IEEE

Power Engineering Society

Tutorial on Harmonics Modeling and

Simulation

C!D

IEEE

TP-125-0
 Tutorial on

Harmonics Modeling and Simulation

IEEE Power Engineering Society Task Force on Harmonics Modeling and Simulation
IEEE Power Engineering Society Harmonics Working Group

Sponsoredby the Life Long Learning Subcommittee of the

IEEE Power EngineeringEducation Committee
IEEE Power Engineering Society

Tutorial On
HARMONICS MODELING AND SIMULATION

Abstracting is permitted with creditto the source. For other copying, reprint, or republication permission,
write to the IEEE Copyright Manager, IEEEService Center, 445 Hoes Lane, Piscataway, NJ 08855-1331.
All rights reserved. Copyright 1998 by The Institute of Electrical and Electronics Engineers, Inc.

IEEE Catalog Number: 98TP125-0

Additionalcopies of this publication are available from

IEEE Operations Center

P. O. Box 1331
445 Hoes Lane
Piscataway, NJ 08855-1331 USA

1-8oo-678-IEEE (IndividuallMember Orders)

1-800-701-IEEE (Institutional Orders)
1-732-981-0060
1-732-981-9667 (FAX)
email: customer.service@ieee.org
 FOREWORD

The problem of power system harmonics is not new. Utilities recognizedthe consequences ofhannonics in
the 1920s and early 1930s when distorted voltage and current waveforms were observed on transmission lines. At
that time, the major concerns were the effects of harmonics on electric machines, telephone interference and power
capacitor failures. Although such concerns stillexist today, harmonics are becoming a serious problem, potentially
damaging consumer loads as well as power delivery equipment because of the substantial increase of harmonic-
producing loads in recent years.

Significant efforts have been made in the past two decades to improve the management of harmonics in
power systems. Standards for harmonic control have been established. Sophisticated instruments for harmonic
measurements are readily available. The area of power system harmonic analysis has also experienced significant
developments and well-accepted component models, simulation methods and analysis procedures for conducting
harmonic studies have been established. Harmonic studies are becoming an important component of power system
analysis and design.

The progress in the area of power system harmonic modeling and simulation and the need of practicing
engineers to upgrade their harmonic analysis skills were recognized by the Power System Harmonics Working
Group of the IEEE Power Engineering Society and the Harmonics Working Group of the IEEE Industry
Applications Society. Under the sponsorship of the Transmission and Distribution Committee in the IEEE-PES and
the Power Systems EngineeringCommittee in the IEEE-lAS, the Harmonics Modeling and Simulation Task Force
of the PES Harmonics Working Group and the lAS Harmonics Working Group have developed this tutorial on
harmonics modeling and simulation. The purpose of the tutorial is to summarize the developments in the area from
both theoretical as well as application perspectives. Latest and proven techniques for harmonic modeling and
simulation are discussed along with case studies. By focusing on the practical aspects of applying harmonic
modeling and simulationtheories,the tutorial is expectedto provide readerswith a sound theoretical background as
well as practical guidelinesfor harmonic analysis.

We begin the tutorial with an introduction to the objectives and key issues of harmonics modeling and
simulation. The theory of Fourier analysis is discussed for applications in power system harmonic analysis. Detailed
discussion of modeling of electricnetworks and components including harmonic-producing devices then follows in
several papers. Various network solution techniques for harmonic power flow and frequency scan calculations are
summarizedand case studies are used to demonstrate the practical aspects of harmonic analysis. Three harmonictest
systems are presented. Finally, the areas that stillneed further research and developmentare discussed in the closing
comments of this tutorial.

This tutorial material has drawn on the considerable expertise of the Harmonics Working Groups and their
task forces. The contributors have generously donated their time and effort to what we believe will be a valuable
reference work on the subject. In addition, guidance and encouragement of Mr. Tom Gentile, Chair of the PES
Harmonics Working Group and Dr. Mack Grady, Chair of the General Systems Subcommittee of the PES T&D
Committee made our task much easier. Dr. M.E. El-Hawary, Chair of Life Long Learning Subcommittee of the
IEEE Power Engineering Education Committee, provided generous support to many aspects of this activity. We
wish to take this opportunityto thank all contributors for their effort in completingthis task.

Mark Halpin, Tutorial Editor

Wilsun Xu, Tutorial Organizer and Chair, PES Harmonics Modelingand SimulationTask Force
Satish Ranade, Past Chair, PES Harmonics Modeling and Simulation Task Force
Paulo F. Ribeiro, Vice Chair, PESHarmonics Modeling and Simulation Task Force

iii
 Tutorial Contents

I~H~ii~~t%rjil~~~iwil;1!'j~1@~j~IIi.4Ji~j~lm~~~~~~m~mlf:{~imt,~lillil~!.1~I~al,.ilijilt~~W1IMJ~~~~~If..:~~;~~~~~~j:~1
Forward M. Halpin, W. Xu,
S. Ranade
1. An Overview of Harmonics Modeling and S. Ranade, W. Xu 1
Simulation
2. Harmonics Theory G. Chang 8
3. Distribution System and Other Elements P. Ribeiro 15
Modeling
4. Modeling ofHarmonic Sources: Power G. Chang, W. Xu 28
Electronic Converters
5. Modeling ofHannonic Sources - Magnetic Y. Liu, Z. Wang 35
Core Saturation
6. Harmonic Modeling of Networks T. Ortmeyer, M.F. 43
Akram, T. Hiyama
7. Frequency-Domain Harmonic Analysis M. Halpin, P. 49
Methods Ribeiro, J.J. Dai
8. Time Domain Methodsfor the Calculation of C. Hatziadoniu 55
Harmonic Propagation and Distortion
9. Analysis of Unbalanced Harmonic Propagation w. Xu, S. Ranade 61
in Multiphase Power Systems
10. Harmonic Limit Compliance Evaluations Using M. Halpin, 67
IEEE 519-1992 R. Burch
11. Test Systems for Harmonics Modeling and W.Xu 71
Simulation
Conclusions M. Halpin 78
Author Biographies 79

v
 Chapter 1
AN OVERVIEW OF HARl\10NICS MODELING AND SIMULATION

s. J. Ranade W.Xu
New Mexico State University University of Alberta
Las Cruces, NM, USA Edmonton, Alberta, Canada

1.1 Introduction The waveform is aperiodic but can be expressed as a

trigonometric series [3]. In this case the components in
Distortion of sinusoidal voltage and current waveforms the Fourier series that are not integral multiples of the
caused by harmonics is one of the major power quality power frequency are sometimes called 'non-integer'
concerns in electric power industry. Considerable efforts harmonics.
have been made in recent years to improve the
management of harmonic distortions in power systems. 1.00 ~
0.90
Standards for harmonic control have been established.
Instruments for harmonic measurements are widely 0.80

available. The area of powersystemharmonic analysis has 0.70 Harmonic spectrum

also experienced significant advancement [1,2]. Well- -8 0.60
::I
accepted component models, simulation methods and i 0.50
analysis procedures for conducting systematic harmonic ~ 0.40
studies have been developed. In this chapter we present an 0.30
overview of the harmonics modeling and simulation issues 0.20 .
and alsoprovide an outline of this tutorial. 0.10
.:-: -.-.
1.2 Fourier Series and Power System Harmonics 1 3 S 7 9 11 13 1S 17 19 21 23 2S 27 29 31
Harmonic Order
Fourier Series: The primary scopeof harmonics modeling Figure 1.1. A harmonic (amplitude) spectrum.
and simulation is in the study of periodic, steady-state
distortion. The Fourier series for a regular, integrable,
periodic function f(t), of periodT seconds and fundamental 1.5
frequency f=lrr Hz, or eo=21tfradls, can be written as [3]: Fundamental
00

