is the fluid


&
conductivity (assumed to be uniform throughout the fluid and invariant with
ionic species), E is the
& electric field, D is the coefficient of diffusion (assumed to be constant for
all species of ions), v is the fluid velocity vector, and q is the net volume charge density. The
effect of these transport mechanisms on the charge density in the fluid as a function of time, is
found from the divergence of the current density, with a substitution for the divergence of the
electric field based on Gauss Law, as [4]:
q q
j = = + D 2 q ( v q ) (2)
t

The solution to this equation gives the full charge density distribution development in space
and time.

Implementation

The driving engine for the entire simulation is the finite difference numerical method. This is
based upon the use of Taylor series expansion to calculate numerical derivatives.

2-84
Modeling and Prediction of Static Electrification

Initial Charge Density Distribution

The general geometry covered in the scope of the simulations performed here is depicted in
Figure 1. The system is assumed to be invariant in the z direction (normal to the page). The
x-axis is horizontal, parallel to the mean flow, with the zero at the leading edge of the channel.
The y-axis is aligned perpendicular to the channel walls with the zero at the channel center. The
region of computation is subdivided into a rectangular mesh, so that derivatives are calculated
along lines of constant y and constant x. To provide for good resolution near the wall, the step
size must be less than the Debye length, although a lower level of resolution may be used in
regions of reduced charge densities, so a variable step size is used to reduce memory
requirements and computation time.

Figure 1
Typical Geometry for Dynamic Streaming Electrification Solver. Darker Shading Indicates
Region of Higher Node Density

In developing a solution, boundary conditions are crucial. This model requires spacial boundary
conditions near the wall, the inlet, and the outlet. The wall boundary condition has been the
subject of considerable theoretical development (e.g. [4-6]), but the exact nature of the
electrochemical processes occurring at the interface is still not fully understood, so a simplified
approximation is used. Due to the nature of the solid surface and its contaminants, there is a
preferential adsorption of one species of the ions existing in the dielectric fluid. The adsorbed
ions form a fixed layer of surface charge density, -Qwall. The remaining ions are distributed
through the fluid in the diffuse layer as dictated by conduction, diffusion, and convection. At the
wall, there is no convective current, although conduction and diffusion continue to operate as
described in equation (1) [7]. This assumes a simple ohmic injection at the wall, and is sufficient
to demonstrate the practicality of the simulation, although a degree of asymmetry is necessarily
incorporated to model the effect of imposed electric fields, as will be shown.

The charge density in the fluid adjacent to the wall is set by a modified Dirichlet condition,
adjusting for the current density through the wall [4]:

2-85
Modeling and Prediction of Static Electrification

Q wall j
q wall = (1 + wall ) (3)
jw

where jw is an empirical constant. The leading edge of the region of computation utilizes a
Dirichlet condition fixed at zero, and all other boundaries utilize Neumann conditions.

The initial pre-flow charge distribution was determined by using the finite difference scheme to
&
solve the steady-state form of equation (2) (i.e. with qt = 0 and = 0 ):

q
= D 2 q (4)

Figure 2
The Log of the Normalized Initial Charge Distribution. Plot Shows Lines of Constant
Charge Density

This may be solved in closed form in one dimension, but must be solved numerically in two
dimensions. A normalized solution may be computed for a given geometry, and then scaled by
wall charge density and Debye length. A typical normalized initial charge distribution contour
plot is given in Figure 2. The dimensions are scaled by the Debye length and the charge density
is scaled by the wall charge density.

Electric Potential Distribution

After the initial charge distribution has been determined, the potential distribution must be
calculated to determine charge injection from the wall into the fluid. The approach is largely the
same as the approach taken in the initial charge distribution calculation, utilizing Poissons
equation as the basic driving equation. The lower wall of the channel is taken to be uniformly
zero potential, and the upper wall is uniformly equal to the applied voltage, i.e. both are Dirichlet
boundaries. This is basically equivalent to defining the wall material to be conducting. Clearly,
future work would ideally include an added level of complexity to account for the characteristics
of insulating surfaces. All other boundaries are Neumann boundaries.

2-86
Modeling and Prediction of Static Electrification

Flow Velocity Distribution

Key to the assessment of streaming electrification dynamics is the inclusion of flow dynamics,
which is one of the unique features of this model. Due to the complexity of the geometry and
boundary conditions, a closed form solution was not possible, so a finite difference analysis was
also used on this portion.

Laboratory experience indicated that the flow transients were more than an order of magnitude
shorter than transients in the charging characteristics, so the details of the fluid velocity temporal
transients were considered to be less important. Nevertheless, the sudden application of full flow
to the system was found to create problematic numerical instabilities. The compromise between
excessive computation time required for a dynamic solution of the flow and this instability was
to solve the steady-state flow velocity profile once and scale it uniformly from a no-flow initial
condition to the full flow steady-state [8], so that
8tv
v ( x, y, t ) = [1 exp( )] v ( x, y, ) (5)
hh

with v the kinematic viscosity and hh the channel half height. The approach for the numerical
solution of the flow velocity profile of the system is based on the time independent Navier-
Stokes equations in two dimensions [9]. The flow boundary conditions assume uniform flow at
an infinite distance upstream and fully developed laminar flow an infinite distance downstream.
A plot showing stream lines for a typical configuration having a Reynolds number of 565 is
shown in Figure 3. It should be noted, that the laminar assumption loses its validity in the wake
of the eddy at the leading edge because the instability tends to induce some degree of turbulence.
Nevertheless, as an approximation, the laminar assumption does shed some light on leading edge
effects.

Figure 3
Streamlines for NRe = 565. Dimensions are Normalized by Channel Half Height

2-87
Modeling and Prediction of Static Electrification

Dynamic Charge Distribution and Currents

Once the initial charge distribution, potential distribution, and flow profile have been calculated,
q
t
is successively computed, from which the new charge distribution and resultant currents are
determined. Because the change with respect to time is computed using a forward difference
calculation, the system is not inherently stable. Therefore, it is necessary that some precautions
be taken to ensure stability. First, the time step must be chosen such that the distance in which an
effect may be propagated in a single time step is less than one spacial step. Second, up-wind
differencing is used for improved numerical stability over a central difference scheme, i.e. the
derivative at a point is calculated using the point for which the derivative is being calculated and
the adjacent point upstream. Other numerical stability issues are addressed in some detail in [2].

The knowledge of the charge distribution in the system is sufficient to visualize the processes of
charge transport and better understand what is happening in the system. However, the technology
currently available does not provide the means to measure the charge distribution with
significant resolution in regions within a few tens of Debye lengths. Therefore, to make useful
comparisons with experimental systems, the simulation must provide other information, in the
form of measurable currents. Two currents are measurable for this system: a leakage current and
a streaming current. The streaming current is simply the integral of the convected current
density, across the trailing boundary. The leakage current consists of two components: a
conduction current, consisting of the integral of the wall current density, and displacement
current, being the current induced on the electrodes due to the removal of charge from the region
of interest.

Model Performance

Base Case

For purposes of comparison, it is necessary first to establish a base case. The base case here will
be referred to as Case A, for which the chemical, geometric, and operating parameters are given
in Table 1. The currents produced for these operating conditions are plotted in Figure 4 as a
function of time from the initiation of flow, where the leakage current is the sum of wall
conduction current and displacement current. Kirchoffs current law indicates that with zero inlet
charge, the total leakage current (into the fluid from both walls) should be equal to the streaming
current (exiting the outlet of the channel). A sharp peak in the leakage current results from
computational error as the charge undergoes a redistribution with the initiation of flow in a non-
uniform finite difference grid. The shift of charge from a region of high density to a region of
low density results in an apparent annihilation of charge when the charge of the entire system is
integrated, hence the narrow spike seen in the displacement and leakage currents. Aside from the
presence of that spike, the leakage current from both walls (only one wall is plotted) matches the
streaming current within 6%. This mismatch would be reduced by a finer subdivision of the grid.

The cause of the current transient is illustrated in Figure 6. These plots show charge density
contours in the vicinity of the wall, the line closest to the wall (bottom edge of frame) being
0.01 C/m3, and each subsequent contour dropping by one order of magnitude. With the initiation
of flow, charge swept off the leading edge towards the center of the channel provides a front that

2-88
Modeling and Prediction of Static Electrification

propagates down the channel. Behind the front, the thickness of the diffuse layer is reduced and
the net charge entrained in the fluid is decreased, resulting in an overall decrease of streaming
current, even though the charge density in the center of the channel is much larger than it was
prior to initiation of flow. Similarly, the wall current increases as the concentration gradient
increases, and the field due to the charge in the oil decreases, while the displacement current falls
from an initial peak as conditions tend toward the steady-state.

Table 1
Summary of Fluid, Operational, and Geometric Parameters for Case A. The
Meaning of Dimensional Parameters is Illustrated in Figure 1


2 -5
Kinematic viscosity (m /s) 1.77 X 10

3
Fluid Density (kg/m ) 900
2 -11
D Diffusion Coefficient (m /s) 1.28 X 10

-12
Conductivity (S/m) 4.42 X 10

-11
Permittivity (F/m) 1.86 X 10
2 -7
Qwall Surface Charge Density (C/m ) 1.0 X 10

-6
Debye Length (m) 7.33 X 10
hh Channel Half Height (m) 0.001
hhs Spacer Half Height (m) 0.0005
Nhh Subdivisions of hh 30
Nhs Subdivisions of hhs 20
L Channel Length (m) .1
NL Subdivisions of L 200
Lup Upstream distance (m) .002
Nup Subdivisions of Lup 24
h0 Smallest Step Size /2
Vmean Mean flow velocity (m/s) 0.5
NRE Reynolds Number 56.5
3
Qinlet Inlet Charge Density (C/m ) 0.00
Eapp Applied Field (kV/m) 0.00
F Frequency of Applied field (Hz.) 60

2-89
Modeling and Prediction of Static Electrification

Figure 4
Flow Induced Currents for Example Case A

Figure 5
Streaming Current as a Function of Time for Various Velocities

2-90
Modeling and Prediction of Static Electrification

Operating Conditions

The results of simulations or laboratory experiments are very dependent on a broad range of
issues. Because of the complexity of the phenomenon, an absolute analysis, either in
measurement or simulation is not possible. However, much insight with respect to the nature of
the phenomenon may be gained by comparative analysis. The sections below investigate the
parametric dependence of the phenomenon on the operating conditions, geometry, and chemical
issues, respectively, by comparing some of the various simulated situations with the base case
example (Case A), and then assessing the reason for the differences.

Velocity

Based on Case A, one may deduce the effects of various changes in operating conditions. An
increase in velocity, for example, may be expected to shorten the transient somewhat, due to the
faster movement of the charge front. At the same time, the height of the transient is increased,
because the charge density gradient is higher, providing for higher wall current, as well as
reduced penetration of the diffuse layer into the flow, in the steady-state. Similarly, a reduction
in flow rate will lengthen the transient and shorten its height, so that, in the limit, no overshoot
will be apparent. This was investigated using the computational model on Cases B, C, and D
having all characteristics identical to Case A as outlined in Table 1, except that the average oil
velocities were 1.0, 0.2, and 0.1 m/s respectively, while Case A had a velocity of 0.5 m/s. The
plot of the streaming currents for cases B-D are shown in Figure 5.

Figure 6
Peak and Steady-State Streaming Currents as a Function of Flow Rate

2-91
Modeling and Prediction of Static Electrification

Figure 7
Charge Distribution Development near the Lower Wall for Case A. The Contour line
3
Closest to the Channel Wall is .01 C/m with one order of Magnitude Decrease for each
Subsequent Line

As predicted, the faster flow had a higher and narrower peak, while the slow flow had barely any
overshoot at all. The height of the peak is directly proportional to the flow rate. At the outlet, the
flow and charge profile are fully developed at the initiation of flow, so the convolution of charge
density with the velocity profile will be directly proportional to the flow rate. The steady-state
streaming current magnitude, on the other hand, is much more dependent on the degree of
development of the flow and charge distribution profiles at the outlet, and so will not necessarily
be proportional to flow rate. The peak and steady-state currents for a range of velocities are
plotted in Figure 7. For this system geometry, the steady-state current has a nearly linear
dependence on flow rate, diverging gradually from the peak current curve. Such a near-linear
relationship corresponds to observations of published experimental studies [10, 11]. Clearly the
dependence will change at the transition to turbulence, as this causes an effective change of
diffusivity outside a laminar boundary layer.

2-92
Modeling and Prediction of Static Electrification

Applied Field

The effect of applied electric fields is also significant. Experimental studies have indicated that
applied AC fields enhance streaming electrification [12, 13]. Tanaka [11] attempted to explain
this by a semi-empirical enhancement factor based on a concept of a modulating diffuse layer.

However, the equation for charge displacement is independent of field magnitude, although a
term for electric field is included in the equation for charge transport. Therefore, in a system
having a constant conductivity throughout the fluid, a non-divergent field will not redistribute
charge. Any charge transported by the conductive current density will immediately be
replenished, leaving constant net charge density. The interface, on the other hand, is strongly
influenced by the magnitude of the field, resulting in injection or depletion of charge. However,
this is a boundary condition, which must propagate into the bulk to modify global charge
densities.

In a symmetric system, equal and opposite fields appear at either side of the channel. If the
interfacial charge transfer mechanism is symmetric, then equal charge is transferred across each
interface and no field enhancement to streaming electrification results. To provide a charge for
convective transport, it is necessary that some non-linearity be introduced to provide a DC
component to the leakage current. When an asymmetry was introduced (a field-dependent
injection coefficient for one ionic species), the applied field significantly enhanced the streaming
current, as shown in Figure 8. While the mechanism used to introduce asymmetry into the model
was incomplete, it demonstrates that the enhancement by applied AC field results from the non-
linear interfacial behavior. Further investigation should provide an exact model of the charge
transfer mechanism at the interface, but this mechanism was sufficient for present purposes.

Geometry

Another factor having a strong influence on the streaming electrification phenomenon, is


geometry. The shape of the model was fixed, but the dimensions of the channel half height, the
spacer half height (the cross section of the channel wall), and the length of the channel could all
be adjusted independently. The results of exercising the model for variations in these dimensions
are the subject of this section.

Figure 8
Simulated Steady-State Streaming Current as a Function of Applied Field

2-93
Modeling and Prediction of Static Electrification

Channel Half Height

To compare the effects of channel half height on streaming electrification, three cases, G, H and
I, having half heights of .5, 2, and 5 mm respectively, were run at a velocity of 1 m/s for
comparison against Case B. The aspect ratio of the leading edge and all other characteristics are
as given for Case A in Table 1. The streaming currents are plotted in Figure 9. The steady-state
current decreases with increased channel width for constant average velocity because the
velocity near the wall is lower for the wider channel. The lessening of the overshoot in the initial
transient with increased width is the result of the slower development of steady-state flow
conditions.

There is little previously reported information regarding the effect of channel half height.
However, an investigation by Heydon et al. [14], indicated that for the same flow rate through
sintered metal filters, higher currents were generated for smaller porosities, which supports the
concept of higher currents in narrower channels.

Figure 9
Plot of Streaming Current as a Function of Time for Various Channel Half Heights

Leading Edge Aspect Ratio

The shape of the leading edge has a significant effect on the nature of the flow in that region and
further down the channel. To observe the effect on the streaming current, Cases J and K were
created for comparison with Case I, having a leading edge aspect ratio, hh hs of 10, 1 and 2,
respectively. It was anticipated that the formation of eddies for smaller ratios would result in a
higher streaming current, but the results of the simulation, shown in Figure 10, found that the
highest aspect ratio resulted in the highest current. This is the effect that was seen in both
simulations and experiments by Touchard with a periodic roughness in a pipe [15]. Eddies in the
periodic grooves shielded the wall from the higher charge gradients, resulting in reduced charge
transport. Because the eddies of his system were constrained within a groove, no wake was
formed, and therefore his result was valid. For this model, the flow simulation was based on an
assumption of completely laminar flow, but in reality, a region of turbulence would form in the
wake of the eddy that is formed on the leading edge. This would enhance the diffusion

2-94
Modeling and Prediction of Static Electrification

coefficient, resulting in elevated charge generation. When the region of turbulence is ignored the
eddy appears to shield part of the wall near the leading edge, so that the larger eddy results in a
lower current, as seen in the plot. Thus the counter-intuitive result is a limitation of the model. It
is expected that if turbulence, including the turbulent diffusivities, were modeled, the eddy would
contribute significantly to the streaming current. A similar group of cases (not shown here) was
performed with the same aspect ratios, and a channel half height of 1 mm, in which no eddy
forms on the leading edge. In the absence of an eddy, the overshoot of the transient increased
slightly (a few percent) for the higher aspect ratio, but the steady-state results were the same for
all three aspect ratios.

Figure 10
Streaming Currents as a Function of Time for Channels with Different Leading Edge
Aspect Ratios

The current in Cases I and K indicates a double peak. The first peak occurs because sufficient
charge is swept from the leading edge into the bulk of the flow when the flow is first initiated,
that the streaming current is affected when that charge reaches the outlet. This peak is not present
for higher aspect ratios where the vertical flow component is significantly reduced, limiting the
charge swept into high flow regions. The second peak is due to charge gradients stabilizing near
the wall, as discussed previously.

Channel Length

The influence of channel length on streaming electrification was investigated using three cases
derived from Case A, differing only in length. The lengths of the channels in cases L, M, A, and
N, are 2, 5, 10 and 15 cm, respectively. The values of peak and steady-state currents are plotted
in Figure 11. Because the model does not account for outlet effects, this becomes a plot of the
development of the streaming current along the channel. Both the peak and steady-state currents
asymptotically approach a common limiting value. An exponential streaming current
development was predicted in previous research [16, 17], but here the rate of rise is faster than a
pure exponential because the perturbation in the entry of the channel due to the leading edge
results in faster development of the charge profile. Likewise, the diminishing overshoot results
from the relaxation of the charge in the initial front.

2-95
Modeling and Prediction of Static Electrification

Figure 11
Plot of Streaming Current as a Function of Channel Length

Chemical Characteristics

Perhaps the most subtle of the parametric dependencies of streaming electrification is that of the
inherent chemical characteristics of the oil, such as conductivity, diffusivity, and propensity of
charge to be adsorbed onto the wall of the channel. These characteristics are often strongly
affected by impurities and contaminants, even at immeasurable levels. Nevertheless,
understanding the relationship to these parameters may assist in specifying appropriate
acceptance levels for the oil used.

Conductivity

The conductivity affects the streaming electrification because the charge distribution penetration
into the flow is strongly dependent on the Debye length, , defined as

D
= (6)

where D is the coefficient of diffusion, is the oil permittivity, and is the conductivity. An
investigation of this characteristic entailed running two cases, O and P, derived from case A,
with conductivity increased by a factor of 2 and 4 respectively. The results in Figure 12 show
that both peak and steady-state currents are lower for increased conductivity, because of the
reduction in Debye length. Experimental evidence has shown, however, that for most oil
systems, the streaming current tends to increase with increased conductivity [18, 19]. Yet, the
results of the simulation indicating a decrease as the Debye length decreases, are quite intuitive.
The resolution of this apparent paradox lies in the nature of the conductivity change.
Conductivity is dependent on both the mobility and concentration of ions. Given that the
adsorption of ions is an interfacial reaction, the adsorbed charge density also depends on ion
concentration. The experimental interaction between streaming electrification and conductivity
also include secondary effects of charge adsorbed on the surface. This concept is further

2-96
Modeling and Prediction of Static Electrification

underscored by the fact that although leakage currents usually increase with increased
conductivity, raising conductivity by increasing moisture content results in decreased leakage
current [20, 21], implying that some contaminants may modify the oil conductivity without
increasing the wall charge density.

Figure 12
Plot of Streaming Current as a Function of Time for Cases A, O and P, Demonstrating the
effect of Oil Conductivity on Streaming Current Transient and Steady-State Characteristics

Wall Charge Density

To investigate the effect of change in wall surface charge density, cases Q and R (comparable
to A), were simulated with charge densities decreased by a factor of 2 and 5 respectively. The
streaming current for the three cases changed only by an exact linear scaling, as would be the
case for a wide range of wall charge density. Hence, the net effect of a change in streaming
current with conductivity may be estimated. If the adsorbed charge is assumed to be proportional
to the ion concentration, and thus proportional to the fluid conductivity, the results obtained
above may be scaled, yielding the plot of Figure 13, in which the streaming current is seen to
increase with conductivity.

Figure 13
Plot of Streaming Current as a Function of Conductivity, Adjusted for Changes in Surface
Charge Density, Compared with the Unadjusted Case

2-97
Modeling and Prediction of Static Electrification

Coefficient of Diffusion

Debye length may also be changed by modifying the diffusion coefficient, D. For cases with the
same Debye length, one would expect the initial current to be the same, and the steady-state
current would depend on the development of charge along the wall. Case S demonstrates the
difference in the performance depending on the way the Debye length is achieved. This case has
the same Debye length as Case O, but with half the diffusivity and conductivity. The streaming
currents for the two cases, plotted in Figure 14, start at the same level, having the same initial
charge distribution. However, the reduced diffusion coefficient results in slower development of
the diffuse layer near the wall, so the final charge density does not penetrate as deeply into the
flow, resulting in a significantly lower steady-state current.

Figure 14
Comparison of two Cases, Cases O and S, having the same Debye Length, but Different
Conductivity and Diffusivity

Conclusions
The complexity of the streaming electrification phenomenon has obscured an understanding of it
for a number of years. Nevertheless, through the assistance of numerical modeling, this paper has
brought new light to the problem. With the details made visible by the spatially and temporally
distributed nature of the calculation, the relation of the basic streaming electrification process to
the dependency on operational, geometric, and chemical parameters of the system has been
discussed. The computational model provides the flexibility and refinement to give a detailed
understanding of the material that cannot be obtained by experimental means due to limitations
in measurement technology.

A number of practical implications may be drawn from these results. The presence of an
overshoot in the streaming currents for high flow rates implies that a gradual application of high
flows may avoid some undesirable operating conditions. Electric fields enhance the
phenomenon, and it is desirable to avoid superposition of transients caused by application of
flow and application of electric field. The geometry of the channel has a serious bearing on the
transient nature of the streaming currents. The chemical makeup of the insulating oil also affects
the streaming electrification performance, and this model has indicated that the effects of wall
charge and conductivity must be considered together when modifying oil characteristics. Further
discussion on the practical implications of these findings relating to geometric, design and
operational issues are discussed elsewhere [22].

2-98
Modeling and Prediction of Static Electrification

Acknowledgements

This paper is based on work undertaken under contract for the Electric Power Research Institute,
and the authors are indebted to Mr. S. R. Lindgren who managed the program.

References

1. Palmer, J. A., Dynamics of Streaming Electrification in Large Power Transformers. Doctoral


Thesis, Rensselaer Polytechnic Institute, Troy, New York, 1996.

2. Palmer, J. A. and J. K. Nelson, The Simulation of Short-Term Streaming Electrification


Dynamics. Journal of Physics D: Applied Physics, 1997. 30: pp. 1207-1213.

3. Palmer, J. A. and J. K. Nelson, Parametric Performance of a Transient Streaming


Electrification Model. IEEE Transactions on Dielectrics and Electrical Insulation, 1997.
4: pp. 418-424.

4. Abedian, B. and A. A. Sonin, Theory for Electric Charging in Turbulent Pipe Flow.
Journal of Fluid Mechanics, 1982. 120: pp. 199-217.

5. Watanabe, S., et al., Relationship between streaming current and activation energy. in
Second International Conference on Properties and Applications of Dielectric Materials.
1988.

6. Walmsley, H. L. and G. Woodford, The Generation of Electric Currents by the Laminar


Flow of Dielectric Fluids. Journal of Physics: D., 1981. 14: pp. 1761-82.

7. Kozman, I. and J. Gavis, Development of Charge in Low-conductivity liquids flowing past


surfaces: Engineering predictions from theory developed for tube flow. Chemical
Engineering Science, 1962. 17: pp. 1013-1022.

8. Stavitsky, D. and E. Macagno, Approximate Analysis of Unsteady Laminar Flow. Journal


of the Hydraulics Division, ASCE, 1980. 106: pp. 1973-1980.

9. Wang, Y. L. and P. A. Longwell, Laminar Flow in the Inlet Section of Parallel Plates.
A.I.Ch.E. Journal, 1964. 10(3): pp. 323-329.

10. Higaki, M., et al., Static Electrification and Partial Discharges Caused by Oil Flow in
Forced Oil Cooled Core-Type Transformers. IEEE Transactions on Power Apparatus and
Systems, 1979. 98: pp. 1259-1267.

11. Tanaka, T., N. Yamada, and Y. Yasojima, Characteristics of Streaming Electrification in


Pressboard Pipe and the Influence of an External Electric Field. Journal of Electrostatics,
1985. 17: pp. 215-234.

12. Miyao, H., M. Higaki, and Y. Kamata, Influence of AC and DC Fields on Streaming
Electrification of Transformer Oil. IEEE Transactions on Electrical Insulation, 1988.
23: pp. 129-135.

2-99
Modeling and Prediction of Static Electrification

13. Peyraque, L., et al., Static Electrification and Partial Discharges Induced by Oil Flow in
Power Transformers. IEEE Transactions on Dielectrics and Electrical Insulation, 1995.
2: pp. 40-45.

14. Heydon, R. G., J. Rungis, and R. Sheehy. The Use of Sintered Metal Filters in the Study of
ECT in Transformer Oils. in Conference on Electrical Insulation and Dielectric
Phenomena. 1992.

15. Touchard, G., et al., Static Electrification by Laminar Flows Through Artificially
Roughened Pipes. IEEE Transactions on Industry Applications, 1989. 25(6): pp. 1067-1072.

16. Roach, J. F. and J. B. Templeton, An Engineering Model for Streaming Electrification in


Power Transformers, in Electrical Insulating Oils, H. G. Erdman, Editor. 1988, American
Society for Testing and Materials. pp. 119-135.

17. Touchard, G., et al., Flow Electrification in Power TransformersExplanation of the


Wall-Current Measurements. IEEE Transactions on Dielectrics and Electrical Insulation,
1994. 1(4): pp. 728-733.

18. Higaki, M., et al., Consideration of the Measurement of Static Charges in Insulating Oil and
the Influence of Oil Conductivity on the Charges. Electrical Engineering in Japan, 1984.
104(5): pp. 9-17.

19. Okubo, H., et al., Charging Tendency Measurement of Transformer Oil. in IEEE PES
Winter Meeting. 1979.

20. Poovamma, P. K., R. Jagadish, and K. Dwarakanath, Investigation On Static Electrification


Characteristics Of Transformer Oil. Journal of Electrostatics, 1994. 33: pp. 1-14.

21. Lee, M. J. and J. K. Nelson, Flow-induced Electrification and Partial Discharge


Measurements in Transformer Duct Structures. IEEE Transactions on Electrical Insulation,
1991. 26(4): pp. 739-748.

22. Palmer, J. A. and J. K. Nelson, Practical implications of transformer flow electrification


studies. Journal of Electrostatics, 1997. 40/41 (1997): pp. 693-698.

2-100
Modeling and Prediction of Static Electrification

2.7 Investigation for Standardization of Electrostatic Charging Tendency


Measurement of Transformer Oil in Japan

S. Isaka H. Miyao M. Tsuchie


Toshiba Corporation Hitachi, Ltd. Mitsubishi Electric Corporation
2-1, Ukishima-cho, Kawasaki-ku 1-1-1, Kokubu-cho 651, Tenwa
Kawasaki, 210-0862 Japan Hitachi, 316-8501 Japan Ako, 678-0256 Japan
S. Kobayashi, T. Kobayashi and M. Ikeda H. Okubo
T. Ono Kyushu Institute of Nagoya University
Tokyo Electric Power Company Technology Furo-cho, Chikusa-ku
1-1-3, Uchisaiwai-cho,Chiyoda-ku 1-1 Sensui-cho Nagoya, 464-8603 Japan
Tokyo, 100-0011 Japan Kitakyushu, 804 Japan

Abstract

Several methods for the measurement of ECT of transformer oils were developed in Japan after
the streaming electrification phenomena were confirmed. The demand for standardization of the
measurement has been raised because of the differences of the methods of measurement and
evaluation of ECT. And also many types of measuring apparatus developed up to now have been
evaluated for standardization of measuring methods at CIGRE. There is no method, which
satisfies all the properties desired. In the preliminary draft published recently it is concluded that
the Mini-static Test is suitable for standardization with high reproducibility and repeatability.
Many properties have been studied on the Mini-static Tester, however investigations were not
sufficient. The ECT value by Mini-static Tester is often influenced by oil flow rate and varies
largely even though using filters in the same lot. Recent investigation showed that the negative
properties of ECT against oil flow rate is caused by the property of paper filter. The Mini-static
Test was recommended as the standard method of ECT measurement though it needs to take a
little care of the property of a paper filter in the report by Electric Technology Research
Association.

Introduction

Static electrification in oil-immersed transformers is the phenomena induced by static charge


generated by oil flowing on the surface of insulating solid and is the cause of electrostatic
discharge in large power HV transformers. Electrostatic charging tendency (ECT) of insulating
oils has a close relation to the static electrification. After the first transformer failure was
identified with static electrification in 1974,many types of ECT measuring system have been
developed in Japan and other countries. To date, the following four kinds of charge measuring
apparatus have been used in Japan, paper pipe oil circulating method [1], base oil circulating
method [2], gravity oil flow method [3] and siphon method [4]. The first method has the charge
generation part consist of thin pipe made of kraft pulp through which test oil is constantly
circulated. The leakage current from the electrode covering the pipe to the ground is measured.
ECT is expressed in pA/cm2 by the leakage current divided by the total inner surface area of the
pipe and also evaluated in the unit Acm by the leakage current multiplied by the resistivity of oil
as electrostatic charging potential. The second method has the charge generation part consist of
rolled crepe paper through which base oil is constantly circulated and the test oil is injected into

2-101
Modeling and Prediction of Static Electrification

the base oil using the injector. The leakage current from the electrode containing the charge
generator part to the ground is measured and ECT is expressed in pC/ml of oil. The third method
has the charge generation part consist of many small pieces of pressboard contained in the glass
vessel through which test oil flows downwards by means of gravity. The streaming current from
the charge collection vessel is measured and ECT is expressed in pC/cm3. The last method has
the charge generation part consist of a siphon made of stainless steel through which test oil flows
downward. The streaming current from the charge collection vessel is measured and ECT is
expressed in pA. Although these charge measuring methods have been used for a long time, they
are different each other and so do the evaluation method of ECT. So it has been difficult to
diagnose the static electrification of transformers with the same evaluation method and the
demand for standardization of the measurement has been raised from the point of the
maintenance.

On the other hand, many types of charge measuring apparatus have also been developed besides
Japan, such as Couette flow system [5], duct flow system [6], paper tube oil flow system [7] and
spinning disk system [8], Mini-static charge tester [9] is most commonly used. For the reasons
that Mini-static tester has advantages such as being simple structure, easy to set up, good
repeatability and convenient to handle, it may be assumed that the method is easy to be followed
by other facilities and is widely accepted. In the preliminary draft [10] by TF01 of CIGRE JWG
12/15.13, Mini-static test is recommended for the standard as the initial ECT test method for
transformer materials and this draft was accepted by WG. Mini-static tester has not been
experienced in Japan up to now, investigation was undertaken on characteristics of Mini-static
tester as a candidate of standard ECT measuring method to compare it with the conventional
methods using operating transformer oils by WG on streaming electrification supported by
expert committee of Electric Technology Research Association.

Investigation on Mini-Static Tester

Mini-Static Tester

Schematic diagram of Mini-static tester used in our investigation is shown in Figure 1. It is


basically the same as the tester by Oommen, that is, same paper filter and filter holder is used at
the charge generation part. But some improvement was made to hold the reliability
(reproducibility, repeatability and stability) on the measured data.
1. Test oil should be sampled not to be contacted with air. Apparatus should be filled with
nitrogen gas not to be influenced by oxygen.
2. Oil flowing system should be rinsed by new oil and by test oil two times respectively before
subsequent measurement.
3. Measurement of ECT should be made several times, for example 4 times, for each filter and
repeat the measurement by replacing the filter about three times. The first measurement is not
stable and its data should be omitted. Mean value is calculated from the rest data.
4. Leakage current from the filter holder should be read at the half time point in the waveform
of the current with 50-ml oil flowing.
5. In case of large noise in current waveform, pressure buffer should be attached in the oil
flowing system to suppress it.

2-102
Modeling and Prediction of Static Electrification

Figure 1
Mini-Static Tester

Current Waveform Reading

Leakage current waveforms are shown in Figure 2. There are various current waveforms such as
increasing with time, decreasing with time and saturating with time depending on the measuring
facilities. As one of the reason, it may be thought that transmission of gas pressure in the oil
flowing system differs between them.

Figure 2
Leakage Current Waveforms from Filter Holder

2-103
Modeling and Prediction of Static Electrification

Current waveforms with and without a pressure buffer are shown in Figure 3. Leakage current
value is recommended to be read at the midpoint of period of oil flowing time. It corresponds to
the mean value of the integrated waveform of the current.

Figure 3
Current Waveforms with and without a Pressure Buffer

Algebraic Calculation

Investigation was made on various service-aged oils from operating transformers, which
included oil forced cooling, high voltage over 187 kV power transformers. Some of the deviation
factors (=standard deviation/mean value) of ECT values from the Mini-static tester measurement
are shown in Figure 4 with and without the first measurement of data included for each filter.
The deviations of ECT values without the first data included becomes smaller than those with the
first data included. For the mean ECT values of measurement, it is recommended to omit the first
measurement of data.

Figure 4
Deviations Due to Way of Dealing with Current Waveforms and Measured Data

As for the variation in deviations of ECT values of 75 field samples, 95% of transformers had the
deviation less than 30% and 85% of transformers had the deviation less than 20%.

2-104
Modeling and Prediction of Static Electrification

Reproducibility Assessment by Plural Facilities on the Same Sample Oils

For low charging oils like a new oil, the deviation factor among the facilities was large, however,
for the service-aged oils having high ECT values, the deviation factor is about 30%, which can
be said that it does not matter in practical use.

Table 1
Reproducibility of ECT Values

Sample Oil ECT Level Deviation


new oil 10 pC/ml level 40~60% (5 participants)
service-aged oil 100 pC/ml level 10% (3 participants)
service-aged oil 15000 pC/ml level 32% (5 participants)

Correlation between Mini-Static Tester and Conventional Methods

Correlation of ECT values of in-service oils from power transformers between by Mini-static
tester and by conventional methods are shown in Figure 5.

Figure 5
Correlation between Mini-Static Tester and Conventional Method

2-105
Modeling and Prediction of Static Electrification

Some of them have not so good correlation, but Mini-static tester has good correlations between
the paper pipe oil circulating method and base oil circulating method which are relatively well
equivalent to transformers. Degrees of correlation depend on the measuring methods and their
reason is assumed that in paper pipe oil circulating method, correction for charge relaxation is
made, in Mini-static test method and base oil circulating method, it is not necessary to consider
relaxation, but gravity oil flow method is not considered of relaxation and siphon method uses a
different material at the charge generation part.

Correction factors of ECT values to that of Mini-static tester are about 20 times at paper pipe oil
circulating method, about 10 times at base oil circulating method, ranging from 0.5 to 2.0 times
at gravity oil flow method and 8 times at siphon method.

Characteristics of Mini-Static Tester

Mini-static tester has a paper filter at the charge generation part, so it has a high sensibility. Its
ECT has a temperature dependence and often has an oil flow dependence.

Temperature Dependence

Mini-Static tester shows temperature dependence of ECT like a conventional method. So it needs
to set the standard temperature to measure ECT, for example, at 20 and measured data should be
corrected to the standard temperature. Average temperature dependence and correction factor to
the standard temperature are expressed by next equations.

ECT exp{-5.3 1000/(273+t)}


Figure 6
Temperature Dependence of ECT

2-106
Modeling and Prediction of Static Electrification

ECT(20) = ECT( t)x3.18 exp(-0.0589t)


Figure 7
Correction Factor

Oil Flow Rate Dependence

New oil having small ECT value does not show a strong flow rate dependence as shown in
Figure 8. But it is often observed that service-aged oils show strong negative oil flow rate
dependence as shown in Figure 9 and Sens et al. [8] has experienced this phenomenon.

Figure 8
Oil Flow Dependence of New Oil

2-107
Modeling and Prediction of Static Electrification

Figure 9
Oil Flow Dependence of Service-Aged Oil

It is assumed that this negative flow dependence is due to the structure of a paper filter. That is,
filter is not homogeneous and pores are distributed heterogeneously even if paper filters are used
of the same lot as shown in Figure 10. And as shown in Figure 11, when test oil is pressurized by
nitrogen gas, condition of paper filter changes with enlargement of pore size or increase of large
size pores and test oil shall probably flow through the large pores selectively. When oil flow rate
varies much in repeat measurements, the measurements of data shall be omitted and tried again
by replacing the filter.

0.1mm

a. new filter 1 b. new filter 2

Figure 10
Microscopic Pictures of Paper Filters

At first of the investigation, oil flow rate was set in the order of 1.4 ml/sec in reference to the
value noted in relative papers to Mini-static tester. But it sometimes appeared that gas was likely
to leak by measuring apparatuses and the measurement of data became unstable for the large oil
flow rate as shown in Figure 12. It is not so difficult to control the oil flow rate, it is
recommended to set the standard oil flow rate at the smaller value 1.2 ml/sec and to make the
measurement in the range of 1.1~1.3 ml/sec. The deviation of ECT values remains +-5% for the
region of above flow rates. Mini-static tester is not always applicable to the oil containing sludge
because of the possibility of stuffing the paper filter.

2-108
Modeling and Prediction of Static Electrification

0.1mm

a. picture before use b. picture after used 1 time c. sketch before use

d. sketch after used 1 time e. sketch after used 3 time f. sketch after used 5 time

Figure 11
Surface Conditions of Paper Filter before and after Use

Figure 12
Oil Flow Dependence of Streaming Current

2-109
Modeling and Prediction of Static Electrification

Verification of Oil Flow Dependence on ECT Measurement due to


Heterogeneous Pores

When a paper filter has small and large size pores mixed, oil is easy to flow through the large
pores because head loss of a hole due to the friction increase proportionally to the inverse of
square of hole diameter. Pores are enlarged by pressurizing and higher the pressure, larger the
pore size, oil flows faster through the filter. From above reason, it is assumed that oil flows
selectively through large pores and enlargement of pore size makes its tendency strong and ECT
of oil show the negative oil flow dependence. ECT of oil was measured using filters with
artificial needle holes and its results are shown in Table 2.

From the test results, it can be seen that ECT values decreased with the number of holes
increased or hole diameter enlarged for both oil flow rates, it leads to the selective oil flow
through the large holes. With No.2 filter, gas pressure is proportional to oil flow rate and in this
case, ECT values were almost constant independent on oil flow rate.

To the contrary, for No.1 filter and No.3 filter, ECT values with No.1 filter decreased at faster oil
flow rate and those with No.3 filter increased. In this case, gas pressure is not exactly
proportional to oil flow rate and conditions of filters might be changed by applying pressure
on oil.

Table 2
ECT Values with Needle Holes

No. Needle Holes ECT ECT Ratio*


size x number 1.0 ml/sec 1.4 ml/sec (Flow ratio/Pressure ratio)
1 0.1 mm x 10 548 pC/ml 485 pC/ml 0.89
(0.50 kgf/cm2) (0.75kgf/cm2) (0.93)
2 0.1 mm x 20 406 pC/ml 398 pC/ml 0.98
(0.35 kgf/cm2) (0.50kgf/cm2) (0.98)
3 0.3 mm x 10 196 pC/ml 230 pC/ml 1.17
(0.10 kgf/cm2) (0.20 kgf/cm2) (0.70)
4 no holes 367 pC/ml
(0.80 kgf/cm2

* ; ECT ratio = ECT(1.4 ml/sec)/ECT (1.0 ml/sec)

Microscopic pictures of filters before use and after use with needle holes of diameter 0.3 mm are
shown in Figure 13. The filament of pulps restored a little by using and hole size at flow rate
1.4 ml/sec became smaller compared to that of at oil flow rate 1.0 ml/sec. So it needed the gas
pressure twice to make the oil flow fast 1.4 times and ECT value increased by faster oil flow due
to the restoration of hole size.

As to No.1 filter, restoration of holes at oil flow rate 1.0 ml/sec was large (holes became smaller)
compared to that of at 1.4 ml/sec and ECT value became larger.

2-110
Modeling and Prediction of Static Electrification

From these results, it may be said that if gas pressure is exactly proportional to oil flow rate, ECT
value can be constant and if filter condition varies and pore size becomes large, streaming
current does not increase proportionally to oil flow rate and ECT show a negative oil flow
dependence.

a. before use b. after use (1.0 ml/sec) c. after use (1.4 ml/sec)

Figure 13
Microscopic Pictures of Filters before and after Use

For further reference, ECT value by #4 filter was smaller than that of #1 filter. For this reason,
interference of double layer of charges may be thought not because of the enlargement of hole
size but because of the very small oil flow path in paper filter from the fact that gas pressure with
no holes is certainly high.

Conclusion

ECT of oil by Mini-static tester often shows a negative oil flow dependence. It may be thought
that pore size distribution in paper filter is not homogeneous and the variation of the paper
condition such as pore size enlargement and increase of the number of large pores.
Homogeneous and stable paper filter is expected.

Mini-static test is recommended for the standard as ECT test method for transformer oils. It is
closely correlated between the conventional paper pipe oil circulating method, which has also
close correlation to transformers. Mini-static tester has a better sensitivity compared to the
conventional methods.

References

1. R. Tamura, Y. Miura, T. Watanabe, T. Ishii, N. Yamada and T. Nitta, Static Electrification


by Forced Oil Flow in Large Power Transformer, IEEE PAS, Vol. 99, No.1, 1980,
pp. 335-343.

2. H. Ookubo, M. Ikeda, M. Honda and T. Yanari, Charging Tendency Measurement of


Transformer Oil, IEEE PES Winter Meeting, A79 051-4, 1979.

2-111
Modeling and Prediction of Static Electrification

3. M. Higaki, Y. Tsutsumi, H. Ohtani and H. Tsukioka, Consideration on Measurement of


Static Charges in Insulating Oil and Influence of Oil Conductivity, Trans. IEE of Japan,
Vol. 104-A, No. 6, pp. 329-336, 1984.

4. Maintenance on Streaming Electrification, Technical Report of Electric Technology


Research Association in Japan, Vol. 54, No. 5, pp. 69-155, 1999.

5. A. J. Morin, M. Zahn and J. R. Melcher Fluid Electrification Measurement of Transformer


Pressboard/Oil Insulation in a Couette Charger, IEEE EI Vol. 26 No. 5,
pp. 870-900 1991.

6. M. J. Lee and J. K. Nelson Flow Induced Electrification and Partial Discharge


Measurements in Transformer Duct Structures, IEEE EI Vol. 26 No. 4,
pp. 739-748 1991.

7. T. V. Oommen and S. R. Lindgren Streaming Electrification Study of Transformer


Insulation System Using a Paper Tube Model, IEEE PD Vol. 5 No. 2
pp. 972, 1990.

