New Frontiers of DGA Interpretation

for Power Transformers and their Accessories

Michel Duval

November 16, 2012

1
 Risk of failure of transformers

The risk of failure of transformers in service, as

indicated by dissolved gas analysis (DGA), depends
on three main parameters:

-the type of fault involved (e.g., arcing or hot spot).

-the location of the fault (in oil or in paper).
-the amount of gases formed (concentrations and
rates).

2
 The most dangerous faults:

-high-energy arcing D2 in oil and paper.

-low-energy arcing D1 in paper.

-hot spots of high temperature T3 and T2 in paper.

3
 The less-dangerous faults:

-low-energy arcing D1 in oil.

-hot-spots T3 and T2 in oil.

-hot spots T1 in paper.

4
 The non-dangerous faults:

-hot spots T1 in oil.

-corona partial discharges PD (unless very large

amounts of hydrogen are formed).

-catalytic reactions of free water in drain valves.

-aging of paper at low temperatures.

5
 Identification of faults:

-general faults (PD, D1, D2, T1, T2, T3) can be

identified by several methods (Rogers, Dornenburg,
Key Gas, IEC ratios, Duval Triangle 1).

-faults in paper are more difficult to identify.

CO, CO2 and the CO2/CO ratio are generally used
for that purpose. However, it has been shown
recently at CIGRE that they are not always good
indicators.

6
 Limitations of CO and CO2:

-large amounts of CO (>1000 ppm) and low values

of the CO2/CO ratio (<3) are often observed in
closed transformers without any fault.

-this has been attributed at CIGRE to the low

availability of O2, which favours the formation of CO
over CO2.

7
 Limitations of CO and CO2:

-also, localized faults involving small volumes of

paper often do not produce detectable amounts of
CO and CO2 against the usually large background
of these gases in oil.

-in many cases, local faults with carbonization of

paper are more reliably detected by the formation
of the other gases (H2 and hydrocarbon), using for
example the Duval Triangles 4 and 5.

8
 Duval Triangle 4 (updated 2012)

Zones:
-S: stray gassing of oil (T<200C).

-O: overheating, T<250C.

-C: possible carbonization of paper

(probability of 80%, not 100%).

-PD: corona partial discharges

(change C2H6 zone boundary from 1
(Graph courtesy: Serveron) to 0.6, and verify it with stray gassing
tests in the lab).
9
 Duval Triangle 5 (updated 2012)

Zones:

-T3 and T2: hot spots in oil only

(T>700C and >300C).

-C: possible carbonization of paper

(probability of 90%, not 100%).

-O, S, PD, use rather Triangle 4.

(Graph Courtesy: Serveron)

10
 Use of Triangles 5 and 4:

-Triangles 5 and 4 should be used only after

having identified first the general type of fault (D1,
D2, T3, T2, T1, PD) with Triangle 1.

-Triangles 4 and 5 should never be used in case

of electrical faults D1 or D2.

11
11
 Use of Triangles 5 and 4:

-Triangle 5 should be used only in case of thermal

faults T3 or T2, to distinguish between faults with
carbonization of paper and faults in oil only.

-Triangle 4 should be used only in case of faults PD,

T1, T2 or high levels of H2, to make further
distinctions between these low-energy faults.

- A small algorithm is available for free for Triangles

1 to 7 from duvalm@ireq.c.
12
12
 Recommendations of IEC for CO, CO2

-in IEC 60599, presently under revision:

-in open transformers, high CO values (>1000 ppm)

and low CO2/CO (<3) ratios indicate faults in paper.
-in closed transformers, this also applies, but only if
other fault gases (Cn, H2) are formed.

-high CO2 values (>10,000 ppm) and CO2/CO ratios

(>15) indicate accelerated paper aging.
13
13
 Transformer aging

-the risk of failure of transformers indeed increases

with age and load.

-however, in a majority of cases, this is because of

high moisture content, gas-producing faults,
leaking gaskets and membranes, rusting of tank
and dielectric problems, not so much because of
paper aging.

14
 Transformer aging

-for instance, a large number of transformers have

been reported to operate quite normally with DPs
of paper between 200 and 100.

-also, no or very few convincing cases have been

reported so far, of transformers with a DP of 200
where failure could be clearly attributed to the
mechanical weakness of paper.

15
 Transformer aging

-so, based on present observations, the risk of

failure at low DPs of paper appears to be rather
low, not very high as generally believed.

-transformers with gas-producing faults are more at

risk, and should be replaced in priority.
-for replacements based on DP only, a limit value
of 100 or 150 appears more realistic than 200,
based on present observations.

16
 Transformer aging

-aging paper loses density and thickness, but this

can easily be compensated by re-clamping windings.