f(t)=C o+ :LCacos(nmt+9a) (1.1) 0.5

11=1
where Co is the dc value of the function. ell is the peak
value of the nTh harmonic component and ell is its phase -0.5
angIe. A plot of normalized harmonic amplitudes eric} is
called the harmonic magnitude spectrum, as illustrated in -1
Figure 1.1. The superposition of harmonic components to
createthe original waveform is shownin Figure 1.2 -1.5
o 0.0041 0.0082 0.0123 0.0164
Domain of Application: In general one can think of Time Sees
devices that produce distortion as exhibiting a nonlinear
relationship between voltage and current. Such Figure 1.2. Synthesis of a waveform from harmonics.
relationships can lead to several forms of distortion The waveform is aperiodic where the Fourier series is an
summarized as: approximation [4].
A periodic steady-state exists and the distorted The first case is commonly encountered and there are
waveform has a Fourier series with fundamental several advantages to using the decomposition in tenns of
frequency equal to power system frequency. harmonics. Harmonics have a physical interpretation and an
A periodic steady state exists and the distorted intuitive appeal. Since the transmission network is usually
waveform has a Fourier series with fundamental modeled as a linear system, the propagation of each
frequency that is a sub-multiple of power system harmonic can be studied independent of the others. The
frequency. number of harmonics to be considered is usually small.
which simplifies computation. Consequences such as losses
1
can be related to harmonic components and measures of network or load unbalance. Even for small deviations from
waveform quality can he developed in terms of harmonic balanced conditions at the fundamental frequency. it has
amplitudes. been noted that harmonic unbalance can be significant. In
the unbalanced case line currents and neutral currents can
Certain types of pulsed or modulated loads create contain all orders of harmonics and contain components of
waveforms corresponding to the second category. The third all sequences. Three-phase power electronic converters can
category can occur in certain pulse-width modulated generate non-characteristic under unbalancedoperation.
systems. Some practical situations such as arc furnaces and
transformer inrush currents correspond to the fourth case. 1.3 Harmonics Modeling and Simulation
DC arc furnaces utilize conventional multiphase rectifiers
but the underlying process of melting is not a stationary The goal of harmonic studies is to quantify the distortion
process. When reference is made to harmonics in this in voltage and current waveforms at various points in a
instance it corresponds to the periodic waveform that power system. The results are useful for evaluating
would be obtained if furnace conditions were to be corrective measures and troubleshooting harmonic caused
maintained constant over a period of time. Harmonics problems. Harmonic studies canalso determine the existence
modeling can lend insight into some of the potential of dangerousresonant conditions and verify compliance with
problems but transient studies become veryimportant. harmonic limits. The need for a harmonic study may be
indicated by excessive measured distortion in existing
The Origin of Harmonics: Main sources of harmonics in systems or by installation of harmonic-producing
conventional power systems are summarized below. equipment. Similar to other power systems studies the
harmonics study consists of the following steps:
1. Devices involving electronic switching: Electronic
power processing equipment utilizes switching devices. Definition of harmonic-producing equipment and
The switching process is generally, but not necessarily, detennination of models for their-representanon,
synchronized to the ac voltage. Determination of the models to represent other
2. Devices with nonlinear voltage-current relationships: components in the system including external networks.
Iron-core reactors and arcing loads are typical examples Simulation of the system for various scenarios.
of such devices. When excited with a periodic input
voltage the nonlinear v-i curve leads to the generation Many models have been proposed for representing
of harmonic currents: harmonic sources as well as linear components. Various
network harmonic solution algorithms have also been
Distortion Indices: The most commonly used measure of published. In the following sections, we briefly summarize
the quality of a periodic waveform is the total harmonic the well-accepted methods for harmonic modeling and
distortion (1lID). simulations. Other chapters in this tutorial will expandupon

J~ C~
these ideas and illustrate how to set up studies in typical
TIID = I Cl (1.2) situations.
IEEE Std. 519 [5] recommends limits on voltage and
current THD values. Other indi~ such as telephone
interference factor (TIF) and leT product are used to
measure telephone interference. The K-faetor indices are
used to describe the impact of harmonics onlosses and are
useful in de-rating equipment suchas transformers.
Harmonics in Balanced and Unbalanced Three-Phase
Systems: In balanced three-phase systems and under
balanced operating conditions, harmonics In' each phase
have specificphase relationships. For example, in the case
of the third harmonic, phase b currents would lag those in
phasea by 3x120 or 360, and those in phase c wouldlead
by the same amount. Thus, the third harmonics have no
phase shift and appear as zero-sequence components.
Similar analysis shows that fifth harmonics appear to be of
negative sequence, seventh are of positive sequence, etc.
System impedances must be appropriately modeled based
on the sequences.
The magnitudes and phase angles (in particular) of
three-phase harmonic voltages and currentsare sensitiveto
1.4 Nature and Modeling of Harmonic Sources
The most common model for harmonic sources is in the
form of a harmonic current source, specified by its
magnitude and phase spectrum. The phase is usually defined
with respect to the fundamental component of the tenninal
voltage. The data can be obtained form an idealjzed
theoretical model or from actual measurements. In many
cases, the measured waveforms provide a more realistic
representation of the harmonic sources to be modeled. This
is particularly true if the system has significant unbalances
or if non-integer harmonics are present When a system
contains a single dominant source of harmonics the phase
spectrum is not important However, phase angles must be
represented when multiple sources are present A common
method is to modify the phase spectrum according to the
phase angle of the fundamental frequency voltage seen by
the load. Ignoring phase angles does not always result in the
'worst case'.
More detailed models become necessary if voltage
distortion is significant or if voltages are unbalanced. There
2
are three basic approaches that can be taken to develop source. The devices are sensiuve to supply voltage
detailed models: unbalance. For large power electronic devices such as
HVDC terminals and transmission level SVCs, detailed
Develop analytical formulas for the Fourier series as a three-phase models may be needed. Factors such as tiring-
function of terminal voltage and operating parameters angle dependent harmonic generation and supply voltage
for the device. unbalance are taken into account in the model. These studies
Develop analytical models for device operation and normally scan through various possible device operating
solve for device current waveform by a suitable conditions and filter performance,
i terati ve method.
Solve for device steady state current waveform using Rotating Machines: Rotating machines can be a harmonic
time domainsimulation. source as well. The mechanism of harmonic generation in
synchronous machines is unique. It cannot be described by
Advanced models require design data for the device. using either the nonlinear v-i device model or the power
For example, for a medium power ASD it is necessary to electronic switching model. Only the salient pole
specify parameters such as transformer data, de link data synchronous machines operated under unbalanced
and motor parameters. Apart from potentially higher conditions can generate harmonics with sufficient
accuracy, an important advantage of such detailed models magnitudes. In this case, a unbalanced current experienced
is that the user can specify operating conditions, e.g., motor by the generator induces a second harmonic current in the
speed in a drive, rather than spectra. field winding, which in tune induces a third harmonic
current in the stator. In a similar manner, distorted system
In the analysis of distribution and commercial power voltage can cause the machines to produce harmonics.
systems one may deal with a harmonic source that is an Models to represent such mechanisms have been proposed
aggregate of many sources. Such a source can be modeled [1]. For the cases of saturation-caused harmonic generation
by measuring the aggregate spectrum. It is very difficult to from rotatingmachines, the non1ine3(~i modelcan be used.
develop a current source type model analytically based on
the load composition data. Reference [7] has pointed out High frequency sources: Advances in power electronic
that the aggregate waveforms can be much less distorted devices have created the potential for a wide range of new
than individual device waveforms. power conversion techniques. The electronic ballast for
fluorescent lighting is oneexample. In general, these systems
Harmonie sources' may also exhibit time-varying employ high frequency switching to achieve greater
characteristics. Since standards and practice permit flexibility in power conversion. With proper design, these
harmonic guidelines to be violated for short periods of techniques can be used to reduce the low frequency
time, including the time-varying characteristics of harmonics. Distortion is created at the switching frequency,
hannonic sources can be useful and can present a more which is generally above 20 kHz. At such high frequency,
realistic picture of actual distortions. More research is current distortion generally does not penetrate far into the
needed in this area [8]. system but the possibility of system resonance at the
switching frequency can still exist
Nonlinear Voltage-Current Sources: The most common
sources in this category are transformers ( due to their Non-integer harmonic sources: There exist several power
nonlinear magnetization requirements), fluorescent and electronic systems which produce distortion at frequencies
other gas discharge lighting, and devices such as arc- that are harmonics of a base frequency other than 60 Hz.
furnaces. In all cases there exists a nonlinear relationship There are also devices that produce distortion at discrete
between the current and voltage. The harmonic currents frequencies that are not integer multiples of the base
generated by these devices can be significantly affected by frequency. Some devices havewaveforms that do not submit
the waveforms and peak values of supply voltages. It is to a Fourier or trigonometric series representation. Lacking
desirable to represent the devices with their actual standard terminology, we will call these non-harmonic
nonlinear v-i characteristicsin harmonic studies, instead of sources. Modeling of this type of harmonic sources has
as voltageindependent harmoniccurrent sources. attracted many research interests recently.