8. J. Kedzia Investigation of Transformer Oil Electrification in a Spinning Disk System,


IEEE EI Vol. 24 No. 1 pp. 59-65, 1989.

9. T. V. Oommen and E. M. Petrieet Electrostatic Charging Tendency of Transformer Oils


IEEE PAS Vol. 103 No. 7 pp. 1923-1931, 1984.

10. Static Electrification in Power Transformers, Preliminary Draft 08/97 CIGRE WG. 15.01
August 20 1997.

11. M. A. Sens, J. B. Fernandez and R. Ficara, Benzotriazol Effect in Electrostatic Kraft


Paper/Oil flow Model Experiments, Workshop Proceedings: Static Electrification in Power
Transformers, EPRI TR-105019, May 1995.

2-112
3
OPERATING EXPERIENCE OF TRANSFORMERS

3-1
Operating Experience of Transformers

3.1 A Static Electrification Transformer Survey


J. Keith Nelson
Department of Electric Power Engineering
Rensselaer Polytechnic Institute

Abstract
It has become evident that many of the characteristics involved with streaming electrification
cannot be replicated in laboratory tests either because the phenomena are not amenable to
scaling, or because actual conditions cannot be replicated. As a result an attempt has been made
to garner as much information as possible from known transformer electrification incidents in the
field. This has been done on a worldwide basis through literature searches, a survey
questionnaire and correspondence with individuals who have experienced problems. As a result
of this survey common trends are identified and patterns of failure have been established. The
results are provided in the form of probability density plots from which some conclusions are
drawn and countermeasures identified.

Rationale
In the 1970s the problem of streaming electrification (SE) emerged in Japan as a dangerous
failure mechanism for power transformers [1]. The realization that the forced circulation of oil in
large units could generate charges in the internal structure which were not anticipated by the
transformer designer has fuelled research in the last decade to try to understand and mitigate the
problem. Much has been learned and instrumentation has been developed. However, it is now
very clear that the phenomenon does not scale easily, and so it is very difficult indeed to interpret
laboratory data in the context of an operating transformer even if steps are taken to try to
replicate the actual transformer structure. A simple example of this may be understood from the
fact that history teaches us that only large units seem to be vulnerable despite the fact that
smaller transformers may have very similar construction and operating parameters. This is
further discussed later.

The inability to predict SE problems from laboratory results, and the extreme sensitivity of the
phenomenon to constructional details make it imperative that as much as possible is learned from
field failures known to have resulted from static electrification problems. Included in this
category are the transformers, which are known to have had electrification incidents (usually
static discharges) even though a complete failure did not ensue. In some cases, the incidents are
extremely well documented, but, regrettably, there is a paucity of reliable data on many of the
events. It is the purpose of this survey to try to piece together, from a plurality of sources, as
much as possible of the relevant circumstances behind each incident. It must also be recognized
that some transformer owners and most manufactures are, understandably, reticent to document
such events. This has compounded the task of assembling a comprehensive database.

The provision of a common database on electrification failures does permit trends and common
features to be identified (in later sections of this paper) which it is hoped will allow users to be
alerted of practices and operational areas which may create a greater risk for this type of failure
mechanism.

3-3
Operating Experience of Transformers

Survey Methodology

Although there are thought to be fewer than 50 failures worldwide attributable with certainty to
streaming electrification, it is likely that there are a large number of other failures for which a
cause is not known, but, in fact, failed due to this phenomenon. This is especially so since the
incidence of the mechanism is not known by many field engineers. However, this survey only
includes those units for which SE events have been documented with reasonable certainty. In
most cases this can only be accomplished by an expensive teardown analysis which many
transformer operators are unwilling to undertake. This severely limits the inventory of units that
can be included in the database, but it is necessary in order to separate fact from fiction.

Sources of Existing Information

Prior to surveying the industry with a questionnaire, a literature search was conducted to provide
both a basis for developing the questionnaire and a starting point for the database. The following
sources were used:
A computer-based literature search of the worldwide public domain literature [2-5]. Although
there is a significant body of literature on the phenomenon and laboratory tests, there is very
little archival material relating to field experience.
EPRI Transformer Electrification Workshop reports and contractors reports where available
[6-8].
Relevant Doble client reports [9-17].
Edison Electric Institute Reports [18].
CIGRE Task Force 12/15.13 materials [19,20].
Correspondence with individuals known to have relevant experience.

Survey Questionnaire

For the purposes of garnering data from the transformer community, a survey was developed to
try to identify cases, which were not identified in the literature search, and also to add detail to
those incidents for which there had been published reference. The questionnaire was developed
as a compromise between obtaining comprehensive data and creating a document which would
be seen as too time consuming and difficult to merit returning. The strategy here was to be able
to identify units and information resources and then, if necessary, engage in follow up
correspondence to obtain further or missing information. However, references to specific
Companies have been removed from this paper.

3-4
Operating Experience of Transformers

Database Refinement
Having established the database containing 32 entries, there were several instances where
conflicts were clearly evident and many cases where data was very sparse. The conflicts usually
arose when data returned on a questionnaire contradicted that given in the literature. It was thus
necessary to refine the database in the following way:
Conflicts were resolved by correspondence with the respondent supplying the information.
Relevant sections of the database were sent to representatives of the Companies involved
with an invitation to amend and append additional material where known.
The complete database was sent to Mr. Harold Moore in an attempt to fill in other gaps. This
was done since Mr. Moore, as a consultant to EPRI, has been involved with a number of the
incidents involving SE.

Derived Measures
Data was extracted in a number of different ways in order to try to characterize both the
equipment and the operating conditions, which form the basis for the incidence of streaming
electrification. Since the data is so incomplete, extracted data is normalized only with respect to
the data sheets which include the parameter being discussed (and not in relation to the total
transformer inventory). Probability density plots are also sometimes artificially truncated where
necessary so that the reader does not make the assumption that there are high levels of incidence
in the tails of the distribution. Furthermore, it must be recognized that, in several instances,
sample sizes are so small that statistical inferences are dubious. Nevertheless, they are provided
so that the order of magnitude may be gauged.
In a later section, an attempt has also been made to encapsulate common aspects, which are not
so amenable to quantification, together with characteristics thought to be significant even if not
commonly seen.

Equipment Characteristics
Transformer static problems, although first uncovered in Japan, have been ascribed
predominantly to the United States and to units of shell-form construction. However, the survey
does indicate that, although the majority of cases have involved shell construction, the
phenomenon is also present in core-form units. Indeed, 23% of the units surveyed were
core-form. It has already been noted that SE is prevalent only in large transformer units. Figure 1
underscores this attribute by providing a distribution of the MVA levels of units documented as
having been involved with an electrification incident. In the cases where a unit has a multiple
rating (e.g. OA/FOA/FOA), the highest rating, with all pumps and fans, is used. The figure treats
core- and shell-form units separately, although the differences are not considered significant.
Most of the units involved are Generator Step-up Units (GSU) or large autotransformers. What
is, however, of significance is the inclusion of a unit having a capacity of only 28 MVA. This
very low value is anomalous. It is also very different in that it is a mobile unit with Nomex
insulation and a very high oil moisture content (21 ppm). Since no teardown analysis was
performed, one cannot be absolutely certain that the discharges detected were of static
electrification origin. However, the cessation of activity on changing the pumps certainly is
grounds for suspecting the phenomenon.
3-5
Operating Experience of Transformers

0.8
0.7
0.6
P rob . 0.5
0.4
D en sity 0.3
0.2
0.1
0
50 150 250 350 450 550 650 750 850
R atin g (M V A )

S h ell-fo rm C ore-form

Figure 1
Distribution of Transformer MVA Ratings for Shell- and Core-form Units Involved with
Electrification Incidents

0.7
0.6
0.5
Prob. 0.4
Density 0.3
0.2
0.1
0
100 250 350 450 550 650
Voltage Rating (kV)

Shell-form Core-form

Figure 2
Distribution of Voltage Ratings for Shell- and Core-form Transformers Involved in
Electrification Incidents

Similar information may be obtained from Figure 2 which depicts the distribution of voltage
ratings for the units that have demonstrated SE problems. As previously, a distinction is made
between the core- and shell-form constructions, but the constructions are really indistinguishable
in this regard having a typical voltage rating between about 400 and 600 kV. From Figures 1 and 2
it is very clear that problems are confined to EHV units over 200 MVA.

3-6
Operating Experience of Transformers

Oil Data
Since the streaming electrification problem is intimately dependent on the electrochemistry of
the oil and its impurities, the nature of the coolant is clearly of paramount importance. However,
the survey has not shown any particular oil type to be a common factor in the incidents on
record, although it is clear that one oil type (now no longer on the market) was involved with
several of the early US failures. The issue is also complicated by different practices and additives
in use worldwide. For example, oils have been subjected to ion exchange processes (by Fuller's
earth treatment) by some users and it is well documented that, in Japan, insulating oils have
1,2,3-Benzotriazol (BTA) added to reduce the incidence of electrification [21]. It will also be
appreciated that, in cases where there was major damage (explosion and fire fighting operations),
the results of testing of oil samples after the event lack credibility.

Notwithstanding the difficulty in identifying the nature of the oil, some information can be
obtained from oil analysis for parameters known to affect static electrification. Figure 3 shows
the distribution of oil Electrostatic Charging Tendency (ECT) for the incidence where it was
known. It is seen that there is a spread of values from a low of 125 C/m2 to a high of
1000 C/m2 with one outlying value as high as 3000 C/m2. The ECT test is not standardized
and is notoriously variable in the results obtained in different laboratories [22] for the same oil
sample. Consequently, the wide scatter in the results is perhaps to be expected. However, it is
also clear from the lower tail of the distribution that a low ECT oil is no guarantee of protection.

0.25

0.2

Prob. 0.15
Density 0.1

0.05

0
100

500

900

1300

1700

2100

ECT (uC/m3)

Figure 3
Distribution of Oil Electrostatic Charging Tendency (ECT). All Types

Laboratory studies have also made it clear that the moisture content of an oil (and the dynamic
equilibrium that exists between water in the oil and cellulose) is a factor in both the generation
and relaxation of charge [23,24]. At very low moisture levels, the water can be a source of ions
to augment the surface processes, while at very high moisture levels the leakage of charge is
affected. This occurs both through the oil time constant and, more importantly, through enhanced
pressboard conductivity. If the one outlying data point of 21 ppm (see also comment in Section
on Equipment Characteristics) is censored, all the units exhibited oil moisture levels between 5
and 10% which is typical of the transformer population at large. The data is shown in Figure 4.

3-7
Operating Experience of Transformers

0.4
0.35
0.3
0.25
Prob.
0.2
Density 0.15
0.1
0.05
0
2 7 12 17 22

O il M oisture (ppm)

Figure 4
Distribution of Oil Moisture for Transformers Involved with Electrification Incidents. All
Types

Operational Parameters

From the user perspective, the greatest interest is in the operational parameters over which there
may be some control. Since the phenomena are electrokinetic in origin, there should be no
surprise in the finding, depicted in Table 1 (for both shell- and core-form construction
combined), that the number of pumps in operation has a very strong influence on electrification
incidence. It can be concluded that 90% of the incidents occurred when all the pumps were
running. However, it is still of significance that there were some major problems with units
operating with only half the cooling deployed.

The number of pumps in operation is not an independent variable since it is conditioned by the
electrical load and the ambient temperature; both of which will affect the oil temperature on
which automatic pump control relies. Some data on these variables is provided in Figures 5-7.

3-8
Operating Experience of Transformers

Table 1
Incidence of Streaming Electrification Occurrence as a Function of the Proportion of
Pumps in Operation

Pumps in operation Percentage of incidents


(as % of number available)
50 0.1
100 0.73
>100* 0.18
* Values >100% due to transformers being cooled more aggressively than specified by the manufacturer.

Figures 5 and 6 suggest that the ambient temperature and unit loading per se are not issues that
are strong factors in determining the vulnerability to this mechanism of failure.

0.4
Prob.
0.3
Density
0.2
0.1
0
2

12

22

Amb. Temp. (C) 32


Figure 5
Distribution of Ambient Temperatures at the Time of Electrification Incidents. All Types

0.5
0.4
Prob. 0.3
Density 0.2
0.1
0
10 30 50 70 90
Load (%)
Figure 6
Distribution of Unit loading at the Time of Incident. All Types

3-9
Operating Experience of Transformers

0.3
0.25
0.2
Prob. 0.15
Density 0.1
0.05
0
15 25 35 45 55 65

Oil Temp. (C)


Figure 7
Distribution of Oil Temperature at the Time of Incident. All Types

Figure 6 suggests a bimodal distribution in the loading which is probably a manifestation that
most of the large units are either GSUs supplying base load or transmission substation units
which may often be lightly loaded. Of greater interest is Figure 7, which does show that units
appear more vulnerable at intermediate temperatures in the range between 30C and 50C. This
is significant because this is the range of temperatures in which enhanced electrification activity
has been seen in tests [4].

Other Factors

Study of the sources of information relating to transformer SE failures has also yielded other
information, which, although not amenable to the same quantification, is nevertheless of
significance. These factors are tabulated below in decreasing order of importance together with
the number of instances in which they occur {shown in parenthesis}.
Shell-form construction: Discharge tracking over washers in the H-L space, usually at the
45 point. {13}
The unit was reprocessed (or initially energized) within 1 year of the incident. {13}
Shell-form construction: Failure (or evidence of long discharge) at the top of the winding
usually involving leads or cross-overs. {11}
Pumps were manually overridden at the time of failure or the detection of statically induced
partial discharges. {9}
Incident occurred within 1 week (usually within a few hours) of energization. {8}
Unusual gas generation was detected. {8}
Oil type was implicated. {3}
Failure follows an outage. {3}
Dependence on which pumps were operating. {3}
Textured pressboard was used in the construction. {3}
Failure following a sudden reduction in temperature. {2}

3-10
Operating Experience of Transformers

Appraisal

This phenomena has previously been identified as limited to shell-form construction, but the
finding that about a quarter of the failures world-wide have been of core-form units undermines
that assumption. The sentiment in Europe is that this is a US and Japanese problem due to the
large inventory of large shell-form units. The fact is that core-form units have also suffered from
SE phenomena, but there is a paucity of documentation on these failures in order to establish a
pattern or a mechanism of failure. Activity in the US by the Electric Power Research Institute
has tended to heighten awareness there, and it is very probable indeed that SE failures are more
widespread than this survey would indicate. Indeed, although not included in this paper, several
unexplained failures are known which have many of the same outward signs as those tabulated
here.

The finding that the SE problem is confined to EHV transformers of 200 MVA rating and above
can be understood from considering the competing processes which culminate in SE stress
accumulation. It is likely that charge generation also occurs in high flow regions of smaller units.
However, only in large units is the path to ground (or to the windings which are a static
ground) long enough to provide an impediment to this beneficial leakage. The higher voltage
units are also built, for economic reasons, with a lower BIL margin. In that sense, it is likely that
it is voltage level rather than the MVA (i.e. Figure 2 rather than Figure 1) that is the controlling
factor.

The data in Figure 6 suggests that SE failures can occur over a wide range of loading (from 0 to
100%). The lack of sensitivity to load suggests an interplay between the load and the ambient
temperature (i.e. the oil temperature is reduced by a low ambient for the same load losses).
However, study of the data for individual units (as opposed to the consideration of the ensemble)
showed no consistent pattern with the limited data available.

This survey has also highlighted the large number of units that have presented problems:
i. soon after reprocessing or field redrying.
ii. within a few hours of energization without obvious system disturbances.
Laboratory studies have shown the sensitivity to the condition of the oil/cellulose structure, and
also demonstrated charge accumulation time constants, which are compatible with this field
experience. On this basis, there can be little doubt that moisture migration in the structure is
having a significant effect. Although the survey only provided evidence of two cases, the fact
that some units have failed following a rapid fall in temperature and/or load is thought to be
pertinent. Rapid cooling will create changes in the local moisture equilibrium that will affect the
interface conditions contributing to the phenomenon.

Speculation on the Mechanism of Failure

The number of shell-form failures which involve:


a. Flashover at the top of the windings involving the lead structure, and
b. Discharge damage in the lower part of the H-L duct space (usually on exposed washers at the
45 point)

3-11
Operating Experience of Transformers

is far too great to be attributed to coincidence. A pattern of failure has been established for these
designs, and there had been considerable speculation about the mechanism. The most likely
scenario would appear to involve the accumulation of static charges on the washers in the
H-L space due to the impinging oil flow. Subsequent electrostatic induced partial discharge
activity in this region would then be expected to generate gas. The extensive nature of the
discharge damage seen in some teardown analyses would imply the generation of gases at a
higher rate than they could be reabsorbed in the oil. It is then speculated that buoyancy forces
acting on the bubbles generated could create large gas cavities in the upper structure which could
precipitate failure. The fact that some units have been found to show signs of discharge activity
only at the bottom, without a failure over a long oil path in the upper plenum adds credence to
the theory postulated that failure in the upper oil space is precipitated by static electrification in
the entrance regions.

Data for the core-form units is too scarce to be able to provide even a tentative mechanism.

Countermeasures

The finding for the tests conducted with the unit at Comanche Peak (Texas Utilities), and
inferred for two other units, that SE activity was dependent on which pumps were running
(as well as just how many) supports the view that the flow patterns in the lower entrance regions
are critical to this phenomenon. When this finding is considered, together with the incidence of
discharge activity in those areas where turbulence is thought to be high, much points to the need
for design solutions aimed at this aspect. The provision of baffles or other flow modification
means in the lower plenum might provide sufficient margin to avoid electrification damage
without much impact on the overall cooling of a unit.

It is clear that a large proportion of the failures is attributable to transformer pump control being
manually overridden to provide maximum cooling. Although not surprising, this provides ample
justification for the advice now being provided by EPRI, and others, that this practice is
dangerous. Indeed, comments provided by several of the users surveyed, attributed the lack of
SE failures to adherence to that advice. Similarly, some Companies who had experienced
problems made the comment that such steps taken in the wake of their experiences had prevented
a recurrence. Notwithstanding that, the problem still persists and the inventory of units in the
field thought to be vulnerable is substantial. Field monitoring of transformers is possible either
with conventional techniques (Dissolved gas analysis, acoustic emission, and PD analysis) or
using instruments specially designed to monitor charge generation [25-28]. Such techniques are
very time consuming if implemented on all large units, and the expense of an on-line installation
is difficult to justify for a risk that is seen as small. However, the finding that pump control is of
paramount importance does suggest that more sophisticated, but inexpensive, control may be
used to reduce the risks substantially. Such a scheme of control (to replace the usual simple
pump energization in response to oil temperature) has been researched [29,30], but, as yet, not
commercialized. The challenge is to provide a means to insure adequate cooling without
operating a transformer in regimes known to accentuate risk. For example, the risk peak shown
in Figure 7 needs to be avoided by reducing flow until a highly loaded unit is through the
vulnerable regime. High temperature operation does reduce the risk due to the inherent enhanced
conductivity of the constructional materials. However, the implications for transformer life
dictate that a strategy of undercooling alone is unacceptable.

3-12
Operating Experience of Transformers

Conclusions

This survey has attempted to assemble a comprehensive database of the circumstances relating to
all the known and confirmed cases of transformer static electrification failure. On the basis of
this data, the distribution of the primary parameters known to affect the phenomenon have been
extracted. Many of the indications are not new, but have been quantified from the data in a way
that demonstrates their impact.

On the basis of the survey, an appraisal is provided which includes some speculation about the
mechanisms involved (at least for shell-form units) and suggests fruitful ways to pursue
countermeasures.

Acknowledgements

Thanks are due to all those who returned the survey questionnaires and to Mr. Harold Moore
who provided data from the benefit of his experience with teardown studies. The survey was
conducted under the auspices of EPRI and managed by Mr. S.R. Lindgren.

References

1. Takagi T. et al., Reliability improvement of 500 kV large capacity power transformer


CIGRE Report 12-02, 1978 Session.

2. Crofts D.W., Static electrification in power transformers, in Electrical Insulating Oils ed.
H. G. Erdman, ASTM STP 998, 1988, pp. 136-51.

3. Crofts D. W., The static electrification phenomena in power transformers, Ann Rep. Conf.
Elect Ins. & Diel. Phen., IEEE, 1986, pp. 222-36.

4. Tamura R., Static electrification by forced oil flow in large power transformers,
Trans IEEE, Vol PAS-99, 1980, pp. 335-43.

5. Shimizu T., Electrostatics in power transformers, Trans IEEE, Vol PAS-98, 1979.

6. EPRI, Workshop Proceedings: Static Electrification in Power Transformers,


Report EL-6918, 1990.

7. EPRI, Proceedings: Static Electrification in Power Transformers, Report TR-102480,


1993.

8. EPRI, Proceedings: Static Electrification in Power Transformers, Report TR-105019,


1995.

9. Crofts D. W., Methods of controlling static electrification in large power transformers,


Doble Client Conference, 1986.

3-13
Operating Experience of Transformers

10. Brunson P. W. Jr., Investigation of the failure of a 500-230 kV, 480 MVA Westinghouse
Shell-type power transformer, Doble Client Conference, 1986, Sec 6-1101.

11. Brunson P. W. Jr., The second failure of a 400 MVA, 500-230 kV, Shell-type
autotransformer at South Bessemer transmission substation, Doble Client Conference, 1987,
Sec 6-1501.

12. Lindgren S. R., Progress in the control of static electrification in power transformers,
Doble Client Conference, 1987, Sec 6-1601.

13. Lindgren S. R., Update on progress in the control of static electrification in power
transformers, Doble Client Conference, 1988, Sec 6-12.1.

14. Lindgren S. R., Update on progress in the control of static electrification in power
transformers, Doble Client Conference, 1989, Sec 6-5.1.

15. Lindgren S. R., Update on progress in the control of static electrification in power
transformers, Doble Client Conference, 1990, Sec 6-12.1.

16. Fallon D. J., Update on Public Service Electric and Gas Company experience with static
electrification in power transformers, Doble Client Conference, 1990, Sec 6-13.1.

17. Lindgren S. R., Update on progress in the control of static electrification in power
transformers, Doble Client Conference, 1991, Sec 6-12.1.

18. Crofts D. W., Static Electrification in power transformers, Minutes of the 163rd meeting of
the Edison Electric Institute Electrical System & Equipment Committee, 1987.

19. Lindgren S. R., Static electrification in power transformers, CIGRE Conveners report to
Joint Working Group 12/15.13.

20. Lindgren S. R., Washabaugh A. P., von Guggenberg P., Zahn M., Brubaker M. and
Nelson J. K., Temperature & moisture transient effects on flow electrification in
power transformers, Paper 15/12-02, CIGRE, Paris, France, September 5th 1992.

21. Ieda M. et al., Suppression of static electrification of insulating oil for large power
transformers, Trans IEEE, Vol EI-23, 1988, pp. 153-57.

22. Palmer J. A. and Nelson J. K., Method for repetitive measurement of the electrostatic
charging tendency of liquid dielectrics, IEEE Trans, Vol DEI-3, 1996, pp. 70-74.

23. Morin A. J., Zahn M. and Melcher J. R., Fluid electrification measurements of transformer
pressboard/oil insulation in a couette charger, Trans IEEE, Vol EI-26, 1991 pp. 870-901.

24. Palmer J. and Nelson J. K., Streaming electrification dynamics in oil/cellulose systems,
Ann. Rep. Conf. Elect. Ins. and Diel. Phen., IEEE, Arlington, TX, 1994, pp. 895-900.

25. Melcher J. R., Morin A. J. and Zahn M., An instrument for the detection of electrical charge
entrained in a fluid, US Patent #4873489, 1989.

3-14
Operating Experience of Transformers

26. Nelson J. K. and Lee M. J., Tandem chamber charge density monitor Trans. IEEE,
Vol EI-25, 1990, pp. 399-404.

27. Nelson J. K. Bypass tandem chamber charge density monitor, US Patent #4947468,
August 7th 1990.

28. Gish W. B. and Nilsson S. in Ref. [8].

29. Nelson J. K. and Palmer J. A., System and method for mitigation of streaming
electrification in power transformers by intelligent cooling system control, US Patent
#5838881, November 17th, 1998.

30. Palmer J. A. and Nelson J. K., Intelligent control of large power transformer cooling
pumps, IEE Proceedings on Generation, Trans. and Distrib., Vol 143, #5, 1996, pp. 474-78.

3-15
Operating Experience of Transformers

3.2 A Static Electrification Failure at the Navajo Generating Station

Thomas G. Lundquist, P.E.


Salt River Project

Abstract

On March 13, 1997, a single-phase shell form, 308 MVA, 24.7/303.1 kV generator step-up
transformer failed at the Navajo Generating Station. The failure was judged the result of static
electrification. A presentation of pre-fault information regarding the transformer operation and
incidents that could have impacted the failure are presented. Photographs are used to illustrate
the investigation during a detailed dismantling. These photographs illustrate the findings inside
the transformer supporting the conclusion that static electrification was the primary cause of the
failure. Findings and recommendations of the manufacturer and of consultants are summarized.
A risk assessment and cost analysis of the possible corrective actions is presented. Also
presented is the plan that will be implemented on nine (9) identical transformers.

Pre-fault Information

The location of the faulted transformer was at the Navajo Generating Station (NGS) located near
Page, Arizona. The Navajo Generating Station is comprised of three (3) generators with a total
capacity of approximately 2250 MW. Each unit is connected to the system through identical
banks of three (3) generator step-up transformers (GSU). One spare transformer is maintained at
the station. The transformers are single phase, 308 MVA, 65C rise, 24.7/303.1 kV, shell form
construction. The transformers BIL are HV - 1425 kV, HO 350 kV, LV - 150 kV.
Westinghouse Electric Corporation, Muncie Indiana, manufactured the transformers. Surge
arresters at or near the terminals of each bushing protect each transformer. The transformers
operated with two (2) stages of cooling one stage in operation at all times with the other stage
switched on at a winding hot spot temperature of approximately 75C. Each cooling stage
operated with two (2) coolers and pumps. The transformers are nitrogen gas blanketed at a
regulated nominal pressure of three (3) psi.

The GSUs high voltage windings normally operate with a subsynchronous resonance (SSR) filter
connected in series with the neutral end of the winding before the transformer bank wye is
formed and grounded. At rated current the neutral of the transformer is operating at
approximately 13 kV with respect to ground. Between the transformer and the SSR filter
connection for each phase a surge arrester is applied for protection of the transformer. The
primary windings are connected in delta to the generator terminals through iso-phase bus.

During the failed transformers operating life, one unusual event occurred when the transformer
was operated at rated MVA with the cooling disabled for a period of 27 minutes. An evaluation
of this incident indicated that the transformers have sufficient thermal capacity to operate for the
27 minute interval without serious damage to the insulation system.

3-16
Operating Experience of Transformers

In December 1992 an investigation into oil color changes and elevated oil power factor was
conducted. The oil was previously replaced in 1985 because of reported color change. In 1992
the oil color was > 2.5 with an oil power factor at 0.064 at 25C and 2.7% at 100C. Tests for
metals and other contaminants revealed only normal oxidation products and that the oil could be
restored to high quality by clay filtration.

It should be noted that a sister transformer serial 7002257 operating on unit 3 phase C
exhibited a power factor in the range of 8% to 14% at 100C and the color > 4.5 to 6 for tests
conducted from 1991 to 1994. The oil was filtered and the inhibitor increased from 0.018% to
0.25%. As of September 1997 the oil color remains 5.5 and the power factor is 2.5% at 100C.

The dissolved gas in the transformer 7002104 for 1990, 1992, 1993, 1994, and 1996 are shown
in Table I using six (6) sets of data and stated in parts per million (PPM). In several instances
tests were repeated at least twice to demonstrate errors had occurred in laboratory tests. These
were reported as high concentrations of combustible gas but subsequently found to be
acceptable. Salt River Project now qualifies laboratories and has eliminated laboratories with
unacceptable statistical repeatability. This has minimized the number of retest and provides more
reliable information for analysis and trending studies.

Table I

Serial Number 7002104 Dissolved Gas from May 1990 to February 1996

GAS AVERAGE MAXIMUM Date of Maximum

Hydrogen 43 231 1/6/93

Methane 4 9 1/6/93

Ethane 2 4 2/8/96

Ethylene 3 10 2/8/96

Acetylene 0 0

Carbon Monoxide 17 33 8/12/93

Carbon Dioxide 3892 5900 1/6/93

Nitrogen 71740 75151 Avg. is for 5 tests

Oxygen 5109 8690 1/6/93

Note: In 1993 eight (8) additional tests were conducted during the year and the Hydrogen level decreased to 4 PPM by
December.

3-17
Operating Experience of Transformers

The transformer oil quality is represented in Table II with values provided for 1990, 1991, 1992,
1993, and 1994.

Table II

Serial Number 71002104 Oil Quality Information

Perimeter Average Worst Test Date of Worst Test

Acid Number .014 .023 4/29/94

Interfacial Tension 36 32 12/9/92

D-1816 in kV 40 mil 37 32 12/9/92 - 3 tests

% Power Factor 25C 0.062 0.083 12/9/92

% PF 100C 2.3 2.7 12/9/92

Water in PPM 12 15 max - 9 low 4 tests

Inhibitor in % 0.147 0.104 4/29/94

The Failure

On March 13, 1997 the GSU serial number 7002104 on generating unit #2 C phase failed
violently. The tank was deformed but had not ruptured. There is little doubt the nitrogen blanket
in the headspace of the transformer prevented a tank rupture and subsequent oil spill. The high
voltage bushing porcelain broke allowing it to drop into the tank onto the high voltage lead
connection. The fault involved an arc from the high voltage bushing terminal to the low voltage
winding and ground (tank wall). The fault recorders have 1/8 cycle resolution and with this
resolution there is no distinction between the time the fault involved ground and the low voltage
winding or stated another way the arcs appeared simultaneous. The system 525 kV tie between
the transformer and the substation was interrupted by the substation circuit breakers three (3)
cycles after fault inception. The arcing continued for approximately 59 cycles until the generator
bus voltage dropped to a level where the arc extinguished. The generator bus surge arresters
failed when the HV failed to the LV winding. These arresters saved the generator from damage.
At the time of the failure generating unit #2 was operating loaded at 450 MVA. The loading on
this plant for the previous three (3) months is shown in Figure 1 Further detail showing the last
7 days of the plant load are shown in Figure 2.

3-18
Operating Experience of Transformers

* The load is three (3) phase MVA.


Figure 1

* The load is three (3) phase MVA.


Figure 2

3-19
Operating Experience of Transformers

Dismantling Investigation

The attached photographs illustrate the findings during the investigation. In photograph
number 1 (photographs are attached to the report) a typical installation is shown of the three (3)
single phase transformers connected to the primary isophase bus and the secondary 525 kV
overhead bus. The connection to the SSR filter yard is through the lower bus work from the
HO bushing. In this photograph C phase transformer is the closest.

In early April of 1997 the process was started to evaluate repair versus replacement of the
transformer. The decision was made to replace the transformer after evaluations showed the
replacement costs were lower than the repair cost. A controlled dismantling was planned for the
transformer and included an expert in the industry to help direct dismantling and evaluate the
probable cause of the failure. Evaluation of the data and a preliminary inspection inside the top
of the transformer had led to the conclusion the probable cause was a static electrification (SE)
failure. The bushing failure as the root cause of the transformer failure was eliminated because of
the fault recorders and lack of evidence on the bushing that any fault had occurred where the
porcelain had fractured. The bushing failure was judged the result of the shock of the fault that
caused a sudden displacement of the tank top fracturing the porcelain and allowing the bushing
to drop into the transformer. As shown in photograph number 2 the 525 kV bushing shield had
15 separate location where arcing had burned through the insulation. Several are holes that
ranged in size from 10 mm to 25 mm in diameter. The expert reported that this type of shield
burning with no other immediate arc path was indicative of other failures attributed to SE.

Figure 3 is a top view of the transformer with the top removed. The general area of the burn
marks on the shields is shown as black dots. The path of the arc through the barrier board is
shown in photograph number 6 and is illustrated in Figure 3. The cylindrical barriers around the
shields were partially destroyed in the explosion and from the bushing collapse into the tank.
Sufficient amounts were recovered to determine that partial discharge had occurred at the bottom
of the shields and that the upper portions were essentially partial discharge free. The conclusions
regarding the failure are that the failure originated near the lower end of the HV bushing
including the shield and cylinder around the shield. The failure progressed through the bolt at the
bottom of the cylinder that attached the cylinder to the barrier board shown in photograph 6. The
failure went through the barrier board and over the surface of the barrier to the clamp supporting
the LV lead from coil number 38 and 35 that failed. Another path through the barrier went back
toward the corner of the tank.

3-20
Operating Experience of Transformers

Figure 3

Coil Packet Removal

Dismantling was conducted at NGS to save cost of transport to a repair facility. After the top hat
was removed, the coil and coil packet were removed from the tank and set on timbers for
dismantling. Photographs 3 and 4 show two of the steps of this procedure. A modified T beam
was used to lift the coils and two cranes were used to lay the coils horizontal. An 80-ton
hydraulic crane was used to lift the coils from the tank. The crane system reported a breakaway
force of 73,300 pounds to move the packet from the base. The dead weight of the insulation and
copper packet was 69,900 pounds. Photograph 5 shows the extent of damage to the top of the
coil packet, barrier boards, and the DLTC. The damage in the photograph was the result of the
explosion, not because of the dismantling process. The explosion had destroyed many of the
barriers around the HV bushing base.

Static Electrification Evidence

Because this failure involved multiple points in the washers and coils, only the observations
judged contributory to the failure will be reported in this paper. The coil arrangement and
numbers are shown in Figure 4 as viewed from the top of the transformer. The LV coils are made

3-21
Operating Experience of Transformers

up of four windings in parallel. The packet as viewed in photograph number 4 has coil number
38 on the top which was the first coil removed. The coils and barriers were removed one at a
time with detailed examination made of each as removed. LV coil 38 had three (3) failures on the
corner that was adjacent to arc marking on the tank wall near the top of the packet. LV coil 35
had multiple failure points. The barriers are shown in Figure 5 illustrating the LV to HV barriers
(washers) between coils 35 and 34. Coil 34 is the 525 kV lead connection to the HV bushing.
Washers 16, referred to as the full-length washers had evidence of the failure where burning
occurred during the arc. Evidence of SE was found in the remaining washers 7 15 except for
number 12. Examples of this evidence are shown on photographs number 8, 9, and 10. These
so-called wormholes were found starting at the bottom of the barrier at approximately 45 degrees
toward the corner. Many radiated up as might be observed in photograph number 7. These are a
unique type of discharge path that is characteristic of SE.

Coil Failure at the Neutral

The remaining coils were dismantled with no evidence of SE or other distress until coil 7 was
reached. A coil to coil failure had occurred between coil 7 and 6 and coil 5. Coils 6 and 7 were
connected at the top and as such were at the same potential where the failure occurred. The
failure point on coil 5 was essentially the next to the neutral connection. The cause of this failure
was attributed to improperly placed insulation pieces when the transformer was manufactured.
As previously explained the neutral is not solidly grounded and is subject to transient. This
failure could have existed prior to the loss of the transformer but it probably resulted from the
transients experienced during the failure of the HV winding. An investigation into the possibility
of resonance as the cause was judged impractical since a detailed model of the transformer
would be required and ABBs models were not adequate for such an investigation.

Low Voltage Conductor Heating

During the dismantling of coils 15 to 22, evidence of heating in the form of discolored insulation
was found on the conductors. The evidence indicated this heating was probably the result of
unbalanced flux causing the current to add to the load current in part of the conductors and
subtract from the load current in others. See photograph number 7 that illustrates this heating
aspect. As the insulation was removed from the individual conductors the discoloration appeared
less severe.

This apparent change in discoloration was never explained. One theory is that the oil on the outer
portion of the conductor may have exposure to more oxygen and this results in a higher degree of
oil oxidation resulting in much darker discoloration.

Life Evaluation

Degree of Polymerization Tests

A set of 17 tests following ASTM method D4243 evaluated the degree of polymerization (DP) of
the insulating paper from locations on the copper of the conductors suspected of having the
highest operating temperature. End of life is usually considered a DP of 200 or less for such

3-22
Operating Experience of Transformers

paper, which starts in the transformer with a DP of 1200. The worst DP test results were from
paper removed directly from the copper from the crossover connection between coils 5 and 6.
This DP result indicates that only 30% of the insulation life remains at that location. Paper
removed from the conductor where the most discoloration had taken place revealed that 44% of
the insulation life is remaining. These tests allow for the conclusion that the coils in the other
transformers have not suffered any appreciable harm and that the remaining operating life based
on DP tests is probably close to transformers that have operated for approximately 25 years.

Figure 4

Figure 5

3-23
Operating Experience of Transformers

2-Furaldehyde (2FAL) Analysis

Tests evaluating 2FAL were used in an effort to judge and compare the aging of the
transformers. The furanic compound analysis for (2FAL) indicated very low levels of this
compound in all ten (10) transformers. The highest level reported was 0.02 PPM although the
transformers have operated for nearly 25 years at or near maximum rating. These tests were
judged inconclusive. Recent studies1 indicate that thermally upgraded paper may account for
significantly lower levels of 2FAL when compared to paper that is not thermally upgraded. This
may account for the very low levels found in the ten (10) transformers at NGS as they were
probably constructed with paper containing dicyandiamide. If the overall condition of the
insulation is represented by the DP tests, then according to data from Powertech Labs Inc of
Surrey, B.C. Canada, anticipated levels of 2FAL should be in the range of 0.1 PPM to 1.0 PPM
or at least an order of magnitude greater than found in the transformers at NGS.

Conclusions
There are nine (9) identical units with the same operating histories as the subject transformer and
the major concern is what damage may have occurred or is occurring to the remaining
transformers.

The information and observations found during dismantling allow a judgment that there is a very
high probability the failure in the subject transformer was either because of SE or a result of
damage from SE. The opportunity for flow of oil above a critical velocity over the insulation
system between temperatures of 20C to 40C has existed on the GSU and allows for SE.
Because the velocity of the oil and temperature are two important parameters, it was decided to
review the cooling system used on the NGS transformers. The consultant used to support the
NGS investigation was Mr. Harold Moore. Mr. Moore has considerable experience involving
shell form transformers and SE. Mr. Moores opinion was that if SE is halted and the damage
resulting from SE is not extensive, then the remaining transformers might be expected to provide
normal service. He also suggested that inspection to find evidence of SE in the other
transformers may lead to more damage than the inspection or findings could justify.

Action Plan

The proposed method to stop or eliminate the SE is to modify the cooling systems on the
transformer. The modifications to the cooling are to convert the cooling from a two (2) group
control to a three (3) group control and inhibit the last two (2) groups of cooling unless the top
oil temperature is above 50C. For these GSUs at least one pump must operate to provide
sufficient cooling from excitation and low load levels. The cooling inhibit will operate on a
dropping temperature at about 45C where the 2nd and 3rd cooling stages will be deenergized.
The second stage of two (2) pumps and coolers will start at a hot spot temperature of 70C, and
the third stage of one pump and cooler will start at a hot spot temperature of 75C provided the
top oil has reached 50C. The order of pump energization is important to minimize turbulent
flow at the bottom under the HV windings. The first stage pump that operates continuously is
diagonal from the HV bushing. The second stage is on the opposite diagonal and the third stage
is on the HV bushing quarter.

3-24
Operating Experience of Transformers

Perplexity

A perplexity developed at this juncture because the original transformer design has a high-
pressure (3.5 psi) nitrogen oil preservation system that results in conflicting requirements for the
operation of the cooling pumps. On one hand the pumps are required to run while the oil cools to
ambient temperature. This pump operation is to prevent bubble evolution in the oil that could
result in transformer failure when the transformer is energized. On the other hand operation of
the pumps when the oil temperature is below 40C may produce static electrification that
damages the transformer insulation. Working with Mr. Moore and ABB, a relative risk
assessment for bubble evolution and static electrification versus different operating
configurations, difficulty of implementation, and estimated cost for modifications was developed
and follows:
Risk Configuration Difficulty Estimated Cost
Level 0 Add COP system & stage pumps Great $1,000,000+
Level 3 Stage & time control pumps Moderate $120,000
Level 8 Stage pumps Easy $90,000
Level 10 Do Nothing None $0

Definitions:
COP - constant oil pressure (Oil head only).
Risk level 0 is lowest risk (Eliminates any further damage).
Stage pumps - change from 2 stage to 3 stage cooling.
Time control pumps - operate each pump alternately for 15 to 30 minutes before energizing
the transformer. One or no pumps would run during cool down of the transformer. Four (4)
pumps per transformer.
Estimated cost - Total for nine (9) transformers.

Recommendation

There is clear evidence of static electrification in the failed transformer and steps to minimize
further damage in the other transformers are appropriate. Because conversion to a COPs system
is expensive, requires detailed feasibility studies, and may not be possible, modification of the
cooling system to risk level 3 for all nine (9) transformers was recommended by engineering.
The nitrogen pressure had previously been reduced to 3.0 psi, which helps reduce the risk of free
bubbles forming in the transformer. At this time the modifications to the cooling have been
implemented on Unit #1 and are being implemented on Unit #2 to risk level 8. The timed
operation is only important when the transformer is deenergized and all pumps have been shut
down. The pump operation one at a time is to insure any trapped nitrogen bubbles are moved into
the gas blanket. This has always been a requirement to operate all pumps for 4 hours prior to
energization. Operators have not followed this practice in the past. The modification to risk
level 3 is under evaluation. However these changes when made must be done in a way that
requires no additional interface with operators and, if possible, reduced interface. The staff
reductions at NGS require any changes be done in an automatic and fail safe approach.

3-25
Operating Experience of Transformers

The failure of this transformer at NGS results in the realization that the station transformers are
susceptible to SE and steps to mitigate the SE problem are being implemented. Further steps
under consideration are the additional of multiple combustible gas analyzer for headspace gas
analysis. EPRI is of the opinion these True Gas Monitors TGMs may be able to detect SE
during susceptible periods. This information could be used to remove a transformer from service
should SE be taking place.

Summary

The station has opportunity for SE. Planned modification of the cooling on the remaining
transformers should reduce any risk of additional SE to a very low level. The evidence of heating
in the LV coils is not as severe as appearances as reviewed by insulation life tests and studies
conducted. The transformers appear to have a reasonable remaining life expectancy as related to
the thermal degradation. Because of the transformers age, further loss of transformers should be
expected as the transformers age and the integrity of the insulation systems of the transformers
have been compromised from damage already done by SE. The risk of bubble evolution remains
a concern but is judged the lower risk when compared to the SE that is documented.

Reference

1. K. Muge Bilgin, Andrew B. Shkolnil, Joseph J. Kelly. Effect of Dicyandiamide in


Insulation on 2-Furfaldehyde Concentrations in Transformer Oil, TechCon99, 1999,
pp. 207-216.