-brittle paper squeezed between windings turns

(where it is most needed) will not fall off or tear
away, because of the compression forces applied to
windings, even if its tensile strength is low.

-and it can easily resist compression forces down to

very low DPs, without creeping.
17
 Gas limits in service

-once the type and location of fault have been

identified, the next parameter to look at is the amount
of gases formed.

-gas limits in the IEEE Gas Guide C57.104 are

presently under revision. The most significant change
will be much higher values of CO and CO2 (750 and
7500 ppm) for condition 1. Present values are often a
cause of false alarms.

18
18
 Gas limits in service

-limits for gas concentrations and rates above IEC

typical (condition 1) values have also been
recommended by CIGRE (in TB # 443, 2010).

-CIGRE (WG47) is now investigating the effect of type

and location of faults on these gas limits.

19
19
 Effect of type of faults on gas limits

-the effect on typical (condition 1) values is

obtained by calculating 90% percentile values for
each type of fault.

-the effect on pre-failure (condition 4) values is

deduced from inspected fault cases where high
levels of gases are observed without failure or just
before failure.

-specific types of faults are identified first with

Triangles 1, 4 and 5. 20
20
 CIGRE condition 1 values vs. type of faults:

21
21
 Occurrence of types of faults in service:

22
22
 Examples of high gas formation
without failure

current transformer
with 6400 ppm C2H4!

power transformer with a

fault D1 in oil (on a bakelite
plate, with 480 ppm C2H2!
23
23
 Examples of the effect of fault type
on conditions 1 and 4

24
24
 Effect of fault type on CIGRE condition 4

-to confirm these values, other examples of the following fault cases
would be needed:
-arcing faults D1/D2 in oil with high levels of C2H2 and no failure,
-arcing faults D1/D2 in paper just before failure,
-thermal faults T1, T2 and T3 in paper with high levels of C2H4, C2H6 or
CH4 and no failure.
-please email such examples to duvalm@ireq.ca.
25
25
 Application of DGA to LTCs

-a significant percentage (~30%) of failures of

transformers in service is due to faults in LTCs.

-dissolved gas analysis (DGA) is therefore

increasingly used to detect such faults in LTCs.

26
 Application of DGA to LTCs

-what makes the interpretation of DGA in LTCs

more complex than in transformers is that their
normal operation involves several different LTC
components, depending on their type and
operating conditions.

-normal gas formation resulting from this normal

operation of LTCs must be identified first as
precisely as possible in service.

27
 Normal gas formation/ operation of LTCs

May involve:
-arc-breaking-in-oil between contacts, switching of
selectors and valves, with the formation of arcing
gases D1(C2H2).

-current dissipation in transition resistors, increasing

their temperature and producing thermal gases in oil
(C2H4/ CH4/ C2H6).

28
 Normal gas formation/ operation of LTCs

-depending on the specific type and design of

LTC used (in oil or vacuum, of the compartment
type or in-tank type) and its operating conditions
(at low or high powers or currents), different
mixtures of these gases may be formed during
normal operation of the LTC.

29
Faulty or abnormal gas formation/ operation of LTCs

May involve, among others:

-an increase in the resistance and temperature of
arc-breaking contacts, resulting in carbonization of
oil, coking of contacts and formation of thermal
gases T3 or T2.

-abnormal arcing on metal plates or components,

producing arcing gases D2 or D1.

-abnormal hot spots on LTC enclosures or vacuum

bottles, producing thermal gases. 30
 Identification of faults in LTCs by DGA

Two main methods are used:

-the C2H4/C2H2 ratio of IEEE.

-the Duval Triangles 2, allowing to:

-follow visually the evolution of faults in LTCs,
-detect faulty operation in some specific types
of LTCs, thanks to the use of a third gas (CH4).

31
 The Duval Triangle 2 for LTCs of type I

-the original Duval Triangle 2 applies to LTCs of

type I, representing a large majority of LTCs
presently in service:

-examples of LTCs of type I include those of

Westinghouse, ABB (except some UZBs, which are
of type II), General Electric, McGraw Edison,
Siemens, MR-Reinhausen (RM, T and V), Federal
Pacific, Ferranti, Hyundai, Cooper, etc.

32
The Duval Triangle 2 for LTCs of type I

33
Example of normal and faulty operation
of LTCs of type I

34
Example of normal and faulty operation
of LTCs of type I

35
Example of normal and faulty operation
of LTCs of type I

36
Example of normal and faulty operation
of LTCs of type I

37
Example of normal and faulty operation
of LTCs of type I

38
Example of normal and faulty operation
of LTCs of type I

39
Example of normal and faulty operation
of LTCs of type I

40
 The Duval Triangles 2a-2e
for LTCs of type II

-in LTCs of type II, representing a minority of LTCs

in service, gas formation may appear in zone N or
in another part of Triangle 2, depending on the
specific type and operating conditions of the LTC.