Power Electronic Converters: Examples of power
electronic devices are adjustable speed drives, HVDC
links, and static var compensators. Compared to the non-
linear v-i devices, harmonics from these converters are less
sensitive to supply voltage variation and distortion.
Harmonic current source models are therefore commonly
used to represent these devices. As discussed before, the
phase angles of the current sources are functions of the
supply voltage phase angle. They must be modeled
adequately for harmonic analysis involving more than one
1.5 Network and Load Models
NetworkModel: The main difficulty in setting up a network
model is to determine how much of the network needs to be
modeled. 1be extent of network representation is limited by
available data and computing resources. The following
observations can be made:
For industrial power systems connected to strong or
dedicated three-phase distribution feeders it is generally
3
 sufficient to model two transtormations from the load
point. Generally, transformer impedances dominate.
Branch circuits should be modeled if they connect to
power factor correction capacitors or motors. Although
capacitance of overhead lines is usually neglected,
cable capacitance should be modeled for cables longer
than 500 feet.
Large industrial facilities are served at sub-
transmission and even transmission voltage. In this
case it is important to model at least a portion of the
HVIEHV network if the facility has multiple supply
substations. If it has only one supply substation,
utilities may provide the driving-point impedance seen
by the facility.
Distributionfeeders (at least in the US andCanada) are
unbalanced and loads are often served from single
phase laterals. Shunt capacitors are extensively used.
Thus it becomes mandatory to model the entire feeder,
and sometimes adjacent feeders as well.
The above observations are not guaranteed rules, but are
based on common practice. Perhaps the best way to
determine the extent of network modeling needed is to
perform a sensitivity study; i,e., one can progressively
expandthe network model until the results do not change
significantly. In many harmonic studies involving
industrial plants, the supply system is represented as a
frequency-dependent driving-point impedance at the point
of common coupling.
Overhead Linesand Underground Cables: Modeling of
lines and cables over a wide range of frequencies is
relatively well documented in literature [9]. Typical lines
or cablescan be modeled by muItiphase coupled equivalent
circuits. For balanced harmonic analysis the models can be
further simplified into single-phase. pi-circuits using
positive and zero sequence data. The main issues in
modeling these components are the frequency dependence
of per-unit length series impedance and the long line
effects. As a result, the level of detail of their models
depends on the line length and harmonic order:
In industrial systems and utility distribution systems
where line lengths are short it is customary to use
sequence impedances. Capacitanceis usually neglected
except in the case of long cableruns,
An estimate of line-length beyond whicb long line
models should be used is 1501n miles for overhead line
and 90In miles for underground cable, where n is the
harmonic number.
Skin effect correction is important in EHV systems
because line resistance is the principal source of
damping.
Transformers: In most applications, transformers are
modeled as a series impedance with resistance adjusted for
skin effects. This is because adequate data is usually not
available. Three-phase transformer connections may
provide 30 phase shift. Other connections such as zigzag
windings are used to mitigate harmonics. The phase shifts
associated with transformer connections must be accounted
for in multiplesource systems.
Other considerations include the nonlinear characteristics
of core loss resistance, the winding stray capacitance and
core saturation. Harmonic effects due to nonlinear resistance
are small compared to the nonlinear inductance. Effects of
stray capacitance are usually noticeable only for frequencies
higher than 4 kHz. The saturation characteristics can be
represented as a harmonic source using the nonlinear v-i
model if saturation-caused harmonic generation is of
concern.
Passive Loads: Linear passive loads have a significant
effect on system frequency response primarily near resonant
frequencies. As in other power system studies it is only
practical to model an aggregate load for which reasonably
good estimates (MW and MYAR) are USUally readily
available. Such an aggregate model should include the
distribution or service transformer. At power frequencies the
effect of distribution transformer impedance is not of
concern in the analysis of the high voltage network. At
harmonic frequencies the impedance gf.. the transfonner can
be comparable to that of motor loadS, because induction
motors appear as locked-rotor impedances at these
frequencies.
A general model thus appears as in Figure 1.3. To
characterize this model properly, it is necessary to know the
typical composition of the load. Such data are usually not
easily available. The following models have been suggested
in literature (n represents the harmonic order):
Model A : Parallel R,L with R = V2/ (P); L = V 2/(21CfQ)
This model assumes that the total reactive load is assigned to
an inductor L. Because a majority of reactive power
corresponds to induction motors, this model is not
recommended.
Model B : Parallel R,L with
R = V2/ (k*P), L = V 2/ (21tfk*Q) ; k= .1h+.9
Model C : Parallel R,L in series with transfonner inductance
Ls, where
=
R V21P; L =n R/(21tf 6.7*(QIP)-.74);Ls= .073 b R
Model C is derived from measurements on medium voltage
loads using audio frequency ripple generators. The
coefficients cited above correspond to one set of studies
[10], and may not be appropriate for all loads. Load
representation for harmonic analysis is an active research
area.
4