3-26
Operating Experience of Transformers

Photograph 1

Photograph 2

3-27
Operating Experience of Transformers

Photograph 3

Photograph 4

3-28
Operating Experience of Transformers

Photograph 5

Photograph 6

3-29
Operating Experience of Transformers

Photograph 7

Photograph 8

3-30
Operating Experience of Transformers

Photograph 9

Photograph 10

3-31
Operating Experience of Transformers

3.3 Whitpain #1C Transformer Failure


by
David F. Goodwin
T&S Engineering and Design
PECO Energy
1111 Old Eagle School Rd.
Wayne, PA 19087

Abstract
Early in the morning of October 4, 1996, PECO Energy Whitpain #1C phase transformer
(500-230-13.8 kV, 217 MVA, 1967 installation) failed violently and caught fire. After external
flames were extinguished, the fire continued to burn internally for over a week. Extensive
investigation did not reveal any hard evidence for the cause of failure. However, from events and
operating conditions surrounding the failure in conjunction with the failure history of this
transformer design (Westinghouse 7000 series), the phenomenon of static electrification was
considered to be the most likely cause.

Investigation
The Whitpain #1C transformer failed on the morning of October 4, 1996 at 3:26:32. PECO
personnel had been performing calibration and trip testing on the #1 230 kV Whitpain bus. When
the trip testing was completed, all #1 bus facilities at Whitpain were being placed back into
service. As the system operator attempted to close the circuit breaker for the J220-10 line, the
last 230 kV line to be placed into service, the #1 transformer tripped and caught fire. Digital
Fault Recorders indicate that the transformer tripped four cycles prior to the J105 closing. With
reference to the single line, a summary of those events is as follows:

Whitpain Substation
(Partial Single Line)

500 kV

#1 Transformer

J105 115 125 215

135 655
230 kV

3-32
Operating Experience of Transformers

October 4, 1996 -

00:24:37.42 Whitman #1 transformer 230 kV #115 circuit breaker intentionally


opened, interrupting load.

00:26 Whitpain #1 transformer 500 kV circuit switcher intentionally opened,


deenergizing transformer.

00:28:33.50 Whitpain #1 transformer 230 kV #115 circuit breaker closed, reenergizing


transformer.

00:30 Relay calibrations begin.

02:20 Relay calibrations end.

02:27:30.39 Bus #1 trip test - All 230 kV circuit breakers associated with Bus #1 tripped.
(J105, 115, 125, 135, 215, and 655) (Whitpain #1 transformer deenergized).

03:07 to 03:09 Closed and tripped 125 and 215 for bus tie trip test.

03:12 Whitpain #1 transformer 500 kV circuit switcher closed (energizing


transformer).

03:25:29 215 bus tie closed.

03:25:37 125 bus tie closed.

3:25:45 Whitpain #1 transformer 230 kV #115 circuit breaker closed (transformer


picks up load).

03:25:55 655 CB closed (energizing 220-65 line).

03:26:04 135 CB closed (energizing 220-13 line).

03:26:32.860 Whitpain #1C transformer fails, fault lasts for 6 cycles. The PECO technician
performing tests identified a fire on the top of #1C. Flames were reported 30
to 40 feet above the transformer. The firefighter who arrived described the fire
looking like a petrie dish full of oil on fire.

On Sunday, October 6, a Whitpain failure investigation team was formed and met at Whitpain to
begin work on gathering data and assessing failure scenarios. A major concern was to prevent
any loss of evidence that might prove valuable in the forensic analysis. Subsequent to this
meeting, a failure scenario matrix was created to provide guidance and structure to the
investigation. This matrix was based upon guidelines contained in IEEE C57.125, IEEE Guide
for Failure Investigation, Documentation, and Analysis for Power Transformers and Shunt
Reactors. It listed all the known transformer failure modes (insulation failure, bushing failure,
etc.), their possible causes (moisture, overheating, etc.), symptoms (what to look for), and our

3-33
Operating Experience of Transformers

investigation results (if obtainable). It also included the team consensus based on available
evidence and past experience on the probability that the subject failure mode contributed to the
failure. Upon completion of our investigation, a review of the matrix showed that the only failure
mode considered to have a high probability was that of insulation failure due to static
electrification. Although there was no conclusive evidence of static electrification found, the
circumstances surrounding the event and the propensity of this transformer design to historically
experience failure attributed to static electrification were compelling.

EPRI reports that there have been over 30 suspected failures due to static electrification
worldwide. The vast majority of these transformers are of shell form construction similar to
Whitpain #1C. In 1996, EPRI conducted static electrification tests on a single phase transformer
at Con Edisons Ramapo Substation.

The transformer was of the same manufacturer and similar in design to Whitpain, although it was
rated 500-345 kV instead of 500-230 kV as at Whitpain. PECO witnessed some of this testing.
With all pumps running, static discharges could be audibly heard and became prominent at an oil
temperature of about 25C and continued as the temperature rose through 28C and further.

On the evening of October 3, the load on the #1 transformer was in the mid to low 400 MVA
range with the temperature dropping through the evening to about 38F (3.3C) after midnight. It
has been estimated that the top oil temperature was about 30C when the transformer was
initially deenergized at 00:24:37.42 on October 4. Post failure investigation revealed that one
stage (one half of all pumps and fans) of cooler control on both A and B phase transformers was
on manual (continuous running) mode. Although the cooler control on the C phase was
destroyed in the fire, it is highly probable that one stage of cooling on C phase was also in
manual mode. Manual cooler control is contrary to the normal operating design for this
OA/FOA/FOA transformer where all pumps and fans are normally controlled to operate in stages
depending on the winding hotspot temperature. For low oil and winding hotspot temperatures,
the transformer is designed to operate with all forced cooling deenergized. Therefore, from
00:24:37 until 3:25:45.585 when the transformer was not carrying load, all pumps and fans
would have been off under normal automatic operation. However, with one cooling stage in
manual, one half of the forced cooling was still running. In this operating condition, the oil
temperature would have dropped further below 30C, into the temperature area of static
discharge activity experienced at Ramapo.

It had been hoped that the C phase teardown would shed some light on the failure mode. Harold
Moore was contracted as a consultant to assess the teardown. Mr. Moore found several strands of
copper winding that he felt had burned open and had possibly initiated the failure. The strands
were located near the top of the transformer in the series winding, 4 to 6 coils from the end of the
525 kV winding. The location and appearance of the strands were similar to failures documented
as resulting from static electrification that he had previously investigated. In addition, one coil in
the series winding in the vicinity of the failed strands had the outer turns twisted over each other
in the leg. This twisting is typical of local short circuit currents in windings. The strands were
removed from the transformer and sent to the PECO Corporate Lab for further investigation.
Microscopic examination, however, led the lab to conclude that the strands failed from
mechanical shear, not from an arc. There was no resolidified metal or distinguishable heat
affected zone observed. Mr. Moores response to the labs conclusion was that the breaks were

3-34
Operating Experience of Transformers

similar in appearance to those found in failed transformers that had definitely resulted from arcs.
However, as Mr. Moore stated in his response, it is virtually impossible to determine the cause of
a failure after fire has destroyed the evidence on insulation and coils. The fire burned for over ten
days with temperatures reaching over 700F. There was not a single piece of insulation (a key
indicator for transformer forensics) that survived the fire.

In conclusion, the investigation has not found any hard physical evidence to reveal the cause of
the Whitpain #1C failure. However, considering both the operating parameters surrounding the
failure and the history of this transformer design, static electrification was considered to be the
most likely cause. Based on the circumstantial evidence at hand, PECO has taken actions to
mitigate static electrification at Whitpain and other transformer locations on the system. Primary
among those actions has been the identification of those transformers that might be prone to
experiencing static electrification and then instituting oil pump operating procedures as a counter
measure to the problem.

3-35
Operating Experience of Transformers

3.4 Static Electrification Failure of MPT at Powerton Generating Station

Raymond F. Cameron
Christopher P. Stefanski
Commonwealth Edison Company
1319 S. First Avenue
Maywood, IL 60153

Abstract

A large, shell form main power transformer (MPT) installed at Commonwealth Edisons
Powerton Generating Station failed catastrophically on April 13, 1994. A physical examination
of the accessible portions of the transformer interior revealed extensive damage to the B-phase
high-voltage winding, with flash marks on the winding crossover lead and high-voltage lead, and
partial ejection of the C-phase high-voltage bushing. Analysis of the failure event and preceding
conditions as well as evidence observed during the teardown inspection pointed to static
electrification as the primary cause of failure.

Background

Powerton Unit 6 Main Power Transformer (MPT) was a shell form, 23.7-345 kV,
892/1000 MVA unit originally built by Westinghouse Electric Company in Muncie in 1972
and was assigned serial number 7002150. It was installed at the fossil fueled Powerton
Generating Station in 1973.

After ten years of service, the transformer failed due to a close in through fault requiring a
complete rebuild. The rebuild included conversion of the oil preservation system from a nitrogen
blanket (Inertaire) to a COPS tank system, with some necessary modification of the tank top, and
the replacement of the original oil pumps with those of a more modern design. The rebuilt
transformer was returned to its former Unit 6 MPT service position in April 1984.

Powerton Unit 6 and its associated MPT were taken out of service on December 6, 1993, for the
planned rebuilding of the boiler and a boiler control retrofit. During the time from December 6,
1993 through April 13, 1994, Unit 6 MPT remained de-energized. The transformer insulating oil
was processed by ABB during March 1994 due to concerns following an observed low oil level
and unsatisfactory results of laboratory oil tests. Oil processing and refilling of the transformer
was completed on March 13, 1994.

Incident Description

On April 7, 1994, high vibration alarms on the generator occurred during start up and the unit
was shut down. On April 13, 1994, the unit was again started and synchronized to the grid. Three
hours later, while still at minimum load of 23 MW, Unit 6 MPT failed, resulting in a catastrophic
rupture of the main tank and expulsion of oil. Physical damage occurred to the transformer
bushings and the iso-phase bus duct, resulting from the extensive movement of the transformer

3-36
Operating Experience of Transformers

tank. A physical examination of the accessible portion of the transformer interior revealed
extensive damage to the B-phase high voltage winding, with flash marks on the winding cross-
over leads and high voltage lead, and partial ejection of the C-phase high voltage bushing. A
345 kV, C-phase-to-ground fault resulted from the movement of the bushing. Extensive damage
to the low voltage buswork was visible, including mechanical collapse of the support structure
and arc damage at various locations.

The transformer was adequately protected from lightning, and there was no evidence of lightning
activity at the time. There was also no record of high voltage switching activity at the time that
may have induced high voltage switching surges.

Operations
Powerton Station procedures, since initial installation of the transformer, required the operation
of all pumps and fans at the time when the transformer was energized. Prior to the failure on
April 13, all eleven fan and pump cooler combinations had been in operation for about one-half
hour before the generator reached its operating speed of 3600 RPM.

During the ten years of operation since its rebuild in 1984, Powerton Unit 6 MPT experienced
76 unit start-ups because of the circumstances of this plants operation. Approximately 60% of
these start-ups were made in cold weather.

This MPT was back-fed from the 345 kV system on several occasions to supply plant auxiliary
power. During these periods, the transformer was operated with full cooling and very light
loading. The Powerton generator and its MPT were generally scheduled for base load operation
during its early years. Since 1980, the unit operation was more often in the load following mode.

Teardown Inspection
The core and coil assembly of Powerton Unit 6 MPT was dismantled at the ABB plant in
Muncie, Indiana, during the period of December 1 to 16, 1994. There was substantial evidence of
static electrification in all phases. Although B-phase faulted to initiate the failure, the evidence of
electrification was more severe in A- and C-phases. Subsurface charring or wormholes and
surface level treeing on insulation washers and side insulation panels demonstrated that there
were significant levels of static electrification present in this unit before the actual failure. The
evidence was most pronounced in the bottom corners of the coil washers and insulating panels.

Minor winding damage was found on the outer turns of the B-phase high voltage winding due to
flashover events. There was also some evidence of spacer block displacement and/or
misalignment. Evidence of slippage of spacers, used to direct oil flow, was observed throughout
the inspected B-phase coil package. ABB personnel explained that this had most likely happened
in the vacuum drying process during manufacture when the windings were heated and water was
extracted from the insulation. In this process, the washers shrink as the water is removed and the
coils expand as they are heated. The spacers, which are in two parts, separate, with one sticking
to the washer and the other sticking to the coil. This may result in partially impeded oil flow
causing increased turbulence in certain areas of the transformer.

3-37
Operating Experience of Transformers

Analysis

All of the observed internal damage to the transformer was at the very top of the unit and
involved arcing over a distance of approximately 14 inches through oil, indicating the presence
of an extremely high voltage and/or significant contamination of the oil. Such extreme voltage
can be developed by lightning, switching surges or by electrostatic charge accumulation where
very high DC voltages far in excess of the operating AC voltages are created by movement of the
viscous insulating oil against the solid transformer insulation. With the known absence of
lightning or switching surges and the observed evidence of static electrification damage on the
insulating materials, electrostatic charge accumulation was determined to be the source of the
high voltage.

Static electrification phenomena can also produce carbon contamination and gas bubbles in the
oil. In addition to the high levels of electrostatic charge in the upper oil plenum area,
contamination and/or gas bubbles resulting from electrical breakdown of insulating material
could have been forced by oil flow into this top oil space and may have contributed to the
flashover. The existence of electrostatic charging was supported by many instances of
wormholes and carbonized tracking of the pressboard insulation systems and by carbon
imprinting of fern-like (tree) patterns on the pressboard sheets.

There were at least three operating conditions present which promoted the development of static
electrification in this transformer. First, the oil was very cold, approximately 10 degrees C, and
viscous. Second, the start-up procedure required the operation of all pumps and fans, prior to and
during energization. This transformer had 11 six-horsepower pumps, each rated at 800 gallons
per minute. Third, the transformer had been subjected to an established ABB vacuum
dehydration and degassification process during the foregoing outage. After processing, the
moisture content of the oil was reported as 2 ppm, with a 36 kV dielectric strength. Very dry oil
conditions resulting from the vacuum oil process are conducive to the development of static
electrification.

Conclusions

The failure was initiated by carbon and gases generated by static electrification discharges at the
bottom corners of B phase, in the main high to low space. The carbon and gases were pumped up
through the insulation to the top end. A probable failure mechanism involved gas bubbles
becoming trapped under channels and angles in high electrical stress regions. This resulted in
partial discharges which then led to the ultimate failure involving the top end of the high voltage
coil and the low voltage leads, coils and bars.

References

1. Reynolds, J., et al., Powerton MPT 6 Failure Analysis, An Independent Transformer


Review Committee (ITRC) Report, ComEd/FTI document dated April 28, 1995.

2. Oomen, T. V., et al., Static Electrification Control in Power Transformers, EPRI EL-6081
Research Project 1499-7, December 1988.

3-38
Operating Experience of Transformers

Photograph 1
View of Operating Powerton Unit 5 MPT (Similar Installation to Failed Powerton Unit
6 MPT)

Photograph 2
Evidence of Electrostatic Charge by Sub-Surface Charring and Tracking on High-Voltage
Coil Package Pressboard Insulation

3-39
Operating Experience of Transformers

Photograph 3
C-Phase High-Voltage Coil Package Pressboard Insulation with Sub-Surface Tunneling
(Wormholes) and Carbonization

Photograph 4
A-Phase High-Voltage Coil Package Collar and Wrapping Pressboard with Sub-Surface
Tunneling (Wormholes) and Carbonization

3-40
Operating Experience of Transformers

3.5 Investigation of a Failed Westinghouse 500/345 kV 500 MVA Single


Phase Transformer
By
Don Angell
&
Kent Venosdel
Idaho Power Company

Abstract

Midpoint substation is a 500 kV transmission northwest inter-tie point for Idaho Power. The
500 kV part of the substation was purchased from Pacificorp in 1981. This paper outlines the
Midpoint T501 transformer failure and investigation of a Westinghouse 500/345 kV 500 MVA
shell form autotransformer serial number 7002504.

History

The Midpoint substation 500 kV interconnection is through a 500/345 kV, 1500 MVA
autotransformer bank composed of three single-phase shell-form units with a spare. The
transformers were built by Westinghouse in 1977 and were energized in 1981. Each unit is rated
250/333/416/500 MVA OA/FOA/FOA/FOA at 65C rise with a nitrogen blanketed oil
preservation system.

The transformer has a 250 MVA self-cooled rating and three stages of 65C rise FOA cooling.
Each forced-oil cooling stage has one 5.5 horsepower, 650 gpm pump and several fans controlled
by a winding temperature simulation circuit. Winding temperature simulation for H.V. Hot Spot,
L. V. Hot Spot and T. V. Hot Spot can cause the cooling stages to operate when 75C, 80C and
85C are reached. The cooling is provided with manual/auto switching but records and memories
do not indicate prolonged manual operation of the cooling equipment.

The tertiary of the three units is delta connected and supplies two switched 50 MVAR air-core
reactors and a local service transformer. The 525-345 kV neutral is solidly grounded.

The transformer failed acceptance test in 1977 due to elevated partial discharge measurements
during the factory induced test. The manufacturer rebuilt the transformer, which then passed
factory tests without concern.

Until 1998, the bank had performed without any problems.

3-41
Operating Experience of Transformers

Incident

On March 24, 1998, phase B of the Midpoint T501 transformer bank catastrophically failed at
2:45 P.M. The 500 kV inter-tie to the northwest was out of service for 30 hours due to this loss.

Complications to our failure were that there were only two other utilities in the nation identified
by NERC as having a spare. And only one repair facility in North America existed that could
handle the size, voltage class, and design of this transformer.

Preliminary Failure Investigation

The following data was collected after the fault to determine if any precursors to the fault
existed. The ambient temperature at the time of the fault was around 40F. The temperature
range for March is around a high of 50F and a low of 30F.

The insulation was power factored in 1995 at .38%. The oil quality showed a dielectric of 34 kV
(D1816) and moisture of 7 ppm in October 1996. The DGA results are as follows:
October 96 June 94 April 93

Hydrogen 0 1 2

Methane 0 0 4

Ethane 4 1 1

Ethylene 2 0 3

Acetylene 0 0 0

Carbon Monoxide 9 26 18

Carbon Dioxide 2131 2346 1651

Nitrogen 61035 92968 72572

Oxygen 4204 4477 949

Total Gas 67385 99819 75200

Total Combustible Gas 16 28 29

Equivalent TCG Percent 0.0124 0.0221 0.0251

The five-day loading cycle of the Midpoint transformer bank is shown in Figure 1.

3-42
Operating Experience of Transformers

Plot-0
1400 0.85
MW
1200 0.85
VARS
1000

800

600

400

200

-200
3/19/98 04:00:00 PM 5.00 Day(s) 3/24/98 04:00:00 PM
MPSN T501 WATTS M500 T501 VARS

Figure 1
Midpoint T501 Transformer Bank Five Day Loading Cycle

The Midpoint T501 transformer bank did not have the Doble INSITE online monitoring system
installed at the time of the fault. The only monitor on the unit was a Hydran 201 hydrogen
monitor. The Hydran detected the hydrogen that developed after the transformer faulted as
shown in Figure 2. The hydrogen in the transformer when from zero ppm to 2828 ppm in
15 minutes.
Plot-0
3000 2828.91
PPM

2500

2000

1500

1000

500

0
3/24/98 02:30:00 PM 1.50 Hour(s) 3/24/98 04:00:00 PM
MPSN T501 PHASE B HYDRAN P

Figure 2
Midpoint T501 Phase B Hydran

The transformer had a high resistance fault of 2400A for 5 ms then returns to 550A for 31 ms
then evolves in to a 10,076A fault which causes the transformer differential relay to operate
clearing the fault in two cycles as shown in Figure 3.

3-43
Operating Experience of Transformers

Figure 3
Midpoint T501 Transformer Fault Record

A review of the test data and fault information does not indicate a known failure mode
preexisting in the unit. The failure did not seem to have the symptoms of the well-known failure
modes and may have a problem specific to the design of this transformer. It was decided that a
Shell form transformer expert should be hired to participate in the failure analysis and to
determine the root cause of the failure.

Transformer Teardown

A group consisting of the insurance company, Idaho Power Company and Doble Engineering
Company personnel along with transformer expert Harold Moore met in Bradenton Florida to
conduct the transformer failure investigation. The investigation was complicated by an accident
that occurred after the phase package was removed from the tank. A forklift moving the phase
package dropped the coils and insulation after an extension fork failed. The coils fell to the
floor and were somewhat randomly oriented with some insulation missing or out-of-order. An
experienced worker saw the package before the accident and indicated a failure path was visible
from the high voltage to low voltage across the bottom end of the insulation. This path was not
traceable on the day of the failure investigation.

Harold Moores experience allowed him to reconstruct the pile of coils and insulation. However,
the damage caused by the forklift accident masked damage caused during the electrical failure.
Harold found puncture failures in layers 1,4 and 7 of the high voltage coil that connects to the

3-44
Operating Experience of Transformers

high voltage bushing. He also found failures on a coil at the neutral end of the low voltage
winding and severe burning on the static plate adjacent to the neutral end coil. Abnormalities on
the electrical connection of the static plate to the adjacent coil were also observed.

Throughout the transformers insulation system signs of static electrification were found which
included various sizes of trees and worm hole burning as shown in Figure 4. The forklift
accident had disrupted the orientation and placement of the insulation pieces which made
investigation difficult.

Figure 4
Worm Holes

The available evidence suggests that static electrification was the initiating cause leading to a
failure of the high to low space insulation. A definite cause for the failure could not be
determined. This transformer failure is different from other documented static electrification
failures in that no abnormal gassing was observed prior to the failure and the transformers
FOA cooling equipment is activated by temperature controls.

During the failure investigation the possibility of contamination from a source such as the oil
pumps was discussed. The pumps from this transformer were sent to J. W. Harley for tear-down
and remanufacture. No extraordinary wear was found or signs of bearing problems that may have
generated particles.

3-45
Operating Experience of Transformers

Online Monitoring

We are planning to install the Doble INSITE system on the bank to monitor temperatures,
loading, fans, pumps, bushings and hydrogen gas. We are also installing a PSD Tech partial
discharge system to detect any corona that may be occurring. The system uses acoustic sensors
with a Rogowski coil sensor.

Conclusion

The cause of this failure could not be determined. The most logical explanation at this time is
static electrification.

Nationally, the Westinghouse 7 million series of Shell form transformers has well known failure
modes. T-beam insulation and Static Electrification failures have been documented and
pre-failure symptoms have been identified.

A review of operating conditions and test results of the transformer before the fault indicates that
the transformer was operating properly and normally with no evidence of problems. A review of
the number faults on the 500 kV line reveals that phase B as experienced twice as many faults as
A and C phases. Discussions with industry experts familiar with this type of transformer
conclude that is this not a typical failure mode for this type of transformer. Nationally, this type
of transformer does have a history of failure modes. Idaho Power has never experienced any of
these types of failures.

3-46
4
MONITORING AND MANAGEMENT METHODS

4-1
Monitoring and Management Methods

4.1 Further Analysis of Streaming Current as Recorded from the


Ramapo Transformer

Stig Nilsson, Consultant


Los Gatos, CA 95033
Markus Zahn, Professor
MIT
Cambridge, MA 02139
Stan Lindgren, Manager
EPRI
Palo Alto, CA 94303

Abstract

Tests on a transformer at the Ramapo substation sponsored by Con Edison and EPRI have
yielded a high volume of data about static electrification and on-line monitoring of transformers
in general. This paper deals with streaming currents, electric charges entrained in the transformer
oil and partial discharge data as captured by the static electrification sensors. The data has been
analyzed to see if a prediction for imminent failure of the transformer could be developed from
the data. The conclusion is that this appears to be feasible. Also, the data clearly shows that it is
not the space charge of the oil that is the dominant factor in the failure mechanism but the
charging of the pressboard dielectric system.

Introduction

The tests of Con Edisons Ramapo transformer1 span the time period from early 1995 through
October 1996. The data analysis discussed in this paper was captured by means of a plate sensor,
referred to as the Nilsson Plate (NP), which is a virtually grounded metal plate surrounded by a
guard ring2,3 mounted close to the wall. The signal from the plate was divided into a quasi-dc
component intended to capture the electrokinetic activity (streaming current) in the tank and a
high frequency component4, which would detect partial discharges in the vicinity of the plate. In
addition, a temperature sensor was included in the instrumentation system. An absolute charge
sensor5 (ACS) was also a part of the system. This sensor takes a sample of the oil through a small
nozzle at the surface of the Nilsson plate and measures the charge of the sample volume.

Installation of Nilsson Plates

The installation of the top NP is shown in Figures 1 and 2. The plate was mounted on the side of
the unit below the Nitrogen blanket. (Because the plate has to be in contact with the oil it could
not be mounted on the top of the unit.) The plates area is 0.147 m2. Before the October 1996
tests, a plate was also installed at the bottom of the transformer. However, only a short data
record was ever captured from this sensor because it failed a few hours into the first test6.

This name was given to the plate by the late Professor Jim Melcher, MIT.

4-3
Monitoring and Management Methods

N2
18
Plate
Location 24
Oil flow
48 channel

Figure 1
Installation on the Top Nilsson Plate on the Ramapo Transformer

36 Oil Return
Plate and Pipes
ACS

Figure 2
Ramapo Transformer Side View

Circuit Analysis

The top plate currents have been analyzed assuming a lumped parameter network as shown on
Figure 3. The analysis, which is described in Appendices A and B, assumes that the charges are
spatially stable and that there are no discharges. Both of these assumptions are not true because
there is discharge activity inside the transformer tank and it is reasonable to assume that the
charge distribution varies with oil flow and with time. The magnitude of the errors is not known
either. However, in spite of these errors, the results could still be useful for monitoring purposes.

4-4
Monitoring and Management Methods

U
ICP Plate Top Plate
IS Source Top Plate
IRP Plate Current
Top
IRW Wall Tank
Current
Wall
Source ICW Wall
Zeq. Bottom Wall

Z eq. Bottom Plate

Bottom Plate

Figure 3
Lumped Parameter Equivalent Circuit Used for Analysis of the Plate Current
Measurements

Review of Test Records without Applied Voltage

July 10, 1995

This record is the basis for many of the assumptions so it will be looked at in detail. During this
test, the system was not energized. The temperature during this test was relatively constant
around 27 to 29C. The relaxation times for oil and pressboard dielectric has been estimated
based on available data in the literature as described in Reference 6. However, for this analysis
the data for the pressboard dielectric was adjusted such that the time constant at 35C was
2700-s. That is, it was adjusted to match the measured relaxation time as shown in Reference 6.
For this analysis, the relaxation time therefore is assumed to be about 5,500-s at 27C. The
conductivity of the pressboard ( pb) is given in equation No. 1 below.

Ramapo Tests 7/10, 1995


10000 30

Top oil temperature 25


NP current
Temperature (deg-C)

7500
20
Est. voltage
Plate current in pA

or Voltage

5000 15

10
2500
Est. Icp current
5
Est. Irp current
0 0
16.23

16.65

17.07

17.48

17.90

18.32

18.77

19.18

-5
-2500
-10

-5000 -15
Time (xx.xx Hours)

Figure 4
Ramapo Test during July 10, 1995. Transformer not Energized During this Test

4-5
Monitoring and Management Methods

10
Log(pb ) = 15.3 + 3.88 10 2 T S/m where (1)

T= temperature in C.
The relative dielectric constant for pressboard has been assumed to be 3.2.
The relaxation time for the oil is about 100-s, which is significant in comparison to the sample
interval, which is about 2 minutes. This could have affected the first 4 to 8 minutes of the
measurements after a step change of the streaming current.

For this analysis, the plate current has been used without any smoothing. The streaming current
(NP current), shown in Figure 4, reaches equilibrium at around 18.00 hours. (Note that the time
has a resolution of 1/100th of an hour.) The capacitive component of the plate current (ICP) at
time (t+ t) has been estimated based on Eq. (B6) and the resistive component at time (t=t+ t)
is calculated based on Eq. (B8). As can be seen, the capacitive portion of the plate current is
close to zero after about 1 to 1.25 time constants. At 18.55 the pumps were reduced from four to
three and a few minutes later to two pumps because an audible discharge was heard at about
18.50. The estimated voltage (U), calculated as per Eq. (B7), as shown in Figure 3 had then
reached about 22 p.u. where U is defined as per Eq. (A15).

The capacitive current changes polarity and begins to discharge as expected when the pumps are
turned off at around 19.00. When all of the pumps are turned off, the capacitive discharge current
is balancing the resistive component so that the plate current is close to zero even though, as can
be seen in the figure, the charges trapped in the transformers dielectric system is far from
drained. That is, the system behaves as expected.

Three other tests with the system de-energized were analyzed. One from the March 27, 1995
tests followed by the September 17, 1996 and October 15, 1996 tests.

Figure 5
Ramapo Tests during March 27, 1995. Transformer was last Processed in late 1994 and
had not been Energized since the Reprocessing Took Place

4-6
Monitoring and Management Methods

March 27, 1995


At the start of the test on March 27 as shown in Figure 5, the relaxation time for the pressboard
dielectric was calculated for 8C to be about 30,000-s (over 8 hours). At the end of the test, when
the temperature had reached about 16 degrees, it was estimated to be about 18,000-s (5 hours).
Note that the resistive component of the plate current is not shown in order to reduce the
complexity of the graph. However, it is the difference between NP current and the capacitive (Icp)
plate component. While the current is only 40% of the current measured in the July 10 test, the
voltage builds up and is still increasing at the end of the test. The voltage is estimated to be about
23 p.u. shortly before the test was terminated. Even though the estimated voltage was higher than
what was seen in the July 10 test, no major discharge activity was reported. That is, this analysis
indicates that the PD or corona inception voltage was higher during this test than in later tests.
While this is contrary to conventional wisdom, at lower temperatures more of the voltage
stress would build up across the pressboard and less would be seen by the oil, because the
resistivity of the pressboard is relatively higher than that of the oil. However, in a creep stress
situation, the relatively lower breakdown voltage at the higher relative humidity of the oil should
have led to discharges. Since the stress distribution inside the dielectric system is not known, the
result is not possible to fully explain. However, had the test been run a little longer, breakdowns
would have been expected.

Figure 6
Test During September 17, 1996. The Transformer had not been Operated Since the
July 1995 Test Series

September 17, 1996


The tests on September 17 were run after allowing the system to rest since the July 1995 test
series. The first hour of the test was run without turning on any heaters with all four pumps
running. The relaxation time of the pressboard dielectric system was about 15,000-s at the
beginning of the test and about 8,000-s at the highest temperature reached during the test. Larger
discharges were recorded after about 14.00, at which time the estimated voltage had reached
about 15 p.u.

4-7
Monitoring and Management Methods

Prior to this test, the PD detectors were modified for continuous operation. Figure 6 shows the
PD count readings from the plate. It is readily apparent that the PD Count signal complements
the noise in the plate current readings as an indicator of discharge activity. As can be seen, the
high PD count frequency does not always coincide with the high noise level intervals in the
plate current. Therefore, both types of information are needed to assess the impact of static
electrification of the transformer. However, the PD count could also be a result of ac induced
partial discharges arising in the vicinity of the plate.

The estimated voltage peaks at around 15.00 after which the reduced relaxation time of the
dielectric system combined with what appears to be a slightly reduced electrification activity
begins to discharge the system. This reverses when the cool-down is initiated. The high level
discharges arise at a relatively lower voltage than what was seen in the preceding two tests when
the transformer was not energized. However, this is a transformer which has been exposed to
substantial discharge activity already so permanent damage can not be excluded as a cause for
having a lower discharge inception voltage. As will be seen later, the voltage at which high level
discharges were recorded in this test is close to the level that was observed when the transformer
was energized. Another possible error source is the estimate of the relaxation time for the
pressboard because the estimated voltage is very sensitive to the conductivity versus temperature
function for the pressboard.

October 15, 1996

During the October 15, 1996 test, all four pumps were operated until the initial test run was
ended shortly after 14.00. The data from this test are shown in Figure 7. The relaxation time of
the pressboard dielectric system varied from about 7 hours at the beginning of the test to about
6 hours at the end. Both the resistive and the capacitive components of the plate current are
shown in the figure. In addition, the data from the ACS is plotted in the figure. The shape of the
ACS readings match the shape of the NP current data until some time after 13.00 hours at which
time there is a step increase in the ACS readings. There is no apparent reason for this change.

Figure 7
October 15, 1996 Tests. This was the First Test run after Reprocessing of the Transformer

4-8
Monitoring and Management Methods

High level discharges were reported shortly before 14.00, which coincide with an estimated
voltage at around 15 p.u. The analysis indicates that had the test been continued longer, it could
have resulted in serious discharge activity because the rate of change of the estimated voltage
shows no sign of diminishing at the time the test was terminated.

Review of Test Records with Applied Voltage

July 11, 1995

The first test with applied voltage was run on July 11, 1995. A graph of the test data is shown in
Figure 8. The relaxation time for the pressboard dielectric system was calculated to be about
8,400-s when the test began. The test was begun by energizing the transformer, which did not
result in any streaming current. This proves that the plate is immune to 60 Hz fields. After the
energization, all four pumps were put into operation. The temperature of the transformer
increases as soon as the transformer was energized and the relaxation time for the pressboard
begins to go down.

The NP-current reading increased sharply as soon as the pumps were started. In fact, the
amplifier saturated. The only valid conclusion that can be drawn for this initial transient is that
the current was at least 50% higher than for the same test run without having the transformer
energized. But, the test shows clearly that the applied ac voltage increases the streaming current.
The estimated voltage on the pressboard dielectric system increases sharply. At about 13.00
when the estimated voltage had reached about 14 p.u. a few very loud discharges were heard.
The pumps were shut off right after a very severe discharge. These discharges occurred before
the estimated voltage reached 15 p.u.

Figure 8
First Test of the Ramapo Transformer when Energized

4-9
Monitoring and Management Methods

When the dielectric system is stressed by a combination of both ac fields from the ac system and
dc fields from the static charging activity, the breakdown strength of the dielectric system could
appear to be lower than for the case of only static electrification activity. Thus, breakdown or
PD inception level is a function of not only the magnitude of the fields but also the temporal
characteristics of the field. Since the detailed stress distribution inside the transformer is
unknown, it is impossible to know exactly at what level the discharges will take place. Note also
that the discharge inception level for a stress caused by switching surges when combined with
static charges would be different than with 60-Hz stresses combined with static charges. Thus,
the transformer could still be in jeopardy below the estimated voltage levels at which PD activity
has been recorded in these tests.

The test was restarted with two pumps in operation at around 14.00. The relaxation time for the
pressboard dielectric is also plotted and it can be seen that the plate current is strongly related to
the relaxation time. In particular it appears as if the flat plate current beginning at around 14.00
can be explained as a result of the change in temperature and a lower resistivity of the pressboard
dielectric system. Hence, a faster drain of charge, which appears to be compensated by a higher
charging activity since the estimated voltage stays flat, too.

At about 15.40 three pumps were put into operation. With three pumps in operation, an increased
charge level and a slight voltage increase can be seen even though the resistance of the
pressboard dielectric system is still declining. Discharge activity is reported during this test run
with the most severe discharge occurring at around 17.00. The estimated voltage levels out at
about 16 p.u. The voltage could in fact be prevented from increasing beyond 16 p.u. by a gap that
keeps breaking down at about this level because there is significant hash in the NP-current
data, which indicated significant partial discharge activity.

July 12, 1995

The tests on July 12 began with three pumps running and the transformer energized. It should
also be noted that the transformer had been operated up to about 35C during the 11th so the
moisture level of the oil would be equal to a temperature around 35 degrees. This might explain
why the NP current rapidly drops down to about 4,000 pA, about 1,000 pA lower than seen
during the tests on the 11th for a similar pump configuration. The large amount of hash seen in
the plate current beginning around 11.30 when the estimated voltage had reached about 15 p.u.
could well indicate a high level of discharge activity inside the tank. Other PD sensors recorded
very big discharges beginning at 10.30 AM.

4-10
Monitoring and Management Methods

Figure 9
July 12, 1995 Tests of the Ramapo Transformer

As the temperature begins to increase, the resistive component of the plate current (difference
between NP current and Icp current) begins to increase rapidly as expected. The voltage also
begins to go down so the charge generation also has to go down. It appears as if the reduced
resistivity of the dielectric system results in an actual reduction in the charge because at about
12.15 the capacitive component changes sign. At around 15.15 there is a breakpoint seen in the
NP current. This must indicate the beginning of reduced electrification activity.
When the fourth pump was turned on at about 16.90, there was some but no longer any
significant discharge activity. Just prior to starting the fourth pump, the estimated voltage had
gone down to close to 3 p.u. The plate current analysis indicates that the capacitive charging
resulting from adding the fourth pump was very short because the relaxation time for the
pressboard dielectric at around 50C is only about 450-s (7 to 8 minutes). When the cooling of
the transformer began at 18.10, an increased capacitive component of the plate current can be
seen and the estimated voltage begins to increase, too. When the temperature went down to about
41C, a breakpoint becomes visible in the plate current and the capacitive component of the plate
current begins to go down although the system is still charging. The estimated voltage stress
inside the transformer continues to increase but it had not reached any critical value at the end of
the test. However, the trend at the end of the test was towards higher stress of the pressboard and
problems could have arisen if the test had been continued longer.

July 13, 1995


The tests on July 13 did not lead to any significant discharge activity. The system had been
operated at around 50C during the 12th and the moisture equilibrium of the oil dielectric should
therefore be representative of a 50-degree system. The test with four pumps in operation starting
at around 14.30 gave rise to a plate current exceeding 12 nA. That is the same as had occurred on
July 11th. However, the current drops to 5.4 nA after just 30 minutes and the capacitive
component goes to zero at about that time, too. This should be compared with about 7.5 nA
during the July 11th test. It should be noted that the estimated voltage never went above 13 p.u.
at any time during the test. However, there appears to be a slight increase in the electrification
activity as the temperature is allowed to increase. This can also be seen in the final test of the
day.

4-11
Monitoring and Management Methods

Figure 10
Ramapo Test During July 13, 1995

October 16, 1996

The test, shown in Figure 11 for the October 16 test series, was run with the transformer
energized. Prior to this test series, the amplifiers for the plate had been rescaled to prevent
saturation. Thus, the peak current readings shown in Figure 11 are accurate. All four pumps were
turned on in sequence over a 15 minute interval beginning at 12.30 PM. Shortly before 15.00 the
pumps were shut off and then two pumps were restarted at 15.10. All of the pumps were turned
on again at 16.30 but after about an hours operation all but one pump was turned off. Shortly
after 19.00 all four pumps were again switched on and this was the situation until the end of the
test. Note that the relaxation time of the pressboard is about 480 seconds at 50C. This explains
the short duration of the capacitive component of the plate current. For oil the relaxation time is
estimated to be on the order of 200-s at 20C but it is only 20-s at 50C.

Figure 11
Ramapo Tests October 16, 1996. Transformer Energized During the Test

4-12
Monitoring and Management Methods

The first audible discharge was reported at 1.25 PM (time equal to 13.42) at which time the
estimated voltage had reached about 12 p.u. Severe discharge activity was recorded and led to
the use of only two pumps at about 15.00. Just prior to this, the estimated voltage was up to
about 13.7 p.u. Between about 15.00 and until around 17.00 hot oil was injected into the high
voltage bushing area. This seemed to have reduced some of the discharge activity but nothing
can be seen in the test data to indicate when the hot oil was injected or when the injection
stopped. However, there was much less discharge activity after 15.00. This also corresponds to
an oil temperature at around 40C at which temperature the relaxation time of the pressboard
dielectric system is down to about 20 minutes. That is, charges would drain off fairly fast even
without any higher conducting path in the top of the tank.

The estimated voltage increased sharply when the cool-down period began after 19.00. Just
before the test ended, the voltage had reached about 13.7 p.u. Again, this shows that a cool-down
cycle can lead to substantial stress on the dielectric system.

As was mentioned above, the smart manhole cover (plate) was modified prior to the
September and October 1996 tests. This made it possible to record both the ACS and the
NP current signals simultaneously as well as recording high frequency signals (PD related) as
seen by the plate. The high frequency signals, which should indicate high level partial discharges
in the vicinity of the plate, are shown in Figure 12.

Figure 12
October 16 Tests Showing PD Counts as Recorded on the top NP

The graph shows that virtually all sudden changes in the plate current reading is associated with
a high frequency spike. That is, the noise in the dc component of the plate signal is
associated with discharge activity. As was seen in the September 17th record (Figure 6 above),
there are periods when the PD signal shows discharge activity that is not visible in the NP
current. Thus, the two signals are complementary. Therefore the high frequency signal from the
plate appears to be useful in determining if the static electrification activity has reached a level
that causes frequent discharge activity in the transformer. However, switching operations and
lightning surges were also recorded as spikes, which is to be expected for all electrical sensors.

4-13
Monitoring and Management Methods

Review of ACS Data


There are few records with both ACS data and streaming currents recorded because ACS motor
pulses disturbed the partial discharge function of the instrumentation-system. This was corrected
prior to the October 1996 tests.

Figure 13
ACS Data Captured in the October 16 Event

The ACS data from the October 16th test has been used in an attempt to correlate the NP-data
with the ACS. The assumption for the comparison is that the charge density is uniform. Then
according to Gauss Law for a simple one-dimensional analysis with short circuited electrodes at
x=0 and x=L, the field gradient is:
L
(x )
dE 2
= E(x) = (2)
dx oil oil

The plate current is then:


oil L A LA
I P = oil E(x = L) A = = (3)
2 oil 2 oil

Setting:
L=0.6 m
A= 0.147 m2

4-14
Monitoring and Management Methods

and calculating the relaxation time for oil ( oil) as follows6,

oil
/o 2.20
oil
(at 20C) 4.20E-14 (S/m)
oil
= (oil) / oil 464 s (at 20C)
The results are shown in Figure 14. Note that for this calculation, a smoothed value for the
ACS readings have been used. The conductivity was calculated based on information in the
literature6, but was adjusted by about an order of magnitude to get a reasonable fit between the
plate and ACS readings. The temperature dependence of the conductivity followed the slope
used in Reference 6. The results indicate that this may not be entirely accurate because the
difference between the two quantities is having one sign at low temperatures and the opposite at
higher temperatures indicating that the slope for the conductivity as a function of temperature is
not quite correct. Of course, the location of the charge is not well known either and neither are
the flow velocities in the area of interest. Using the more accurate expression for a moving
medium:
L

U oil U oil U oil
I = oil E(x = L)A = UA[ e ( + 1)] where (4)
P o L L
L
U oil
o = (x = L) e (5)

Setting U (the oil velocity) between 0.1 and 1 m/s in Eq. (4) gave almost the same result as for
the case with an assumed uniform charge distribution because in this range the term L/U oil << 1.
Figure 14 shows an exponential charge distribution with an assumed oil velocity of .01 m/s.
Where the plate is located, there must be a significant oil-flow with unknown velocities. It is
probably more than 1 cm/s. Thus, a uniform charge density is probably the best model.

Figure 14
Analysis of ACS Data using October 16, 1996 Test Data

4-15
Monitoring and Management Methods

This simple, one-dimensional analysis shows that both the ACS and the NP should be possible to
use for monitoring of static electrification activity in transformers.