-the Duval Triangles 2a-2e have been developed

for these LTCs of type II.
41
The Duval Triangles 2a-e
for LTCs of type II

42
 The Duval Triangles 2a-e
for LTCs of type II

-Triangle 2a applies to LTCs such as OilTaps M and

D.

-normal gas formation appears either in zone N or

zone N1 depending on operating conditions, and
should be determined in service first.

-faulty or abnormal operation is indicated by gas

formation in zones X3 or T3 (from zone N), or in
zones T3 or T2 (from zone N1).
43
Example of normal and faulty operation
of LTCs of type II (2a)

44
 The Duval Triangles 2a-e
for LTCs of type II

-Triangle 2b applies to LTCs such as VacuTaps VR.

Normal gas formation appears in zone N2 (not in
zone N so far).

-Triangle 2c applies to LTCs such as VacuTaps VV.

Normal gas formation appears either in zone N or
zone N3 (at high powers), and should be
determined in service first. Faulty or abnormal
operation is indicated by gas formation in zones X3
or T3 (from zone N), or in zone T3 (from zone N3).
45
 VacuTaps VR (2b):
Normal operation in zone T2 is probably due to operation
of selectors and resistors.

Ref: R.Frotscher (Orlando)

46
 VacuTaps VV (2c):

During high power tests,

In service, normal
normal (in blue) and faulty
operation of selectors
operation (in red) of
resistors.

Ref: R.Frotscher, CIGRE 47

 The Duval Triangles 2a-e
for LTCs of type II

-Triangle 2d applies to LTCs such as OilTaps R and

V. Normal gas formation appears either in zone N
or zone N4 (mild overheating).

-Triangle 2e applies to LTCs such as OilTaps G

(operating at high currents), or in a few resistive
UZBs of ABB, depending on operating conditions.
Normal operation appears either in zone N or zone
N5, and should be determined in service first.
48
Oil Taps R (2d):

49
 Oil Tap G (2e)

(OilTaps G are high-currents application models)

50
 The Duval Triangles 2a-e
for LTCs of type II

-To confirm the position of zone boundaries, other

examples of normal and faulty operation of the
following types of LTCs of type II would be needed;
please send them to duvalm@ireq.ca.
-resistive MR-Types (M, D, VR and VV),
-Types G and some UZBs (operating normally
in the X3 zone).

51
Triangle 2 applied to large databases by J.Dukarm

52
Triangle 2 applied to large databases by J.Dukarm

53
 Equivalence between
Triangles 2 and 2-gas ratios

-the red lines in previous Figures have been

shown by J.Dukarm to correspond to constant
values of the C2H4/ C2H2 ratio (vertical red line)
and C2H2/ CH4 and C2H4/ CH4 ratios (horizontal
red lines).

54
 On-line gas monitors

There are two main types of on-line gas monitors:

-hydrogen monitors: they are less expensive but

may miss the early signs of arcing, the most
dangerous fault.

-multi-gas monitors: they allow to get fault diagnosis

on-line and take immediate action on the equipment.

55
 On-line gas monitors

-the recommendation of CIGRE (in TB # 409) is

to use hydrogen monitors for relatively healthy
transformers, and multi-gas monitors for
transformers where problems are suspected
(already gassing or critical transformers).

-the main advantage of on-line gas monitors is to

detect faults occurring between regular oil
sampling intervals.

56
 Reliability of DGA results

-it has been shown by CIGRE that the accuracy of

laboratories and on-line monitors is acceptable on
average, at ~ 15%.

-some laboratories and monitor units, however,

may be much less accurate than average, and it is
the responsability of DGA users to detect them.

57
57
 Reliability of DGA results

-this can be done by comparing with results from

other laboratories and with samples of gas-in-oil
standards.

-the latter are available commercially, or can

easily be prepared by the laboratories, following
existing standard methods.

58
58
 Reliability of DGA results

-a recent RRT has shown that DGA results

obtained with Method A are more accurate on
average than with Methods B and C.

-this does not mean that B and C are inherently

less accurate, but suggests that they are more
difficult to apply accurately (because of calibration
and contamination).

59
59
 Reliability of DGA results

-a procedure for verifying the accuracy of on-line

gas monitors is available in TB # 409.

-for calibrating the headspace method, the IEC has

recently recommended to use gas-in-oil-standards
rather than published partition coefficients.

60
60