Figure 1.3: Basic Load Model.
Large Rotating Loads: In synchronous and induction
machines the rotating magnetic field created by a stator
harmonic rotates at a speed significantly different from that
of the rotor, Therefore at harmonic frequencies the
impedance approaches the negative sequence impedance.
In the case of synchronous machines the inductance is
usually taken to be either the negative sequence impedance
or the average of direct and quadrature sub-transient
impedances. For induction machines the inductance is
taken to be the locked rotor inductance. In each case the
frequency-dependence of resistances can be significant.
The resistance normally increase in the form n' where n is
the harmonic order and the parameter 'a' ranges from 0.5-
1.5. Most motors are delta-connected and therefore do not
provide a path for zero-sequence harmonics.
1.6 Harmonic Simulation
It is appropriate to note that a large number ofharmonic
related problems encountered in practice involve systems
with relatively low distortion and often a single dominant
harmonic source. In these cases simplified resonant
frequency calculations, for example, can be performed by
hand [5] and distortion calculations can be made with a
simple spreadsheet For larger systems and complicated
harmonic producing loads, more fonnal harmonic power
flow analysis methods are needed. In this section,
techniques presently being used for harmonics studies are
reviewed. These techniques vary in terms of data
requirements, modeling complexity, problem formulation,
and solution algorithms. New methods are being developed
and published. .
Mathematically, the harmonic study involves solving
the network equation for each harmonic written in matrix
form as
[Iml = [YnJ[VnJ m=I, 2 ... n (1.3)
where [Yml represents the nodal admittance matrix, [lml is
the vector of source currents and [Vml is the vector of bus
voltages for harmonic number m. In more advanced
approaches thecurrent source vector becomes a function of
bus voltage.
Frequency Scan: The frequency .scan is usually the first
step in a hannonicstudy. A frequency or impedance scan is
a plot of the driving point (Thevenin) impedance at a
system bus versus frequency. The bus of interest is one
where a harmonic source exists. For simple system this
impedance can be obtained from an impedance diagram.
More formally, the Thevenin impedance can be calculated
by injecting a 1 per unit source at appropriate frequency into
the bus of interest. The other currents are set to zero and
(1.3) is solved for bus voltages. These voltages equal the
drtving-point and transfer impedances. The calculation is
repeated over the harmonic frequency range of interest.
Typically, a scan is developed for both positive and zero
sequence networks.
If a harmonic source is connected to the bus of interest,
the harmonic voltage at the bus is given by the harmonic
current multiplied by the harmonic impedance. The
frequency scan thus gives a visual picture of impedance
levels and potential voltage distortion. It is a very effective
tool to detect resonances which appear as peaks (parallel
resonance) and valleys (series resonance) in the plot of
impedance magnitude vs. frequency.
Simple Distortion Calculations: In the simplest harmonic
studies harmonic sources are represented as current sources
specified by their current spectra. Admittance matrices are
then constructed and harmonic voltage components are
calculated from (1.3). The hannomc'" current components
have a magnitude determined from the typical harmonic
spectrum and rated load current for the harmonic producing
device.
where n is the harmonic order and the subscript 'spectrum'
indicates the typical harmonic spectrum of the element To
compute indices such as THD the nominal bus voltage is
used.
For the multiple harmonic source cases it is important to
also model the phase angle of harmonics. A fundamental
frequency power-flow solution is needed, because the
harmonic phase angles are functions of the fundamental
frequency phaseangleas follows:
9n = 9D-spedIUm + n(81 -91-spectzum}
where 91 is the phase angleof the harmonic source currem. at
the fundamental frequency. 9n-specllUm is the phase angle of the
n-th harmonic current spectrum. Depending on the phase
angles used, the effects of multiple harmonic sources can
either add or cancel. Ignoring phase relationships may,
therefore, lead to pessimistic or optimistic results.
Harmonic Power Flow Methods: The simple distortion
calculation discussed above is the basis for most harmonic
study software and is useful in many practical cases. The
main disadvantage of the method is the use of 'typical'
spectra. This prevents an assessment of non-typical
operatingconditions. Such conditions include partial loading
of harmoruc-producing devices, excessive distortion and
unbalance. To explore such conditions the user must develop
5
typical spectra tor each condition when using the
simplified method. The disadvantages have prompted the
development of advanced harmonic analysis methods. The
goal is to model the physical aspects of harmonic
generation from the device as a function of actual system
conditions.
The general idea is to create a model for the harmonic
producing device in the form
Here V t, V2, , Vn are harmonic voltage components, It, 12,
..., In, are corresponding harmonic currentcomponents and
C represents multiple operating and design parameters.
Equation (1.4)permits the calculation of harmonic currents
from voltages and includes power flow constraints. The
total procedure is to simultaneously solve (1.3) and (1.4).
One of the well-known methods is the so called
"harmonic iteration method" [11,12]. Equation (1.4) is first
solved using an estimated supply voltage. The resulting
current spectrum is used in (1.3) to calculate the supply
voltage. This iterative process is repeated until
convergence is achieved. Reliable convergence is achieved
although difficulties may occur when sharp resonances
exist Convergence can be improved by including a
linearized model of (1.4) in (1.3). A particular advantage of
this "decoupled" approach is that device models in the
form of (1.4) can be in a closed form, a time domain
model, or in any other suitable form.
Another method is to solve (1.3) and (1.4)
simultaneously using Newton type algorithms. This
method requires that device models be available in closed
form whereinderivatives can be efficiently computed [13].
The various methods above can be extended, with a
significant increase in computational burden, to the
unbalanced case. Both (1.3) and (1.4) are cast in a multi-
phase framework [11,14]. Such an approach can have
several advantages. The first is the modeling of zero
sequence current flow. Second is the capabiltty of
addressing non-charaeteristic harmonics.
Finally, it is appropriate to note that harmonic studies
can be performed in the time domain. The idea is to run a
time-domain simulation until a steady state is reached. The
challenge is first to identify that a steady-state has indeed
been achieved. Secondly, in lightly damped systems
techniques are needed to obtain the steady-state conditions
within a reasonable amount of computation time.
References [14,15] provide examples of such methods.
1.6 Summary
Harmonic studies are becoming an important
component of power system planning and design. In using
software to analyze practical conditions it is important to
understand the assumptions made and the modeling
capabilities. Models and methods used depend upon system
complexity and data availability. The purpose of this tutorial
is to suggest what is required to set up harmonics studies
with emphasis on modeling and simulation.
This overview has attempted to summarize key ideas
from chapters that follow. The propagation of harmonic
current in a power system, and the resulting voltage
distortion, depends on the characteristics of harmonic
sources as well as the frequency response of system
components. Characteristics of various harmonic sources
and consideration in their modeling have been summarized.
Component modeling has been described. Different
approaches to conduct analysis werediscussed in a common
framework. Subsequent chapters of this tutorial will expand
upon each of these topicsandprovided illustrative examples.
Acknowledgments
This chapter was adapted from a paper developed by the
Task Force on Harmonics Modeling and Simulation [1].
References
1. Task force on Harmonics Modeling and Simulation,
"The modeling and simulation of the propagation of
harmonics in electricpower networks PartI : Concepts,
models and simulation techniques, IEEE Tranasactions
It
on Power Delivery, Vol.l l, No.1, January 1996, pp.
452-465.
2. Task force on Harmonics Modeling and Simulation,
"The modeling and simulation of the propagation of
harmonics in electric power networks Part II : Sample
systems and Examples," IEEE Tranasactions on Power
Delivery, Vol.I 1, No.1 January 1996, pp. 466-474.
t
3. A. Guillemin, The Mathematics of Circuir Analysis,
John Wiley and Sons,INC., NewYork, 1958.
4. Corduneanu, Almost Periodic Functions, John Wiley
(Interscience), New York, 1968.
5. IEEE Recommended Practices and Requirements for
Harmonic Control in Electric Power Systems," IEEE
Standard519-1992, IEEE, New York, 1992.
6. Emanuel, A,E, Janczak, J., Pillegi, D.O., Gulachenski, E.
M., Breen, M., Gentile, T J., Sorensen, D.,"Distribution
Feeders with Nonlinear Loads in the Northeast USA:
Part l-Vojtage Distortion Forecast, IEEE Transactions
on Power Delivery, Vol.10,No.1, January 1995, pp.340-
347.
7. Mansoor, ~ Grady, W.M, Staats, P. T., Thallam, R. S.,
Doyle, M. T., Samotyj, " Predicting the net hannonic
currents from large numbers of distributed single-phase
computer loads:' IEEE Trans. on. Power Delivery, Vol.
10, No.4, Oct.. 1995, pp. 2001-2006.
8. Capasso, ~ Lamedica, R, Prudenzi, A, Ribeiro, P, F,
Ranade, S. J., .. Probabilistic Assessment of Harmonic
Distortion Caused by Residential Loads," Proc. ICHPS
IV, Bologna, Italy.
6
9. Dommel, "Electromagnetic Transients Program
Reference Manual (EMTP Theory Book)", Prepared
tor Bonneville Power Administration, Dept. of
Electrical Engineering, Universityof British Columbia,
Aug. 1986.
10. ClORE Working Group 36-05, "Harmonics,
Characteristics, Parameters, Methods of Study,
Estimates of Existing Values in the Network," Electra,
No. 77, July 1981, pp. 35-54.
II. W. Xu, J.R. Jose and H.W. Dommel, "A Multiphase
Harmonic Load Row Solution Technique", IEEE
Trans. on Power Systems, vol. PS-6, Feb. 1991, pp.
174-182.
12. Sharma, V, Fleming, R.I., Niekamp, L.,"An iterative
Approach for Analysis of Harmonic Penetration in
Power Transmission Networks," IEEE Trans. on
Power Delivery, Vol. 6, No.4, October 1991, pp.
1698-1706.
13. D. Xia and G.T. Heydt, "Harmonic Power Row
Studies, Part I - Fonnulation and Solution, Part IT -
Implementation and Practical Application", IEEE
Transactions on Power Apparatus and Systems, Vol.
PAS-lOl, June 1982, pp.1257-1270.
14. Lombard, X., Mahseredjian, J., Lefebvre, S., Kieny,
C.,"Implementation of a new Harmonic Initialization
Methodin EMlP,n Paper 94- 8M 438-2 PWRD, IEEE
Summer Power Meeting, San Francisco, Ca., July
1994.
IS. Semlyen, A., Medina, A., "Computation of the
Periodic Steady State in Systems with Nonlinear
Components Using a Hybrid Tune and Frequency
Domain Methodology," Paper 95- WM 146-1 PWRS,
IEEE Winter Power Meeting, New York, NY, Jan.
1995.