Summary and Conclusions

The simple RC model used for the analysis of the Nilsson Plate data captured during the Ramapo
tests seems well suited to explain the basic results of the tests. That is:
The static electrification failure of the Ramapo transformer appears to be directly
correlated with charging of the pressboard dielectric system because there is no significant
discharge activity taking place until the pressboard dielectric system has been charged to a
critical voltage.
The voltage at which larger partial discharges take place appears to be equal to or lower
when the transformer is energized with ac voltage and with the pumps running than with the
transformer de-energized.
The relaxation time for the pressboard dielectric system determines the rate of charging of
the system and indirectly for how long the transformer can be operated before failure caused
by static electrification. Naturally, the higher the streaming current the higher the risk of
failure. With high streaming currents, the time to failure will be shorter, too.
The critical voltage at which dielectric breakdowns will occur can not be calculated directly
from the data and will no doubt vary from transformer to transformer and be different for
different designs, material combinations and flow rates. However, the estimated generic
voltage does correlate reasonably well with reported partial discharge activity for the studied
transformer.
It might well be feasible to calculate the critical voltage in the winding entrance region more
accurately by combining ACS and plate readings.
The high frequency signal content of the plate currents appeared to capture discharge activity
from the volume in the vicinity of the plate. The requirement for seeing PD events appears to
be that the capacitive coupling between the PD location and the plate must be good.
However, it will probably require more than one plate for complete monitoring of PDs in one
transformer tank.

It appears from the data that some of the tests were terminated prematurely. Had the test been
allowed to continue some serious discharges could have been triggered.

The main conclusion from this analysis must be that by using the plate current alone, it would
now be possible to predict when the Ramapo transformer would be close to failure as a result of
static electrification when energized with normal fundamental frequency ac system voltage.
Based on these results, it may be possible to set up a test to determine the critical voltage for an
unknown transformer so that the plate can be used for on-line monitoring of static electrification
of any transformer. Since a monitor using the streaming current captured by a winding would not
be suitable for on-line monitoring, the plate offers a practical alternative. Also, the plate possibly
combined with an ACS might be used to determine the propensity for static electrification failure
of transformers by running a test in the factory prior to shipment. This has been a long sought
after goal for the transformer buyers.

4-16
Monitoring and Management Methods

One can not conclude that the Ramapo transformer would safely withstand a switching or
lightning type impulse stress with the levels of static charging seen during the Ramapo tests. This
kind of test would have to be run in the factory prior to delivery of the unit as a design test. The
plate could be used for such a test to establish what level of static charging was present during
the test.

References

1. EPRI Report TR-111386, Ramapo Static Electrification Tests.

2. W.B. Gish: Static Electrification Monitor Acceptance Testing; Proceedings: Static


Electrification in Transformers, Report EPRI TR-102480, June 1993, Pages 1-7-1
through-7-12.

3. William B. Gish and Stig Nilsson: Static Electrification Monitoring on Union Electric
Mason Substation Transformer, Proceedings: Static Electrification in Power Transformers,
Report EPRI TR-105019, May 1995, Pages 4-1-1 through 4-1-11.

4. William B. Gish and Stig Nilsson: Monitoring DC Partial Discharges with Nilsson Plate,
Proceedings: Static Electrification in Power Transformers, Report EPRI TR-105019,
May 1995, Pages 4-3-1 through 4-3-10.

5. A.J. Morin II, M. Zahn, J.R. Melcher, and D.M. Otten: An Absolute Charge Sensor for
Fluid Electrification Measurements, IEEE Transactions on Electrical Insulation, Vol. 29,
No. 2, April 1991, pp. 181-199.

6. Stig Nilsson, Markus Zahn and Stan Lindgren: Ramapo Tests: Results from Streaming
Current Monitoring Using a Nilsson Plate and an Absolute Charge Sensor. Accepted for
publication in the IEEE Transactions on Power Delivery, paper number 99 TD 351.

Appendix A. Equations Used for the Analysis of the Equivalent Circuit

The following equations describe the circuit shown in Figure 3. Note that subscript (P) stands for
plate, (W) for wall and (S) for source.
I = I + I + I + I (A1)
S CW RW CP RP

Q Q Q +Q
U= CP = CW = CP CW (A2)
C C C +C
P W P W

RPRW
U = I RP R P = I RW R W = (I RP + I RW ) (A3)
RP + RW

4-17
Monitoring and Management Methods

The following definitions are used below.


I R = I RP + I RW (A4)

I C = I CP + I CW (A5)

I P = I RP + I CP (A6)

For a step function at time zero in the static electrification charge transport (IS,0), the current
distributes between the plate and wall components at time (t) as follows:
t

IC = IS,0 * e
pb
(A7)

or for the plate component:


t

CP
= IS,0
pb
ICP *e (A8)
C P + CW

RP RW
where: pb = (CP + CW ) is the relaxation time for the pressboard (A9)
RP + RW

Note this simple model ignores the time constant of the oil, which at lower temperatures is
relatively long but still much shorter than the time constant of the pressboard dielectric.
Therefore, for long time periods and lower temperatures, the dominant component of the time
constant ( ) is that of the pressboard dielectric system. For higher temperatures the relaxation
time for pressboard is approaching that of oil. But, for temperatures below 60C, the relaxation
time for pressboard is still at least 20 times that of the oil.
Assuming that CP is much smaller than CW gives:
t t

C
IS,0 P * e pb = ICP, (t = 0) * e pb

ICP (A10)
CW

Again, assuming that there is a step in the streaming current (IS,0), the difference between the
capacitive current and the streaming current is then at time (t):
IR,t = IS,0 IC,t (A11)

However, the current flowing to the walls of the tank is not measured. Only the capacitive
current flowing to the plate is known at t=0. That is, the current through C P is known but the total
capacitive current is unknown. Substituting plate current functions for the total currents and
using these in (A11) gives:

4-18
Monitoring and Management Methods

CW RW
I RP, t = (ICP, 0 ICP, t ) (A12)
CP RP + RW

Based on the results of the July 15, 1995 tests, the factor:
RW C
W =K (A13)
R P + R W Cp

was estimated to be about 0.36. This is the plate current at time t= after a step in the streaming
current. However, this does not produce a zero streaming current when the pumps are shut off. If
K is set to one, the result matches the actual data fairly well. This indicates that RWCW is equal to
RPCP (note that the product CPRW can be assumed to be small). That is, the time constant for the
plate is the same as the time constant for the wall components, which is not an unreasonable
expectation. Thus, the plate resistance can be estimated as follows:
pb
RP = (A14)
KC P

It follows from (A2 and A33) that by setting CP = 1 p.u. and K = 1


U U Q CP 1
pb
I RP = = = = I Cp dt (A15)
R p pb pb

However, the measured plate current is not a simple step function and therefore, the signals have
to be processed assuming a time varying convection current.

Appendix B. Equations Used for the Analysis of the Nilsson Plate Currents

Differentiation of (A10) around a time t=tN gives:


tN

pb t
ICP,(t = t N ) = ICP,( t = 0 ) * e *( ) (B1)
pb

By Taylor series expansion of the exponential function in (B1) and using the first term of this
expansion one gets the following approximate expression for the capacitive plate current at time
(tN+ t) under the assumption that the total plate current is constant:
t
ICP,(t = t N + t ) = ICP,(t = t N ) (1 ) (B2)
pb

The resistive current is then equal to the resistive current at time t N plus the change in the
capacitive current between time t N and (tN+ t) or:
t
I RP,(t = t N + t) = I RP,(t = t N ) + ICP,(t = t N ) (B3)
pb

4-19
Monitoring and Management Methods

The resistive current at time tN is calculated from (A15) as: U t=t(N)/ pb where the relaxation time
for pressboard at the actual temperature has to be used. As is shown in (A15), the voltage across
the capacitor can be calculated by integrating the capacitive current component to get a measure
of the accumulated charge. However, the source current (streaming current) can not be assumed
to be constant but varying in time. Using the fact that if short sample intervals are used any
change in the source current will primarily flow through the capacitors, one can write:
CP
d(IS ) / dt = d(IC ) / dt = d(I P ) / dt (B4)
CW

The sampling frequency is 120 seconds or less and the estimated relaxation time for pressboard
is about 500-s at 50C. or about four times longer than the sampling frequency. Thus, the errors
introduced by this simplification should therefore be reasonably small even at the maximum
temperatures reached during the tests. For higher temperature operation, a higher sampling
frequency would have to be used or a more accurate analysis method will have to be used. Using
(B4) the difference between the predicted capacitive component of the plate current and the
actual change in the plate current is a measure of the change in the streaming current. That is
(based on Eq. A1)
~ ~
IP, (t=tN +t) ICP,(t=tN +t) IRP,(t=tN +t) = ISP (B5)

~
where IP is the actual measured plate current and I is used to indicate the predicted value. Thus,
the actual capacitive current at time tN + t equals:
~
ICP, (t = t N + t) = ISP + ICP, (t = t N + t) (B6)

The voltage across the capacitor has been estimated by integrating the estimated capacitive
component of the current by means of a trapezoidal integration function as follows:
1 (ICP,(t = t N t) + ICP,(t = t N ) )
U(t = t N ) = t + U(t = t N t) (B7)
CP 2

This integration has been done assuming that CP=1 p.u. The resistive plate current component is
then calculated as follows:
U(t = t N ) t
I RP, (t = t N + t) = + ICP, (t = t N ) (B8)
pb,(t = t N ) pb

This calculation follows equation (B3) above where the resistive component at time t=tN has been
estimated using (A15).

4-20
Monitoring and Management Methods

4.2 Static, The Ramapo Experience and What can be Done

C.M. Cooke
Massachusetts Institute of Technology
155 Massachusetts Ave.
Cambridge, MA 02139 USA

Abstract

The reliable detection of incipient discharge events induced by static electrification was an
important goal of the EPRI, Con Edison Ramapo Test effort. The well defined operational
conditions and extensive instrumentation of a full size energized power transformer, afforded a
rare and valuable opportunity to quantify static phenomena in a full scale practical environment.
The subsequent teardown of the unit enabled correlation of measurement and actual damage.
This paper reports on the clear, unambiguous identification of static phenomena by electrical
partial discharge measurements during the Ramapo experience and provides interpretation as to
implications for other power transformers.

Introduction

Ramapo provided three major opportunities in regard to static events induced by flowing oil:
(1) evaluate performance of static event detection in a full size energized test, (2) tear down
examination to better quantify static event discharge processes and (3) establish operating
conditions which control static event processes. The combined results are broadly applicable to
other transformers. Further, additional measurements by other groups (see other papers at
conference) provided valuable supporting information about the coincidence of acoustic PD
signals, changes in static collection currents, etc.

It is important to remember that the static process has its own circuit which is NOT the same
as the power frequency circuit, Figure 1. Thus, the static circuit has its own sources, transport
paths, collection and power flow which can be very different from those in the power frequency
circuit. Furthermore, these different static paths are often inter-dependent and difficult to control
by practical means. While static charging may be slow and continual, sudden static induced
discharge events can also occur.

The static circuit includes non-copper paths such as ducts, oil flow paths internal to a winding,
and paths through external oil gaps. Solid dielectric surfaces can also be static-active, and may at
one time store charges and at another be a preferred conduction path. External factors such as
temperature can change static conduction (usually not ohmic) by orders of magnitude. The
result is that in contrast to power frequencies, static charges can migrate and accumulate in
non-copper regions within the transformer structure, and solid dielectric surfaces often provide
preferred static charge sites.

4-21
Monitoring and Management Methods

The Static Circuit


Not Confined to Copper Paths
NU HV LV path across
oil flow oil gaps oil flow
OUT OUT

path across / to
solid dielectrics

path in
ducts path thru
windings

oil flow oil flow


IN IN

Figure 1
Static Circuit Paths Differ from Power Frequency Circuit Paths in a Transformer

The static circuit and power frequency circuits, although different, do interact. This interaction
may include both real conduction and capacitive induced coupling (steady or pulsed) between
charges and conductors, Figure 2. The result can be damage to materials and even failure. For
example, a static charge accumulation on a dielectric interface may suddenly discharge and
create a conduction path that is then followed by energy from the power frequency circuit.
Detection and control of static events thus requires great caution and careful identification.

Conductor [ Coupling ] Static Charge


Circuit Circuit
coupling by coupling by
exposed transport induction
conductors
charges in
oil volume
conductors
in
dielectrics
charges on
dielectric
surfaces
windings

Figure 2
Conductor Circuit, Static Circuit and Coupling

4-22
Monitoring and Management Methods

Electrical Detection of Static Events

Fortunately, a benefit from the coupling between the static circuit and the power frequency
circuit is to provide a useful and reliable means to observe static events. A static event is a
sudden redistribution of charge, and this sudden change will also induce corresponding changes
in the induced image charges on any nearby conductors. Thus, there is inherently a close
coupling between static events and induced pulse currents on power frequency conductors.

Furthermore, this close coupling between static and power circuits enables good detection of
induced pulses and thus discerns not only the occurrence of static events, but also their location.
In effect, the power frequency circuit is a distributed antenna that couples the released pulsed
electric energy from a static event to the detection devices. Detection and location of static
events can therefore be accomplished by pulse measurements from detectors coupled to the
conductors.

The bushing taps provide convenient access to static event data because their internal
capacitance make good high-fidelity coupling to the conductor circuit for pulse events.
Furthermore, with access to multiple bushings, improved location of the event origin is achieved
by effectively a triangulation of multiple signals recorded from the same event. This approach
is illustrated in the block diagram of Figure 3, where signals emanating from an event source
location site i travel to each of three separate bushings nodes, (1), (2), and (3), according to
their coupling represented by the transfer functions Hik.

v (t) v (t)
1 2

H i1
H i2

v (t)
i
H i3

v (t)
3
Figure 3
Static Event Pulse-Wave Propagation, Coupling from Source Site to Detection Sites

Hence the ability to measure individual static discharge events as electric waveforms at more
than one detection location enables an opportunity to interpret both the what and where of
each event. It is this electrical PD detection approach of waveform capture and multiple signal
detection, called the PD-PWA (partial discharge pulse wave analysis) method which was
employed to quantify and locate the static events in these tests.

4-23
Monitoring and Management Methods

Measurement Configuration, Ramapo Test Site

An electrical static discharge detection system based on the pulse wave analysis method
(PD-PWA) was constructed at the Con Edison Ramapo Static Test site. It was an on-line
measurement system, since the transformer could be energized by back-feed to a special
bushing at the transformer test tap. Figure 4 illustrates the main internal features of this Ramapo
Test transformer (a 500/345 kV, 333 MVA, 1-phase auto, shell-form design unit removed from
service) including vertically oriented pancake-style windings, bushings, and main oil-flow
connections (in at the bottom and out at the top). There is a large superstructure of kraft
materials at the top used to provide support for various leads and isolation of the HV and
LV bushings as well as for the tap-changer leads. It extended over the oil-flow regions.

Electrical pulse detection was made at the bushing taps and used a high-sensitivity, inductively
based, fast pulse sensor connected in series with the grounded tap lead. Four channels of pulse
waveforms were captured digitally from the sensors. These signals were the high-voltage
bushing tap, the low-voltage bushing tap, the neutral bushing, and in some tests the back-feed
bushing tap. The neutral bushing connection sometimes made connection by passing the neutral
ground lead directly to a sensor, or alternately via the bushing diagnostic tap to ground. The
capacitances of the bushings on the auto transformer were reported to be those given below in
Table 1.

Table 1
Bushing Capacitance Values
Ramapo 500/345 kV Auto Transformer

550 kV HV-Bushing: C1 Stud to Tap 438 pF


C2 Tap to Flange 6396 pF

362 kV LV-Bushing: C1 Stud to Tap 367 pF


C2 Tap to Flange 6565 pF

25 kV Neutral Bushing: C1 Stud to Tap 585 pF

To better model and interpret the measured PD-PWA signals the winding equivalent circuit
connections are illustrated in Figure 5. This structure corresponds to a baseline L-pad circuit
equivalent of the impedances of LV and HV windings. There is allowance for stray capacitance
to ground, and a high-low space capacitance from neutral to high side. The LV winding is split
into two sections, LV-1 and LV-2, one on each side of the HV winding.

4-24
Monitoring and Management Methods

HV
Pressboard LV
superstructure
back-feed

NU

Flow Out
x4
LV-1 Winding

Pancake
HV Winding

Coil
taps

High - Low Space

Windings
LV-2 Winding

Flow In
x4

Figure 4
Illustration View of Ramapo Test Transformer, 500/345 kV Shell-Form Auto

4-25
Monitoring and Management Methods

(W1. W2, and W3 = HV, LV-1 and LV-2 pancake windings, see Figure 4)
Figure 5
L-Type Equivalent Circuit Structure for Ramapo Auto-Transformer

The measured results confirmed that substantial signal differences did occur at the bushing tap
detectors for each event. The differences were in accordance to discharge event location at
different positions, and hence with different paths and different efficiencies of signal coupling.
For example, a static event, Figure 6, which originated within or near the second low-voltage
winding, LV-2, may be efficiently coupled to the HV bushing (via the high-low space
capacitance, Chigh-low) and also to the Neutral, but with less signal to the LV bushing due to the
cross-over lead inductance and the LV-1 winding impedance.

In contrast, an event located near the top of the HV winding is efficiently coupled to the HV
bushing with some signal at the Neutral bushing and to the LV bushing, Figure 7. This event was
obtained with all four pumps ON and no applied voltage, that is the bushings were grounded. A
very similar event, but much larger in magnitude occurred a short time later, Figure 8, when the
four pumps were ON and the transformer was energized with 500 kV HV voltage ON. Note that
in general, the with-voltage and pumps events were typically ten times larger than those with
pumps only. This difference is associated with the impact of additional energy available from the
copper circuit. The importance of high oil flow was seen by the observed major reduction in
static events corresponding to when there was a reduction in the number of pumps ON. With 1 or
2 pumps ON, there was very small activity and no large static events were maintained.

4-26
Monitoring and Management Methods

Partial Discharge Pulse Wave Event


(Ramapo 500/345 kV Auto Transformer, Energized, Four Pumps ON)
Detector Signals:
HV-tap
0.600

0.400

0.200

-0.000

-0.200

-0.400

-0.600
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00

LV-tap
0.600

0.400

0.200

-0.000

-0.200

-0.400

-0.600
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00

Neutral
0.600

0.400

0.200

-0.000

-0.200

-0.400

-0.600
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00

microseconds
E071195.043
Figure 6
Static Induced Event, ~10,000pC, T = 25C (coincident acoustics: bottom region)

4-27
Monitoring and Management Methods

PD-PWA Signal Waveforms, Ramapo Auto Transformer


Detector Signals:
HV-tap
0.600

0.400

0.200

-0.000

-0.200

-0.400

-0.600
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

LV-tap
0.600

0.400

0.200

-0.000

-0.200

-0.400

-0.600
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

Neutral-tap
0.600

0.400

0.200

-0.000

-0.200

-0.400

-0.600
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

microseconds

R101696.035
Figure 7
Static Induced Event, Cascade-Type, ~ 8,000pC
500 kV/345 kV Auto Transformer, T=12C
All Four Pumps ON (880GPM rated), Voltage OFF (coincident acoustics: top region)

4-28
Monitoring and Management Methods

PD-PWA Signal Waveforms, Ramapo Auto Transformer


Detector Signals:
HV-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

LV-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

Neutral-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

microseconds

E101696.060
Figure 8
Static Induced Event, Cascade-Type, ~ 80,000pC
500 kV/345 kV Auto Transformer, T=25C
Voltage ON (HV=500 kV), All Four Pumps ON (880GPM rated)
Shortly after pumps-only test of Figure 7)

4-29
Monitoring and Management Methods

The Distinctive Electrical PD Signal, The Cascade - Event

An important distinctive type of electrical PD event readily identified by the PD-PWA method is
the occurrence of a rapid succession of repeated short duration sub-events, such as seen above in
Figures 7 and 8. The time interval between these sub-events is short, typically only a few
microseconds. This rapid succession of short sub-pulses is called a Cascade type static event.
The specific ability of the PD-PWA method to preserve waveform details enables their clear
discrimination. A classical PD detection systems, in contrast to the PD-PWA approach, would
average these sub-pulses and yield a single value no different from that of a single event of equal
accumulated size.

The cascade-type event is a signature, fingerprint, associated with a specific kind of PD


activity. In particular, the cascade event progresses in time steps. This repeated stop-and-go
process is linked to a long discharge along a solid interface. An example is a pressboard
(or paper) surface in oil. If the applied electric stress is high along a solid-oil surface, then
breakdown can develop rapidly and completely without long quiescent intervals. On the other
hand, when the surface is in a nominally low tangential electric field, then breakdown is not
expected to develop at all.

The combination that establishes conditions for a cascade event is when there is (1) an extended
surface at low applied tangential electric stress; and (2) an additional source of energy available
from static electrification with significant surface charge accumulation, such as occurs with all
pumps ON at cool temperatures, say about 10-30C. Experimental conditions for cascade-events
are depicted in Table 2. The relation between cascade-events and surface damage was seen from
the teardown results, see later.

Table 2
Transformer Operating Conditions for Cascade-Type Events
(top oil temperatures in the 10-30C range)

State Voltage / Pumps Status Cascade-Event Activity

1) Voltage OFF and Pumps OFF NO (background level very low)

2) Voltage ON Only NO (none detected, many tests)

3) Pumps ON Only YES (notably with 4-pumps ON,


beginning within an hour, seen with
many tests)

4) Voltage ON and Pumps ON YES (larger compared to pumps only)

Hazardous, Loud Audible Bang Discharge Events

On many occasions the static electrification tests were run until large, loud audible BANG events
occurred. These events clearly were at the limit of safe operation, and typically pumps were shut
off or reduced, or voltage was switched off when events of very high magnitude developed,
except in the final test. The BANGs always corresponded to high intensity, near 100,000pC

4-30
Monitoring and Management Methods

equivalent PD size, and would saturate the instrumentation (even at the highest settings). Some
other nearby electronic measurement apparatus latched or failed from the transients produced by
some huge BANG events.

The recorded pulsewaves for two different type of BIG BANG events are shown in Figures 9
and 10. The first is for the case of especially large HV region energy while the second is for the
case of especially intense HV and NU region energy, perhaps even a momentary connection
between these terminals.

Discharge Track Marks After Cascade-Type Events

Visible tracking damage was observed during inspection after final tests which induced major,
(very large ~100,000pC) cascade events. Upon teardown, long path discharge surface and
subsurface (worm-like) tracks were found on pressboard layers located within the high-low
space, Figure 11. Long-path discharge surface tracks were also found along the broad area
elements of the super-structure just above the high-low space insulation, Figure 12. From the
construction arrangement of this transformer, there is substantial oil flow within the high-low
space, and this oil flow also emerges into the pressboard super-structure. In both these discharge
locations the nominal electric stress in the direction of the discharge channels was low, but the
oil-flow, if charged, could be expected to yield static charge accumulations to large values over
these extended area insulating surfaces.

The occurrence of oil-flow induced cascade events, and the occurrence of long extended length
discharge tracking damage at the regions where flow could cause static induced charges provides
strong support for the link between cascade-type events and discharges along cellulose-oil
interfaces. The recorded electric pulse waveform signals also correspond to the locations of these
regions of significant damage. It is most likely the unit would have catastrophically failed if
connected directly to a 500 kV transmission grid, since in that case much more low-impedance
energy would be available to feed the discharges.

Conclusions

The PD-PWA measurement technique was shown to identify distinctive forms of electrical PD
signals that can occur in large power transformers when flowing-oil static-induced events occur.
The PD-PWA measurements were consistent over many tests and provided early warning signals
before powerful damage causing events occurred. One particular form, the cascade-type event,
has been clearly linked to visible tracking damage and hence to important degradation within
power transformers. Static events were associated with high oil flow at cool temperatures,
especially 10-30C, and when large area dielectric surfaces were present to accumulate static
charges. In general, both operation and design to avoid these conditions would greatly limit or
eliminate the possible exposure to such damage.

4-31
Monitoring and Management Methods

BIG BANG Event


Detector Signals:
HV-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

LV-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

Neutral-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

microseconds

P091896.093
Figure 9
Big BANG Event, Large Cascade HV Region, >100,000pC
500 kV/345 kV Auto Transformer, T=19C
All Four Pumps ON (880GPM rated), Voltage ON

4-32
Monitoring and Management Methods

BANG Event
Detector Signals:
HV-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

LV-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

Neutral-tap
6.000

4.000

2.000

0.000

-2.000

-4.000

-6.000
-1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00

microseconds

E101696.061
Figure 10
Large BANG Event, Large HV, NU Region, >100,000pC
500 kV/345 kV Auto Transformer, T=25C
All Four Pumps ON (880GPM rated), Voltage ON

4-33
Monitoring and Management Methods

Figure 11
Static - Event Track Mark Damage on Paper Surface from Within High-Low Space
(After Big BANG Cascade Events)

4-34
Monitoring and Management Methods

Figure 12
Static-Event Induced Track Damage on Top Pressboard Superstructure
(Above High-Low Space, After Big BANG Cascade Events)

Acknowledgement

This work benefitted from the substantial involvement of Con Edison personnel, Donald Chu and
Leo Savio, and from Harold Moore of Harold Moore Associates. Important support for the
development of the PD-PWA method used in this work was provided by ESEERCO of NY.
EPRI with Stan Lindgren was a major sponsor of the static electrification tests on the HV power
transformer at the Con Edison Ramapo Substation. Other member of the Ramapo Measurement
Team contributed greatly to the net knowledge by providing a jovial spirit and good comparison
about coincidence of events based on acoustic detection by ABB and Harley and collected static
current by ERM.

Reference

Cooke, C.M., Electrical PD Measurements, Contribution to Ramapo Paper, EPRI Substation


Equipment Diagnostics Conference, New Orleans, February 17, 1997; Detection of Internal
Surface Discharges via Electrical PD-PWA Methods, EPRI Substation Equipment Diagnostics
Conference, New Orleans, February 17, 1998; and EPRI Final Report, Ramapo Static
Electrification Project, 1998.

4-35
Monitoring and Management Methods

4.3 Implementing a Static Electrification Diagnostic Model

Dan Dahlgren
Omaha Public Power District
444 South 16th Street Mall
Omaha, NE 68102
Jack Harley
J.W. Harley Inc.
9177 Dutton Drive
Twinsburg, OH 44087

Abstract

The beta version of an automated software model to detect and alert users to the possibility of
static electrification in large power transformers has been installed in an on-line transformer
monitoring system at Omaha Public Power District. The model is based on the analysis of tests
sponsored by EPRI. The tests were conducted in the factory, and in the field, on transformers
owned by Texas Utilities Company and Consolidated Edison Company of New York. The
monitoring system detected acoustic signals indicating probable partial discharge (PD). The
signals had a high correlation with known temperature/partial discharge characteristics. Utilizing
this data, Omaha Public Power District made modifications to the FOA cooling system on the
test transformer in an effort to mitigate the occurrences, and reduce the duration of the internal
electrical disturbances. Examples of data from the 17.1 kV 345 kV, 600 (672) MVA
Westinghouse (7000000 series) shell form transformer and subsequent changes to the cooling
system will be discussed. The algorithm being used will be described. Work is continuing to
better understand the boundaries of the model.

History

Nebraska City GSU #1 is a 17.1 kV - 345 kV, 600/672 MVA Westinghouse shell-form
transformer, built in 1978. This transformer is part of a family of Westinghouse generator
step-up transformers, which have exhibited two known problems:

1. Uninsulated T-beams

Core laminations can come into contact with the vertical members of the T-beam that are not
insulated. The subsequent hot spot involving the magnetic circuit in the transformer will produce
a variety of hot metal gases, including acetylene. The severity of the heating and the associated
insulation breakdown is directly related to the rate and composition of the gases detected in the
insulating oil.

(Note: The uninsulated T-beam situation has been recognized, for several years, as the principal
cause of problems associated with these transformers. However, Omaha Public Power District
(OPPD) tests indicate that the failure mode described below may be a more serious concern.)

4-36
Monitoring and Management Methods

2. Open Winding Failures

Turn-turn failures in the high-voltage coils of these transformers have been occurring at an
alarming rate. The root cause of the problem is usually difficult to determine since the failure is
typically violent, and causes substantial coil and core damage.

The uninsulated T-beam problem will cause low level gassing, but has not shown itself to be
progressive. The open winding problem, however, is a serious situation that should be
immediately addressed and continuously monitored.

With these problems known, an extensive monitoring package was added to the OPPD
transformer in October 1997. The monitoring package installed consisted of a J.W. Harley
T-MAP 3100 system (Figures 1 and 2). This system was designed and installed to monitor the
following quantities:
A, B, and C phase 345 kV Amps
A, B, and C phase Generator Bus Volts
Cooling Group 1 Amps (Fans + Pumps)
Cooling Group 1 Contactor Status
Cooling Group 2 Amps (Fans + Pumps)
Cooling Group 2 Contactor Status
Top & Bottom Oil Temperature
Ambient Oil Temperature
Partial Discharge Detection System
Various Transformer Alarms also monitored by Power Plant

Figure 1
J.W. Harley T-MAP 3100

4-37
Monitoring and Management Methods

Figure 2
J.W. Harley T-MAP 3100 PD Sensor

Early Detection

At the time that transformer monitoring system was being installed, an upgrade was being
implemented at the Nebraska City generating plant to increase capacity by ~30 MW. Increasing
the maximum net GSU transformer output from 660 MVA to 692 MVA required OPPD to
operate the transformer above its original nameplate rating of 672 MVA. Given the transformers
history of thermal problems, the decision was made to replace the original coolers with newer,
larger, more efficient units. This work was completed in late fall of 1997. With the new coolers
installed, the GSU transformer operated at full power throughout the summer of 1998 without
generating the thermal alarms that had previously plagued the unit.

Routine checking of data brought random acoustic partial discharge (PD) events to the attention
of OPPD and customer service personnel at J. W. Harley Inc. With experience gained through
participation in the tests of a TU Electric transformer and tests at the Ramapo Substation of the
Consolidated Edison Company of New York, the combination of low oil temperature and PD
events signaled the possibility that the static electrification cycle of charge build-up and
discharge was occurring.

The on-line partial discharge system detects the highest level of acoustic emission activity each
second and records the highest level in each minute. Six piezo-electric sensors (Figure 2) with a
frequency range centered at about 150 kHz are installed externally on the transformer tank wall.
An additional ambient sensor is installed separately to determine whether acoustic activity is
due to a source external to the transformer.

The number and level of PD events are shown in Table 1. The data are shown as the number of
five-minute intervals that have had acoustic activity. The activity at 100 counts/second and
below is regarded as noise. There are no very high level events in the data.

4-38
Monitoring and Management Methods

Table 1
Acoustic Event Levels vs. 5-Minute Intervals. Total of 373,752 Intervals
(6 Channels, 216 Days.)

Acoustic energy level Number of 5 min. intervals


(counts/second)
25 221,927
50 5,763
100 645
250 150
500 76
750 26
1000 13
1500 4
2000 0

Figure 3 shows the pattern of acoustic activity 250 counts/second and greater between
June 30, 1998 and January 26, 1999. The most frequent and the highest levels of activity
occurred on channel 6 which is located on the south side of the transformer tank close to the top.
There was no activity in the Channel 7 ambient sensor, indicating the acoustic source was within
the transformer.

OPPD Nebraska City GSU


PD channel activity vs date
7
6
PD channel

5
4
3
2
1
0
06/30/98

07/30/98

08/29/98

09/28/98

10/28/98

11/27/98

12/27/98

01/26/99

Figure 3
PD Channel Activity vs. Date

4-39
Monitoring and Management Methods

Correlation with Oil Temperature

The pattern of acoustic activity increased in intensity of count level in several cases at about
50C and noticeably at temperature below about 40C as shown in Figure 4. This occurred even
with only one bank of cooling in operation.

O P P D Ne b ra s k a C ity G S U
C o o lin g & A c o u s tic 2 5 0 + v s . T e m p

2000 3
1800
Ac o u s tic c o u n ts

1600

C o o lin g b a n k s
1400
2
1200
1000
800
1
600
400
200
0 0
32

34

35

36

41

42

50

52

56

58

59

60

60

61

62

63

63

64

65

66

68

70
P D c ounts T e m p e ra tu re d e g . C

Cooling s tate

Figure 4
Acoustic Activity 250 Counts/Second and Greater and Cooling State vs. Temperature
between June 30, 1998 and January 26, 1999

Cooling Modifications

It was decided by OPPD to uniformly reduce the cooling on the four manually controlled
coolers. These coolers are considered the base cooling group, and operate anytime the
transformer is energized. Consequently, these coolers would be the only cooling operating during
the light load, low ambient conditions. Each cooler assembly consists of one oil pump forcing oil
through an oil cooler that has four fans forcing air across the radiator fins. In order that the
reduction of cooling was uniform, one fan from each of the four manually controlled cooler
group assemblies was disconnected (Figure 5).

4-40
Monitoring and Management Methods

Figure 5
Blocks Indicates Fans Removed from Service

Diagnostic Model

The purpose of this section is to discuss the automation of the process of identification,
verification and warning of potential static electrification using the J. W. Harley Inc. diagnostic
software SAGE.

The Diagnostic Editor module permits SAGE users to set up algorithms or models of their choice
for diagnostic purposes. In this case, it is used to determine if the various symptoms associated
with static electrification exist. The three-step process is:
Determine the algorithm
Install the algorithm (and variations desired) in the SAGE formula screen
Test the algorithm using real or simulated data

Figures 6, 7, and 8 show the three steps implemented. The model used is based on the
interpretation of tests at the Consolidated Edison Ramapo Substation. Results have been
published in various EPRI papers by Harold Moore, Stan Lindgren, Donald Chu et al.

4-41
Monitoring and Management Methods

Any PD channel above 500 counts,


top oil 45C or less, and both pumps on.

AND (IF
(MAX ({PD 1 (cnts) [1]},{PD 2 (cnts) [1]},{PD 3 (cnts) [1]},
{PD 4 (cnts) [1]},{PD 5 (cnts) [1]},{PD 6 (cnts) [1]}) <500,0,1 )

,(IF ({Top Oil C (degC ) [1]}>45,0,1) )

,{Cooler Group 1 ( ) [1]},{Cooler Group 2 ( ) [1] } )

Figure 6
One Algorithm used Includes Top Oil Less than 45C and the Pumps Associated with both
Cooling Banks in Operation

Using variations of the algorithm in Figure 6 to account for one or both banks of pumps, it was
determined that the acoustic activity was occurring even with one bank of pumps in operation.
Other variables can be introduced and run concurrently, such as the acoustic count level as
shown in Figure 7. Mathematics and statistical functions are also available.

Figure 7
The SAGE Diagnostic Formula Screen

4-42
Monitoring and Management Methods

Testing the algorithm consisted of selecting data from several days when the PD sensors were
known to be active and using this in the test diagnostic screen, Figure 8.

Figure 8
SAGE Test Diagnostic Screen

The warning from a diagnostic comes in the form of an alert if SAGE is running at a place
remote from the substation or from an alarm or alert if at the substation. Figure 9 shows an alert
from data that has exceeded the algorithm limits.

Figure 9
SAGE First Screen Showing an Alert and the Cause

4-43
Monitoring and Management Methods

This diagnostic is being implemented as a template in SAGE for EHV transformers. Future work
is directed to adding time and trend elements to the algorithm.

Acknowledgements

J. W. Harley Inc. thanks The Consolidated Edison Company of New York, TU Electric and
EPRI for permission to participate in the tests that led to the implementation in SAGE of the SE
diagnostics discussed in this paper. Working with Omaha Public Power District, Jeff Parmelee
and Michael Keresztesy recognized the symptoms and implemented the diagnostic.

4-44
Monitoring and Management Methods

4.4 Mitigation of Streaming Electrification by Intelligent


Control of Oil Circulation1

J. A. Palmer
Division of Engineering
Colorado School of Mines
Golden, CO 80401
J. K. Nelson
Electric Power Engineering Department
Rensselaer Polytechnic Institute
Troy, NY 12180

Abstract

The paper describes a novel approach to the control of large power transformer cooling. An
intelligent controller is proposed taking elements of fuzzy logic and neural networks, together
with recent advances in sensor technologies. This controller drives the cooling system to operate
in flow and temperature regimes which limit the problem of streaming electrification while
minimizing thermal degradation of the insulation. A simplified example is given to demonstrate
the viability of the proposed controller.

Introduction

Streaming electrification is a phenomenon which occurs in large power transformers as cooling


oil is convected past the solid insulation. This phenomenon has been responsible for over thirty
incidents, including catastrophic failures, over the past 15-20 years since industry became aware
of the problem [1]. As a result, it has been studied extensively in the laboratory with research
being focused principally on the parametric characterization of the steady-state performance of
various streaming electrification models. Such studies have enhanced understanding of the
phenomenon and its dependence on operating conditions. For example, it has been observed that
the generation of charge is roughly proportional to flow rate. The dependence on transformer
temperature is such that, for lower temperatures, charging increases with temperature, up to a
peak in the midrange (typically 50-65C) and decreases as temperature rises further. The
potentials resulting from the surface charge accumulation are also dependent on temperature,
being most serious at lower temperatures when the resistivity of the solid insulation is highest.
Furthermore the distribution of moisture within the solid insulation, which is dependent both on
temperature and changes in temperature, has an effect both on the rate at which streaming
electrification occurs, as well as the amount of damage that it can do.

1
This contribution is reproduced with the permission of the Institution of Electrical Engineers. It is a reprint of the paper entitled
Intelligent Control of Large Power Transformer Cooling Pumps which appeared in IEE Proceedings: Generation, Transmission
& Distribution, vol. 143, September 1996 [10]

4-45
Monitoring and Management Methods

Nevertheless, despite the abundant research that has been conducted, recommended solutions to
the streaming electrification problem have been limited to the use of chemical additives
(e.g. benzotriazole) and the modification of the cooling system to provide the same rate of
cooling at a lower flow rate. However, utilities have expressed considerable reluctance to use
chemical additives, based on the incomplete understanding of how they work and the concern
that once added to the system, the chemicals cannot be removed. Physical modification of the
cooling system configuration can be expensive and time consuming. This paper proposes an
intelligent controller for the operation of large power transformer cooling pumps [2], which has
been developed to provide an alternative, non-invasive solution to the problem of streaming
electrification. Rather than significantly altering the configuration of the cooling system, the
controller, utilizing the base of parametric understanding of the phenomenon, together with
recent advances in instrumentation for the on-line monitoring of streaming electrification [3,4],
changes only the operation of the system. While this controller does not eliminate the
phenomenon, it does force the cooling system of the transformer to operate in flow and
temperature regimes which pose the smallest streaming electrification risk.

Design Approach

A number of issues must govern the design of a controller for this function. In order to be
practical and applicable to a variety of transformers, the controller should be computationally
simple and independent of the details of the transformer design. Because of the strong electric
and magnetic fields found in the environment of a large power transformer, the controller must
be able to reject noisy inputs from the instrumentation. Instrumentation is currently available to
provide inputs of transformer load, transformer top oil temperature (which in turn may be used to
compute the hot spot temperature), flow rate, charge density in the oil, neutral leakage current,
electrostatic charging tendency, moisture, partial discharge rate and magnitude, and others.
Ideally, the controller should be able to take any number of these sensor inputs, according to
availability, and determine the optimal operation of pumps.

Some particular issues relating to the pump operation must also be considered. For example,
recent studies have shown that the transient in streaming electrification resulting from the
initiation of flow includes an overshoot that increases for higher flow rates [5]. Consequently, it
is desirable that pumps be switched on as infrequently as possible, and only one at a time. The
order of pump energization should also be considered by the controller because this has been
observed to significantly affect the streaming electrification in the lower plenum of some
transformers [6].

In consideration of all these details, it was decided to base the controller on a combination of
fuzzy logic and neural networks. The selection of fuzzy logic was based on the continuous nature
of the inputs and the likelihood of high noise levels, the absence of a complete and simple model
of the processes involved, and the availability of laboratory data on which to formulate a set of
operational rules, as shown in [7]. A neural network implementation was chosen because of the
simplicity of retraining to provide for the necessary adaptations associated with a variety of
transformers and to accommodate advances in technology. Additionally the neural network
inference mechanism provides the power to nonlinearly interpolate along a five- or six-
dimensional control surface, while minimizing computational complexity [8].

4-46
Monitoring and Management Methods

The conceptual design of the controller, consisting of a module for each component of
instrumentation, one for the main controller, and one for the pump operator, is shown in
Figure 1. A modular approach was taken to maximize adaptability. For example, if one
component of instrumentation becomes unavailable, the main controller could be retrained
without that input. If a transformer is specifically designed to operate at higher or lower
temperatures, the temperature input module could be modified appropriately without changing
the controller.

Figure 1
Block Diagram Showing the Modular Design of the Controller

Controller Design

The software implementation of the controller may be described by the flowchart shown in
Figure 2. After the neural network weights are obtained, either from a file or through training, if
a file with the correct configuration of weights is not available, the program enters a control loop,
which consists of data acquisition and fuzzification, computation and pump operation. These are
performed in a continuous cycle until an interrupt halts the program.

4-47
Monitoring and Management Methods

Figure 2
Flowchart of Intelligent Controller Software

Input Preprocessing

An instrumentation structure is created for each sensor. That structure includes instructions for
polling the equipment via D/A board, serial communications etc. as well as any instrument-
specific preprocessing. The instrumentation structure also includes five user-specified level
designations, which are used to direct the module to interpret the data according to the desired
fuzzy membership functions as are generally used in typical fuzzy logic systems. Figure 3 shows
the typical arrangement of membership functions for an arbitrary distribution of temperature
level designations. Based on this example, if the input indicated a temperature of 65C, the
module outputs would all be equal to zero, except for a one on the high level output, while for a
value of 55C, the medium level output would register 0.5 as would the high level output,
while all others would be zero. This approach for input preprocessing serves a dual purpose.
First, it breaks up the analog range of an input into five nodes for easier training and better
performance of the main controller network. It also allows minor modifications in operating
conditions without retraining or reprogramming.

Figure 3
Example of Fuzzy Membership Functions for Temperature

4-48
Monitoring and Management Methods

Main Processor

The computational unit of this controller is based on a typical fully connected feed-forward
neural network (see Figure 4). The number of inputs is five times the number of sensors, and the
four outputs consist of two control outputs, requesting more or less cooling, respectively, and
two alarm outputs, one indicating a streaming electrification danger and the other indicating a
sensor malfunction. There is one hidden layer. The processing for the controller uses the
customary approach of each node giving an output based on a sigmoid activation function
operating on the weighted sum of its respective inputs. The program includes a routine for the
training based on a standard back propagation algorithm, if the correct weights are not available
upon initialization of the program.

Figure 4
Diagram of Main Processor Neural Network

Pump Operator

The pump operator interprets the instructions to increase or decrease cooling into a relay
operation for the appropriate pump. The selection of which pump to energize is based on a user-
specified sequence which may exclude any pumps which, due to construction or location, may
have the effect of enhancing the streaming electrification. The pump operator includes a user-
specified time delay to allow the contribution of one pump to stabilize before another pump is
permitted to be energized. This also avoids the superposition of multiple transients in streaming
electrification effects [5]. In addition, forcing a time delay limits the cycling of pumps and the
resultant wear on relay contacts and pump motors.

In addition to outputs controlling the pump relays, the pump operator includes two alarms. One
alarm signals that more cooling has been requested when all pumps were already energized,
suggesting an over-temperature condition. The other alarm signals that the main processor is
requesting more and less cooling simultaneouslysuggesting a malfunction or unexpected
operating regime.