7
 Chapter 2

HARMONICS THEORY

Gary W. K. Chang
Siemens PowerTransmission & Distribution
Brooklyn Park, MN, USA

2.1 Introduction for all~. The smallest constant T that satisfies (2.1) is called
Fo~ most conventional analyses, the power system is
the penod of the function. By iteration of (2.1), we have
f(t)=f(t+hT), h=O, l, 2, .... (2.2)
essentially ~odeled as a linear system with passive
elements excited by constant- magnitude and constant- Let ~ function .f(t) be periodic with period T, then this
frequency sinusoidal voltage sources. However, with the function canbe represented by the trigonometric series
Widespread proliferation of power electronics loads
nowadays, significant amounts of harmonic currents are 1 00

being injected into power systems. Harmonic currents not f(1) ="2 lJo + L {ahcos(h~) +~sin(haJol)}, (2.3)
only disturb loads that are sensitive to waveform distortion, h=l
but also cause many undesirable effects on power system w~ere (i)o = 2nl T . A series such as (2.3) is called
elem~nts. As a result, harmonic studies are becoming a
trigonometric Fourier series. It can be rewritten as
growing concern.
00

Harmonics are usually defined as periodic steady state

distortions of voltage and/or current waveforms in power f(t) = Co + LChSin(h&1~+ ;h)' (24)
systems. In the harmonic polluted environment, the theory h=l
regarding harmonic quantities needs to be defined to where Co =lZo /2, ch =~a~ +/i ' and ;h = tan-leah I~).
distinguish from those quantities defined for the
fundamental frequency. Obs~rvi~g (2.4~, we see that the Fourier series expression
The purpose of this chapter is to present basic of a pen~c function represents a periodic function as a sum
harmonics theory. Initially, the Fourier series and analysis of SInUSOIdal components with different frequencies. The
methods that can be used to interpret waveform component of hlDo is called the h-th harmonic of the
phenomenon are reviewed. Some fundamentals of Fourier periodic function. Co is the magnitude of the de component.
transforms used in today's harmonics measurement The component with h = 1 is called the fundamental
component Ch and /Ph are known as the h-th order harmonic
techniques are also introduced. The general harmonics
theory, the definitions of harmonic quantities, harmonic magnitude and phase angle, respectively. The magnitude
indices in common use, and power system response and and phase angle of each harmonic determine the resulting
solutions to harmonics are then described. wavesbape of.f(t).
2.2 Fourier Series and Analysis Equation (2.3) also can be represented by its complex
fonnas
The theory of the Fourier series was first introduced by 00
~e Fr~nch physicist and mathematician, Joseph Fourier, in
~s article 'Analytic Th~ of Heat' which was published
f(t) = Lc,.ejh~, (2.5)
In 1882. The theory Involves expansions of arbitrary h=--
functions in certain types of trigonometric series. It proves where for h =0, 1, 2, ...,
that any periodic function in an interval of time could be
represented by the sum of a fundamental and a series of
higher orders of harmonic components at. frequencies
cia =- I
T 2
1 ' !(t)e-jhmoldt.
T -T/2
. (2.6)
which are integral multiples of the fundamental
component The series establishes a relationship between Orthogonal Functions
the function in time and frequency domains. Today, the A set of functions {tph (t)} is called orthogonal on an
theory has become the famous 'Fourier series' and it is one
of the most important tools for engineers and scientists in interval a < t < (J if all groups of any two functions tp.(t)
many applications. and tp j V) in the set {fJh (t)} satisfy I

Fourier Series
A periodic function can be defined as any function for
1"i
a
fJ
(t)" i (t)dt ={Or, 1=)
t
~ ~,
'.
(2.7)

which
/(t) =f(t+D (2.1)
where , is a nonzero value. It can be shown that
{1, cosliJol, ..., cosh tiJot, ...,sintlJol, .., sinhfiJol
t is an
}

orthogonal set of sinusoidal functions on interval

8
- T / 2 < t < T / 2. Using the orthogonal relations, we can Discrete Fourier Transform
show that Fourier coefficients an, ai, and bh of (2.3) are
When the frequency domain spectrum and the time
domain function are both periodic sampled functions with
=-f
2 T/2
an f(t)dt, (2.8) N samples per period, (2.14) and (2.15) can be represented
T -T/2 by the following so-called discrete Fourier transform
(DFf) pair:
ah =-T2 fTI2
-T12
f(t)cos(h(iJol)dt, and (2.9)
N-l
F(ktill.) = Lf(n~T)e-j2111cnIN (2.16)
2 JTI2
bh = - f(t)sin(h{j)ot)dt,. (2.10) n=O
T -T12 and
N-l
where h =1,2, .... f (n~T) =L F(ktill.)e j2111cn1N (2.17)
For the set of complex valued functions {tph(t)}, it can k=O

be shown that (2.7) holds when tp j t t) is the complex where k, n =0, 1, ..., N-I Jill = 211,/ aT and ~T =T / N.
t

conjugate of q; i (1) The OFf is often used in harmonic measurement because

Waveform Symmetry
A function f(t) is called an even function if it has the
property
f(-t)=/(t), (2.11)
and it is calledan odd function if
f(-t) =- /(t). (2.12)
An even function is symmetrical to the vertical axis at the
origin, and an odd Junction is anti-symmetrical to the
vertical axis at the origin. A function with a period of T is
half-wave symmetry if it satisfies the condition .
/(1) =-:-f(tT /2). (2.13)
Iff(t) has half-wave symmetry and is eitheraneven or odd
function, then it has even or odd quarter-wave symmetry.
The use of symmetry simplifies the calculation of Fourier
coefficients in (2.8) - (2.10).
Fourier Transform
The Fouriertransform of a functionf(t) is defined as
F(O) = [f(t)e-i-dt. (2.14)
and f(t) is called the inverse Fourier transform of F ( OJ) ,
which is defined as
/(1) =1- j- F(OJ)eJO%dOJ
. (2.15)
2n ....
Equations (2.14) and (2.15) are often called the Fourier
transfonn pair; and they are used to map any function in
the interval of (-00,00) in time or frequency domain into a
continuous function in the inverse domain. 'The key
property of the Fourier transform is its ability to examine a
function or wavefonnfrom the perspective of both the time
and frequency domains. A given function can have two
equivalent modes of representations: one is in the time
domain and is calledf(t), and the other is in the frequency
domain and is called F (m) . Equation (2.14) transforms
the time function into a frequency spectrum, and (2.15)
synthesizes the frequency spectrum to regain the time
function.
the measured data is always available in the form of a
sampled time function. The sampled time function is
represented by a time series of points of known magnitude
separated by fixed time intervals of limited duration.
Fourier analysis can be done by DFfs. The DFfs are
often calculated by the use of fast-Fourier transform (FFf)
algorithm [1]. FFr techniques are very fast methods for
performing the DFf calculations (Z.,iD) and (2.17) which
allow the evaluation of a large number of functions. There
are a number of available FFr algorithms that can be easily
used in harmonic analysis.
2.3 Basic Definitions of Harmonic Quantities
Conventionally, the definitions used to describe electric
quantities for power system study are for systems operating
in sinusoidal steady state. However, when harmonics are
present because of system nonlinearities, the definitions of
these electric quantities need to be modified from those
appropriate for single-frequency systems.
This section gives the basic definition of power system
harmonics and describes some useful defin1tions associated
with voltage, current, instantaneous power, average (active)
power, apparent power, reactive power, and power factor
computations undernonsinusoidal situations.
Definition of Power System Harmonics
In thepower system, the definition of a harmonic can be
stated as: A sinusoidal component of a periodic wave having
a frequency that is an integral multiple of the fundamental
frequency. Thus for a power system with fo fundamental
frequency, the frequency of the h-th order of harmonic is
hfo. Harmonics are oftenusedto define distorted sinewaves
associated with currents and voltages of different amplitudes
and frequencies.
One can compose a distorted periodic waveshape of any
conceivable shape by using different harmonic frequencies
with different amplitudes. Conversely, one can also
decompose any distorted periodic waveshape into a
fundamental wave and a set of hannonics. This
decomposition process is called Fourter analysis. With this
technique, we can systematically analyze the effects of
nonlinear elements in power systems.

9
 Most elements and loads in a power system respond the We see that each harmonic makes a contribution, either plus
same in both positive and negative half-cycles. The or minus, to the average power. There are no contributions
produced voltages and currents have half-wave symmetry. to the average power from the voltage at one frequency and
Therefore, harmonics of even orders are not characteristic. the current at another. The average power generated by
Also, triplens (multiples of third harmonic) always can be harmonics is usually very small in comparison with the
blocked by using three-phase ungrounded-wye or delta fundamental average power.
transformer connections in a balanced system, because
triplens are entirely zero sequence. For these reasons, By applying orthogonal relations, the nns values of (2.18)
even-ordered and triplens are often ignored in harmonic and (2.19) are proved to be
analysis. Generally, the frequencies of interests for
harmonic analysis are limited to the 50th multiple. Vrms = 1 T 2
T 10 v (t )dt = ft(i
L vl ,
h=1
(2.23)
One majorsource of harmonics in the power system is

JF
and
the static power converter. Under ideal operating
conditions, the current harmonics generated by a p-pulse
line-commutated converter can be characterized by
lh = III h and h =pn 1 (characteristic harmonics) where
t.; = -1
T
1 i T 2 (t )dt
0
= L/i '
h=l
(2.24)

n = 1, 2, ... and p is an integral multiples of six. If 1) the respectively.

converter input voltages are unbalanced or 2) unequal The apparent power is given by
commutating reactances exist between phases or 3)
unequally spaced firing pulses are present in the converter S=Vrmslrms (2.25)
bridge, then the converter will produce non-characteristic
harmonics in addition to the characteristic harmonics. Non- A widelyaccepteddefinition of apparentpower is
characteristic harmonics are those that are not integer S2 = p2 + (jl + D 2 , (2.26)
multiples of the fundamental power frequency.
The harmonic frequencies that are not integral multiples where Q is the reactive power defmed:as
of the fundamental power frequency are usually called 00

interharmonics. A major source of interharmonics is the

cycloconverter [2]. One special subsetof interharmonics is
called sub-harmonics. Sub-harmonics have frequency
values that are less than that of the fundamental frequency.
lighting flicker is one indication of the presence of sub-
harmonics. A well-known source of flicker is the arc
furnace [3].
Electric Quantities Under Nonsinusoidal Situation
When steady-state harmonies are present, instantaneous
voltage and current can be represented by Fourier series as
follows:
00 00
Q= I, Vl& l l& sin( 8/a - 81&)' (2.27)
h=l
and D is defined as the distortion voltamperes which
correspond to the products of voltages and currents of
differentfrequency components in (2.18) and (2.19).
When harmonics are not present in (2.25), S is equal to
VtI I which is the conventionally defined apparent power at
fundamental frequency. Under the sinusoidal situation, the
power equation relates mutually the average, reactive, and
apparent power, and it is defined as

v(t) =~>h(t) =I,.J2Vhsin(hfl'ot+ ~), (2.18) <Vi 11) 2 =if + G2 , (2.28)

h=l h=l
where Q1 = Vi /1sin( ~ - ~) is the fundamental reactive
i(t) =
00 00

Lih(t) = I,-Iit, sin(hOJot +~), (2.19)

power defined in (2.27) for h 1. =
h=l h=l
At present, there is still no consensus in the definitions
and physical meanings regarding reactive power and
where the de terms are usually ignored for simplicity, V h distortion power among researchers and scientists [4-7]. In
and t, are rms values for h-th order of harmonic voltage [8], some alternate definitions with interpretations on power
and current, respectively. definitions other than the above are described.
The instantaneous power is defined as The conceptof power factor originated from the need to
p(t) = v(t)i(t), (2.20) quantify how efficiently a load utilizes the current that it
draws from the ac power system. Regardless of sinusoidal
andthe average power over one period T of p(t) is defined or nonsinusoidal situation, the total powerfactor is defined
as as
lITp(t)dt.
P =- (2.21)
p!=-,
p
(2.29)
T 0 s
If we substitute (2.18) and (2.19) into (2.20) and make use where P is the average power contributed by the
of the orthogonal relations of (2.7), it can beshown that fundamental frequency Component and other harmonic
00 00 components, as shown in (2.22). In the next section, we also
p= I,VIJhcos(8h-8i1) =LJ;. (2.22) will show the relationship between the power factor and
h=l h=l some harmonic distortion indices.
10
Phase Sequences of Harmonics
For a three-phase balanced system under nonsinusoidal
conditions, the Iz-th order of harmonic voltage of each
phase can be expressed as
vaJ,(t) =.fiv,r sin(hCtJot + ()h)'
vbh (I) =.fivh sin{hCtJot - 2hn /3+ 6h ) ,
vch (t) =.fivh sin(haJol + 2hn /3 + 6h ) .
Therefore, the harmonic phase sequence in a balanced
three-phase system has the patternshown in Table 1.1.
Table 1.1. Harmonic PhaseSequences in a Balanced
Three-Phase Power System
Harmonic Phase
Order Sequence
1 +
2 -
3 0
4 +
5 -
6 0
zero. Similarly, the measures of individual harmonic
distortion for voltage and current at Iz-th order are defined as
vh IV I and lhlll' respectively.
Total Demand Distortion
(2.30) The total demand distortion (TOO) is the total harmonic
currentdistortion defined as
(2.31)
(2.32) TDD =V
~
6. lh (2.34)
IL '
where I L is the maximum demand load current (15- or 30-
minute demand) at fundamental frequency at the point of
common coupling (Pee), calculated as the average current
of the maximum demands for the previous twelve months.
The concept of TOO is particularly relevant in the
application of IEEE Standard 519.
Telephone Influence Factor
Telephone influence factor (TIF) is a measure used to
describe the telephone noise originating from harmonic
currents and voltages in power systems. TIF is adjusted
based on the sensitivity of the telephone system and the
human ear to noises at various frequencies. It is defined as
00 00

L(WhVh)2 L(wh l h)2

Observing Table 1.1, we find that the negative and zero TlFv = h=l or TIF] = h=l , (2.35)
sequences are also present in the system, and all triplens Vmu Inns
are entirely zero sequence. The above simple phase
sequence pattern does not hold for the unbalanced system, where wh is a weighting accounting for audio and
because harmonicsof each order contain the threedifferent inductive coupling effects at the h-th harmonic frequency.
sequences. It requires a more complicated analysis [9]. Obviously, TIF is a variation of the previously defined
THD where the root of the sum of the squares is weighted
The definitions in (2.18) - (2.24) are also suitable for using factors that reflectthe response in the voice band.
three-phase balanced system. However, for the unbalanced
system, the apparent power needs to be redefined and the VeT and IT Products
consensus has yet to be reached. Reference [10] provides Another distortion index that gives__ a measure of
some practical power definitions under unbalanced harmonic interference on audio circuits similar to TIF is
conditions. the V-Tor IT product, where V is rms voltage in volts, I is
2.4 Harmonic Indices rIDS current in amperes, and T is the TIF. In practice,
telephone interference is often expressed as V T or IT,
In harmonic analysis there are several important indices which is defined as
used to describe the effects of harmonics on power system _ 00

components and communication systems. This section

describes the definitions of those harmonic indices in V T= L(W"V,,)2 or I T= L(whlh)2, (2.36)
common use [11-13]. h=1 h=l

Total Harmonic Distortion (Distortion Factor) where wh is the same as previously described. If kVT or
kl-T is used, then the index must be multiplied by a factor
The mostcommonly used harmonic index is of 1000. EQuation (2.36) refers to the fact that the index is
a product Of harmonic voltageor harmonic current and the
Jf~2 Jill corresponding telephone influence factor. Observing
(2.35)and (2.36), we fmd that
THn-
~
= h=2
~
or THD
1
h=2
11'
(2.33) TIFv Vmrs = V T and TIFf I rms I T.= (2.37)
C-Message Weighted Index
which is defined as the ratio of the rms value of the
harmonic components to the rms value of the fundamental The C-message weighted tndexis similarto TIF, except
component and usually expressed in percent This index is that each weighting Ch is used in place of wh. The
used to measure the deviation of a periodic waveform weighting is derived from listening tests to indicate the
containing harmonics from a perfect sinewave. For a relative annoyance or speech impairment by an interfering
perfect sinewave at fundamental frequency, the THD is
11
signal of frequency f as heard through a -SOO-type"
telephone set. This index is defined as
cv =- - - - or C/ =--------
!he relation between TIF weight and C-message weight
by substituting (2.23) and (2.24) into (2.33). The total
power factor in (2.29) becomes
p
pltot = 1 (2.43)
Yi/lv l + (THDy /1(0)2 ~1 + (THD[ /1(0)2
(2.38)
~n most ~ases, only very sm~l portion of average power of P
IS contributed by harmonics and total harmonic voltage
distortion is less than 10%. Thus (2.43) can be expressed as