4-49
Monitoring and Management Methods

Training Considerations

One major part of preparing the controller is obtaining an appropriate training set. In preparing
this, one must consider both thermal and streaming electrification aspects, as well as the
interaction between the two. The purpose of the controller is to maintain a temperature condition
that is the lowest reasonable without posing streaming electrification danger.

The first step is the consideration of fuzzy memberships and what their meaning should be. For
example, consideration should be made at what power level a minimal number of pumps should
be on, i.e. if natural convection no longer suffices at 5% of full load, this level may be designated
low load, and the fuzzy training table should contain instructions requiring at least minimal
cooling when the power approaches or exceeds that level, and so forth.

One may expect that if the streaming electrification is low, as indicated by both the charging
tendency and the charge density, then a lower temperature is permissible. On the other hand, if
either streaming electrification characteristic indicates high, then it would be prudent to operate
at a higher temperature, because the flow rate would be lower, and the temperature would be
driven beyond the characteristic peak that is seen in charging characteristics in transformers
(typically around 55C). The streaming electrification characteristics are particularly important
in considering the operation of the alarm function of the controller. By considering a plot of the
charging characteristic of the transformer, as determined empirically or by a model such as that
developed by Brubaker [9], one may determine the expected charging levels for various
operating conditions, and train the controller to sound an alarm when those levels are exceeded.

For the simplified model used in this example, the charging characteristic is plotted in
Figure 5 as a function of flow rate and temperature. The level designations for all three variables
are indicated on the plot. From this, it may be seen, for example, that the maximum expected
charging, designated as high, would be approximately 900 (arbitrary units) at very high flow,
and medium temperature, while medium charge density may be that which would be expected
for high flow at the temperature peak etc. If the charge density detected were significantly higher
than expected, the alarm would sound.

4-50
Monitoring and Management Methods

Figure 5
Plot of Charge as a Function of Temperature for Different Flow Rates in the Example Case

The sensor malfunction alarm should be trained to trigger for any set of input conditions which
could not be encountered under normal circumstances. For example, if the charge density
measurement indicates high when there is no flow present, the alarm would be energized.

Example Case

In order to demonstrate the viability of the intelligent controller, a simulation was performed.
The case used was artificial, although some effort was made to approximate typical
characteristics. For example, power was made to fluctuate sinusoidally like a daily load cycle,
and charge density was made to increase with flow and exhibit a peak in midrange of
temperature. The controller configuration used for the example cases consisted of five sensors:
power, temperature, flow, electrostatic charging tendency, and charge density. A transformer
with six pumps was assumed for both cases considered.

During the first case, the ECT remained constant, so that the charge density characteristic was
only a function of flow and temperature. This allowed for consideration of regular operation for a
relatively low charging system. The second case included a linearly increasing charging
tendency, with an equivalent linear increase in charge density. This permitted observation of how
the controller modified the operating point for different streaming electrification conditions.

4-51
Monitoring and Management Methods

Figure 6
Output Variables for Case 1 where ECT is Maintained Low and Constant

Results

The results from the first case are shown in Figure 6. As expected, the temperature stabilized at a
maximum just above the medium designation of 55 degrees, and the pumps were all off at the
lower temperatures, except when the load was sufficiently high to require it, holding the
minimum temperature just below the low designation of 35 degrees (ambient temperature was
taken here to be 25 degrees). Because the ECT was held low and constant, streaming
electrification was not a significant concern, and the controller operated much like a simple
thermostat. During the course of this run, no alarms were energized.

4-52
Monitoring and Management Methods

Figure 7
Output Variables for Case 2 where ECT is Linearly Increasing

The value of the controller is demonstrated in the second case, Figure 7. Here the linearly
increasing ECT, Figure 8, has basically no effect in the beginning of the run, but towards the end
of the run, the controller compensates for the streaming electrification problems by operating
with fewer pumps, and a higher temperature. With this compensation, the charge levels at the
end of the run are nearly the same as those at the beginning of the run, whereas without the
intelligent controller, the charge levels would be higher by an order of magnitude. The streaming
electrification alarm is triggered as the ECT rises to levels that may be considered dangerous.
While the controller has successfully limited the charge despite the rise in ECT, the station
operator would be alerted by this alarm before the condition worsens beyond the controllers
ability to compensate.

4-53
Monitoring and Management Methods

Conclusion

Industry is now in a quandary as to the appropriate action to take for the prevention of failures
and damage to transformers as a result of streaming electrification. While much laboratory
research has been undertaken to explain the phenomenon, few practical remedies have been
proposed. This initiative has been taken with the hope of filling the obvious gap. It is believed
that a controller which performs the functions described above may have very significant
applications in industry. As transformer failures, either due to streaming electrification or
insulation failure, involve substantial costs in down time and repair, the investment which would
be required for the hardware implementation of such a controller would be very justified.

Figure 8
Increasing Electrostatic Charging Tendency and Resultant Alarms for Case 2

It is even conceivable that the costs of such a controller may well be compensated by the more
efficient operation and extended life of the transformer due to the lower temperatures, while
operating only as many pumps as necessary. There has been some discussion that over-cooling
of transformers would increase efficiency, due to the lower resistivity of the windings. Making
that consideration or other operational issues a part of this controller would be as simple as
generating a new training set, and retraining the controller.

Acknowledgements

The authors are indebted to the Electric Power Research Institute under whose auspices this
study was conducted, and to Professor M. J. Embrechts for his contribution to the neural network
issues. Gratitude is also expressed to the Institution of Electrical Engineers for permission to
republish this work[10].

4-54
Monitoring and Management Methods

References

1. Crofts, D. W., The Static Electrification Phenomena in Power Transformers. IEEE


Transactions on Electrical Insulation, 1988. 23(1): p. 137-146.

2. Nelson, J. K. and J. A. Palmer, Intelligent Control of a Large Power Transformer Cooling


System, US Patent No. 5,838,881, 1998.

3. Furey, M. E., J. A. Palmer, and J. K. Nelson. A New Reciprocating Apparatus for the
Measurement of Electrostatic Charging Tendency of Transformer Oil. in EPRI Workshop:
Static Electrification in Power Transformers. 1994. Milwaukee, Wisconsin.

4. Morin, A. J., et al., An Absolute Charge Sensor for Fluid Electrification Measurements.
IEEE Transactions on Electrical Insulation, 1991. 26(2): p. 181-199.

5. Palmer, J. A. and J. K. Nelson. Streaming Electrification Dynamics in Oil/Cellulose


Systems. in Conference on Electrical Insulation and Dielectric Phenomena. 1994.
Arlington, Texas.

6. Ugo, J. H. and H. R. Moore. Static Electrification in a 700 MVA 23.7/345 kV Generator


Step Up Transformer. in Proceedings: Static Electrification in Power Transformers. 1995.

7. Cox, E., Fuzzy Fundamentals. IEEE Spectrum, 1992(October): p. 58-61.

8. Patrikar, A. and J. Provence, Control of Dynamic Systems Using Fuzzy Logic and Neural
Networks. International Journal of Intelligent Systems, 1993(July): p. 727-748.

9. Brubaker, M. A. and J. K. Nelson. A Full-Scale Simulation of Streaming Electrification in a


Large Core-Form Transformer Winding. in EPRI Proceedings: Static Electrification in
Power Transformers. 1995: EPRI TR-10501.

10. Palmer, J. A. and J. K. Nelson, Intelligent Control of Large Power Transformer Cooling
Pumps. IEE Proceedings-Generation, Transmission, and Distribution, 1996. 143(5):
p. 474-478.

4-55
Monitoring and Management Methods

4.5 Transformer Reliability: Management of Static Electrification


in Power Transformers

A C Hall, Consultant
Power Transformers & Reactors
Guildford, England GU1 2NB

Abstract

Arising from the studies begun by EPRI and others in the USA into the cause of several unique
transformer failures in the US during the 1980s, CIGRE Study Committee 12 (Transformers),
formed an international Working Group to investigate the incidence of static electrification (SE)
in power transformers.

The Working Group set up two Task Forces. Task Force 01 concentrated on those aspects of the
phenomenon that appeared to be dependent on the transformer material characteristics and their
environment. In parallel, Task Force 02 was required to examine the transformer design and
operation aspects. The purpose in each case was to determine the extent of the apparent SE
problem in transformers, to accumulate and distil other SE related internationally derived
evidence and experience and of course, to determine cause, effect and any mitigating measures.
Because of the complexity of the problem not all of these objectives were achieved or
achievable. One result of the Task Force 02 work has been to identify means to detect and avoid
SE effects in transformers and to provide transformer engineers engaged in manufacture and
power system operations with the best information and guidance possible.

Task Force 02 listed some eighteen factors for the attention of transformer manufacturers and
users. Along with other work being done by industry and research organisations it has been
possible to formulate some general guidelines for both transformer manufacturers and users.
These are highlighted in this paper, especially the more controversial factors, in order to
encourage audience contributions and obtain if possible, some practical consensus to benefit the
Symposium closing report.

Background

The subject matter of this Symposium and of this paper, namely, the occurrence of static or
streaming electrification in power transformers, has been described in detail on several occasions
at similar previous Symposia and it is suffice to say here, that the number of reported transformer
SE incidents is almost but not totally, confined to the USA and Japan. The US experience began
in the 1980s and that in Japan in the 1970s. Reports from other countries of similar experiences
have been very sparse in comparison. It is difficult from the few data available to understand
precisely why this should be. Obviously differences in transformer design, construction and
operating practices are possible reasons. It is also possible that the customary features of SE
related faults were not otherwise recognised in the course of routine transformers fault
investigations and this could have led to wrongly attributed fault causes. However, the subject
has been widely aired and it still surprising that despite repeated appeals from the Working
Group for reports and other evidence of SE incidents, there has been very little response, with the

4-56
Monitoring and Management Methods

notable exception that is of the USA. Reports of isolated SE incidents or suspected incidents
were received from South Africa, Australia and France during the course of these investigations,
but no others. The Task Force 02 findings are therefore based almost entirely on a composition
of actual fault investigation evidence from the US and Japan, with additional evidence
supporting research from other manufacturers, utilities, research institutes, universities and
individual investigators world-wide. Since the 1970s when the initial Japanese investigations
commenced more than 300 technical papers, reports and other references on the subject have
been published.

Within the context of background information, it is important not to overlook or over-emphasis


how few statistical data exist.

For example:
The SE fault incident rate for transformers in comparison to the transformer populations is
very small. It appears to be less than the transformer fire incident rate or in some cases the
transformer short-circuit failure rate of equivalent rated transformers. On any utilisation
scale, the risk of a transformer outage due to SE if not failure is very small.

The fault evidence also suggests,


SE incidents are confined to large rated transformers of several hundred MVA capacity.
The transformers operate at high voltage e.g. at 300 kV and above.
Within a geographical area, the type of transformer construction is statistically significant i.e.
in the two areas of the world which are known to have experienced this type of transformer
fault, SE failures of shell-type transformers appear more common in the US than in Japan.
SE faults are related to certain transformer manufacturers designs or products.
SE incidents occur more frequently in the field than in the factory.

Again, it has to be emphasised that because so few data has been reported, the total available
may not be representative. The above deductions should not be relied upon to be universally
applicable. Instead, what is important, is to continue to compile, collate and examine the data, as
it becomes available, such as,
Accounts of fault experience, service history and design details
Research results
Development of improved
operating practices
design tools
protection devices
measurement techniques
monitors and detection equipment
analysis and interpretation techniques

4-57
Monitoring and Management Methods

All these aspects are concerned with the management of SE and transformer reliability, however
that may be judged. Both manufacturers and users are affected but not equally. The risk burden is
on users at practically all stages of transformer procurement and operation. The only possible
exception is when the transformer is within the manufacturers guarantee period but not
necessarily entirely.

Static Electrification in Power Transformers

Users should have knowledge and an understanding of the SE mechanism and how it occurs in
power transformers in order to judge what measures are needed to avoid it. Reference has
already been made to the large amount of documented and published information available and
its value as a source of reference.

In general terms, the mechanism comprises several components involving the two principal parts
of the transformer, the main tank contents and the external cooling circuit. The common factor
linking the two is the oil. All three components are relevant to the SE problem.

The evidence from transformers believed or suspected as having failed due to SE effects,
indicates that in some forced oil cooled transformers, the oil can acquire an electrostatic charge
as it is forced around the cooling circuit and that this charge can separate and accumulate on the
solid insulation within the transformer, when the rate of charge accumulation there exceeds the
charge leakage rate. This charge separation and accumulation is of special concern when it
occurs within the winding assemblies where the electric stresses are greatest. The consequence
is, depending on the magnitude and rate of change of charge, the electric fields stress in the
charged region can be altered to the extent that partial discharge inception or insulation
breakdown occurs. So far, reports suggest the problem is associated with large rated, high
voltage transformers. This is probably due to this type of unit requiring,
Large mass oil flows
Greater solid insulation volumes
Higher electrical stresses
Reduced winding oil passage way dimensions and more intricate oil flow paths
The use of compact high performance coolers

The effect of these requirements depends on the materials and equipment used, how they are
incorporated into the design by the manufacturer, what stresses are imposed on them, either as
part of normal operation or as a result of their withstand capabilities having been exceeded by
operating practices beyond those prescribed by the manufacturer or otherwise intended.

The following components should be included in any assessment of transformer SE risk or


measures to avoid it,
Oil
Insulation
External coolers
4-58
Monitoring and Management Methods

The cooler circuit


Pumps
The transformer operating conditions

Present knowledge of the SE problem in transformers indicates these factors are interdependent
to some degree otherwise their individual relevance can be summarised briefly, as follows.

Oil

Laboratory tests suggest the electrostatic charging tendency (ECT) of oil depends on the crude
oil source and the oil refining treatment. [Ref: 1] Oxidation, particle content and viscosity also
affect the oil ECT.

Table 1
Relationships between Oil Condition and ECT

Oil Characteristic Tendency Oil ECT

Temperature

Moisture

Electric stress

Flow rate - laminar flow -


- turbulent flow

Oil oxidation

Electrostatic charge separation with increased temperature

Electrostatic charge dissipation with increased temperature

One of the techniques that emerged from the work done in Japan to reduce the oil ECT and thus
reduce the risk of SE, was to add a quantity of benzotriazole (BTA) to the oil [Ref: 2]. This
solution is not one that has been universally adopted, particularly by those manufacturers and
users who prohibit the use of oil additives in general.

Insulation

Reference was made above to the risk that arises when SE phenomena occurs at the interface
between oil and solid insulation within the transformer and how it is dependent on the
electrochemical behaviour of the oil and solid insulation combination and the surrounding
common environment. The effects are usually limited to local regions in the windings and the
associated major insulation and not distributed throughout the whole of a winding, though they
may be evident in more than one place.

4-59
Monitoring and Management Methods

The principal factors involving the solid insulation systems and affecting ECT include,
Electric stress
Conductivity
Temperature
Moisture content and dispersion
Insulation surface texture
Age

Moisture content and the changes that occur at the oil-insulation interface appear to be a major
factor. There is some evidence that a critical and rapid risk can arise when a transformer starts-up
or is shutdown or during transient load changes. The mechanism involved is complex and
clearly, determining and controlling a localised effect such as this within the transformer is very
difficult if not impossible. Consequently, the best means of avoiding any impairment of
transformer reliability from this cause is to avoid the problem in the first instance. This means
ensuring the transformer design, construction and recommended operating practices are
ECT free.

ECT Free Transformer Design and Construction

It has already been observed that SE effects occur in large rated, high voltage, forced oil-cooled
transformers, possibly as a result of their complexity and compact and efficient design e.g.
throughput MVA, highest voltage, losses, weight etc., expressed as a ratio of cost or volume. The
US experience includes a significant number of shell-type transformer designs, but in the overall
context of the reported SE incidents or problems, there is no firm evidence to suggest that one
form of core or winding construction is more or less prone to SE than another. This has been
generally confirmed by laboratory research studies of the factors affecting the oil ~ interface
behaviour.

Irrespective of the type of core constructions involved, it is apparent that most of the reported SE
related faults occurred during normal operation i.e., when the forced cooling was in operation. In
this context it is very important to note that a number of SE faults have occurred when the forced
cooling equipment was under manual control and sometimes when the transformers were not
energised.

So far as it is known, no SE incidents have occurred involving transformers that were operating
with natural oil circulation. The reason for this is simply that the oil velocities in the windings
and cooling circuit under these conditions are much less than those occurring with forced oil.
Under natural oil cooling, the typical winding oil velocity may be in the region of 0.03 m/s as
opposed to figures in excess of 1-2 m/s used in SE studies in the laboratory.

The reporting of SE incidents is vital to everyones understanding of the types and design of
transformer at risk, including the families involved. Whether or not a particular manufacturers
product or batch is a factor or that a users operating practice or particular components are a
contributory cause, is important information that can be vital to the operation of existing

4-60
Monitoring and Management Methods

transformers and for evaluating new designs, especially if they involve electrical, mechanical and
thermal criteria not previously experienced. Any departures from previously proven intrinsic
electrical, mechanical and thermal criteria should be regarded with concern.

The principal construction features that emerge from the reported evidence as affecting SE
include,
Oil e.g., type, velocity, temperature [Refs: 3, 4, 5, 6]
Insulation e.g., type, condition, surface texture, moisture content [Refs: 7, 8]
External coolers e.g., pumps, cooler design, pipelines, earthing, thermal performance.
Transformer operating procedures

The critical oil factor appears to be the oil velocity across the surface of the insulation but other
related considerations referred to above, are not be overlooked i.e.,
temperature
moisture content
insulation surface texture
external oil circuit features

As a function of transformer reliability assurance, ensuring that winding oil velocities do not
present an SE hazard is not easy, either at the design stage or later during factory testing or in
service. On balance, decisions about what level of oil velocity is acceptable in a particular design
should be made at the design stage. This should be done using computer analytical software
capable of resolving the dynamic thermal and hydraulic performances of the winding and major
insulation systems and applying it not only to the new design but also to earlier designs that have
a proven ECT free service history.

Analysis of fault investigation records indicates SE usually commences as a random incipient


partial discharge event, often audible as a distinct clicking noise. It may or may not develop into
a persistent discharge behaviour or even into an electrical breakdown event. The general view is
that once the fault mechanism has been initiated the insulation is permanently affected i.e.
physically damaged and that further SE activity can only progress the deterioration further. The
reports indicate discharge sites may occur where,
The oil velocities are greatest
The oil is most turbulent
The leakage path conductivity for charge induce current is least
The AC and DC combined electric field stress exceeds the local insulation intrinsic voltage
withstand strength.

In practice, it is probable two or more of these effects, if critical, will combine and promote a SE
activity collectively. Monitoring and protection devices have been developed to detect and locate
partial discharges and the presence of charges in transformers [Refs: 9, 10]

4-61
Monitoring and Management Methods

Otherwise, the winding oil velocities will be a function of the mass oil flow rate and the
geometries of the oil passages. Account should be taken of their dimensions and shape together
with the hydraulic, thermal and boundary surface conditions. In practice the geometries and oil
flow conditions vary throughout all parts of the winding assemblies and relying on average
conditions derived from external measurements is unlikely to be a reliable indicator of reliability
or ensure freedom from potential SB hazards, even if the data compares favourably with that
obtained from other successful and reportedly SE free transformers.

Solid Insulation

All reports examined concerned transformers that had conventional oil, paper and board
insulating materials. The steady state condition of the solid insulation at a fault location can only
be surmised at present. Condition assessment techniques e.g., polarisation and recovery voltage
measurement (RVM) are being applied but there is no confirmation yet that the condition of
electrical insulation systems can be evaluated by these means, especially those in the winding
critical areas. However, these techniques offer interesting possibilities, have a potential for
further development and are being more widely applied. The aim being to assess the
electrochemical condition and behaviour of the insulation system at the oil-insulation interface.

External Circuit

This area of transformer design and construction was not subject to same in-depth investigations
and reported upon to the same extent as the main tank contents. Most of the investigation reports
were not explicit about the external cooling circuit. The SE events appeared to have been
confined to forced oil cooled transformers fitted with forced air coolers i.e., OFAF type, having
minimal or no ON ratings. There is also little information in the reports detailing the pumps,
coolers: pipelines etc., or how these components are arranged or related physically or earthed
(grounded). The details that are required in order to assess the suitability and reliability of the
external cooling equipment in respect of possible SE hazards include,
oil velocity
oil flow conditions
pump charge generations
pump bearings
cooler head loss
leakage current
electrostatic charge pickup

Any extremes of oil velocity, pump blade speed, multiplicity of pumps or cooler head loss tend
to promote oil shearing and charge generation. Where multiple coolers and pumps are employed
they should be regulated to start in sequence and match the throughput load. It has to be noted
that not all reports concerned transformers operating on load when the SE incidents occurred. In
some cases the transformers were being shutdown or, restarted after shutdown or after a period
of processing. A number of transformer SE incidents were reported to have occurred when

4-62
Monitoring and Management Methods

additional coolers i.e., spare or auxiliary coolers, had been initiated. The overall conclusion from
the reports is that a cooling circuit should be selected which is capable of delivering the required
oil volume to the main tank without creating excessive charge density and that the pump capacity
at any stage should match the load and the required winding and oil temperatures as specified by
the transformer manufacturer. Unless details are known about the oil paths through windings and
coolers, some care has to be taken when comparing and evaluating data about mass oil flows of
different transformers.

Pumps should be arranged to be operated individually or in groups and be automatically


controlled to operate in sequence according to predetermined winding temperatures and time
constants. Extreme caution should be taken if the cooling is under manual control at any time. In
no circumstances should the manufacturers instructions concerning the operation of the cooling
equipment or its control be varied without consultation first with the manufacturer.

Summary and Conclusions

Any summaries or conclusions concerning SE in transformers have to be treated as interim for


the present, particularly, when considering how best to ensure the safe and reliable operation of
existing equipment and how to avoid any potential hazard in new plant. Condition assessment
techniques, monitoring and detection equipment and their associated control systems are being
continuously developed and improved. Detailed knowledge of winding assemblies, oil flow
patterns and hydraulic behaviour, electrical and thermal stresses etc., is required at some stage
and will need the involvement of the transformer manufacturer.

However, the consensus includes the following, that SE,


Occurs in the windings and associated insulation assemblies
Is predominantly a function of oil condition and the electrochemical condition prevailing at
the oil ~ insulation interface
That oil velocity, moisture content and solid insulation condition are major factors
Occurs only when forced cooling operates
May occur during steady-state or transient load conditions or when the transformer is
off-load
Is a function of oil and insulation characteristics
Depends on the transformer rating, voltages, core construction (shell or core-form), external
cooler design and performance including cooler ancillaries.

Finally, amongst the possibilities for managing SE in transformers in order to assure reliability
the following are recommended.
The internal oil velocity and stresses on insulation systems under working conditions do not
exceed proven values
That the external cooling systems are operated at all times in accordance with the
manufacturers instructions, in particular, when operated under manual control.

4-63
Monitoring and Management Methods

Contracts for new large rated high voltage transformers should include procedures for
pre-contract design and technical review to assess the SE risks.
Only proven mitigating measures should be adopted.
Departures from existing service proven practices should be treated with caution
Where transformers are identified as potential SE risks due to
type
manufacturer
batch history or operating circumstance etc.,

then,
consideration should begin to mitigating the risk by,
reducing or avoiding the cause of charge generation
fitting on-line monitors and detection equipment
modifying existing operating regimes and control procedures

Recommendations

In practice, as these investigations world-wide have demonstrated, when a SE problem arises in


service, the costs to a utility, including the consequences for the power systems and other similar
transformers, can be substantial. The consequences for transformer manufacturers involved can
be equally onerous not only for existing equipment but also for future contracts. In both cases
there is a need for open reporting and exchanges of information. Therefore, anything that can be
done to further awareness, improve transformer engineering and avoid further disruptive
incidents of this kind has to be actively and wholeheartedly supported. To this end it is
recommended that some central provision be made for the continued reporting and record of
future SE incidents using a formal and confidential protocol that can be accessed freely.

Also, these investigations have highlighted the need for the revision and improvement in
international standards e.g., IEC Standards. For instance, it is recommended that IEC 296 be
revised to include a standard test technique to determine the ECT of transformer insulating oils
and to specify maximum acceptance criteria.

Acknowledgements

This paper has been prepared for this EPRI Symposium using data and other information
obtained by Task Force 02 of CIGRE Joint Working 12/15.13 and from other sources, including
individual investigators and these contributions are respectfully and gratefully acknowledged.

The invitation from EPRI to present this paper and to participate in the Symposium is gratefully
acknowledged and appreciated.

4-64
Monitoring and Management Methods

References
1. Final Report EPRI TR-104973, EPRI Project RP 3334-01, May 1995.
2. Ieda M., Suppression of Static Electrification of Insulating Oil for Large Power Transformers
IEEE, Annual Report, CEIDP, pp. 246-256, Claymont, Delaware.
3. Oommen T.V., Pressboard Tube Flow Model Study EPRI Report, EL-6918, Project 1499-98,
Proc., July 1990.
4. Miyamoto T., Effect of Insulating Oils on Streaming Electrification in Transformers Ibid,.
5. Kamata Y., Power Transformer Insulation IEEE Trans. On Electrical Insulation, vol. EI-21,
no. 6, pp. 959-962, December 1986.
6. Higaki M., Investigation of Static Electrification in Large Power Transformers in Japan EPRI
Report, EPRI TR-102480, Project 1499-99, Proc., June 1993.
7. Moser H. P., Influence of Transformer Board and Nomex Board on the Static Electrification
of Power Transformers Ibid,.
8. Nelson J. K., The Role of Moisture Distribution in the Generation of Static Charge in Model
Core-form Transformer Structures Ibid,.
9. Feero W. E., Static Electrification Field Monitoring System EPRI Report, EL-6918, Project
1499-98, Proc., July 1990.
10. Gish W. B., Static Electrification Monitoring on Union Electric Mason Substation
Transformer EPRI Report, EPRI TR-105019, Project 3334-78, Proc., May 1995.

4-65
5
PANEL SESSION: ARE WE THERE YET?

5-1
Panel Session: Are We There Yet?

5.1 Mitigation of Static Electrification Discharges in Power Transformers

Harold R. Moore
Harold Moore & Associates
1062 Lakeway Drive
Niceville, FL 32578

Abstract

An examination of the known static electrification failures shows that a high percentage of the
failures could have been prevented if certain mitigation procedures had been followed. In fact,
there are only one or two out of the total population, which appeared to fail under usual operating
conditions. However, the exact history preceding the failure of these two units is not known. All
of the others failed under conditions that were conducive to the generation of static discharges.
This paper will discuss the enhanced learning from the Ramapo project and the actions that can
be taken to minimize the probability of static electrification failures in transformers that have a
propensity for this phenomenon.

Introduction

Laboratory studies have demonstrated that the electrification process is maximum at low
temperatures with high flow into and through the windings. Experience from circumstances
involved in the failures of transformers in service contributes to the understanding of procedures
that can be used to minimize failures from static electrification. The work conducted in the
Ramapo project conducted by Consolidated Edison and EPRI enhanced the understanding of the
steps that should be taken to minimize such difficulties in service.

Enhanced Understanding from the Ramapo Project

The laboratory studies have been documented in various papers, and Dr. Nelsons paper1
summarizes the failures and the circumstances surrounding the known failures. This section will
emphasize the enhanced understanding of the mitigation process that was obtained in the
Ramapo project.

Impact of Temperature

The Ramapo tests were conducted over a range of temperatures to again investigate the effects of
temperature on the electrification process. As expected, the following characteristics were
experienced.
There were essentially no measurable discharges below 10C.
Some low-level discharges were measured around 15C, and the magnitude increased as the
temperature was raised to 20C.

5-3
Panel Session: Are We There Yet?

The magnitude of the discharges increased and the time between discharges decreased as the
temperature increased from 20C. The maximum discharge activity occurred in the range of
20 to 29C. Very high magnitude, audible discharges occurred in this range.
Some high level discharges occurred as the temperature was increased from 30 to 40C, but
the magnitudes started to decrease in the high 30s.
There was discharge activity in the 40s, but there were no high magnitude events.
The discharge activity essentially stopped as the temperature reached 50C.
When the temperature was decreased from 50C, the discharge activity was less as the
temperature decreased through the 40s, but the magnitudes increased to rather high values as
the temperature decreased from 40C. High level discharges again occurred as the
temperature decreased from 30C into the high 20s.
It should be noted that the very high level, audible discharges having magnitudes in the tens of
thousands of picocoulombs occurred in the range of 20 to 29C when similar conditions existed
in the insulation system. This will be discussed in more detail later in the paper. We were not
able to perform intensive investigations of the discharge characteristics as the temperature
decreased after being heated due to test time availability. However it was clear that the discharge
activity was somewhat less, but high level discharges still occurred when the temperature
decreased into the danger zone.

Effect of Oil Flow Volume and Velocity


A number of experiments were conducted to determine the effect of oil flow volume and velocity
on the electrification process. There were measurable discharges with pumps only after the
temperature reached the middle teens, and the discharge activity increased with pumps and no
voltage as the temperature increased into the 20s. However, the high level discharges only
occurred with pumps in operation with voltage on the windings. The following was observed
during these tests.
Discharges did not occur when voltage only was on the windings.
There were no measurable discharges when only one pump was in operation with voltage on
the windings.
Some discharges were measured with two pumps in operation. However, the magnitude was
not high.
High level discharge occurred when three pumps were in operation with voltage.
Very high level, audible discharges occurred when all four pumps were in operation with
voltage. Some were very loud and could be heard 100 meters from the transformer.
During periods when the very high level, audible discharge were occurring, the discharge
activity decreased by magnitudes when the pumps were deactivated or the number of pumps
in operation was reduced.
No high level discharges occurred with all pumps in operation with voltage on the windings
when the oil temperature was above 40C.

5-4
Panel Session: Are We There Yet?

The Ramapo experiments again confirmed that the most significant way to minimize the
probability of damaging discharges from static electrification is to operate the pumps with
automatic temperature controls so that high flow does not exist when the temperature is below
40C.

Effect of Temperature Cycling

The Ramapo experiments demonstrated that the discharge activity decreased significantly after
the transformer had been through a temperature cycle up to 50-55C. After going through one
such cycle, it was not possible to produce the high level discharges under the ideal conditions
when the temperature was in the 20s and all pumps in operation with voltage on the windings.
This has been observed in other operating transformers. This reaction, which is believed to result
from a change in the moisture distribution, is another way to mitigate the possibility of static
electrification failures after transformers have been dried or processed in the field.

Effect of Running Pumps before Applying Voltage

During the Ramapo project, we learned to produce the high level discharges in minimum time.
One important component of this procedure was to run all pumps for an hour before applying
voltage. The discharges were of a higher magnitude and occurred in less time by running the
pumps in advance of applying the voltage on the windings. Conversely, the discharges were
lower in magnitude if the pumps were not run before applying voltage.

Recommended Mitigation Procedures


There are no known cases of static electrification in shell form or other transformers below
345 kV so that the following procedures are recommended for shell form transformers 345 kV
and above. Although the known cases in core form transformers is less, it is recommended that
the procedures involving activation of pumps with temperature controls be used on core form
also since it is difficult to predict which core form designs are susceptible to this phenomena.
The documented static electrification events have occurred with large pumps in the range of
800 gpm. However, the data suggests that the phenomena can be produced if a number of smaller
pumps are discharging oil into the bottom of windings so that the total flow is the same.
The Ramapo and other experiences demonstrate that the following actions have a high
probability of preventing failures from static electrification.

Control of Oil Flow


Flow of Oil into the Bottom of Windings. A number of experiences have demonstrated that
the oil flow into the bottom of individual windings is important in the process of preventing
static electrification. In more than one instance, discharges were found in the windings of
three phase transformers that had high flow rates into the bottom of the phases whereas there
were no discharges in adjacent phases with less oil flow. See Figure 1 for an example.
Therefore, it is important to change the sequence of pump activation so that there is never
high flow into the bottom of any phase at low temperatures when the probability of
electrification is highest.

5-5
Panel Session: Are We There Yet?

Staging of Pumps on FOA Transformers. Many of the FOA cooled transformers have only
two stages of cooling such that the oil flow is high with only one group of coolers in
operation. If is recommended that the pumps be divided into three groups so that the oil flow
will be low when the oil temperature is low. The following table can be used as a guide.

Number of Pumps Group 1 Group 2 Group 3

3 1 2
4 1 1 2
5 1 2 2
6 2 2 2
7 2 2 3
8 2 3 3
9 2 3 4
10 3 3 4
11 3 4 4
12 3 4 5

Control of Pumps as a Function of Temperature

The pumps must always be activated with the temperature controls. Almost all of the failures
have occurred because all pumps were in operation at low oil temperatures.
Never run all of the pumps before energizing the transformers. If the manufacturers
instructions specify running of oil pumps to move gas bubbles, etc, run the pumps in stages.
Stop the pumps and let the unit set for a minimum of 2 hours before applying voltage.
Never start operation of the transformer with the pumps in the manual mode. If there is an
emergency in which the controls are inoperative, special precautions must be taken to bring
the pumps on in stages commensurate with the temperature controls.
It is recommended that lockout circuits be installed so that the pumps cannot be activated
manually if the temperature is below the automatic activation temperature without special
actions.
It may be advisable to install a warning plate on the control cabinet door to warn operators
that the pumps are not to be operated in the manual mode.
Since some of the failures appear to be the result of a lack of knowledge about the static
electrification process, it may be advisable to add a section with instructions on this subject
in operating manuals.

5-6
Panel Session: Are We There Yet?

Special Precautions after Drying, Oil Processing or Long Term Idle Periods

A significant percentage of the failures have occurred after transformers have been dried or the
oil has been processed at the operating site. During the Ramapo experiments, it was proved those
long-term idle periods, when the transformer is not excited or loaded, is the same as drying or
processing. In other words, the discharge activity was the same after an idle period of several
months as it was immediately after performing a field drying operation. It is recommended that
special precautions be taken if any of these conditions exist.
If possible, take the transformer through a temperature cycle at reduced load with a limited
number of pumps in operation.
Generator Step Ups: Operate the transformer at reduced load and reduced flow with only
part of the pumps in operation so that the transformer can be taken through a temperature
cycle up through 60C and hold at this temperature of a minimum of 2 hours. Reduce the
temperature slowly by removing the load. Then initiate the pumps at full load in the
normal manner with the temperature controls activating the pumps.
Autotransformers: It is more difficult to perform such operations on autos since they are
normally energized and loaded by closing breakers on both sides of the auto which loads
the transformers. Most autos are made triple rated so that they should always be loaded
starting in the OA or self cooled mode so that no pumps will be in operation during the
heat up period. If feasible, maintain the temperature in the range of 50-60C for several
hours. For the few autos with only FOA ratings, it is recommended that the transformers
be operated at no load with reduced cooling such as with all fans off with one
pump/phase to circulate oil until the temperature has reached 50C for two or more hours.
This may be difficult to perform when the ambient temperatures are low.

Change of Cooling

The precautions listed above are in reality rather straightforward and require little special effort if
the systems are set up properly. However, a few utilities have chosen to change out the cooling
on transformers having only forced oil ratings. This requires changing the coolers as well as the
pumps because the cooler rating is a function of oil flow so that this process is quite expensive.
The coolers should still be arranged in groups and activated by temperature controls.

Design of New Transformers

The following guidelines should be followed when designing new transformers.


Use radiators and coolers that have acceptable output with low flow rate pumps such as
500 gpm or less.
Avoid insulation designs that result in barrier effects, which produce turbulence in the bottom
of the windings.
Arrange the pumps so that the flow into each phase is equal to avoid high flow rates into any
one phase.

5-7
Panel Session: Are We There Yet?

The pumps should be activated by temperature controls. Lock out circuits should be used so
that the coolers cannot be manually started if the temperatures are less than the temperature
control activation temperatures.
Install an instruction plate on the control cabinet with precautions that should be followed
such as never running pumps immediately before applying voltage.

Summary
The following is a summary of the recommendation above. The recommendations are simplified
for use in operational procedures.
Arrange pumps so that there is equal or near equal flow into each phase.
For transformers having FOA ratings only, arrange the pumps in the following sequence.

Number of Pumps Group 1 Group 2 Group 3

3 1 2
4 1 1 2
5 1 2 2
6 2 2 2
7 2 2 3
8 2 3 3
9 2 3 4
10 3 3 4
11 3 4 4
12 3 4 5

Activate pumps at all times with temperature controls.


Lock out circuits are recommended to prevent manual operation at low temperatures.
After drying or processing of oil, perform a temperature cycle at reduced load and with
reduced oil flow so that the temperature of oil reaches 50C minimum for 2 hours for FOA
cooled units. For OA/FOA/FOA or OA/FA/FOA, always start in the OA mode so that the oil
will be at elevated temperatures before pumps are activated.
For extreme cases where control of pumps with the temperature controls may not be
practical, changing of the coolers so that low flow pumps can be used is recommended.
Design of new transformers:
Use coolers with low flow rate pumps.
Avoid insulation designs with large barriers to oil flow.
Flow should be equalized into the phases.
Temperature controls with lock out circuits are a must.

5-8
Panel Session: Are We There Yet?

Reference

1. A Static Electrification Transformer Survey, J. Keith Nelson presented at the EPRI Static
Electrification Symposium, May 19-21, 1999.

Original Grouping
Group 1: Pumps 1,3,5,7,9,11
Group 2: Pumps 2,4,6,8,10,12

Modified Grouping
Group 1: 5,7,9
Group 2: 4,6,8,10
Group 3: 1,2,3,11,12

Observations During Dismantling


Much erosion resulting from
discharges and failure in phase
with 6 pumps
Some discharges in phase with 4 pumps
No discharges in phase with 2 pumps

5-9
Panel Session: Are We There Yet?

5.2 Record of Discussion from TRANSFORMER RELIABILITY: Management


of Static Electrification in Power Transformers Symposium

Edited by
Stig Nilsson
Consultant
http://www.stig-nilsson.com

Session: Panel Session for Presentation of Ramapo Test Results

Paper: Application of In Service Transformer Monitoring Systems during Tests


at Consolidated Edisons Ramapo Station, by Harold Moore and Donald Chu

Comment: Just a comment about the Ramapo unit and electrification charging tendency (ECT)
of the oil. As Harold mentioned, the charge density was about 80 to 90. During experiments with
a pressboard tube flow model, conducted under an EPRI project, I produced discharges very
similar to what Donald Chu reported. The system would charge up for awhile and then it would
discharge, and then the cycle would repeat. The ECT in my experiment was also 80 to 90, which
is very significant. What it means is when you have a high oil velocity, you dont need an oil
with a high ECT but if you have low velocities, you may need a much higher ECT of the oil. As
you probably will hear later, in our recent ABB studies we didnt have that kind of a velocity so
we had to increase the ECT.

Question: What was the velocity of the hot oil injected at the top and what was the proximity of
that injection point to the bottom of the high voltage bushing?

Answer: The capacity of the setup was 200 gallons per minute (gpm). It was not a very big
heater, neither very high velocity nor high output. We were injecting the oil through a two inch
diameter hose. So from 200 gpm, the area of the hose, you can conclude that the velocity was
fairly low. The regards to the proximity of the bushing, it was almost on the centerline of the
vertically mounted bushing, very near the bottom where the shield is mounted on the bottom of
the bushing.

Question: The temperature dependence of the oil resistivity is very important. Did you measure
or evaluate the temperature dependency of the oil used in the transformer?

Answer: It will be discussed later.

Question: You made these tests over a long time period. How did the oil change over this time
period? I understand that the behavior and the static electrification activity didnt change much
but some change of the oil could have taken place during the test period. Some of these changes
may not have had any impact on the static electrification activity. Can you to give us some idea
about any change in the characteristics of the oil?

Answer: Lets go back to the first field failure. One of our theories when the first failures were
seen was that one of the reasons was a change from acid refining to hydro refining for the oil
processing in this country. The Ramapo project disputes this because, as you heard, the Ramapo

5-10
Panel Session: Are We There Yet?

transformers were oil filled in 1971 with acid refined oil. So it was so-called old oil.
Furthermore, during the Ramapo project, the transformer was filled with the original oil. In other
words, they took the oil from the transformer, stored it and then processed it into the transformer.
This would have been in early 95. The oil was not changed through any of this. I dont believe
that we got the oil hot enough to see any oxidation. It was well protected with a nitrogen blanket.
So, as far as we know, from a practical standpoint there were no changes to the oil.

Question: One thing that I would have liked to have seen tried on the Ramapo unit was the
addition of a charge modifier like BTA or something similar to see if this truly has an effect on
the static electrification phenomena. Can you comment on that a little bit?

Answer: If wed had another quarter of a million, we would have done it. We wanted to, and it
was included in the original project plan but we simply ran out of time and money. It would have
been very interesting had we been able to do it.

Question: The BTA comment triggered another question on my part related to the original
moisture level in the unit and after the processing. Was there a dramatic change in the moisture
in the oil when the situation first started taking place? And what were the measured moisture
levels later?

Answer: As Don said, ABB Raleigh actually helped us a great deal in many ways, one of which
was to bring a state-of-the-art oil titration setup. By changing the agent almost daily, we could
get consistent results. I believe that the moisture level was consistently between two and four
parts per million. Thats a pretty dry transformer.

Question: I just want to ask a question concerning the PD measurement sensors. You said that
the transformer was equipped with photo diodes and photo multiplier tubes. Since you got a very
large discharge during the test I would like to know if you detected the discharge with the
sensors?

Answer: Lets wait with this discussion because it will be addressed later.

Paper: Acoustic Partial Discharge and Leakage Current Measurement Results of


Ramapo Tests, by George K. Frimpong, Michael J. Walden and Stan Lindgren

Question: Some of your reported leakage currents were a little positive but most were large and
negative. Do you have any explanation or some understanding of why it reverses polarity?

Answer: Im not sure. I know that in the second series of tests, it seemed like there was a large
offset, which I attributed to be a problem with the instrumentation. In this case, the shape of the
current was as expected but the amplitude might have been wrong. This may have been a
problem with the instrumentation. I believe that once the transformer is at rest, you may see
slight positive current from the charges relaxing inside the transformer, but Im not quite sure
about this.

5-11
Panel Session: Are We There Yet?

Question: When the instrumentation shut down after the first discharge, what did you do
differently to the instrumentation or the connection set up to preclude that from reoccurring? Did
you do anything different?

Answer: Lets see, there was a large discharge that really knocked out our computer and the
recorder and we had to reset. That just took a couple of minutes. I may have taken a little bit
longer to replace some chips in the acoustic PD circuits, although I dont believe that the leakage
current measuring instrument was ever damaged. It was just a matter of time to reset it and get it
back going again.

Comment: As you will see later on, the plate measures something and the winding current, or the
leakage current through the windings, measures something else. We noted that there was a ratio
of between 1000 and 2000 most of the time but a few times there was a change of winding
leakage current polarity.

Comment: Because there are two sources of current to that winding; one is the charge, which is
left on the surface after the separation and the other is the countercharge, which is pulled off the
surface and plating out somewhere else in the system. The nature of the polarity and, indeed,
the polarity changes, tell us something about the relative importance of those two processes. I
believe weve got a lot to learn from that. I would encourage you certainly to think more about
that aspect.