=- 11 -;==================
IS
1
pltot
}\.'h =5Chfh' (2.39) VIII ~1 + (THD[/lOO)2
where fh is the frequency of the h-th order harmonic. =cos(61 - ~1) pIdist (2.44)

Transformer K-Factor In (2.44), the first term, cos( 61 - t51) , is known as the
displacement power factor, and the second term, P!dist' is
Transformer K-factor is an index used to calculate the defined as the distortion power factor. Because the
derating of standard transformers when harmonic currents
are present [14]. The K-factor is defined as displacement powerfactor is always not greater than one, we
have
00
P/'ot s pIdist (2.45)
L,h2 (lh I /})2
Obvi~sly, . for single-phase nonlinear loads with high
K =h=~ (2.40) current distortton, the total power factor is poor. It also
L(lh lll )2 should be noted that adding power factor correction
h=1 cap~~rs to such load is likely_~.9 cause resonance
conditions. An alternative to improve-the distortion power
where h is the harmonic order and lhlll is the factor is using passive or active filter to cancel harmonics
corresponding individual harmonic current distortion. produced by nonlinear loads.
(2.40) is calculated based on the assumption that the
transformer Winding eddy current loss produced by each 2.5 Power System Response to Harmonics
harmonic current component is proportional to the square In comparison with the load, a power system is stiff
of the harmonic order and the square of magnitude of the enough to withstand considerable amounts of harmonic
harmonic component currents without causing problems. This means that the
The K-rated transformer is constructed to withstand ~ystem impedance is smaIler compared to the load
1~ A powersystem ~tself is not a significant source
more voltage distortion than standard transformers. The of harmomcs. However, It becomes a contributor of
K-faetor actually relates to the excessive heat that must problems by way of resonance whensevere distortion exists.
be dissipated by the transfonner. It is considered in the
design and installation stage for nonlinear loads, and it is Assuming all nonlinear loads can be represented as
used as a specification for new or replacement power barm~nic current injectio~ the harmonic, voltage at each
source equipment Table 2 shows typical commercially bus In a power system can be obtained by solving the
following Impedance matrix or nodal admittance equations
available K-rated transformers, where all regular for all orders of harmonics under consideration:
transformers fall into K-l category.
Vh = Z. III (2.46)
Table 2: Commercially Available K-RatedTransfonners or
Cate20ry Ih = Yb Vh t (2.47)
K-4 where Vh is the vector consisting of the h-th harmonic
K-9 voltage at each bus that is to be determined. Z. is the system
K-13 harmonic impedance matrix, Yb is the system harmonic
K-20 admittance mattix, and I. is the vector of measured or
K-30 estimated harmonic currents representing the harmonic-
K-40 generating loads at connected busses.
In (2.46), Z. can be obtained by using a Z.bus building
Distortion Power Factor algorithm for each harmonic of interest or from the inverse
When voltage and current contain harmonics, it can be of Yh in (247). but the harmonic effects on different power
shown[15] that system components and loads need to be properly modeled
[16]. Approaches for harmonic analysis based on (2.46) or
Vnn.r =VI ~1 + (THDv /1(0)2 (2.41) (2.47) are commonly called curreet injection methods.
and 1bese approaches are usually - used in conjunwon with
fundamental frequency load flow computations. Through
providing the network harmonic impedance or admittance
and harmonic currents injected by nonlinear loads for all

12
harmonics under consideration, the individual and total frequency. When parallel resonance exists on the power
harmonic voltage distortions at each bus can be system, significant voltage distortion and current
determi ned. Reterence [ 16] also describes some other amplification may occur. The highly distorted bus voltage
harmonic analysis methods. may cause distorted currents flowing in adjacent circuits.
The amplitied current may result in equipment failure.
Observing (2.46), we see that system harmonic
impedance plays an important role in the system response XL
to harmonics, especially when resonance occurs in the Power
system. Resonance is defined as an amplification of power System
system response to a periodic excitation when the
excitation frequency is equal to a natural frequency of the
system. For a simple LC circuit excited by a harmonic
current, the inductive and capacitive reactance seen from Figure 2.2. Parallel Resonance
the harmonic current source are equal at the resonant
When parallel resonance occurs in the circuit of Figure
frequency Ir = 1/ (2rc.J LC). 2.2, the resonant frequency can be determined by
In a power system, most significant resonance
problems are caused by a large capacitor installed for
displacement power factor correction or voltage
regulation purposes. The resonant frequency of the
hr = J
Xc
XL
= MVAsc
MVAR cAP
, (2.49)

system inductive reactance and the capacitor reactance
often occurs near fifth or seventh harmonic. However,
resonant problems occurring at eleventh or thirteenth
harmonic are not unusual. There are two types of
resonances likely to occur in the system: series andparallel
resonance. Series resonance is a low impedance to the
flow of harmonic current, and parallel resonance is a high
impedance to theflow ofharmonic current
Series Resonance
As shown in Figure 2.1, if the capacitor bankis in series
with the system reactance and creates a low impedance
path to the harmonic current, a series resonance condition
may result. Series resonance may cause high voltage
distortion levels between the inductance and the capacitor
in the circuit due to the harmonic current c-oncentrated in
the low impedance path it sees. Series resonance often
causes capacitor or fuse failures because of overload. The
series resonant condition is given by
where MYA sc is the short-circuit MVA at the harmonic-
generating load connection point -~ to the system and
MYAR CAP is MVAR rating of the capacitor. It should be
understood that this approximatiqI!:!s only accurate for
systems with high XIR ratios.
Another resonant scheme is shown in the distribution
network of Figure 2.3. If some of the feeder inductance
appears between groups of smaller capacitor banks, the
system may present a combination of many series and
parallel resonant circuits, although the resonant effects are
somewhat less than that caused by one large resonant
element. For this type of resonance problem, more
sophisticated harmonic analysis programs must be
employed to predict the harmonic characteristics of the
system.
Substation