Comment: Maybe I can give you some information. In the past we have made some
measurement on a series of transformers and the polarity of the measured current was the same
on all the transformer except one. So I was waiting for the answer to understand this phenomena.

Comment: Part of my EPRI sponsored work has to been to develop physical models to verify
laboratory experiments. Ill present a little of that this afternoon, where Ill show some polarity
reversals. But I wont present any analysis here, because were only allowed 15 pages in our
report. But we have published various theoretical formulas and the simplified forms, as well as
very complete forms do predict polarity reversals as a function of various parameters such as
conductivity, temperature etc. So I am always very interested in having some physical data that
we can use for testing of our models to see if there is some sort of agreement between the theory
and the experiments.

Paper: Ramapo Static Electrification Project Test Results from J. W. Harley Inc.
by J. C. Harley

Question: I think you indicated to us there was a correlation between the discharges that you saw
and the 60-cycle waveform. Is that correct?

Answer: Yes.

Question: Did I not hear Harold Moore just tell us that there wasnt any correlation between the
60-cycle wave and partial discharges? This is important, because I believe that there is
something to learn from that. In regards to your photo multiplier tube, can you say something
about its spectral response? Im thinking that the emission from these events is at the ultraviolet

5-12
Panel Session: Are We There Yet?

end of the spectrum and the aromatics in the oil absorb very strongly at that point of the
spectrum. Maybe we shouldnt be too surprised that they are not very sensitive, even though
theyre counting a few photons.

Answer: Maybe were trying to cover too much detail, but until we got permanent damage, the
discharges were not correlated with the 60-cycle waves. The loud boom or bang, whichever term
you use, not only knocked out the sensors, but dimmed the lights in the trailer and knocked out
some other equipment in the substation. So it was pretty severe. After that event, and considering
the information gotten from the dismantling, I believe that we did have permanent damage. For a
while we did see discharges near the crest of the sixty hertz wave, as one would expect.
Concerning the second question of the spectral response of the photo multiplier tube, I do not
have any data about that. The reason is that these tubes were calibrated using a helium neon
laser, so we only know the response at one particular wavelength. Because we did not have a
good spectral light source, we have no way of determining the spectral response. Also, the way
that we were interpreting the data from the photo multiplier tubes did not really allow us to get a
good time signature. It was just basically an impulsive response. And, again, I do not know
whether we got a response caused by electromagnetic interference or if, in fact, we recorded a
light event within the tank.

Question: You mentioned that you believe that the partial discharge activity was related to the
relative humidity. I am basically looking for a little clarification. If the water content of the oil
was approximately constant, then the relative humidity is basically a temperature indicator
because we know there is a correlation with temperature. Do you believe that the measured
activity is related to moisture, or is it simply a temperature effect that were seeing?

Answer: I actually look at it the other way around. I believe that the most significant variable is
the state of the oil and that the relative humidity (RH) of the oil is probably even more significant
than the temperature. The temperature variation is rather forcing an RH change, and that may be
affecting the static electrification propensity of the oil but this is pure speculation. The
temperature is also important because of the conductivity of the pressboard is a strong function
of the temperature. So, even though the moisture dynamics across the pressboard/oil interface is
very important and related to the relative humidity, I believe that ultimately the real cause of the
partial discharges is due to the surface charge accumulation on a dry interface of the pressboard.
There must be very little leakage at low temperatures with very low moisture content of that
interface. As moisture is driven off into the oil with rising temperature, surface charge will
accumulate and it will keep accumulating until it exceeds the withstand level of the dielectric
system at which time a discharge will take place. So the temperature is probably also very
important.

Comment: I dont mean to say that it is not important but I do feel that the percent RH is
important as well.

Answer: Youre exactly right. Clearly the moisture dynamics across the oil/pressboard interface
is thats an extremely important variable and so is the conductivity of the interface. Both vary
with temperature.

5-13
Panel Session: Are We There Yet?

Addendum to Paper by Dr. J.C. Harley:

The graphs shown below depict response of the Harley Moisture Sensor during the Ramapo
Static Electrification Test. The sensor responds to changes in % Relative Humidity. For
reference, the corresponding temperature of the oil surrounding the sensor (loop oil temperature),
and the calculated PPM concentration of water in the oil are included. The PPM concentration of
oil is calculated based on the following formula1:

PPM = % RH e (0.347T 1.328 )

Loop Oil Temp, deg. C.


% Relative Humidity March 27, 28, 29, 1995
Water, PPM
14 16
12 14

Loop Oil Temp, C


10 12
% RH and PPM

10
8
8
6
6
4 4
2 2
0 0
7:07 19:45 7:45 3:44 15:44
Time

1
H. Azizian et. al., Ontario Hydro Research Division Report # 92-77-K. 1992

5-14
Panel Session: Are We There Yet?

Sept 17, 18 1996


10.00 25.00
9.00
8.00 20.00

Loop Oil Temp, C.


% RH and PPM
7.00
6.00 15.00
5.00
4.00 10.00
3.00
2.00 Loop Oil Temp, deg. C. 5.00
% Relative Humidity
1.00 Water, PPM
0.00 0.00
0:00 6:00 12:00 18:00 0:00 6:00
Time

Loop Oil Temp, deg. C.


% Relative Humidity Oct 15,16,17,18 1996
Water, PPM
8.00 60.00
7.00
50.00

Loop Oil Temp, C


6.00
% RH and PPM

5.00 40.00

4.00 30.00
3.00 20.00
2.00
10.00
1.00
0.00 0.00
13:41 1:41 13:41 1:41 13:41 1:41
Time

Question: Different groups, ABB, RPI and MIT and others have been looking at partial discharge
counts. Other groups have also been looking at leakage currents and the actual charge density in
the oil. Has anybody basically plotted on the same plot different peoples data to show, lets say,
that all the partial discharge counts peaked at the same RH and the same temperature? And also
that the leakage current did something interesting at the same time as the partial discharge counts
were up and the charge density in the oil also peaked out at the same time. Is there any overall
correlation?

5-15
Panel Session: Are We There Yet?

Answer: In the EPRI report from the Ramapo tests youll find all of this information. All the PD
data are plotted on the same graph so you can see how each sensor responded. Our conclusion is
that we have to reach our own conclusions about the correlation. Every conclusion you want is in
there.

Question: I would like to once again confirm that there is prominent damage caused by the
tremendous number of partial discharges or static discharges. Is this the reason for the discharge
activity with no pump operations?

Answer: Yes, there was permanent damage, as well show when we get into the dismantling
report, which contains information about the subject that we are discussing. The damage is
probably the cause of the discharges that we measured as a function at the 60 Hertz or at the crest
of the 60-cycle wave.

Question: This is the reason for the discharges under no pump operation?

Answer: Its conceivable. Its certainly related to the energization state.

Answer: The way we started the each test series was by energizing the transformer with no
pumps in operation and there were never any discharges. We did not apply a voltage-only-test
after the big bang that John describes. The next day, after the transformer had sat, which
probably moved a little carbon or whatever, the PDs stopped. So we believe that there was
considerable damage. It was definitely related to the damage, which was extensive.

Answer: Thats a good point. I guess I need to expand on that. There was definitely an effect.
However, it was still coupled to the pump state. I have data from some small periods of quiescent
data, where no pumps were operating and the transformer was energized, and we simply did not
see anything. So it was not only voltage related PDs that we were detecting.

Question: I guess I need a point of clarification. Figure 9 is identified as Maximum Bottom on


PD Activity versus Number of Pumps. In this figure we see PD activity. Could you give us a
time reference? The note I took for this PD activity was that no pumps were on. I wrote down
transformer energized after the failure event. Is all this data from the same time reference?

Answer: This figure shows data with energized transformer and with pumps on. Unfortunately,
on these graphs I didnt have room to put in a legend. However, this is data for the entire set of
energized tests. So this graph combines data from seven or eight different data sets. So it is
covers the entire time span. The individual time dependent data sets would be in the report that
Donald Chu mentioned.

Question: Okay, so based on that comment, can we assume that Harolds statement that the
PD events with no pumps on were all occurred after the internal bang in the transformer? Is that
correct?

Answer: This is an extremely lengthy, complex test program. Some of the discharge events
shown with zero pumps, occurred after we shut off the pumps for a few minutes. We believe that
the transformer was very close to failing. So that explains the zero pump PD events after we shut

5-16
Panel Session: Are We There Yet?

off the pumps when we still had discharges with voltage applied on the transformer. However, as
I said, within a short time the discharge activity decreased. It did not go away completely,
because we had created the event and we started pumps again. This is also true for the last series
of tests when we had repeated loud bangs. And when we reduced it back to one pump, we still
had discharges that originated from the events with the full number of pumps. So, as John says,
hes trying to represent all of this. The Ramapo report contains the data for the individual events.

Comment: One comment thats not included in this particular volume of papers has to do with
the factory testing that we did on a GSU from our Comanche Peak station. A unit that had been
in service for several years was shipped back to the factory and subjected to a lot of tests. We did
not produce the loud physical discharges, etc., etc. but the transformer did have show tracking of
the insulation system, where you could see there had been streaming electrification activity.
Likewise, the detectors picked up PD activity. We learned a lot from that. But we did give the
transformer full electric test, reproducing what was done when it was new and it passed these
tests without failure.

Question: Was there any thought given to varying the oil flow rate to develop a correlation
between the oil flow rate and the development of charge buildup?

Answer: We did vary the flow rate simply by changing the number of pumps in operation. That
is, by selecting one, two, three and four pumps. We did try to vary the flow rate by valving from
each pump but this is a triple rated transformer so these are big pumps and consequently the
pressure drop through the cooling system is very low. We were actually running over 900 GPM
per pump. We were just not successful in reducing the flow rate per pump. We had to manage
the overall flow by reducing the number of pumps in operation.

Comment: I guess my point was that we have had very, very high-speed FOA systems.

Answer: Well that is covered. Theres a paper on mitigation, but the facts are that with two
possible exceptions, all of the transformers that failed have failed with light load, low
temperature and all pumps running. In other words, they were not controlled with the thermal
control. But you are right that the secret is controlling flow.

Comment: This is after the results from this test. You can modify the pump controls to mitigate
static electrification. We did it and I believe others have done this, too. We took an existing unit,
modified the controls and put some temperature control into it, so all of the pumps can never be
on unless the winding temperature and the oil temperature are above a certain number. In fact,
we did that for all the FOA units. We also added lockouts, so that the operators could not run the
pumps in manual. Even if they do that, its still temperature controlled because we discovered
that they were having problems with the controls on one unit so they let all the pumps run all of
the time in manual. So we also interlocked that.

Comment: I think its very important to know how the phenomenon of static electrification
started. We experienced problems with a South African transformer. There was no monitoring
system on the transformer. The people in the station heard some noise. When we make some
measurement with just the pumps running and detected this charge. We checked the gas analysis
results from past oil samples and discovered that the gas analysis from three months prior to the
event showed an increase of gases. So we are studying how to avoid discharge in a quite dead

5-17
Panel Session: Are We There Yet?

transformer. We must understand how the phenomenon appears to be able to avoid it. Not only
to avoid discharge when the transformer is injured. In regards to model experiments, Im very
surprised to see how many model experiments we have with no apparent discharge activity. It is
very surprising that in operating transformers we have a lot of discharge activities including very
loud discharges. So load that you can hear of them a hundred meters away from the transformer.
Its crazy not to have any discharges in model.

Comment: We have in Japan already reproduced the static discharge in experiments. And we
have already identified maybe three or four types of static discharge in transformer oil.

Paper: Ramapo Tests: Results from Streaming Current Monitoring Using a


Nilsson Plate and an Absolute Charge Sensor, by Stig Nilsson, Markus Zahn and
Stan Lindgren

Comment: Maybe just a little more information. The absolute charge sensor measures the charge
density locally at the point of the probe. So it gives you sort of a local measurement. The Nilsson
plate gives you some sort of average current over a larger area and that current has got multiple
components to it. Part of it is the convection, the streaming current of net charge in the flow
being convected along by the pumped liquid. But thats not all of it. Theres an ohmic current,
due to the fact that the charge creates an electric field that creates and ohmic current somewhere
else. There is also an oil flow by the Nilsson plate itself. And that undergoes some streaming
electrification. In regards to the bottom plate, there might be charges stripped off the bottom
plate. There are also diffusion currents, because you have a charge density gradient and there is
some diffusion. So youve got numerous mechanisms for measuring current. The Nilsson plate is
records some sort of average of all these currents. But you cannot expect that that Nilsson current
to exactly equal the charge density times the velocity times the area because there are these other
components also.

Paper: Dismantling of Ramapo Transformer by Harold Moore

Question: First, Id like to know about any tracking in the bottom of the Ramapo transformer.
This seems to be the well-known wormhole pattern. But I wonder about the bushing shield, or
bushing barrier, which seems to be a bit different discharge pattern. Or is it also the same? Were
there any wormholes at the top?

Answer: There were some but they were fairly light. But remember, we didnt have the power.
It was a very weak power source. We did have some wormholes in the barriers but most were on
the surface. Now, again, this rectangular barrier system of the Ramapo unit is in a minority of the
ones that we now had static electrification problems. The others had a cylindrical barrier, but it
was a very similar type discharge. On one of the walls, where it really progressed over to the
neutral, there are some wormholes, but not a great amount.

Question: I suppose that the partial discharge pattern is slightly different for bottom insulation
discharges and upper stage discharges.

5-18
Panel Session: Are We There Yet?

Answer: Oh yes. The bottom discharge is caused by the local turbulence. We had some
wormholes but we did not have the degree of discharge nor the severity. In the other two
Ramapo units, the bottom washers were literally eaten up. In these most of the discharges were
interior to the surface of the wormhole.

Question: Why was there not the same type of damage in this unit?

Answer: I remember a transformer a long time ago where we found almost nothing at the bottom
and a failure at the top. I believe that the reason is that were separating charges. Were getting a
so-called cloud of charge, creating a potential gradient around the bushing, which could lead to a
discharge. This relates back to problems in other countries.

Comment: From my experience, we observed some type of wormhole patterns at the lower part
of the insulation structure but we didnt find any upper insulation tracks. Therefore, this
information is very interesting.

Comment: Well, in all weve dismantled, at least the ones Ive been involved in starting with
Palo Verde, at the top of the Palo Verde unit, we found one small little discharge, about three
inches long. We never knew whether it was there or secondary to the flashover. In the three
ConEd units, including Ramapo, there was no indication of discharging at the top in the two field
failures. In this one, we found one place. The discharging we found on the winding insulation
was actually connected to this. There was none of the other kind of discharge.

Question: And the other question Id like to ask, just clarify my understanding is if the barrier
near the bushing shield in the Ramapo transformer is a special design or general?

Answer: It was used by Westinghouse and other shell-form manufacturers in this country until
about 1974 or 75. At that time, Westinghouse went to the circular, cylindrical barrier. The other
manufacturer basically stayed with the type used in Ramapo. In later versions there was no
barrier. It had the coated pressboard on the shield. But there have been problems in all of them.
So it was not the barrier system that was the problem.

Question: So the question had just come up as to why there might be some partial discharges at
the top of the transformer. Can I offer another scenario, which is that the tracking at the lower
part of the transformer creates some gassing. This gas rises up to the top and that creates some
partial discharges at the top. Would you agree with that?

Answer: I think I created a lot of misunderstanding after Palo Verde with that suggestion. I think
its two different things. Depending on how we condition the transformer during these tests, we
could have high discharges at the top and very low at the bottom. So discharges at the top do not
appear to be connected with discharging at the bottom. And then the other thing of blowing the
oil I wish that some of you would explain to me. The velocity of the oil was certainly not high
enough to move bubbles. So I think these are two different things. Now, from an ego standpoint,
Id like to say youre right because thats what I may have led too many people to believe after
Palo Verde.

5-19
Panel Session: Are We There Yet?

Question: I am kind of curious about the small amount of winding damage that you found. Was
there any impedance shift of the winding? Is this a way to detect that these things are manifesting
themselves in the winding groups?

Answer: No, there was not enough damage for an impedance shift.

Question: How about a very, very subtle shift? On units that are still in service, is this a direction
we could go to detect whether were having winding damage if it indeed is manifesting itself in
the way that you showed over the course of ten, 15, 20 years.

Answer: I dont think you could do it with impedance monitoring. When a unit fails, the
dimensional change is almost non-existent. You will need some dimension change to get an
impedance change. You fellows are working on some things that I believe is a better approach
for detecting it. The third known failure was a transformer that flashed over much the same as
this one. This was a failure of a crucial auto-transformer for the period of the year when they just
could not do without it. Knowing there was a risk, a crew was sent out and repaired the damage
on site. Fortunately, where the arc went into the low voltage, the coil is accessible. They were
able to repair this tiny little puncture hole. They were also able to remove the burning on barriers
and to repair the lead. That was in 1984. That autos still cooking. They knew there was risk.
The only point in bringing it up is, with this minimal damage, that there probably was some
damage at the bottom, which has not caused a problem. Could it at some point in the future?
Sure, theres a probability.

Question: Why did the pump stop? My understanding was that the transformer was not
de-energized or did it trip off-line?

Answer: No, it did not trip off-line. We stopped the testing. In the Ramapo project, with all of
these discharges, we didnt have enough current to trip the unit. Its probably because the source
was so weak so we didnt have enough energy in the discharge to cause any protection to
operate.

Question: So the discharge self-extinguished?

Answer: Yes. The one photo we dont have is the damage in that neutral lead. It was so tiny so I
would believe that of this entire crowd, probably only half of you could find it. It was pretty low
energy.

Question: I just wanted to clarify something I thought you said. On the two Ramapo units, which
failed, did you see any evidence of the failure or any discharges at the bushing shield at the top
end?

Answer: They were blown apart so badly we couldnt have seen any. There was a flashover from
one of the shields. In the first of these failures we also had some fire. The damage was so great
that we couldnt have seen it. In a couple of the others that we described, there were big holes in
the shield.

5-20
Panel Session: Are We There Yet?

Session: Modeling and Prediction of Static Electrification

Paper: Static Electrification at Solid/Liquid Interfaces of Power Transformers,


G. Praxl and K. Neuner

Comment: Like St. Paul on the Damascus Road, I think Ive just had a conversion. There was a
transformer failure in Europe that was announced at a CIGRE meeting. It was put down to the
large particle content in the transformer. I never believed this explanation because we have tried
in the lab to create this kind of failure by dumping enormous amounts of particles into the oil and
it doesnt have any effect on electrification. However, from what youre saying, I think if that
particle content was due to sludge, which is described in your paper, due to hydrocarbon
oxidation, then the interfacial tension would have been changed. So the particles may well just
have been an indication but not the source of the electrification.

Answer: Yes, of course. That is correct.

Question: I find your Figure 6 in your paper interesting because you show the leakage current at
four different positions in your model, of which three of the currents are positive and the fourth
one starts off negative and then reverses sign to positive. Can you give me some insight, why it
reverses?

Answer: Its more for the principle. Its not clear when we have a positive or to some
components a negative charging current.

Question: But it is the same at the same time, its just a different position?

Answer: Yes.

Question: Strange effects if its not the same polarity everywhere. That it in some regions it is
positive and in other regions negative.

Answer: It could be related to some additional additives in which case some polarity may be
changed. But what we are looking at are very large molecules. It is not one in the oil or in the
paper. Is not a homogeneous molecule like hydrogen so it could react very, very differently.

Question: In your paper you presented a table with a lot of parameters and you classified these
parameters. Its very interesting to see the correlation but I was surprised to find the quality of
the nature of the oil at the end of the table. It this not a large correlation because from our
experience with our transformer in South Africa the quality of oil appears to be very important.

Answer: Well, of course. I agree. Oil works at very, very different places. It works with
conductivity, permittivity, with viscosity and so on.

Comment: Because we have transformer with the same design, the same turbulence, everything
similar. One bank is OK and the other fails. The only difference is the oil. So that the reason for
my question.

5-21
Panel Session: Are We There Yet?

Answer: Of course. Paper is mostly the same and the variations are found in the oil. But there are
different characteristics of the oil, which are involved in the electrification process. The
plus/minus shown in Table 2 means that the oil could work to increase or decrease the
electrification activity. It relates to a lot of other additional parameters.

Paper: Study on Increase of Electrostatic Charging Tendency of Insulating Oil


and the Diagnosis on Streaming Electrification, by Motoo Tsuchie, Noboru
Hosokawa, Susumu Isaka, Hiroshi Miyao, Toshitsugu Ishii, Shunichi Kobayashi,
Takayuki Kobayashi, Takahiro Ono, Hitoshi Okubo

Question: I was involved in ECT for awhile. There are three or four methods for measuring ECT
so we should be careful looking at these numbers. You are saying that we dont have a
standardized method for measuring ECT. So I need to ask you, what your method for
measurement of ECT is?

Answer: The ECT measurement method is universal meaning the ministatic tester.

Question: In your older 1980 publications you showed a flow system with crape paper so I just
wanted to make sure you are using the same apparatus we are using.

Answer: Its a question for tomorrow morning when there will be a presentation about the
ministatic tester.

Comment: In your Figures 6 and 8, you show ECT after adding BTA. My explanation for how
the BTA helped is this: In aged oil you have copper ions. Because there is copper in
transformers, there should be copper ions. The primary function of BTA is to combine with
copper to reduce the copper ions. Therefore, I would expect the charging tendency to come
down. Most of the problems discussed here are related to new oil, which contains very few
copper ions. We still had very high ECT readings. In our lab we did add BTA to the new oil with
high contaminant content, and it had no effect because there were no copper ions. So I can
understand how BTA can be useful in an aged transformer. However, even there, Im not very
sure because you showed the dissipation factor going up, which means that the DC conductivity
is going up and the leakage will then also be higher. So it may not be a big problem with aged
oil. So we are having different situations here. I just wanted to point that out.

Answer: We show that BTA both affects new and aged oil.

Comment: I did not find that to be the case for us. We added BTA to our oil with ECT value
equal to 600. The ECT did not come done.

Question: I was very happy to see this paper because we can test a new transformer in the factory
with a certain oil, design, flow rates etc. Assume that in the factory test we will not find any
problem with this transformer. Then we put it in the field. The ECT test method may give us the
tool to ensure that the transformer is charged with oil thats equivalent to what we had in the
factory. And then 10, 20, or 30 years passes and the transformer oil ages. If we dont pay
attention, the transformer could fail from static electrification because of the aging of the system.
Even if we start out with a very low charging tendency oil, how does the system change with

5-22
Panel Session: Are We There Yet?

age? Because its not just the aging of the oil, its the aging of the oil in the transformer with all
the material interactions taking place. This paper should be subject to study in detail to see if it is
not pointing the way to a useful long term monitoring method. My specific question is, when you
have reached 500 pico-coulombs per milliliter and .5 percent dissipation factor, will you just add
BTA or will you change the oil or reprocess the oil? What is the policy in Japan on that?

Answer: We adopted the 500 pico-coulombs per milliliter criteria in 1998.

Paper: The Electrostatic Charging Tendency of In-Service Oils and the


Evaluation Method of the Leakage Current from the Transformer Winding, by
Hiroshi Miyao, E. Mori, S. Isaka, M. Tsuchie, K. Takamoro, S. Kobayashi,
T. Koybayashi, T. Ono and H. Okubo

Comment: Im glad to see some new data coming out of these things. One comment is the
ECT you found that ECT times the flow rate gives a better correlation with the leakage
current. Its not surprising, because ECT itself is obtained by measuring the leakage current from
the electrometer divided by the flow rate. You also mentioned that for aged oil there was a
relationship between copper content and the ECT, which is to be expected, because in aged oil
the copper ions are responsible for the oil degradation whereas in the new oil you can have
copper particles. These copper metallic particles dont do anything. You can measure the copper
particle contents just as you can measure the iron content etc. but they dont participate in the
electrification process.

Question: Could you explain exactly what you mean by flow rate. In a three-phase unit, is that
the total oil flow rate divided by three, or is it just the flow going through the coil and core
structure, or what is it?

Answer: In this case, we used the total flow rate. The measured leakage current is only for the
high voltage windings. Since the flow rate of the high voltage windings is proportional to the
total flow rate we used the total flow rate.

Question: So you have to know what proportions of fluid that is going through the high voltage
windings in order to interpret your numbers? Is that correct?

Answer: We used the total flow rate divided into one leg.

Question: So the total for a three-phase unit, it would be the total flow rate divided by three.

Answer: Thats right.

5-23
Panel Session: Are We There Yet?

Paper: Flow Electrification Measurements of Transformer Insulation, by Markus


Zahn, Andrew P. Washabaugh and Darrell E. Schlicker

Question: What was the ECT of the oil?

Answer: We dont measure ECT. We measure the charge density directly. So the maximum
charge in the oil in the experiments described here was maybe around 35 micro-coulombs
per cubic meter. And thats in the volume. At the interface we can calculate, but not measure the
charges. Its milli-coulombs per cubic meter. And then it dies off very, very quickly within the
double layer and gets mixed in the volume. And basically the reason that it dies off is because
over the double layer, the milli-coulombs are distributed over maybe 20 microns whereas, out in
the volume, all that charge is distributed over maybe two centimeters. So the total charge is the
same but the density of it goes down because the volume over which the charge is spread is
much larger.

Comment: Well, even though you mentioned the leakage current and all that stuff, I think as a
benchmark, I believe that you need an ECT value for the oil.

Answer: Its true. We did measure it at the beginning and the end of the various experiments. But
the reason I said we didnt use it is, because we couldnt measure it continuously because we are
rotating the cylinders and things like that. But we did measure the ECT although I just dont have
the numbers at my fingertips. But generally it would vary on the order from 10 to maybe
100 micro-coulombs per cubic meter.

Comment: If you had put that value in the paper, it would have been meaningful to some of us.

Question: I would like to ask you that in equivalency between the transformer phenomena and
your Qouette charge apparatus. In case of transformer electrification, we have the relaxation
length, which is the product of the relaxation time and the oil velocity. And in some case the
relationship between the relaxation lengths and the total lengths of the oil flow path inside the
transformer windings. In your Qouette charge system I suppose the relaxation length is infinity
because the generated positive charge influences the next generation of charge.

Answer: It is a re-entrant flow, is the phrase I used. It closed on itself.

Question: Then in the ECT measurement apparatus used by Dr. Oommen, using a membrane or
filter, only the charge generation is measured. But in your case, both generation and relaxation, is
measured. Do you have any comment on this problem?

Answer: Well, we think this is the prime advantage of the Qouette facility. In terms of modeling,
its one-dimensional. The variations are only in the radial direction whereas in a pipe flow, you
have variations across the diameter of the pipe and along the length of the pipe. We thought this
one-dimensional geometry offered great advantages both with respect to modeling as well as in
trying to understand the experiments since it is only a one dimensional variation. So that was one
of the decisions we made very early, that a re-entrant flow was better than a pipe flow where we
had to consider the length. Although I have other measurements with flow in metal pipes so
weve tried to explore different manifestations. But I prefer the Qouette facility because its so
easy to model and understand the results.

5-24
Panel Session: Are We There Yet?

Question: Then I suppose, if we cancel the effect of the relaxation or leakage from the resistivity
data, then your data will be in good agreement with Dr. Oommens ECT data.

Answer: Yes, I would say the quantities of charge are comparable to what wed expect for
generally new oil and weve done measurements with aged oil. Weve done some measurement
with TU oil and also with the Ramapo oil. But whats different is relaxation in space in a pipe
flow, becomes a relaxation in time with the re-entrant flow. That is, if you had some charge due
to rotation and we suddenly stopped the rotation, the charge would decay exponentially with
time whereas in a pipe flow it would decay exponentially with distance. So there is a one-to-one
correspondence between relaxation lengths and relaxation time.

Question: In your discussion on the effect of AC field, you showed in your diagram, how the
charge comes from the wall, intersects with the turbulent layer and therefore would be entrained
within your turbulent flow.

Answer: Thats the model.

Question: My question is what about the other side of the turbulent layer? That is, when you
have negative potential on one side and youre pushing negative charge into the flow, then on the
other side you have a positive potential and therefore are pulling negative charge out of the flow.

Answer: That also happens, by the way, from the same charge but on the opposite half cycle. So
our model includes that phenomena but we do assume only one polarity charge coming out. So it
comes out from the inner wall on the plus side of the voltage and on the negative side of the
voltage, it comes out from the other wall. So we essentially multiply by two, if you like although
its not quite a factor of two, because the radius of the outer cylinder is different. So theres some
small geometric correction factors. But we include, in the turbulent core, the charges coming
from the inner and outer walls. You have to include both of those. If you have different walls, for
instance pressboard on one wall and bare metal on the other, the parameters, which are the
injection parameters, the double layer thickness etc. might be different. We try to account for all
that.

Question: Are you assuming then, that youre only going to have one polarity of charge crossing
the interface per se?

Answer: That s our model. Because what our model assumes that the wall selectively absorbs
one polarity charge; lets say minus. Then the double layer can only have positive charge on the
liquid side. I agree that the true world is probably more complicated although you did remind me
of something. I believe that Mike Brubaker in his thesis used the exact same model, which had
been previously published. And he got extremely good fits in his thesis, using this model. I
believe the model is generally a very, very good one. Its got enough independent parameters so
that you can fit your data reasonably well. Actually there are in general just three parameters.

5-25
Panel Session: Are We There Yet?

Paper: Measurement of Oil Charge Densities in a Transformer Model, by George


K. Frimpong, Michael J. Walden and Stan Lindgren

Question: In your plots, especially the earlier plots, you had the quite counterintuitive result that
at four pumps you had higher charge density than at six pumps. Do you have any explanation for
that?

Answer: No but we saw that as well. There are a lot of things that we still cant quite explain.
These results were very repeatable. We did it several times and got pretty much the same. In the
Ramapo unit we had pumps one, two, three and four as in represented in our model. Thats
regular the normal layout. We just added two extra pumps between each set of pumps to bring
the total to six. But, if you look at the way the six pumps are configured, it is possible that there
wasnt enough distance between pumps. With all six pumps running, we may have had a
canceling effect between the pumps.

Question: When you said that you changed the oil and cleaned out the dielectric system, did you
use a vapor phase or any solvent in that process? And, I guess if you didnt, do you believe that,
if you did use a solvent, you would be able to clean the pressboards sufficiently so that you
wouldnt have those residual effects?

Answer: Yes, I believe so. We were pressed for time and money so all we did was to drain the
tank. We then poured some Shell Diala A from the top to try and wash the insulation down a
couple times. After this, we filled with the tank with Nyns oil. If wed used some chemicals we
could probably have taken out most of those contaminants.

Question: So I guess you never did get an electrical breakdown in your test setup?

Answer: No. But we did detect some discharging. Although, it was not severe enough to interfere
with the electricity supply of our building.

Question: One thing that Ive done to add more charge into my system is to put in some filter
paper. That is, an oil filter with paper into the flow path. You can then multiply the charge by
ten. That might be one way to try to really get super high charge levels and hopefully get some
sort of discharge activity. I guess you were also simultaneously measuring the Nilsson plate
currents, but you didnt report them here.

Answer: That is true, but we wanted to stay as close to reality in a transformer as possible. That
is why we decided not to add artificial charging sources. We will report about the Nilsson current
measurements in the final EPRI report. There was just not enough room in this paper to cover all
the data.

Question: Were the signatures with temperature the same? That is, when the ACS had a lot
charge did the Nilsson plate have a lot of current?

Answer: The Nilsson plate currents were monotonously increasing as a function of temperature
most of the time. We couldnt really get any correlation between the plate currents and the ACS.

5-26
Panel Session: Are We There Yet?

Comment: I have a benefit of a little bit more data than what is shown in the paper as presented.
This is really very, very interesting data. But there are some questions. For instance, when the
Nilsson plate current goes down, the ACS current goes up. The hump that is present in the ACS
readings is not visible in the Nilsson plate current. There appears to be a lot of things in this test
data that differs from what we saw at Ramapo. There seems to be a correlation between some of
these currents and switching of heaters because there were some steps clearly visible in some of
the currents that appear to be directly correlated to turning on the heaters. Since we have both
positive currents at the bottom and the top of the unit, there might be some re-circulation of
charges. Looking at the traces of the paper, it looks like the streaming current to the winding
goes up with temperature. This leads me to believe that the test setup may describe the
generation function of the static charge more closely than the relaxation component. This data
should be looked at in much more detail. Have you got any comments on these aspects?

Answer: In these tests we really were not looking at the physics, so to speak. We were just trying
to come up with a way of measuring the streaming current or static electrification. At a later date
I hope that we can look more closely at the data, write some papers and delve a little bit more
into the physics of what is going on. So, at this point, we cant really answer these questions.

Question: There are some curious differences between the charge density measurements with the
different plate positions. Is there any explanation for that sort of behavior difference?

Answer: Other than, if you increase the volume then the charge density increases but only for a
small temperature range. We cant really explain everything as of now. We are obviously
stripping charge off the pressboard washers that go down into the lower plenum. In the normal
position, the oil comes into the lower plenum almost at the level where the washers stop. When
we moved the washer up, we change the point of incidence of the oil on the washer tips. As a
result, we are probably not stripping as much charge from the tips of the pressboard washers. So
weve altered the geometry of the lower plenum in some sense.

Comment: The strange thing is that with the high charging tendency oil, the raised case position
data is a little bit higher. Thats strange.

Answer: Yes it is strange. I cant explain everything as of now. We may have time later on to try
to understanding this phenomenon.

Comment: OK. The table you showed for a few seconds is very interesting, because you have
discharge measured by hundreds of gasses. You show hydrogen, only in one column. So with
that you know what you have to do to avoid any discharge.

Answer: Thats true.

5-27
Panel Session: Are We There Yet?

Spontaneous Contribution by Professor Dr. Hitoshi Okubo

I would like to introduce you to some Japanese activities and provide a little bit new data. Some
new activity and some old activity as it were 20 years ago. We have in Japan one committee
called the Electric Technology Resource Association. Almost all utilities and almost all
transformer manufacturers in Japan joined this association and we investigated everything about
transformer maintenance. Not only for oil immersed transformers but also for gas insulated
transformer. We had published a report like this here.

The work includes gas-in-oil analysis and other diagnostic methods for transformers including
gas insulated transformers. In this association we investigated the streaming electrification
phenomena in transformers, too.

5-28
Panel Session: Are We There Yet?

Process of SE
Charge transfer
Charge separation
Charge relaxation
Influence of Influence of operating
properties of conditions and structure
insulating oil of transformer
Moisture content of oil Oil temperature
Ionic compounds Flow rate of oil
ECT of oil Flow conditions of oil
Resistivity of oil SE of each part in Insulator
Dissipation factor of the transformer

Increase of local DC potential


Diagnosis by in transformer
insulating oil Increase of negative DC potential
properties by increase of negative charge
generated in the lower part of the
transformer
Increase of positive DC potential Diagnosis by
by increase of positive space leakage current
charge in the upper part of the from windings
transformer

Excess of local DC
electric field over
breakdown electric field
of oil Diagnosis by
Generation of
DC partial dissolved gas
discharge analysis or
acoustic
Damage of insulation measurement
Phase of DC
Amount and distribution
of accumulated charge
Insulation structure
Trigger of
AC
breakdown

Flow of Streaming Electrification from Process to AC Breakdown

5-29
Panel Session: Are We There Yet?

Data from all over the world, including of course the United States, Europe and other countries
was investigated. We added information about insulating oils and design criteria for the
transformers etc. As is shown in the figure above, we investigated everything from charging to
breakdown including AC breakdowns in transformers. Yesterday Mr. Miyao and Mr. Tsuchie
reported on parts of this investigation. And today Mr. Isaka will introduce to you other results of
this investigation. And, finally, as shown in the Table below, we have developed so-called
criteria or guidelines for maintenance of the oil immersed transformer to protect against static
electrification.

Table Showing Maintenance Control Criteria

Measurement Time ECT Power (Reference Value)


Time Interval pC/ml Factor Leakage Current
A
o 1
%, 80 C
2 3
Core form Shell form
Installation After oil filling Below 200 Below 0.3 Below Below 1.75
Measurement 0.3 ~ 1.5
Periodic 1 time/ 1 ~ 3 Below 500 Below 0.5
Measurement years
Leakage Current When either Below Below 1.75
Measurement tan or ECT 0.3 ~ 1.5
exceeds criteria
Temporary 1 time/ 1 week Below 500 Below 0.5 Below Below 1.75
4
Measurement ~ 6 months 0.3 ~ 1.5
Trace Investigation Below 500 Below 0.5
1
Ordinary temperature, no applied voltage, rated flow rate
2
Per leg of winding
3
Per one coil group
4
Details are determined in consultation between utilities and manufacturers

This takes into account the stage of the transformer operation, the construction or installation,
and periodical measurements and so on. It includes the measurement criteria or power factor and
leakage current levels of transformer. It is not only experiments data but also data from
75 transformers in actual operation. We sampled insulating oils and at the same time we
measured the leaking current in operating transformer.

Now I would first like to introduce the old data but it may still be very interesting because we
have experimentally reproduced static discharge in insulating oil by means of a model. Up to
now, we reproduced four different types of static discharges in experiments. In a core type
transformer the oil comes from the bottom and goes up inside the winding. We generated
discharge traces by using an experimental model. We have used three types of experimental
models. One of the models is called a large model. In that we simulate the windings part. The
transformer oil flows inside of a duct model made by pressboard or insulating paper. And then
comes this observation tank with a flow meter. There are three pumps and a relaxation tank to
relax the charges generated up to and from the pumps. Neutralized oil comes into the model and
flows through model. This model is very big. The oil flow velocity, may be over one or two
meters per second.

5-30
Panel Session: Are We There Yet?

We got discharges without energization the model. There was no applied voltage. We got a
breakdown between the central conductor and ground. We recognized that these discharges
propagated not only this radial direction but also in the direction of the axis to maybe several ten
centimeters or maybe to one meter or so. The inlet part shows tracking or damage from the
breakdown or discharge. This is of course, not an AC breakdown, but a partial discharge caused
by electrostatic charging. The discharge magnitudes are from several thousand to several tens or
hundred thousand pico-coulombs. The number of discharges ranges from 120 to 600. The
hydrogen gas contained in oil increased at first. And, of course, also the C2H2 because of the
discharges in the oil.

Different inlet designs were investigated one of which is shown in the figure below.

Countermeasures for Streaming Electrification

5-31
Panel Session: Are We There Yet?

Some results are shown in the graph below.

Confirmed flow rate at which static


discharges did not occur with
distributed oil inlet structure
Oil Flow Rate ( pu)

Lowest flow rate


at which the static
discharge
occurred with
concentrated oil
inlet structure

Oil Temperature (oC)


Graph Showing Improvement of Static Generation Flow Rate by Changing Oil Inlet
Structure

5-32
Panel Session: Are We There Yet?

The second model system is similar. There is the generating part, pump and relaxation tank. The
generating part has a cylindrical insulator with several millimeter oil ducts. We say that this is a
lead model. We simulate the high voltage lead or low voltage lead from the winding to the
bushing. We also have big observation window where we can observe discharges. We covered
the lead by a transparent PMMA cylinder so we can see discharges from outside the tank. An
inner cylinder is covered by crape paper and there is oil duct made by pressboard. Outside all of
this is the PMMA cylinder. The oil flows very fast between the PMMA cylinder and the crape
paper so discharges occur. One was maybe 40-50 centimeter in oil.

Some of the data is very interesting. We have every minute constantly one maybe two or three
discharges. At this point we started to filter the insulating oil. But this filter intended to remove
ions from transformer oil. Then, just after the start of the filtering, the discharge activity reduced
and eventually stopped. The filtering, the conductivity reduced and so did tan-delta and the
charging ECT reduced, too. I dont know what kind of ions or impurities that we removed.

The third type of discharge we had is just in oil space. This is like the top space above the
transformer winding. We have incorporated Teflon insulating tape in this space. In this oil space
we created discharges from the positive charge cloud. We also triggered AC breakdowns from
the static discharge.

Question: In your model when you did create the flashover, several of them, at the outlet of the
duct, were there any bubbles present prior to the flashover?

Answer: We did not see any. There could have been bubbles but we could not look for them at
the time.

Comment: I think this goes to help explain some of the theories that weve heard as to whether or
not bubbles create a void in a flashover. I think youve proved that it does not. Thank you very
much.

Paper: Streaming Electrification Dynamics - Duct Modeling and Analysis, by


J. Keith Nelson and J. A. Palmer

Question: The sampling intervals we had at the Ramapo unit, was one sample every two minutes
for the plate currents and the ACS. What would the conclusions be if you extended your model
to run for, say, a few hours instead of a few ten seconds?

Answer: Because of the dimensional scale being very different from the time scale, this
particular model, if you ran it for a few hours would give you the same steady state performance
as you get in a few seconds. And is as the streaming current falls off in the matter of a few
seconds. The time scale you see is a total of ten seconds. However, because of the intricacies of
bends and changes in direction and this network of ducts that we have in the transformer, the
initial peak is going to be inserted into the next duct, which will enhance it. The enhanced peak
will then be inserted into the next duct etc. What I anticipate that you would see, if you were to
build a network of these short ducts, is a peak a half-hour or an hour after the initiation of the
flow. There would be some kind of peak that would be falling off in a way that you showed in
the leakage current measurements yesterday from the Ramapo experiment. This is, I believe,
contained in this model but on a much larger time scale and a much larger spatial scale.

5-33
Panel Session: Are We There Yet?

Comment: I agree with your model up to a point. When I started my studies of this phenomenon,
I first went to the US Naval Research Lab because they were playing with their jet fuel. They
had discharging problems. The Navy used model compounds to produce charge, which was like
the sulfonate. The Navys solution was that if you add a little bit of sulfonate, you will get a lot
of charging. So, to dissipate the charge, they would increase the concentration of the component
and then, after a certain point, the conductivity would be so high that the charges will relax very
fast. I didnt see your curve going up to a point from which it would be coming down if you keep
on adding the concentration. Thats the only thing I didnt see in your model.

Answer: Thats a good point. In fact, I would expect that as you get to a certain point, you would
also basically saturate your surface. Then an additional increase in the concentration of ions
would no longer increase the wall charge proportionately. And that would start bringing it down
because the conductivity still goes up without an increase of the wall charge. So youre
absolutely right. I didnt include that in my model. But I do believe that it could be explained by
the fact that youre not going to be able to always keep increasing the wall charge. You are going
to reach some point of saturation on the wall, at which point youre no longer going to be
increasing the wall, youll be decreasing your Debye length and therefore youve dropped that
off.

Question: Did you consider the effect of the charge accumulation on the surface?

Answer: No, I did not.

Question: Is this the reason for the very short the time?

Answer: Youre correct in that we are looking, basically, at the entrance effect. Within the
entrance region there isnt as much of an accumulation effect as further down, where the charge
starts relaxing out on the insulating surfaces. I was basically assuming a metal duct with a
constant potential along the wall. You are correct in that observation.

Question: Do you believe that in the near future it will be possible represent a full transformer?

Answer: I would very much like to do so if I can find the right sponsorship.

Paper: Investigation for Standardization of Electrostatic Charging Tendency


Measurement of Transformer Oil in Japan, by Susumu Isaka, Masama Ikeda,
Hiroshi Miyao, Motoo Tsuchie, Toshitsugu Ishii, Shunichi Kobayashi, Takayuki
Kobayashi, Takahiro Ono and Hitoshi Okubo

Question: Ive got question and one comment. In Figure 7 you show the temperature dependence
of ECT. It is a nice uniform curve from ten degrees to fifty degrees. Could you give us a bit more
detail of how you got that data?

Answer: Previous field data has been used; i.e. data as shown in Figure 6. The correction factors
plotted in Figure7 are derived from the ratio of the regression value at 20 degrees C as calculated
by each equation (shown below the figures) divided by values from the measured data in
Figure 6.

5-34
Panel Session: Are We There Yet?

Question: So this not a measurement with the ministatic from ten degrees to fifty degrees?

Answer: The measurements were made at oil temperature between about ten to fifty degrees and
converted to the ratio at standard temperature of twenty degrees.

Question: I want to understand whether this is experimental data or whether this is theoretical
data.

Answer: It is experimental data; 50 degrees should be 10 to 50 degrees.

Comment: You say that relaxation is not to be considered in the ministatic tester but in
93 Dr. Professor Praxl published a CIGRE paper which shows that the conductivity is very
important. Taylor based that on work in 1974. We have certainly found as the conductivity of the
oil increases the efficiency of the ministatic decreases. In fact, above about 0.3 percent power
factor you cannot measure using the ministatic because you dont get any charge. Do you have a
comment on that?

Answer: No.

Question: I just wanted to respond to Dan Crofts comment about the standardization. In CIGRE
we developed a series of models for determining the ECT of oil. In the end those of us who are
responsible for the transformers decided to put those models into the dustbin and we ended up
with TVs ministatic tester. In regards to standardization, you can argue all day in the ASTM and
other standard organizations. But, from an engineering point of view, all of us require a standard
procedure to enable us to have a language for conversation on these topics. And for that purpose,
in the UK the British standard now has an ECT requirement for oil in order to differentiate one
oil from another when we purchase oils. The tester specified in the standard is in fact the
ministatic tester. We in the UK have tried, like TV has tried with ASTM, to get this standard
adopted by IEC as a revision of standard 296 but we failed to get this accepted. The reason was
that nobodys convinced that (a) its necessary and (b) that this is the right system. Were hoping
very much that venues like this will encourage people to press your representatives to make
forward progress. It may be important to have a truly accurate test method but if we havent got
that, we have to use the next best thing. So far were getting consistent results from the ministatic
tester. I can thoroughly recommend that we use that for the present.

Session: Operating Experience of Transformers

Paper: A Static Electrification Transformer Survey, J. Keith Nelson


Question: I would like to ask what is the definition for your electrostatic transformer. Does this
mean a failure, breakdown in operating condition on site or do you have any experience
recognizing the problem before breakdown?

Answer: I defined incident as being either a failure of a unit which was demonstrated,
unambiguously, to be due to static electrification or an incident. There have been some incidents
where pumps have been run when the transformer has not been energized with pinging sounds
being heard. As far as Im concerned, thats an electrification incident too. I didnt require
failure. In many cases when a failure did occur, of course, the evidence often was destroyed. So
its one or the other.

5-35
Panel Session: Are We There Yet?

Comment: I agree with your conclusion that there are probably far more electrification failures
than have been recognized in the industry.

Answer: Yes. That may be. I have to say that the units that are in the database are only those
where either the collective judgment is that they are static electrification failure or that the
evidence points to it.

Comment: A transformer guy that tested transformers in the factory can look at wormholes and
say, Well, thats a wet transformer. Thats what you see in a transformer thats excessively wet
and fails on test. So then you can speculate that you can get into a situation where your
transformer is too wet, and you step up the drying process. By making the transformer dryer and
dryer and you increase the problem of static electrification rather than reduce it. Its one of those
ironic things.

Comments: Ive heard of cases on GSU transformers that are thumpers. When they had a drop
in the load, theyd have this strange thumping that would go on for awhile. Weve had
pingers and bangers and this is about thumpers.

Comment: I can comment on the wormholes. We have, actually, artificially grown wormholes in
the lab. By having a divergent field that pulls moisture into pressboard in a preferential region
and then putting power frequency across that wet area. You can actually artificially grow a
wormhole. So I am convinced that wormholes, although they may have to do with electrification,
because the static stresses may well cause moisture migration locally. Theres no doubt in my
mind they are driven by the 60- Hertz voltages. Or 50- Hertz as the case may be.

Paper: A Static Electrification Failure at the Navajo Generating Station, by


Thomas Lundquist

Question: I couldnt see from the photo of the corona shield on the high voltage bushing if it had
an epoxy coating on it. Do you remember?

Answer: It doesnt look like it did and Mr. Moore said it did not.

Comment: Im just suspicious every time I see a flashover to bushing, because of the Ramapo
experience. In that transformer we had static discharges appearing at the high voltage bushing. In
the Astoria Compact HVdc station we had problems with the bushing shields, too. Roy Nakata,
the GE Project Manager, reported that he sat on top of the reactor and that he could turn on and
off the discharging at the top of the reactor by starting and stopping the pumps. The solution in
that case was to increase the insulation on the bushing shield. If we have static charging in a unit,
we could get a flashover started from an area with a highly nonuniform, which exists at the
bottom end of the bushing. I dont know what the typical fields are in this region. However,
where we have a nonuniform field and we get a small discharge, this can create a streamer into
the oil-insulated system. Also, I would expect that the dc stress would build up across the epoxy
layer at the busing shied if such a coating is used. This could lead to a pitting of the surface as we
saw in the Ramapo unit as a result of breakdowns of the epoxy, which could be the origin of the
flashover.

5-36
Panel Session: Are We There Yet?

Question: We bought a replacement transformer. Im trying to recall, outside of the thermal


problem that we dealt with, what changes were made in the insulation system from the time this
unit that failed was built.

Answer: The biggest change was to change the cooling system and the pump. The cooler was
changed to a design with a turbulator. The others did not have turbulators, which was the reason
for such big pumps. Also reduced flow pumps at the transition, where we have our jacuzzi effect,
were selected. The insulation was stepped to reduce that jacuzzi effect. The high-voltage bushing
shield was designed with similar stresses. However, I believe it had molded pressboard on the
bottom shield of the high voltage bushing. Those were the basic changes. In regards to the
comment about the bushing stresses it is hard to tell from this picture because you can not see the
actual shape. The shield was so damaged by the failure. However, the stresses are very low
considering that it had a quarter of an inch of high-density crape paper tape and two polished
copper toroids. The field was plotted to get information about the stress in the entire region,
including where the lead comes up. The stresses into the oil in the region of the lead as well as
the shield are very low.

Paper: Whitpain #1C Transformer Failure, by David F. Goodwin

Comment: If you compare your situation to Ramapo, we know that when we cool down a
transformer we get an increasing electrification. You were operating this transformer in a cooling
down mode so that electrification activity could be expected to increase. Then youre switching
voltage on and off. We never did any switching of the Ramapo transformer when we had the
transformer exposed to a lot of static electrification stress. But if you have a dc bias of the
dielectric system from the static electrification field and you add switching surge stresses nobody
can predict what the actual withstand level of the dielectric system really is. Conceivably there
might have been enough static charges to make the transformer no longer able to withstand the
switching stresses. This is pure speculation. However, it would be a perfect crime because we
could not see the evidence after the failure.

Answer: Thats correct.

Question: Just a simple question. What was the type of oil preservation that you had?

Answer: It was a nitrogen blanket.

Paper: Static Electrification Failure of MPT at Powerton Generating Station, by


Raymond F. Cameron and Christopher P. Stefanski

Comment: Critically important data came from this inspection. Its not obvious from the slide of
the coolers, but the number of pumps or coolers entering each phase was quite different. The
phase that failed was exactly as described. For the other two, it was definitely a function of the
number of pumps, or the oil entering the bottom of the phase. One end had the most pumps.
Although it didnt fail, it had by far the most severe static electrification evidence. The other had
some evidence, but it had the fewest number of pumps entering. So the static electrification
evidence was a function of the number of pumps entering each phase. In my opinion, this

5-37
Panel Session: Are We There Yet?

transformer failed due to gas bubbles. There was a definite carbon path from the area that Ray
described region up to the failure region. So I dont think theres any question that bubbles were
rising along this path. One of the ConEd units had the same evidence. So I think failures can
happen either at the top or bottom. But this one definitely was traceable to the carbon path from
the bottom to the top.

Comment: In regards to the coolers, on one end of the transformer there are three coolers coming
in, but there are only two entries into the tank at that point. Youve got three coolers funneling
down, so on those two entry points you actually have 150 percent of the total volume of one
cooler. So it probably created more turbulence in that region.

Answer: Correct.

Question: Did we not put the coolers in three stages?

Answer: Not on this particular transformer. We did that in another station, which has a 600 MVA
GSU, but it has a generator circuit breaker so it is energized at all times. Whenever the generator
unit in service or out of service, the transformer is still energized from the 345 kV system.

Question: A question related to these coolers. You had mentioned on our transformer that we
went to a high turbulent cooler. But the high turbulent cooler apparently allows for a drop of the
flow rate. Is there any possibility that that can be used as a retrofit method to reduce the oil flow
rate through the transformer or are there other design considerations that have to be taken into
account before you can do something like that?

Answer: It is definitely a possibility. The standard cooler with the turbulators, however, is wider
so it would be difficult to retrofit without other modifications. However, there are other coolers
available, with turbulators that you can use to reduce the flow or you can use a reduced flow
pump.

Comment: I would like to make one comment on carbon paths. In investigating a failure of a
transformer, once you do have an arc inside a transformer, you do have carbon. Then, if your
pumps are operated from an external power source you can have carbon distributed throughout
the winding. If your pumps are operated and de-energized at the moment your transformer fails,
then you dont have that as a source of the carbon.

Answer: Yes, in this particular case, the pumps are interlocked with the field breaker so that
when we did have the failure and the unit tripped, the pump shut down.

Paper: Investigation of a Failed Westinghouse 500/345 kV 500 MVA Single Phase


Transformer by Don Angell and Kent Venosdel

Question: Ive been dying to ask you this question. How many other transformers do you have
like this?

Answer: Well, Ive got three others that are single phase. That makes up this 1,500 MVA bank.

5-38
Panel Session: Are We There Yet?

Question: If I heard you right, the on-line gas-in-oil analyzer confirmed the failure rather than
detecting a precursor to the failure?

Answer: Thats correct.

Comment: This points out some of the things that have been done through the research project in
trying to develop a precursor to try to indicate some of the activity that goes on and can be used
to give forewarning of failure. That is the reason I wanted to bring it up. On-line gas-in-oil
analyzers alone is not the answer.

Answer: Yes, we agree. One of the things were doing now is going back and looking at our
transformer units. Were going to install partial discharge detectors, probably the Nilsson plate,
and monitor the cooling in three or four different locations on the radiators and the plumbing.
We will also monitor the pumps. Were going to put this thing in a highly monitored state
starting next month.

Question: I read a mention of the seven million series. I just want to make sure that there is
nothing special about the seven million series. When shell form units were built in Sharon they
were given certain serial numbers. And when manufacturing was moved to Munci, the
transformers were given serial numbers in the seven million range. And when we became ABB
that was changed to another number series. So I just want to make sure that we do not do
anything to give a black eye to the seven million series. That is everything that was built in
Munci until we became ABB was a seven million series.

Answer: When the first transformer was shipped out of Munci in 1961 they decided to change
the serial number from the system used in Pennsylvania. It was continued until roughly when
ABB bought the plant. And somebody counted out the number of seven million and, having gone
through the high production years of the 60s and 70s, there was about 2,800 of them.

Session: Monitoring and Management Methods

Paper: Static, the Ramapo Experience and What Can Be Done, by Chatham
Cooke

Question: In regards to the surface discharge signal, high voltage signal tends to have high
frequency oscillation. Could you give us some explanation or interpretation?

Answer: Ive made the simplest model of the physical elements that I could. Basically, Ive
lumped all of the coupling between the high voltage and the neutral into a single capacitance.
Which is not bad, based on actual measurements from the terminals. But each of these windings
Ive lumped as a single impedance. That means that there is both capacitance and inductance, as
well as losses in that impedance. There is also a capacitive coupling from that to the tank. The
windings are physically similar even if some of the details are different. But the basic idea is the
same. The pulse signal that were looking at is sub-microsecond. We are looking for a much
higher frequency component than normally is encountered in the design of a transformer where
you typically dont look at nanosecond response. When you look at these windings from such a
fast rise-time point of view, its actually a traveling wave that propagates on that winding. The

5-39
Panel Session: Are We There Yet?

inner core has very little time to respond to that. But the physical structure does alter the signals.
Most of the discharge is very much a surface current, etc. but there is inductance in the system
and inductance does cause ringing characteristics of the response. When there is a sudden
discharge, you do excite the inductive modes that present in the windings. And thats what we
observe here. We made some measurements where we looked at the transient response, too. We
even injected transient signals and compared the resonant frequencies and found them to be very
similar. Some of them are up around 13 MHz. Some are higher and some lower depending on the
exact locations involved. But it really is because this winding structure is, in fact, a very
complicated RF structure. Its not a simple winding inductance at all.

Question: So the lack of high frequency component in the neutral terminal signal means some
sort of separation?

Answer: Yes, exactly. Depending on the distance traveled, the high frequency can be absorbed
by skin effect losses. If you start a pulse at one end of a long wire it can be quite attenuated by
the time it gets to the other end. Your interpretation is correct. In the cases where there are very
high frequencies clearly visible, the event took place much closer to the high voltage sensor.
There are examples of just the opposite, where there were very clear high frequency signals at
the neutral and much suppressed at both high voltage and the low voltage terminals.

Question: Is it possible to calibrate how much will be the pico-Coulombs of the discharge or
something like this?

Answer: With partial discharges, calibration is a very difficult problem because the event takes
place somewhere inside the tank. What we did was we injected a known pico-Coulombs signal
into each of the bushings and looked at the relative signal strengths. From that we developed a
matrix to interpret the signals. That is, if I see a signal strength of such-and-such a ratio, the
corresponding magnitude is of such a strength. This is of course only an approximation. Its not
exact. But it seems reasonably good. However, I did have a chance to actually inject signals.
When they were finished with the transformer inspection, I went inside with a partial discharge
pulser, connected it to a few locations within the windings, cut open the windings, and looked to
see what I would observe at the outer surfaces. And it was very consistent. I introduced a
500 pico-Coulombs or 50 pico-Coulombs signal and looked at what was observed at the
terminals. However, at least at this point we wouldnt know how to relate if something were
5,000 versus 4,000 pico-Coulombs. That degree or level of resolution in magnitude at this point
is not as important as knowing where it is. Also if it occurred regularly, and the conditions under
which it occurred. But eventually we will learn more. With more experience I hope to get a more
accurate number. But youre absolutely right, it is not an exact number at this point.

Question: You should know that as a part of the calibration of this unit, he had available the
no-load taps of the winding. So he also injected impulses into those taps and studied the
distribution at the terminals. I guess one of the other things, which probably everyone is asking,
is we recreated partial discharges outside the transformer to see the sensitivity of this system to
external interference. Do you have those results?

Answer: I dont have them in a view graph. The bushings on top of the transformer, which are in
the open air, had placed on them, one at a time, corona points. Thus, when the unit was energized
there was significant amount of air corona occurring. So at one series of tests there was a halo of

5-40
Panel Session: Are We There Yet?

corona points on the high voltage bushing while the unit was energized and while we were
measuring. These external corona signals were very clearly distinguishable and notable. And
then, the points were moved over in a different test to the low voltage bushing and again the
signals were very distinguishable.

Question: Suppose you didnt know where they were at? Could you have told me that they were
on the outside of the transformer?

Answer: Yes.

Question: I believe that there was some 60-cycle discharge that was created after the big bang
was acoustically detected. Were you able to differentiate that partial discharge from the
discharge that was resulting from the static electrification?

Answer: Lets take that in two parts. In fact, while the corona points were on the bushings, the
pumps were run simultaneously. So the unit was energized, it was in a temperature range where
there were some internal static events taking place. And the two signatures are clearly the way
we think they should be. They were different and very distinguishable. So the answer is that the
signal that comes from outside definitely looks very different from the signal that comes from
the inside. And if theres a very characteristic single pulse associated with the events, theyre
associated with the corona points from the external area. We had another example to see that. On
one morning, it was after a rain the night before, the bushings were quite wet. And for the first
period of time, we saw small external events as the bushing dried off. So there is a signature that
is distinctive. We were clearly able to say that these are external and these are internal to the unit.
Now, if you see only one event in your life, you had one record in front of you, you could do a
fairly good job of saying I know what this is once we know a reasonable representation of the
electrical circuit. On the other hand, if I can do calibration measurements where I actually inject
a signal, a low voltage signal, pico-Coulombs range signal into the bushing, its the same thing
as if there was a corona point there. So those two signatures were very similar. I could use that as
another way to confirm that the external corona was indeed from an external corona point. Now,
as to the evolution of the discharges within the transformer, as I said, theres probably a
thousand, more than a thousand of these records accumulated over a period of time under many
conditions. And one thing is clear. There was an evolution of the discharges, both in space and
time. They didnt just stay at one place. They did move a lot. And sometimes, after a big bang,
there was clearly a series of other kinds of events that took place. We had a number of periods
where the type of activity was reasonably distinctive and separate from the large bangs or the
cascade events, as I call them. I believe that I classified at least six different kinds of events,
which occurred repeatedly enough so that they became a class. Maybe it was only during one set
of operating conditions, maybe it was early in the test series, but in any case there was a number
of classifications. And some of the events were very similar throughout the tests. We ran into
cascade events the very first time. At first I thought there was something wrong with the
instrumentation, because this was not something that had been reported previously. And then we
came to know them very well and truly they became a very distinctive characteristic associated
with static processes on surfaces.

Question: Could you speak to your explanation physically of whats happening with those
cascade events?

5-41
Panel Session: Are We There Yet?

Answer: Yes. Ive actually simulated them in the lab. Not in oil, but in air. You take a dielectric
surface and spray charges on it. Then if you start a discharge at one end, that discharge will
propagate along the surface. Now it doesnt propagate necessarily, and usually not, as a
continuous process. It takes steps. You might refer to it as a mini set of leader events or
something like this. Its very much the same. And, in fact, you can go to the literature associated
with lightning simulations and find people who have spent a lot of time using the surface
discharge as a means to try to understand the lightning process in air. You can do it in a lab
prettily easily. You dont need to wait for a storm to come by. Ive also simulated it in a deep
volume by implanting charges within a solid dielectric and developing a discharge through that
solid dielectric. And, again, you can see a stepped process as the leaders go out and propagate.
However, there are a lot of unknowns. We dont know what the real charges were on these
surfaces although there is enough commonality between the various experiments to say that this
really isnt so strange under these conditions, where we have very long surfaces. Clearly the
charge propagated in varied fingered structures, indicating that it was a surface process and very
likely a stepped process.

Question: Would it be difficult to envision a cascade event in a transformer where there was not
a static electrification phenomena?

Answer: Yes.

Question: So this is, in fact, a very indicative signature.

Answer: Ive made similar measurements on transformers that do not have static issues, but have
PD locations. In these cases there was some void or some other defect within the winding
structure. Those PD events are never cascades. They are always single events separated by
maybe 20 microseconds. In a case where you have a void theyll come repeatedly.

Paper: Mitigation of Streaming Electrification by Intelligent Control of Oil


Circulation, by J. Keith Nelson and J. A. Palmer

Comment: I guess more of a statement than a question. I guess I went through the presentation
fairly quickly. The decisions we made were done in a period of a few hours. However, things we
did were not taken lightly. I would caution anyone, if you dont have a self-cooled rating on a
transformer, if you start changing around when your manual groups start, or if you start changing
flow rates, these are areas where you can easily create problems. I did not want to be responsible
for doing any of that. ABB did an analysis of our transformer before the power plant upgrade.
One of the suggestions was to reduce the flow rate of the pumps. This is something we decided
not to do. And that was coming from the people that had the design information. For the 20-year-
old, we want to leave them unchanged. The reason we chose to knock the fan out is because I
didnt want to change the flow rate or the dynamics inside the transformer. All I wanted to do
was reduce the Delta-T. I increased the Delta-T when I put the unifins on there. The designers
designed it with a certain Delta-T and a certain flow rate. I kept the flow rate the same but I
changed the Delta-T and I got myself in a little situation in the winter. So all I did was tweak the
Delta T. I dont know if were getting ourselves in a situation where we might have problem
hotspots and some other things in the windings if we begin to vary the flow rates.

5-42
Panel Session: Are We There Yet?

Answer: Thank you. Certainly your concern is a reasonable one. You wouldnt want to be going
way out of bounds. Basically, the intent here is not to modify in any extreme way the flow rates
and so forth. These are subtle changes. It would be largely a matter of how the system was
programmed and what operating temperature youre actually trying to reach. And you can
include things in the logic like anytime we have a voltage on the transformer, we have to have
at least one pump running just to maintain that circulation. All of those rules can be defined at
the outset. Certainly it can be done in collaboration with manufacturers or with the planning
department or the mechanical engineers or whatever.

Panel Session: Are We there Yet?

Paper: Mitigation of Static Electrification Discharges in Power Transformers,


Harold Moore

Question: In terms of heating the oil ahead of time, has any thought been given for GSUs to
integrate the transformer with the entire plant operation. We could provide an external source of
heat to bring that oil temperature up to where its in the correct operating range. We do that on a
turbine generator, we watch hydrogen purity, water conductivity, water purity etc. Why cant we
do that on transformers?

Answer: Its a great idea!

5-43
Panel Session: Are We There Yet?

Comment: We had an 800 MVA, 345 to 22 kV GSU that had been out of service for an extended
period of time. It was at a lignite power plant. The stack collapsed and did a lot of damage to the
plant. In going through the forensic evaluation of why the stack collapsed, etc. it was a long
period of time before the unit was returned to service. We were concerned about this very thing,
on what do we do. We brought out our degassification equipment and simply connected top and
bottom and circulated the oil through a degasser. We degassed the oil and we put it through a
cycle before we brought the unit back on line. Thats one way. Most of us have existing
equipment so we can do something like that. We didnt even have to remove the oil, it was just
circulating.

Answer: That can be particularly easy when youve been field drying because youve got the
equipment there anyway.

Comment: I like to make a comment or two. First, temperature control is possible but the
desirable temperature range varies with pressboard and oil conditions. Unless we know the
criteria or what governs the situation clearly, we dont know to what extent we can correctly
select the temperature control range. I may be OK for a Ramapo type transformer construction
and oil. But I am not sure to what extent we can safely extrapolate from some known conditions.

Answer: Temperature control of the cooling system has been used at least for the 48 years Ive
been fooling around in this country. Thereve been a variety of control methods. Most utilities
depend on hotspot indicators. Theyre certainly not perfect but theyre easy to do as long as
theyre calibrated and as long as the manufacturer has set them properly. In my experience, they
are good enough for this kind of control. Were really not trying to change fifty or sixty years of
practice. All were doing is urging people to use it. So I dont have much concern about damage.
I know of no problems. There are people who use hot oil temperature to bring on cooling, even
though the hotspot temperature may be better used to assess the aging of the transformer.
However, I dont know of any aging problems that have been caused by use of hot oil for cooling
control, either. The only question I have is not from the aspect of damage to the transformer
because there may be a narrow band in real cold climates where if the hotspot is used for cooling
control, the oil temperature will not be up to our magic numbers.

Comment: Maybe moisture works to give a high relaxation over 50 degrees Centigrade. But
50 degrees Centigrade is generalized. For some oil it may be 40 degrees. But if it comes up 70 or
80, that sort of temperature control may be very risky if we dont know the design criteria.

Answer: I didnt listen carefully. We do not control down in the 30-40 degree range. The
temperature control normally at the first stage on the hotspot will be up around 70C or possibly
even 80. So that all of the controllers I know of would cover this wide band. Im not suggesting
you do it at 45 or even 50. So bring on the control has been the practice for years at temperatures
beyond any of these ranges. If you bring on the first phase at 70, the second one would come on
at 75 or 80. Oil temperature is normally 55 to 60 on the first stage and at the second stage around
65 to 70. So I apologize. Im not suggesting we do it in line with these Ramapo temperatures.

Question: In regards to the temperature control range from Ramapo experiment, I agree with
your conclusion. My understanding is its mainly related to the discharge on the top side of the
transformer.

5-44
Panel Session: Are We There Yet?

Answer: It was both.

Comment: I have got sort of an impression that the main concern from the manufacturers point
of view, is that even small discharges at the bottom are not tolerable.

Answer: Everything is at the top. And the higher temperatures, is at the top. Is the difference
between the top and bottom oil temperature the concern?

Comment: But I feel that the mitigation of the top discharge only is not at all tolerable. The main
investigation should be concentrated on the small discharges at the low bottom part of the
transformer.

Answer: Well, we agree and that is the reason for suggesting that we control temperature and get
it up where we dont have any discharges. Although, the data says that we may have a few, but
certainly they should not be gas generating. But, in general, we agree to bring it on so you
prevent these discharges.

Question: Im a little confused about after drying or processing, if we recirculate our oil through
the degasser, does that meet this criteria? Assuming that we raise the oil temperature on the inlet
to at least 50C after processing, does that meet the criteria for energizing the unit?

Answer: Well, youve got to get it up and hold it.

Comment: Well, it will cool down after you stop, of course.

Answer: Thats the problem. Dont let it cool down to ambient and let it sit for a time.

Answer: It should be done a few days before.

Question: Before you energize it?

Answer: Some of the manufacturers instruction books, in the final processing of the unit,
suggest circulating the oil three times, with the temperature at 50 degrees or something like that.
You can just put that part of the procedure in place.

Question: OK but this is within a few days of energizing it. So if I do it within a month or two of
energizing it, then Im not meeting the criteria.

Answer: Youve got to go back and do it again.

Comment: The other thing that the manufacturers tell you do in your instruction manual is to run
the pumps after you do the oil fill. That is a practice, which the field personnel really should
consider changing. Theyve been doing it this way ever since, you know, Hector was a pup, and
they want to turn them all on and run them, because the manufacturer said to do so. And that is
something that we have to change. We either run one pump at a time or something like that.

5-45
Panel Session: Are We There Yet?

Comment: I know that when we vacuum fill or dry out a unit, we run the pumps a few hours
after were finished. And then we bleed the unit.

Question: All the pumps or just one at a time?

Answer: We run about half of them. And then we bleed the unit. So I think you still have to do
that.

Answer: But you dont want to run all of them.

Answer: No.

Answer: Or I dont know.

Answer: I would recommend following this table: One at a time or certainly in three or four
groups.

Comment: But we probably do need to think about somehow changing that practice.

Question: If a transformer is filled with degassed oil and processed, would you from
manufacturing side require running of the pumps to remove gas from them?

Answer: I believe that when we have to run the pump it is a problem. Its not a problem of static
electrification but it is problem of particulate matter. It is to clean the transformer before
energizing. When you filter, you circulate the oil but not through the transformer pumps but
through an external circuit. So you have to run the pump to make the particles move through the
external filter. I think it is a completely different reason.

Answer: I believe that if you have used the degasser and circulated the oil, its not necessary to
do anything more.

Comment: We are talking about having the authority and control to get into these manual groups
and control the procedures that the power plants operate under. Again, ignorance is bliss. Im
sure there are a lot of people in this room, who are not quite sure exactly about the procedure
used to bring a transformer into operation. I told you about one case where they ran for 51 hours,
which Im assuming is probably not the correct thing to do so we jumped right on that. Just on
the opposite side of the spectrum, we energized and back fed the 648 MVA GSU unit and went
three days without any cooling. Nobody ever went through that part of the procedure. We had
temperatures above 100 degrees C.

Comment: I want to make a couple of comments about this oil circulation. We circulate the oil
well for about three volumes, whatever is in the tank, using the rig. We then immediately run the
pumps to circulate the oil through the radiators but you are not just circulating through the
radiators if you have a closed oil system. When you have to reconnect it, and youre circulating
the oil, youre circulating the oil outside the oil box and the coils. This is not absolutely true in a
shell form, but in a core form the only way youre going to get any good circulation of oil in the
coils and the oil box is to run the pumps.

5-46
Panel Session: Are We There Yet?

Comment: I believe that there is one point that has to be hammered in time after time again. And
that is that, for those of you that are familiar with converter transformers, you do not get a DC
distribution, a resistive distribution, of the stress in any unit for quite some time after you
energize the system with DC. And that is what you do when you start the pumps. You put in a
DC stress in the system, and you get the dielectric distribution right from the get-go. You have to
wait for the stress on the capacitive-resistive system to build up. You have to develop the
resistive distribution before you actually get a significant electric stress on the paper dialectic in
ways that are difficult to anticipate. So if you start the pumps and you run them for, say, a time
constant of the paper which can be, at low temperatures, hours, at 20 degrees, it may be an
hour or so, you get two-thirds of the resistive distribution. If you run the pumps for a tenth of the
time constant, you get hardly any resistive distribution and I dont think you have to worry about
the electric stress. So if you tie it to the time constant of the dielectric system, you can run all the
pumps for a while. But thats probably too complicated for an operating procedure. At least this
group should understand these aspects.

Answer: I would but it is too hard to control.

Question: We have a large population of transmission tie-line autotransformers, a lot of which


are made in Muncie. Most of these have two FOA cooler groups with a specified carry; I think,
70 percent of load with one group. Is that kind of arrangement something where we might want
to look at the reconfiguration to three groups or two groups?

Answer: I would. These units have got a total of four coolers, dont they? At least the single
phase units.

Answer: I think so.

Answer: I would seriously consider reconfiguring.

Answer: Particularly auto transformers that have got coolers on them.

Comment: I know at this point of no incidents that weve identified as being related to
electrification. We do have strong procedures that forbid operation of all of the coolers on
manual. We wait and the second group has to be thermally controlled.

Answer: Well, as long as you do it that way, the odds for having problems are certainly
diminished. But over the long haul, the safe thing to do is to reconfigure.

Comment: One older, large transformer was shipped back to the factory for some problem not
related to static electrification. It went through factory drying and went back in service and failed
with all pumps running. I believe that there were one or two others that were field dried. From
one of these they heard discharges and took it out of service. The other one may have failed.
These were not the new but older transformers.

Question: The point I was making yesterday in my presentation was not that there had been no
static electrification failures in the factory, but that there had been none reported. That was in the
CIGRE survey. The question Id like to ask about the recommendation of the pump size is if you
are you still considering using ten or twelve pumps? If youre going to have ten pumps, all at
500 gallons per minute, it seems to me to be a very large amount of oil flowing through the

5-47
Panel Session: Are We There Yet?

transformer. I can understand that you need high flow rate for compact coolers. It really isnt
necessary to go as far down on the hot spot curve as the spot I showed you on my curve
yesterday. From a design point of view, theres no credibility in having to reduce the hot spot
that much. I can appreciate, as I say, that you need a large flow rate for compact coolers but do
you actually need to design the new transformers to have such a large flow rate through the
windings?

Answer: Well, on a triple rated unit, the OA/FA/FOA, it is not necessary to have the big pumps.
What I meant by the 500 is the maximum. On the unit cooler, which you call a compact cooler, it
depends on the turbulator and it depends on the output of the cooler. For most commercial
coolers, if you go below 500 the efficiency goes down. And for some utilities, due to value of
real estate of existing land, you are almost required to use that type cooler. Therefore, youve got
to use a pump of that size.

Answer: Its not uncommon with some manufacturers who use compact coolers to require large
oil flows through those coolers. Some manufacturer, when producing large rated high voltage
transformers, are not actually to direct the oil into the phases; that is into the windings. I know
one manufacturer, which does nothing more than enter the forced oil into the tank. The actual
cooling takes place with what might be describes as a quite natural, thermosiphon action. Theres
no actual forced oil through the windings.

Comment: That is quite a different subject of directed flow versus non-directed flow and the
differences in temperature hotspots that we could keep discussing all day. But that is not what
were here for.

Comment: Weve done a limited amount of testing of transformer cooling systems in our area.
Based on the limited experimental work weve done, we have found that the cooling seems to be
more a function of the cooler design and the surface area of the heat exchanger in contact with
the hot oil and the air flow, than it is with the rate of the oil flows through the cooler itself. In one
instance, we used exactly the same cooler, but we changed the flow rate by a factor of three, and
the amount of energy we were taking out of the oil didnt change. We were removing about
150,000 BTUs of energy out of the oil at a flow rate of 8 gpm and then at 25 gpm. But the BTUs
that were being removed from the oil was the same. So, is the flow rate important for trying to
force large amounts of oil into the bottom reservoir. Or are you just getting a smaller temperature
drop from top to bottom?

Answer: The efficiency of a forced oil cooler is a function of the oil flow. Youre cooling the oil
and not the internal of the transformer. Thats the first objective. Second, you can reduce
hotspots from top to bottom. Now some comments in regards to cooler performance. One that I
am familiar with was rated at 800 gpm. If you pumped 900 gpm, you got less than one- percent
additional heat removal. If you reduced the flow from 800 to 700 you lost ten percent but if you
went from 700 to 600 gpm you lost 50 percent. So you can pump more oil with no effect. It all
depends on the cooler.

Question: In the Ramapo transformer what was the, what maybe you could call the critical
velocity, of the oil in the oil ducts at say 35 degrees Centigrade. What is that the speed of oil that
either gives or doesnt give rise to the static electrification? Is it the same for a core type or is it
different?

5-48
Panel Session: Are We There Yet?

Answer: The answer to the last one is no, no and no. But the average and not the maximum
velocity at the coil was around 0.65 feet per second.

Comment: Pardon me for interjecting, but one aspect with the shell form transformers is that the
failure is in the high-to-low space and not on the coils where the flow is directed. I dont know
that anybody knows what the flow through the high/low space really is. The flow can be through
incidental leakage paths and thats part of the variability. That may not be applicable to a core
form transformer. Im not sufficiently familiar with core form transformers to know whether they
have paths that are not carefully designed. But in the shell form, in the high/low space, in those
areas with large masses of insulation, the flows are not well identified.

Panel Session. From left to right: Harold Moore, Jean Poittevin, Hitoshi Okubo, Tom
Prevost, Ray Cameron, Stan Lindgren and Dan Crofts

Panel Presentation by Ray Cameron

I dont have a formal presentation but I do have a couple comments I just want to mention.

Referring back to Stans comments on the driving force at the power plants for greater reliability
in our main power transformers or generator step-ups, were seeing a greater direction coming
from our insurance companies. Not so much applicable to the transformer failure itself, for which
in most cases we would be self-insured. But as applicable to whatever collateral damage may
occur within the plant on connected equipment. This can be the iso-phase bus, the turbine
generator, or the turbine itself or whatever else may occur during a major through fault on that
equipment. Its more predominant in the nuclear stations. Again, they have greater direction to
keeping those units on line but were also seeing it more in the fossil plants.

5-49
Panel Session: Are We There Yet?

I believe that what we need to do is to make certain that whatever instrumentation and
monitoring equipment that we install is accurate and gives correct direction as to what to do to
the unit. If were saying that a unit needs to be tripped or a transformer taken out of service, if
its at a substation, we need to be fairly on the order of 98, 99 or 100 % certain that there really is
something wrong with that unit. I have seen too many incorrect instrumentation directions inside
generating stations with core monitors and for turbine generators with some partial discharge
monitors. Some early monitors were really influenced by the surrounding environment more than
problems with the equipment itself and we spent days to two weeks trying to find a problem that
wasnt there.

In terms of cooling issues, weve pretty much tried to address that in the generating stations by
not putting all the cooling pumps on. By only operating the first stage cooler initially when the
field breaker is closed and then the second stage coming on as the temperature rises.

I see, in our case, three distinct transformer areas. One is the base loaded GSU at our major
nuclear sites as well as at our more efficient fossil plants. Those units will come online and
theyll stay online continuously for, say, six months, twelve months or maybe even eighteen
months if all goes well. And they will be operating at full load during that entire time period.
Then you also have the GSU at the fossil plants that are load following or cyclic. They may
operate at partial load or they may operate at full load during summer months but there is no
assurance that youre going to have greater than 40 to 50 percent load on the transformer during
the fall, winter or spring months. I also see the 345 to 138 kV autos that we have on our system
as falling into that same category. You close the breakers and whatever load is there, its there.

We also have another situation. About a year and a half ago, we converted our Zion nuclear
plant. We essentially retired it as a nuclear facility and converted both of the 1,120 MW turbine
generators into synchronous condensers. We accelerate these units to synchronous speed using a
variable speed drive connected to one end of the turbine shaft or one end of the generator shaft.
These two units during the summer months are typically running at 300 to 400 MVA. Thats the
total power. We have got two GSUs connected in parallel to the unit. But the loading can
increase instantaneously up to 800 MVA total per unit during transient system conditions. Or,
when the load is cut back, it can come down to and sit at zero MVA for long periods of time.
During the fall, winter and spring months, weve actually seen those units operate anywhere
from minus 300 MVA up to plus 200 to 300 MVA. During the middle of the day its not unusual
to see it sitting at zero MVA. You dont have a whole lot of load there to warm up the oil. And
what weve done in those cases is for the summer months to operate the cooler with the fewest
number of pumps. There are nine total per transformer. Well bring four pumps on initially and
then the second stage comes on thermally.

During the winter months weve cut back down to only two pumps for normal operation and the
second stage coming on thermally with only three pumps. Weve found that to be acceptable. We
see the temperatures sitting in the 40 to 50 to 60 degree C top oil temperature range, and we
havent experienced any problems on those units. But that does offer to us and whoever else who
is considering a similar type of conversion another set of circumstances that they have to address.

5-50
Panel Session: Are We There Yet?

Panel Presentation by Professor Dr. Hitoshi Okubo

I would like to have comment on the operation of existing transformers and static electrification
of transformers at first. For detection of such problems, it is very important to monitor. So the
monitoring of transformers is very important. We, in Japan, have established procedures for
monitoring of transformers and in particular electrification activities.

At first, we check the ECT of the transformer oil. This is a periodic activity. Above a certain
value of ECT, we measure the leaking current of the transformer in operation. We have
developed a flow chart to guide our diagnostic monitoring activities. Most important is the
criterion or guidelines to check the ECT (below 200) or power factor. The guide addresses
leaking current for core type and shell type, for which we have set critical values that vary
depending on the transformer operation. Our data are based on 75 operating transformer and
tremendous amount of experimental data, including the reproduction of static discharge in
models.

In regards to aging of the transformers and transformer oils, the important thing is not, in my
opinion, the total degradation but the initial degradation of the insulating oil. This means the
initial peak of the dissipation factor as a function of the temperature and oxygen content but also
the copper dissolved in oil and so on and so on. To suppress this initial peak, we add the BTA
additive in transformer oil.

In relation to the design of new transformer, I would like to introduce you to one experimental
data set for a core type transformer. You know, the typical data for the discharge area as a
function of the flow rate and temperature. You also have the leakage current. The partial
discharge probability is the inverse of the curve. We are using core type transformer models with
different type of transformer structures; specifically different oil inlet structures. Some have the
concentric oil inlet structure. This means high oil velocities. One is a distributed structure but the
coil winding is the same. In the models we measure the flow rate and temperature. In case of
concentrated oil inlet structure, we have found some discharge areas. But when we introduce
distributed oil inlet structures, this changed. So improved structure for the oil flow can be
important. One important aspect is that we were able to reproduce the static discharge in this
model. There was a relationship between discharge occurrence, or discharge probability and the
improvement of structure. This points out the importance of being able to reproduce static
discharges in transformer models. As was pointed out by Keith Nelson, this is very difficult,
which we also discovered. We have tried, tried, and tried again and it is very difficult to do. But
after we succeeded, we could experimentally produce four types of static discharges in oil. As
was pointed out by Dr. Hosokawa, it is very important to recognize that there are a lot of
different kinds of discharges in transformers. There are discharges in the generating part and in
the accumulating part. So we have to understand the mechanisms leading to partial discharge
initiation. Using this kind of model, we have improved the oil flow pattern and oil flow rate or
local velocity.

5-51
Panel Session: Are We There Yet?

Panel Presentation by Tom Prevost

When Stig asked me to participate in this panel session, I was thinking about things that
Weidmann has done over the years, in regards to participating in static electrification related
work. But in this weeks discussions, a lot of those things have been covered. So I dont want to
go over that. But one of the things that came out to me as I was reading through the Ramapo
report, when Don was kind enough to send it to me, was that I kept coming back to the fact that
there was so much pitting on the shield of the high voltage bushing. And that really got my
interest. Over the years, Weidmann has done a lot of work in high-voltage lead exit systems, and
we do a lot of analysis in that area. So thats kind of what I just want to briefly talk about.

Whats interesting about the bushing shield thats been used as Harold and others have talked
about on many of these units is that there is an epoxy coating surrounded by either paper or
pressboard. Well, crepe paper in most of these types of bushings. If you look at the relative
difference between the conductivity between oil, paper and epoxy, there are orders of magnitude
at ambient temperature. Where pressboard being one, epoxy being ten and oil being one hundred.
So what youre getting is a difference. Since we have talked about time constants, you may
include the permittivity as well. What youre seeing is a different time constant. And also, if you
just look at resistive distribution, youre also looking at a different resistive distribution across
this installation system on the barrier. It would make more sense to me, and I think weve also
discussed and Harold mentioned that in later designs the epoxy coating was not used. It was
simply switched to pure pressboard insulation.

So when Im thinking about what we can do for units in the field we could change out the lower
bushing shield in the field and replace it with simple pressboard insulation. One of the
suggestions that Harold gave earlier was removal. I agree with that. If the neutral lead is placed
where it normally would be placed, the barrier doesnt really improve the strength of the
structure a whole lot because of the big oil space.

One of the other aspects that I noted, are discharges that take place in the upper plenum. Are they
a result of accumulated charge or of gasses being generated? If theres accumulated charge, I
believe changing the insulation arrangement on the top corona shield can help. If gasses are
being generated, one way of strengthening this dielectric structure would then be to put
subsequent barriers around this corona shield. Its a well-known phenomenon that the strength of
an oil gap increases exponentially with distance. So where the stress is high around the high
voltage bushing, you could grade that stress by putting concentric barriers and help to alleviate
any overstress situation. If you look at the design process in the factory, its probably good
enough without these barriers. But now you introduce gas bubbles coming up in this space and
this can help to increase the strength of the gap.

Panel Presentation by Jean Poittevin

Its quite difficult to be the last presenter and to say something about what has been said during
these two days. The presentation by Harold was a very good summary. So I will go very fast on
the things which have been said before. The manufacturers point of view is very important. It is
that if no static discharge has ever occurred in transformers, the static electrification phenomena
would never have been taken into account. That is the reason that in the end we have, as

5-52
Panel Session: Are We There Yet?

manufacturers, to avoid static discharge and, of course, to avoid failures in service. To do that we
need good knowledge of the phenomena. During these two days we have heard a lot about
experiences.
So, to summarize what has been said during these two days, I believe we have heard about some
static electrification experiences. The Ramapo transformer provided a very nice experience
because you have got a lot of information from that. Some other presentations about new failures
shows that the phenomena still exists. My conclusion from the experience is that the cases
presented are very similar. And the main reason, from which we conclude that it is static
electrification, is the comparison with Ramapo. Because I believe that if you present a failure as
the ones, which were presented yesterday, it would be very difficult to say that we had a static
electrification problem. But when you compare these to Ramapo, it becomes more clear.
The study of the phenomena has been presented many times in this session. We have heard about
the interface between solid and liquid. We have been told about the influence of charge
generation, charge separation, charge transfer, relaxation, the influence of ECT of oil, the
measurement of leakage current and streaming electrification. And what is very important that
the discharges have been reproduced in the models. That is very important. That means that we
are very close to understanding the phenomena we are looking for. Therefore, the conclusion of
those presentations is that the main parameters, which have been listed by many authors, and the
phenomena are confirmed. These are oil turbulence, oil velocity, temperature and moisture.
Everybody seems to be in agreement on these parameters.
One more point is that monitoring and management methods have been demonstrated. This is
very important for finding some solution in the future. In the future we should try to find some
solution to decrease the risk of static electrification discharges. But we have not spoken about the
economic point of view, which we cannot ignore. If you change something in the design of a
transformer, you will of course increase the cost of the transformer. Because of the specification
of the transformer has been met before, we have to think carefully about the problem before
changing anything inside of the transformer. For instance, if you change the position of the
winding in the tank, you will change the volume of the tank and you might not be able to
transport the transformer after that. So we have to consider all of the ramifications of the
problem.
And also, sometime a change could decrease the performance of the transformer. If you reduce
the speed of oil, I am sure that in the end, the performance of the transformer will be lower. The
MVA rating of the transformer will be lower. We are never sure that the solution we propose will
be very at a hundred percent effective.
I want to address the proposed solutions for all transformers. Because you have two kinds of
transformers and you have transformer in service and new transformers. So for all transformers,
included in service as well as new one, we can use a monitoring system. This is a very good way
to reduce the risk of static electrification failures. Because all of the presentations made here
show that its possible to detect the phenomena before failure. That is very important. Of course,
its possible also to put some additive in oil, like BTA, to reduce the electrification activity.
These two things can be applied on all transformers. For a transformer in service, the only thing
that can be done in addition to what has been discussed above is to reduce the speed of oil. So if
we reduce the speed the oil, sure it will lower the risk of static electrification but the
performances of the transformer will also decrease.

5-53
Panel Session: Are We There Yet?

For a new transformer we will probably see an increase of the cost. It has been shown that to
reduce the speed of oil is one possibility when we design the transformer. But up to now we
dont know if there is a limit on speed. You can always reduce the speed of oil. But what is the
limit? The limit must be correlated with the inception of discharge somewhere. We can try to
avoid turbulence. Many details inside of the transformer can be studied to avoid turbulence. To
avoid charges in the oil we might be able to increase relaxation in the oil volume. That could be a
possibility. We can also chose material such as oil and paper to lower the propensity for static
electrification. Finally, we have to test the transformer with the eye towards static electrification.
That means to measure the ECT of oil. It seems that the first oil you put in the transformer is
very important because afterwards it is difficult to lower to ECT of oil. We can measure the
leakage current. The resistive leakage current is very interesting to measure, but it is like a
reduction of the oil speed. What is the acceptable limit. If you find some nano-Amperes, is that
good or not. It is very difficult to make a correlation between the limit of leakage current and the
inception of partial discharge. We also have to test the transformer with the pump running. That
could be a solution but the problem is you dont know exactly how long you must test because
the inception of discharge are sometimes coming shortly after the energization of the transformer
and other time it is seven years later. We heard about a transformer that ran properly for seven
years, and then static electrification appeared. So you can understand that it is a problem to
interpret the factory test results.

Panel Discussion

Comment: I would like to make some comment about the barriers around the bushing. We heard
that these should be removed. I dont agree at all with that because the aim of the barrier is to
improve the dielectric withstand. Sometime a transformer in service has some other voltages to
withstand. So the barrier is important. The arrangement that Tom Prevost presented is a good
arrangement, because its better to have the barriers.

Comment: This is a barrier, which Tom explained is three-four hundred millimeters away from
the shield. Thats what were drawing a picture of.

A note on the pockmarks on the high-voltage bushing shield as discussed in paper on


Dismantling of Ramapo Transformer by Harold Moore.

A dielectric withstand problem related to operation of oil pumps was experienced at the
EPRI Compact HVdc Link project in Astoria, New York. The Astoria problem was
encountered in a dc smoothing reactor. The project manager reported that he heard
discharges when testing the system. The tests might have been associated with test
energization with up to 300 kV-dc. The report is that the discharges could be turned on
and off by turning the oil pumps on and off. Thus, the discharge activity, although not
recognized at the time as a static electrification problem, probably was such a problem.
The project manager also reported observing discharges in the oil space above the
winding through a Plexiglas cover on the tank. (He had replaced the steel manhole cover
in order to be able to see what was going on in the tank.) He could see the discharges and
thought that they originated from the bushing shields. Note that this was a reactor
connected to SF6 buses so there was no problem with safety clearances at the top of the
tank because the bushings were inside the SF6 gas buses.

5-54
Panel Session: Are We There Yet?

The bushing shields may have had a semi-conducting epoxy coating since most shields
for the compact terminal had such shields. In any case, the bushing shields had molded
paper insulation on the outside. The solution in this case was to increase the thickness of
the shield insulation. That stopped the discharges.

The same phenomena could have been at work in Ramapo although the dielectric field
around the bushing shield in the Ramapo case was a combination of an ac field with the
dc field caused by the space charge. The field around the bushing shield may not be high
but it is a rather non-uniform field. DC charges outside the shield would penetrate the
paper dielectric and the stress would be across the epoxy, which probably would have a
higher resistivity and block the charges. The time constant for this is the one for the oil
paper dielectric at the temperature of the system. That is, it can vary from several hours to
a fraction of an hour. At around 25 to 30 degrees, it is probably between 60 and
45 minutes. Once the stress has built up on the epoxy layer to the degree that it breaks
down, the spitting will be initiated, which will leave a pockmark in the epoxy coating of
the shield. Such pockmarks were seen on shields coated with semi-conducting epoxy
inside the SF6 equipment at Astoria. These pockmarks in the case of Astoria appeared to
originate from locations with some imbedded metallic sliver or a burr on the surface of
the shields, which would lead to field enhancements. But, they had the same appearance
as the ones seen on the bushing shields of the Ramapo unit.

A flashover through a long gap is often caused by a streamer developing where the field
gradient is high, although the average field across the main gap is low. This is well
known for air gaps and gaps in SF6 equipment. In these situations, the streamer creates
its own field and eventually may propagate across the entire gap, which leads to power
arcs. In the Astoria situation, there was not enough energy in the discharge path to leave
any permanent damage behind even if there were flashovers. The same could have been
the case for the Ramapo test. If this failure mode is assumed to be true, then a possible
improvement for existing units could then be to retrofit with thicker bushing shield
insulation.

Question: We heard yesterday about the Ramapo testing and then it looked like the dangerous
time-temperature range was about 20 to 30 degree-C. So if you bring the oil temperature to about
that level, it should be OK. In the late 1970s in Japan, all the testing showed that the maximum
PD activity was in the 45-50 degrees C. So what Im wondering is, does each transformer have a
signature temperature range. So should we not be doing a little more testing to find out what is
going on? Rather than to say that if you run up the temperature of the oil to 30 degrees or above
we are safe? So thats my question.

Answer: Thank you for your comment. I agree with you, that 20 years ago we saw in Japan
mainly a peak of the discharge occurrence with a temperature around 50-55 degree. But this is
depending on the oil as well as of course, on the structure; shell form or core form. Also in core
form, we see a different behavior but mainly we saw a maximum between 40 and 60 degrees. In
the Ramapo case you found much lower levels, for which we at present have no explanation. We
have to check. We have to control the conductivity and ECT of oil and understand the
relationship between the transformer design and the oil characteristics. This is very important.

5-55
Panel Session: Are We There Yet?

Audience

Question: Im directing this to anybody who wants to pick up the ball and run with it. Is there a
case in the context of static electrification for going to adjustable speed drives for the pumps?
My guess is the cost of power electronics has now come down to the point to where they would
probably pay for themselves in the life of a transformer. So in the context of streaming
electrification, is that an option thats meaningful?

Answer: Ill try it. If you had a commercial pump that we could tie to the temperature, the
answer is yes. I dont think anybody is in a position to commercially develop one. But its a
solution if you had the equipment.

Comment: Life is never quite as simple as you think. These are centrifugal pumps, typically used
for transformer cooling. And the characteristic curve of a centrifugal pump, as you decrease the
speed, the flow and the pressures drop off at a square of the speed. Youre talking about being
able to get flow through there by dropping the speed. Youd have a fairly narrow range of control
before youd drop off to almost no flow. The electronics, to the best of my knowledge is still
pretty expensive. Wed sure like to try it with somebody. It looks like a whole new business.

Comment: As I was listening to the discussion, I thought of the same possibility. But then the
costs associated with that are very exorbitant. Were having problems just getting some of our
stations to retrofit from a two-stage to a three-stage control for thermal operation, let alone a
variable speed drive. Although with the variable speed drive and PLC logic, just about
anythings possible.

Comment: Another subject Id like to bring up, which may trigger somebody to make some
comments from the floor or from the panel, has to do with utility specifications. Does anybody
have anything in their specification that addresses the static electrification phenomena? Stick
your hand up if you do. (One hand was up.)

5-56
Panel Session: Are We There Yet?

Audience Other Half

Comment: The other aspect is industry standards. Can standards associations produce procedures
for eliminating or minimizing the effects of electrification? A lot of these needs must be captured
in some standards. Are there some working groups and can information be passed on to
appropriate people. We dont need the thoughts and the knowledge of all of this research to just
end in a report that goes into somebodys filing system. It needs to be an ongoing activity.

Comment: Id just like to reiterate the point that the purpose of the CIGRE work, which weve
all been involved in, was to do just that. In other words, study the topic, produce what you might
call sensible report, with good guidelines for the standards organizations such as IEEE, ASTM,
or IEC. We need to get started and try interpreting those findings and turning this into standards.
Some work has been already been done in this. For instance, I mentioned I think yesterday that
the UK has put into its own national standard an ECT requirement for oil. It has specified the
ministatic tester as the method to be used for the tests. It has no other purpose than to make sure
that we can make a comparison between oils and keep out the bad ones and use the good ones.
The difficulty was raised earlier as to the inertia in the standards area. I believe that TV made the
point supported by others Lets keep away from that subject, were engineers, were scientists,
lets leave it to the politicians or whoever they might be. But in actual fact, bearing in mind
what Im saying to you, the impetus is now coming out of CIGRE. Weve all got to contribute to
the next stage, which is to get those findings into the standards. Now there are two documents in
the standard areas. One is what we call standards, the other is an application guide. If you cant
get it into the standards, lets put it into an application guide. There are application guides in the
IEEE system as well as in the IEC system. And theres no reason why these things should be cast
in concrete so that they cant be altered. We can put in information in appendices. We can draft
something relatively quickly.

5-57
Panel Session: Are We There Yet?

Audience, Another View

Comment: Im a member of the IEEE Standards Board responsible for producing all kinds of
standards. From IEEEs point of view, I would like to encourage you that if you really want to
develop guidelines or a new standard, help is available. If anyone here is interested, please feel
free to contact me and I will put you in contact with the right people. Number two comment I
have: During my two and half day stay here, I found the symposium to be quite interesting,
rewarding and challenging. I got a lot of information but my basic concern is how do you convey
this information to the utilities? They are the people who are involved. If you look around here,
you dont have more than five utility members. I would like to invite to attend a Nuclear Power
Engineering Committee meeting. The Nuclear Engineering Committee represents the nuclear
utilities. We recognize that static electrification does not discriminate between conventional
plants or nuclear power plant. Our meeting is in October in Seattle. It is in conjunction with the
nuclear symposium and if someone wants to come there and make a presentation the invitation is
open. I will make it happen. The last point I want to make, from a nuclear power point of view,
failure of transformers has far-reaching considerations and implications. We cannot take them
lightly. We may not have a jurisdiction over the safety significance and the requirement related
to balance of plant. But nevertheless NRC is taking a different approach recently. We are looking
at the grid as the whole. We are looking into the operation of all equipment so that the plant can
operate safely. In that respect, that information should also go to an organization like NUMARC.
And NUMARC is also a member of NPAC. So my point is, you have done a good job but get it
to the right people.

5-58
Panel Session: Are We There Yet?

Also Audience

Comment: In listening here, it sounds as if several people have indicated that the criteria of the
oil are very important. But I really havent seen anything that tells me what criteria is important
or what limits the place on that. I did hear that this ECT is an important aspect. And there has
been a lot of bashing of ASTM. And, as chairman of D2705 electric test, I found it curious that
there was a round robin that was conducted, apparently before my time, which would be at least
ten years ago. That didnt develop into anything. Its my observation at ASTM that weve got a
lot of work to do. And there is a lack of volunteers. If somebody wants to have an ECT test
standard developed, then they have to take it on as a crusade. The equipment has to be available
and there has to be a need for it. ASTM will not, and D27 isnt going to undertake developing a
standard for a test just because two or three people want to see it happen. There has to be more
justification than that. Now, when I go back to my next meeting, Ill try to find out what
happened to that round robin. As far as particle count testing, theres some difficulty in
standardizing the sample that you bring to particle count. I dont think particle counting is a big
problem. Theres commercial equipment out there to count it. The problem is how do you get the
sample so that the particle count size remains constant from the time you collect it until you get it
into the machine. Im not sure ASTM is in a position to address that. I know we looked at TVs
proposal, which is what kind of ended it, was the fact that we cant get standardized samples. So,
if ASTM needs to develop some standards, Ill try to find out what happened to the round robin.
But I also would like to solicit some volunteers. I can guarantee you that we got a lot of work to
do. And if someone here is willing to work on a standard, provided it meets criteria thats
acceptable in the industry, then well develop a standard for it. In fact, were actively seeking
what is needed in the industry. And this sounds like it might be one of them.

5-59
Panel Session: Are We There Yet?

Comment: I pointed out that it is important to have criteria or a guideline for a limit of charge
intensity or ETC. But what you pointed out is true. Do we measure ECT? If there is no
acceptance limit it has no meaning. We introduced the 200 pico-coulomb limit or something like
this, based on data from operating transformers and data from modes. Of course, this part is not
perfect, but the important thing is to set a guideline or limit. Otherwise there is no meaning with
measuring the leakage current either. So, from now on we have to make an effort to share data
between Japan, the United States, Europe and all over the world. And my question is, is there
any such a similar guideline or level for ECT value or leakage current in the United States or
Europe or any other countries? Is this a first time a value or a guideline has been proposed?

Comment: We had eleven 765 kV transformers in the Muncie plant ready to be shipped. And
then we found out by just starting the pumps that five of them had discharges from just operating
the pumps. And that got us curious so we did some investigation. We found out that these
transformers had been filled up with two different oils. At that point we thought the oil might
have something to do with the phenomena. We were immediately interested in finding out what
is the charging factor and this is how the ministatic tester came about. We identified the units
with the higher discharge, and these had a much higher ECT; about four times that of the other
ones. And at that point we could only determine approximately the value that might produce the
static activity but only for that sort of transformers. Since then I have found out that there are
units out in the field that had electrification problem with much lower ECT. My conclusion is
that the design is very important in determining the acceptable ECT. So this is going to be a lot
more complicated than we thought. It only means that we may need something like a PD sensor
or something like that to determine at what level you have activity. And then relate that to the
ECT. So there is no magic number for ECT. That is my opinion. However, when you look at new
oil, again for EPRI work, I screened seventeen oils that were available in the US at that time,
including Canada. We found out there were very low ECT oils and a couple of them were very
high; by a factor of, I would say, thirty. So it was easy for me to pick a number for what is low
and what is high. But these are all based on relative measurements. So we recommended that we
changed all the high ECT oils with the low ECT oils. And so this is kind of a learning experience
for all of us. And, as I said again it is very difficult to come up with a guideline for utilities that
says that above this number you should go below this number, and that sort of thing. We need to
do a lot more thinking. We took samples from the fields and then applied a statistical analysis.
What was the average number. That is the approach I took for the particle count. We had about
200 transformers from which we collected the samples and did the analysis. So we need a lot
more statistical data from different types of units.

Comment: One other thing that I would like to get thrown into the pot for everybody to think
about a little bit has to do with some of the IEEE papers that were presented in the 70s and early
80s; particularly from the researchers in Japan. These showed some curves that scared me in
regards to BTA. It was some curves of the dielectric strength of oil versus BTA content. It
showed a decrease in dielectric strength for low values of BTA and for high values of BTA.
There was a funny peak in one of the papers, which showed that the maximum dielectric strength
was in the order of ten to fifteen parts per million. With BTA in it, the dielectric strength
diminished if you had less than that number or more than that number. I wonder if anybody in
this group has any experience with that?

5-60
Panel Session: Are We There Yet?

Comment: Dr. Tsuchie knows the characteristics of BTA. And you mentioned the dielectric
strength of the oil with BTA addition. His comment, if BTA is entirely dissolved into the oil, the
dielectric strength doesnt change at all until up to 50 ppm of BTA content.
Comment: A word of caution to the users in the States and others too, is in particular in the
States weve considered insulating the oil pretty well pristine. And additives to the oil havent
been a generally accepted practice here. Although in the manufacture of insulating oils there are
a lot of things that go into it. There are certain additives in the way of inhibitors and pour point
depressants and things like this we get into. But before any of us start just saying were going out
and start putting BTA in our transformers, I think we need to look at it. And I think we need
some help possibly from ASTM on that too,

Closing Remarks by Stig Nilsson

A couple of things come to mind as I listen to this mornings discussion. One is that we are
looking for perfection when there isnt any. We can go in and model and calculate what the field
distribution will be, based on what Keith and John Palmer presented. For a given design we may
come up with an expected distribution of charges. We can go in and measure ECT of
transformers. And if that is not the ECT that was assumed in the calculations that Keith made,
then we are going to have an error in our estimate of the charge distribution. I think what we can
do, as an industry, is to agree that once a transformer design is completed, and you put it on the
factory floor, we can run a test.
We know how to detect if there are any problems in the transformer. And we know that we have
to run the test with a certain time-temperature regime. If we run and pass this test we should not
have any problems in the field if we have an ECT in the oil of a certain maximum value. Then
what we can do is to charge that transformer in the field with oil with the same kind of ECT. We
can at least make sure that it is not higher. We know that if we dont get any change in the ECT
of the oil then we probably will not have any operating problems, as long as we follow the
assumptions made for the design. So the ECT is relative to a given design. But there are so many
problems when making the ECT test.
Thanks to the Japanese colleagues, we have now a very good basis for making a good test. The
round robin and the work that was done earlier by Doble and others in earlier EPRI projects
provided a good basis for how to sample and store the oil and also about what kind of bottles to
use for collecting, transporting and storing the sample. If you combine all of this, we will have a
useful guide for how to make accurate ECT tests. We should not delay agreeing on how to do the
test because we dont have all the answers.

I would also like to take this opportunity, because it may not come back Stan may shut me
off to say thank you to all of those here who have contributed papers. I know I trapped some of
you. You agreed to be a presenter and then you found out that you had to prepare something.
Now, I only told you that you had to do about two pages, and some of you went ahead and wrote
six or seven pages. Some even complained about the15 page absolute limit. If you wrote more
than I asked you to do, you can only blame yourself. And those of you who agreed to chair
sessions, I snared you maybe at the tail end just before the symposium and got you to agree to sit
here and keep track of time and do things like that. I appreciate your help and the very good
contribution that you have made. And the panelists that are now here, I want to thank you for
your contributions, too.

5-61
Panel Session: Are We There Yet?

Chairmans Remarks

Thanks Stig. Stig has been with this project since the beginning. I can remember when we
brought the topic up to EPRI at an EPRI Task Force presentation following our first failure in an
attempt to get EPRI involved in research on this. And the only experience that was available at
the time was the ConEd problems that they were having with the Compact HVdc Link and the
experience in Japan. Within the industry it was difficult to build a lot of interest in electrification
because it was largely thought that it was one manufacturer and one design. And as the truth
became known and more and more people became aware of the phenomena and the problems
created by it, we began to find more and more manufacturers involved, both shell form and core
form, and a lot of other things that weve discussed here today. So the project has come a long
way. Stigs been with it just about all the way and Stans done an excellent job in bringing
everything together. I do want to thank our panelists for the job they did today. And lets give
them our thanks.

Closing Remarks: Stan Lindgren, EPRI


I want to thank Dan Crofts for chairing the final panel this morning and for his dedicated
participation starting in 1983 with the first transformer failures in the US recognized to have
been caused by SE.
Likewise, Stig Nilsson for putting this symposium together and his comment this morning
that were looking for perfection where there is none. Since becoming the EPRI project
manager for SE in 1986 and the convenor of CIGRE JWG12/15.13 for SE in 1989, there has
been a constant need to whittle things down to the possible.
This morning, someone mentioned an early round-robin in the US, measuring oil ECT in
pursuit of a standard, and wondered what happened. It could be interesting but is probably
meaningless. The CIGRE JWG exhausted much effort to find a standard test method for
ECT, one that is simple & economical, gives consistent results, and would be used around the
world. This ended up to be TV Oommens mini-static test using a syringe to push the oil
through a filter paper. This is much the same as used in the US round-robin back in the 80s,
however the adopted CIGRE method includes refinements and automated procedures needed
for consistent results. I greatly appreciate Japans contributions at this symposium and their
concurrence on the mini-static test method. Although it took 10 years to agree on this test,
that is far from perfect, if all use it ECT numbers will become increasingly meaningful over
time. And, someday a better method will evolve.
Monitoring for SE can work. If the SE activity stops at an early stage, the transformer can
live on for another 40 years. SE is quite unique in this respect, and there are several known
instances in the US where transformers are operating today that had SE activity 1015 years
ago. I also believe that during a GSU transformer start-up, an early stage of SE activity can
be specifically recognized for what it is and the unit tripped with confidence.
Fritz Will, a surface-chemist of some renown, personally tackled the transformer SE problem
for several years while at EPRI, going all the way back to Helmholtz. I recall his saying that
he found SE in transformers to be the most complex problem he had ever encountered or was
ever aware of. The more that is understood, the greater the unknowns seem to be.

5-62
Panel Session: Are We There Yet?

So, much about SE remains a mystery with the potential for committing the perfect crime
(to use Stigs words). Build-up of DC stress can contribute to a dielectric breakdown without
leaving the slightest trace of evidence. Similar mysterious problems occur in all insulated
moving processes (printing, spinning, weaving, and flying through the air), not just
transformers; even lightning is not understood as well as it perhaps could be. SE will always
be present to some extent in very large directed-flow power transformers when pumps are
running, and someday a certain set of conditions may occur to cause a catastrophic failure,
especially when it is believed there is no problem because there has never been a problem
(that was recognized) without really understanding why.

This closes an EPRI chapter on static electrification in power transformers but not the book.
I thank you all very much.

5-63
A
FINAL PROGRAM FOR EPRI SYMPOSIUM ON STATIC
ELECTRIFICATION

Wednesday May 19, 1999

08.00 Panel session for presentation of Ramapo Test Chair: Bruce Gavioli,
Results General Manager
Con Edison

Application of In Service Transformer Harold Moore, Consultant


Monitoring Systems during Tests at Consolidated and Donald Chu,
Edisons Ramapo Station, by Harold Moore and Consolidated Edison Co.
Donald Chu

Acoustic Partial Discharge and Leakage Current J. Michael Walden, ABB


Measurement Results of Ramapo Tests, by
George K. Frimpong, Michael J. Walden and
Stan Lindgren

09.30 Break - Refreshments

Ramapo Static Electrification Project Test Dr. J. C. Harley,


Results from J. W. Harley Inc. by J. C. Harley J. W. Harley, Inc

Ramapo Tests: Results from Streaming Current Stig Nilsson, Consultant


Monitoring Using a Nilsson plate and an Absolute
Charge Sensor, by Stig Nilsson, Markus Zahn
and Stan Lindgren

Dismantling of Ramapo Transformer by Harold Moore, Consultant


Harold Moore

Discussion

12.00 Lunch

13.30 Modeling and Prediction of Static Electrification Chair: Jean Poittevin,


Research Director, Alstom,
France

A-1
Final Program for EPRI Symposium on Static Electrification

Static Electrification at Solid/Liquid Interfaces Dr. G. Praxl, Technische


of Power Transformers, G. Praxl and K. Neuner Universitt Graz, Austria

Study on Increase of Electrostatic Charging Dr. Motoo Tsuchie,


Tendency of Insulating Oil and the Diagnosis on Mitsubishi Electric
Streaming Electrification, Motoo Tsuchie, Corporation
Noboru Hosokawa, Susumu Isaka, Hiroshi
Miyao, Toshitsugu Ishii, Shunichi Kobayashi,
Takayuki Kobayashi, Takahiro Ohno, Hitoshi
Okubo

The Electrostatic Charging Tendency of Dr. Hiroshi Miyao, Hitachi,


Inservice Oils and the Evaluation Method of the Ltd.
Leakage Current from the Transformer Winding,
Hiroshi Miyao, E. Mori, S. Isaka, M. Tsuchie,
K. Takamoro, S. Kobayashi, T. Koybayashi,
T. Oono and H. Okubo

15.00 Break and Refreshments

Flow Electrification Measurements of Dr. Markus Zahn, MIT


Transformer Insulation, Markus Zahn, Andrew
P. Washabaugh and Darrell E. Schlicker

Measurement of Oil Charge Densities in a Dr. G. K. Frimpong, ABB


Transformer Model, George K. Frimpong,
Michael J. Walden and Stan Lindgren

Discussions

Thursday May 20, 1999

08.00 Modeling and Prediction of Static Electrification - Chair: Jean Poittevin,


(continued) Research Director, Alstom,
France

Streaming Electrification Dynamics - Duct Dr. John A. Palmer


Modeling and Analysis, J. Keith Nelson and
J. A. Palmer

Investigation for Standardization of Electrostatic Susumu Isaka, Toshiba


Charging Tendency Measurement of Transformer Corporation
Oil in Japan, Susumu Isaka, Masama Ikeda,
Hiroshi Miyao, Motoo Tsuchie, Toshitugu Ishii,
Shunichi Kobayashi, Takayuki Kobayashi,
Takahiro Ono and Hitoshi Okubo

A-2
Final Program for EPRI Symposium on Static Electrification

Discussions

09.30 Break and Refreshments

09.45 Operating Experience of Transformers. Chair: Don Rose, Manager,


Transmission & Substation
Operations, TU Electric

A Static Electrification Transformer Survey, Dr. J. Keith Nelson, RPI


J. Keith Nelson

A Static Electrification Failure at the Navajo Thomas G Lundquist, Salt


Generating Station, Thomas Lundquist River Project

Whitpain #1C Transformer Failure, by David F. Dave Goodwin, PECO


Goodwin

Static Electrification Failure of MPT at Raymond F. Cameron,


Powerton Generating Station, by Raymond F. Technical Services,
Cameron and Christopher P. Stefanski, Commonwealth Edison
Company

Investigation of a Failed Westinghouse Don Angell, Idaho Power


500/345 kV 500 MVA Single Phase Company
Transformer by Don Angell and Kent Venosdel

Discussion

12.00 Lunch

13.30 Monitoring and Management Methods Chair: Mike Lebow,


Manager, Research and
Development,
Con Edison

Further Analysis of Streaming Current as Stig Nilsson, Consultant


Recorded from the Ramapo Transformer,
Stig Nilsson, Markus Zahn and Stan Lindgren

Static, the Ramapo Experience and What Can Be Dr. Chatham Cooke, MIT
Done, Chatham Cooke

A-3
Final Program for EPRI Symposium on Static Electrification

Static electrification diagnostic model Daniel Dahlgren, Omaha


implemented in an on-line monitoring system, Public Power District
Daniel L. Dahlgren and J. W. Harley

15.00 Break and Refreshments

Static Electrification Mitigation Using Dr. J. Keith Nelson, RPI


Intelligent Pump Controls, J. Keith Nelson and
J. A. Palmer

Transformer Reliability: Management of Static A. C. Hall, Consultant


Electrification in Power Transformers,
A. C. Hall

Discussion

Friday, May 21, 1999

08.00 Panel session: Are we there yet? Chair: Dan Crofts,


Consultant

Mitigation of Static Electrification Discharges in Harold Moore, Consultant


Power Transformers, Harold Moore

09.00 Break and Refreshments


Operation of existing transformers Panelists:
Aging and the risk of electrification failure Harold Moore, Consultant
Design of new transformers Raymond Cameron, CEC
Prof. Dr. Hitoshi Okubo,
Factory testing of new transformers
Nagoya University
Monitoring of transformers to avoid static Tom Prevost, Weidmann
electrification failures Jean Poittevin, Alstom
Material evaluation for new transformers Stan Lindgren, EPRI

Closing remarks Stan Lindgren, EPRI

A-4
B
LIST OF REGISTRANTS BY NAME

Satish Aggarwal Nelson Amy


U.S. Nuclear Regulatory Commission Coordinator, Training
Reliant Energy HL&P
12301 Kurland Dr
Houston, TX 77034-4898
Phone (713) 945-8373
Fax (713) 945-8381
nelson-amy@reliantenergy.com

Don T. Angell Yves Bertrand


Apparatus Supervisor Dir., des Etudes et Recher
Idaho Power Co. EDF Electricite de France
P.O. Box 70 Les Renardieres- Cima 8
Boise, ID 83707 Moret-sur-Loing, F77818, France
Phone (208) 388-2767 Phone (331) 607-36994
Fax (208) 388-6906 Fax (331) 607-37477
donangell@idahopower.com yves.bertrand@edf.fr

Raymond F. Cameron Donald Chu


Equipment Specialist Senior Engineer
Commonwealth Edison Co. Consolidated Edison Co. of N.Y.
1319 S 1st Ave 4 Irving Plaze, Rm 1508
Maywood, IL 60153-2496 New York, NY 10003-3598
Phone (630) 410-5214 Phone (212) 460-3456
Fax (630) 410-7227 Fax (212) 387-2106
raymond.f.cameron@ucm.com chud@coned.com

C. M. Cooke Dan W. Crofts


MIT Massachusetts Institute of Technology Retired
Cambridge, MA 02139 Texas Utilities Co.
Phone (617) 253-2591 1411 Moss Rose Circle
Fax (617) 258-6774 Irving, TX 75061
Phone (214) 502-2950
dcrofts1@tuelectric.com

Daniel L. Dahlgren Alan W. Darwin


Lead Electrical Engineer Chief of Development
Omaha Public Power District Alstom T&D Transformers Ltd.
444 S 16th Street Mall P.O. Box 26, Lichfield Rd.
Omaha, NE 68102-2247 Stafford, ST17 4LN, England
Phone (402) 636-2558 Phone 44 (178) 527-4870
Fax (402) 636-3947 Fax 44 (178) 527-4028
ddahlgren@oppd.com alan.darwin@tda.alstom.com

B-1
List of Registrants by Name

Fred E. Elliott J. Alan C. Forrest


Senior Electrical Engineer Project Manager
Bonneville Power Administration (BPA) Teshmont Consultants, Inc.
P.O. Box 491 1190 Waverly Street
Vancouver, WA 98666-0491 Winnipeg, Manitoba R3T OP4-, Canada
Phone (360) 418-2607 Phone (204) 284-8100
Fax (360) 418-2958 Fax (204) 475-4601
feelliott@bpa.gov aforrest@teshmont.mb.co

Keith Frazier George K. Frimpong


System Engineer Principal Consulting R&D Engineer
South Texas Project Nuclear Operating Co. ABB Electric Systems Tech. Institute
P.O. Box 289 1021 Main Campus Drive
Wadsworth, TX 77483-0289 Raleigh, NC 27606
Phone (512) 972-8329 george.k.frimpong@ustra.mail.abb.com
Fax (512) 972-8197
kefrazier@stpcgs.com

Bruce J. Gavioli Ramsis Girgis


General Manager Mgr., Dev. Eng.
Consolidated Edison Co. of N.Y. ABB Power Transformers
4 Irving Place 4350 Semple Ave.
New York, NY 10003 Saint Louis, MO 63120
Phone (212) 338-4455 Phone (314) 679-4803
Fax (212) 338-4457 Fax (314) 679-4810
ramsis.s.girgis@ustra.mail.abb.com

David F. Goodwin William C. Graham


Engineer Crew Leader
PECO Energy Co. Reliant Energy
1111 Old Eagle School Rd. P.O. Box 8298
Wayne, PA 19087 Bacliff, TX 77518
Phone (610) 971-7148 Phone (281) 316-4343
Fax (610) 971-7175 Fax (281) 316-4333
t207dfg@peco-energy.com

A. C. Hall Jack W. Harley


Consultant President
Power Transformers & Reactors Ltd. J. W. Harley, Inc.
29 Meads Road 8941 Dutton Dr
Guldford, GU1 2NB, England Twinsburg, OH 44087-1944
Phone 44 (148) 353-6651 Phone (330) 425-1838
Fax 44 (148) 353-6651 Fax (330) 425-1812
achall<101561.1624@compuserve.com> jack@harleyinc.com

John C. Harley Ray Holgate


President Sr. Engineer
J. W. Harley, Inc. Salt River Project (SRP)
425 Golden Oak Drive P.O. Box 850
Portola Valley, CA 94028 Page, AZ 86040-0850
Phone (650) 520-8738 Phone (520) 645-6532
Fax (408) 522-8729 Fax (520) 645-6234
jch@fluent.com rjholgat@srpnet.com

B-2
List of Registrants by Name

Noboru Hosokawa Susumu Isaka


Mgr., Dev. Section, Shell-Form Transformer Dept. Toshiba Corp., Hamakawasaki Works
Mitsubishi Electric Corp. 2-1 Ukishima-cho, Kawasaki-ku
651, Tenwa, Ako City Kawasaki, 210-0862, Japan
Hyougo Prefecture, 678-0256, Japan Phone 8 (144) 288-6408
Phone 8 (179) 146-2233 Fax 8 (144) 270-1451
Fax 8 (179) 146-2367 susumu2isaka@toshiba.co.jp
hosokawa@ako.melco.co.jp

Mike A. Lebow Stanley Lindgren


Mgr., Research & Dev. Project Manager
Consolidated Edison Co. of N.Y. EPRI, Palo Alto
4 Irving Pl 3412 Hillview Ave
New York, NY 10003-3598 Palo Alto, CA 94304-1395
Phone (212) 460-4000 Phone (650) 855-2308
Fax (212) 529-8511 Fax (650) 855-8997
lebowm@coned.com

Thomas Lundquist Thomas Manthe


Sr. Principal Engineer Siemens AG
Salt River Project (SRP) EV LT TWN PG1 T3, Katzwanger StraBe 150
P.O. Box 52025 Nurnberg, 90461, Germany
Phoenix, AZ 85072-2025 Phone 49 (911) 434-2465
Phone (602) 236-8617 Fax 49 (911) 434-2454
Fax (602) 236-8640 thomas.manthe@nbg8.siemens.de
tglundqu@srpnet.com

Joe Martin Hiroshi Miyao


Supervising Engineer Sr. Researcher
Public Service Electric & Gas Co. Hitachi, Ltd.
80 Park Plaza T-12A 1-1, Kokubu-cho 1-chome
Newark, NJ 07102 Hitachi-shi, Ibaraki-ken 316-8501, Japan
Phone (973) 430-8444 Phone 8 (129) 436-8412
Fax (973) 643-0978 Fax 8 (129) 436-8399
jmartin@pseg.com miyao@erl.hitachi.co.jp

Harold Moore J. Keith Nelson


Consulting Engineer Professor & Chair
Harold Moore & Associates Rensselaer Polytechnic Institute
1062 Lake Way Dr 110 8th St
Niceville, FL 32578-1776 Troy, NY 12180-3590
Phone (904) 729-1391 Phone (518) 276-6329
Fax (904) 729-1391 Fax (518) 276-6226
moorhr@aol.com nelsonj@rpi.edu

Stig L. Nilsson Hitoshi Okubo


Consultant Prof., Dept. Electrical Engineering
Stig Nilsson Consulting Nagoya University
20715 Brush Rd Furo-cho, Chikusa-ku
Los Gatos, CA 95033-9138 Nagoya, 464-8603, Japan
Phone (408) 353-6572 Phone (81) 52 789-3625
Fax (408) 353-6572 Fax (81) 52 789-3141
stig_nilsson@compuserve.com

B-3
List of Registrants by Name

T. V. Oommen John A. Palmer


ABB Electric Systems Tech. Institute Assistant Professor
1021 Main Campus Drive Colorado School of Mines
Raleigh, NC 27606 Division of Engineering
Phone (919) 856-2437 Golden, CO 80401-1887
Fax (919) 856-2448 Phone (303) 273-3686
tv.oommen@ustra.mail.abb.com Fax (303) 273-3602
japalmer@mines.edu

Donald W. Platts Jean Poittevin


Sr. Engineer Research Director
PP&L, Inc. Alstom
2 N 9th St 25 Rue Des Bateliers
Allentown, PA 18101-1139 Saint Oven, 93404, France
Phone (610) 774-4686 Phone 33 (14) 965-7163
Fax (610) 770-6388 Fax 33 (14) 965-7379
dwplatts@papl.com jean.poittevin@tde.alstom.com

G. Praxl Tom Prevost


Prof. Dr., High Voltage Engineering Technical Manager
Graz University of Technology ENV Weidmann
Inffeldgrasse 18 1 Gordon Mills Way
Graz, A-8010, Austria St Johnsbury, VT 05819
Phone 43 (316) 873-7425 Phone (802) 751-3458
Fax 43 (316) 465-5780 Fax (802) 748-8029
praxl@hspt.tu-graz.ac.at tprevost@ehv.mhs.compuserve.com

James Radford Don W. Rose


Crew Leader Mgr., Transmission & Substation Ops
Reliant Energy HL&P TU Electric
12301 Kurland Dr P.O. Box 970
Houston, TX 77034- Fort Worth, TX 76101-0970
Phone (713) 945-7740 Phone (817) 215-6171
Fax (713) 945-7111 Fax (817) 215-6059
drose1@tuelectric.com

J. Rungis Michele Samoulides


Research Specialist Conference Manager
CSIRO EPRI, Palo Alto
P.O. Box 218 3412 Hillview Ave
Lindfield, NSW 2070, Australia Palo Alto, CA 94304-1395
Phone (612) 413-7045 Phone (650) 855-2127
Fax (612) 413-7261 Fax (650) 855-2166
msamouli@epri.com

Greg Schnaubelt Paul Stiller


Crew Leader Dir., Business Development
Reliant Energy J. W. Harley, Inc.
Cedar Bayou Plant, P.O. Box 1700 8941 Dutton Dr
Houston, TX 77251 Twinsburg, OH 44087-1944
Phone (281) 383-4242 Phone (330) 425-1838
Fax (281) 383-4300 Fax (330) 425-1812
phstiller@harleyinc.com

B-4
List of Registrants by Name

Ken Stricklett Hakan Swahn


Physicist Electrical Engineer
National Institute of Standards & Technology (NIST) OKG Aktiebolag
100 Bureau Drive Stop 8113 Simpevarp
Gaithersburg, MD 20899 Oskarshamn, 572 83, Sweden
Phone (301) 975-3955 Phone 46 (491) 78-7616
Fax (301) 869-6846 Fax 46 (491) 78-6865
stricklett@nist.gov hakan.swahnatokg.sydkraft.se

Motoo Tsuchie Andre W. Van Boetzelaer


Mgr., Transformer Section KEMA Nederland B. V.
Mitsubishi Electric Corp. 6800 Et Arnhem
651 Tenwa, Ako City The Netherlands
Hyogo Prefecture, 378-0256, Japan Phone 31 (26) 356-6064
Phone 81 (79) 146-2397 Fax 31 (26) 351-3683
Fax 81 (79) 146-2388 a.w.vanboetzelaer@kema.nl
tsuchie@ako.melco.co.jp

J. Michael Walden Markus Zahn


Sr. R&D Engineer Prof., Electrical Engineering
ABB Electric Systems Tech. Institute Massachusetts Institute of Technology
1021 Main Campus Drive 77 Massachusetts Ave Rm 10-174
Raleigh, NC 27606 Cambridge, MA 02139-4307
Phone (919) 856-3907 Phone (617) 253-4688
Fax (919) 856-2448 Fax (617) 258-6774
mike.j.walden@ustra.mail.abb.com zahn@mit.edu

Matt Zersen
Elec. Specialist
Reliant Energy HL&P
12301 Kurland
Houston, TX 77034
Phone (713) 945-7853
Fax (713) 945-7722
matt-zersen@reliantenergy.com

B-5
C
LISTS OF REGISTRANTS BY COMPANY

ABB Electric Systems Tech. Institute


George K. Frimpong
T. V. Oommen
J. Michael Walden

ABB Power Transformers


Ramsis Girgis

Alstom
Jean Poittevin

Alstom T&D Transformers Ltd.


Alan W. Darwin

Bonneville Power Administration (BPA)


Fred E. Elliott

Colorado School of Mines


John A. Palmer

Commonwealth Edison Co.


Raymond F. Cameron

Consolidated Edison Co. of N.Y.


Donald Chu
Bruce J. Gavioli
Mike A. Lebow

CSIRO
J. Rungis

EDF Electricite de France


Yves Bertrand

ENV Weidmann
Tom Prevost

EPRI, Palo Alto


Stanley Lindgren
Michele Samoulides

Graz University of Technology


G. Praxl

Harold Moore & Associates


Harold Moore

C-1
Lists of Registrants by Company

Hitachi, Ltd.
Hiroshi Miyao

Idaho Power Co.


Don T. Angell

J. W. Harley, Inc.
Jack W. Harley
John C. Harley
Paul Stiller

KEMA Nederland B. V.
Andre W. Van Boetzelaer

Massachusetts Institute of Technology


Markus Zahn

MIT Massachusetts Institute of Technology


C. M. Cooke

Mitsubishi Electric Corp.


Noboru Hosokawa
Motoo Tsuchie

Nagoya University
Hitoshi Okubo

National Institute of Standards & Technology (NIST)


Ken Stricklett

OKG Aktiebolag
Hakan Swahn

Omaha Public Power District


Daniel L. Dahlgren

PECO Energy Co.


David F. Goodwin

Power Transformers & Reactors Ltd.


A. C. Hall

PP&L, Inc.
Donald W. Platts

Public Service Electric & Gas Co.


Joe Martin

Reliant Energy
William C. Graham
Greg Schnaubelt

C-2
Lists of Registrants by Company

Reliant Energy HL&P


Nelson Amy
James Radford
Matt Zersen

Rensselaer Polytechnic Institute


J. Keith Nelson

Salt River Project (SRP)


Ray Holgate
Thomas Lundquist

Siemens AG
Thomas Manthe

South Texas Project Nuclear Operating Co.


Keith Frazier

Stig Nilsson Consulting


Stig L. Nilsson

Teshmont Consultants, Inc.


J. Alan C. Forrest

Texas Utilities Co.


Dan W. Crofts

Toshiba Corp., Hamakawasaki Works


Susumu Isaka

TU Electric
Don W. Rose

U.S. Nuclear Regulatory Commission


Satish Aggarwal

C-3

Vous aimerez peut-être aussi