J
h; = Xc , (2.48)
&i
XL
Figure 2.3. Distributed Resonance
where h; is the harmonic order of resonant frequency.
2.6 Solutions to Harmonics
XL x, Passive harmonic tilters are an effective mitigation
L--- Power method for harmonic problems. The passive filter is
~Syste~ generally designed to provide a path to divert the
troublesome harmonic currents in the power system.
Two common types of filters are the series and the shunt
filters. The series fIlter is characterized as a parallel
Figure 2.1. Series Resonance resonant and blocking type which has a high impedance at
its tuned frequency. The smoothing reactor used in power
Parallel Resonance electronics device is an example. The shunt filter is
characterized as a series resonant and trap type which has a
Figure 2.2 shows the circuit topology in which parallel low impedance at its tuned frequency. The single tuned
resonance is likely to occur. Parallel resonance occurs LC filter is the most common design in power systems.
when the parallel inductive reactance and the parallel More detailed information on harmonic tilter design and
capacitive reactance of the system are equal at certain applications can be found in [12,17].
frequency, and the parallel combination appears to be a
very large impedance to the harmonic source. The Harmonic currents in a power system canalso be reduced
frequency where the large impedance occurs is the resonant by providing a phase shift between nonlinear loads on
13
different branches. One popular method called phase 4. A. E. Emanuel, "Powers in Nonsinusoidal Situations - A
multiplication is to operate separate six-pulse static Review of Definitions and Physical Meaning," IEEE
converters (12-pulse and higher) in series on the de side Trans. on Power Delivery, Vol. 5, No.3, July 1990,
and in parallel on the ac side through the phase-shifting (~ pp.1377-1389.
~ and L\-Y) transformers [18] so that there is self- 5. A. E. Emanuel, "On the Definition of Power Factor and
cancellation of some harmonics. Sometimes, a specially Apparent Power in Unbalanced Polyphase Circuits,"
designed transformer (zigzag) is used to trap triplen IEEE Trans. on Power Delivery, Vol. 8, No.3, JUly
harmonic currents and to prevent the currents flowing back 1993, pp.841-852.
to the source from the nonlinear load. This Zigzag
transformer is usually designed to provide a low harmonic 6. L. S. Czarnecki, "Misinterpretations of Some Power
impedance between its windings compared to the source Properties of Electric Circuits," IEEE Trans. on Power
harmonic impedance. Thus there are circulating harmonic Delivery, Vol. 9, No.4, October 1994, pp.1760-1769.
currents between the nonlinear load and the transformer. 7. P. S. Filipski, Y. Baghzouz, and M. D. Cox, "Discussion
of Power Definitions Contained in the IEEE
Active filtering techniques [19] have drawn great Dictionary," IEEE Trans. on Power Delivery, Vol. 9,
attention in recent years. By sensing the nonlinear load
harmonic voltages and/or currents, active filters use either No.3, July 1994, pp.1237-1244.
1) injected harmonics at 180 degrees out of phase with the 8. "Nonsinusoidal Situations: Effects on the Performance
load harmonics or 2) injected/absorbed current bursts to of Meters and Definitions of Power," IEEE Tutorial
hold the voltage waveform within an acceptable tolerance. Course90 EH0327-7-PWR, IEEE, New York, 1990.
These approaches provide effective filtering of harmonics
and eliminate some adverse effects of passive filters such 9. K. Srinivasan, "Harmonics and Symmetrical
as component aging and resonance problems. Components," PowerQuality Assurance, Jan/Feb 1997.
Harmonic standards provide useful preventive solutions 10. IEEE Working Group on Nonsinusoidal Situations,
to harmonics. Recent standards such as IEEE 519-1992 "Practical Definitions for Powers in Systems with
[11] and lEe 1000-3-2 [20] emphasizeplacement of limits Nonsinusoidal Waveforms and Unbalanced Loads: A
on harmonic currents produced by nonlinear loads for Discussion," IEEE Trans. on P-ower Delivery, Vol. 11,
customers and network bus harmonic voltage distortion for No.1, January 1996,pp. 79-101.
electric utilities. 11. "Recommended Practices and Requirements for
2.7 Summary Harmonic Control in Electric Power Systems," IEEE
Standard519-1992, IEEE, New York, 1993.
For harmonic studies, Fourier series and Fourier 12. J. Arrillaga, D. A Bradley, and P. S. Bodger, Power
analysis are fundamental concepts. Many FFr algorithms
have been implemented for DFf computations on System Harmonics, John Wiley & Sons, New York,
measuring harmonics. 1985.
In nonsinusoidal situations, the conventional electric 13. G. T. Heydt, Electric Power Quality, Stars in a Circle
quantities used in sinusoidal environment need to be Publications, West Lafayette, IN, 1991.
redefined. However, power definitionsas well as harmonic 14. "IEEE Recommended Practice for Establishing
phasesequences under unbalanced three-phase systems are Transformer Capability When Supplying Nonsinusoidal
still under investigation. Several hannonic indices have Load Currents," ANSIllEEE Standard C57.110-1986,
been defined for the evaluation of harmonic effects on IEEE,New York, 1986.
power system components andcommunication systems.
15. W. M. Grady and R. J. Gilleskie, "Harmonics and How
To predict precisely the power system response to They Relate to Power Factor," Proceedings of PQA93,
harmonics requires accurate models for power system San Diego, CA, 1993.
elements and harmonic-generating loads. A simple
technique for hannonic aDalysis is the current injection 16. Task Force on Harmonics Modeling and Simulation,
method, which is perfonned in the frequency domain. "Modeling and Simulation of the Propagation of
Other analysis nletbods include time domain and Harmonics in Electric Power Networks Part I :
fr~ency/timedomain techniques. Solutions to harmonics Concepts, Models and Simulation Techniques," IEEE
can be classified as remedial and preventive. Passive and Trans. on PowerDelivery, Vol.l l, No.1, January 1996,
active filters are widely-used remedial solutions, and pp. 452-465.
harmonic standards provide the best solution before actual
harmonic problems occur. 17. E. W. Kimbark. Direct Current Transmission, Vol. 1,
John Wiley & Sons, New York, 1971.
References 18. N. Mohan, T. M. Undeland, and W. P. Robbins, Power
1. A V. Oppenheim and R. W. Schafer, Discrete-Time Electronics - Converters, Applications, and Design,
Signal Processing, Prentice-Hall, lnc., Englewood JohnWiley & Sons. New York. 1995.
Cliffs, NJ, 1989. 19. W. M. Grady, M. J. Samotyj, and A H. Noyola, "Survey
2. R. F. Chu and J. J. Bums, "Impact of Cycloconverter of Active Power Line Conditioning Methodologies,"
Harmonics," IEEE Trans. on Industry Applications, IEEE Trans. on Power Delivery, Vol. 5, No.3, July
Vol. 25, No.3, May/June 1989,pp.427435. 1990,pp.1536-1542.
3. R. C. Dugan, "Simulation of Arc Fmnace Power "limits for Harmonic Current Emmisions," International
Systems," IEEE Trans. on Industry Applications, IA- Electroteehnical Commission Standard lEe 1000-3-2, March
16(6), Nov/Dec 1980, pp.813-818. 1995.
14
 ClL\.PTER 3

DISTRIBUTION SYSTEM AND OTHER ELEMENTS MODELING

Paulo F. Ribeiro
BWX Technologies, Inc.
Naval Nuclear Fuel Division
Lynchburg, VA 24505-0785

3.1 Introduction the transmission and distribution systems should be

One difficulty in calculating harmonic voltages and
currents throughout a transmission system is the need for an
adequate equivalent to represent the distribution system and
consumers' loads fedradially fromeachbusbar.
It has become evident that the use of equivalents
without a comprehensive check on the effect of all
impedances actually present can lead to inaccurate estimation
of harmonic voltages and currents in the transmission system.
On the other hand, it is not practicable to obtain and represent
all thesystem details.
A detailed analysis of distribution systems, loads and
other system elements is carried out, models discussed and a
simple but more realistic approach adopted. It consists
basically of representing the dominant characteristics of the
network using altemativeconfigurations and models.
Simpler equivalents for extended networks are also
suggested.
used only for remote points.
(7) For distribution system studies all the elements may
be assumed to be uncoupled three-phase branches
with no mutuals, butallowing unbalanced parameters
per phase.
A distribution system comprises a number of loads
conveniently supplied by circuits from the nearest distribution
point. The distribution circuit configuration depends on the
particular load requirements. IB general, a considerable
number of loads are located so close together and supplied
from the main distribution point thattRey-can be considered as
a whole. For the majority of installations, whether supplying
a small factory, domestic/commercial consumers, or a large
plant, a simple radial system is used [I]. A typical distribution
network is shown inFigure 3.1

3.2 General Considerations

Although further considerations leading to simpler --+~I---'-------- 230kV
equivalents are given later, the basic assumptions used in this
chapter are as follows:
. . . . . . -. :l :l ,lloads 1 1 .
69kV

(1) Distribution lines and cables (say, 69-33kV, for

example) should be represented by an equivalent pi.
. .. + p.f,c,
can.
~
For short lines, estimate the totalcapacitance at each identical circuits
voltage levelandconnect it at thetermination buses. 69kV
(2) Transformers between distribution voltage level ;-T:r1

---.--.- 13.ski: ,..l ,..~ 11

~
should be represented by anequivalent element
(3) As the active power absorbed by rotating machines
does not correspond to a damping value, the active
and reactive power demand at the fundamental
: ::
. . :r
frequency may not be used straightforwardly.
Alternative models for load representation should be - I3.StV
-..-, .
used according to their cOmposition and
characteristics.
- ..........-.-.....----38OkV
(4) Powerfactor correction (PFC) capacitance should be
estimated as accurately as possible and allocated at
thecorresponding voltage level.
(5) Other elements, such as transmission line inductors, Figure 3.1. Typical distribution system configuration
tilters and generators should be represented
A simplified dominant configuration can be derived as
according to their actual configuration and
composition. illustrated in Figure 3.2,based OD the basic assumptions. This
(6) The representation should be more detailed nearer arrangement would represent the dominant characteristics
the points of interest. Simpler equivalents, either for (impedances) of the supply circuit fed radially from each
transmission busbar.

15

230kV
60kV 69kV 13.8k
Figure 3.2. Dominant arrangement
In order to simplify the manipulation of the distribution
system, load and other element data, the following procedure
based on the configuration of Figure 3.2 is suggested. The
dotted lines in Figure 3.3 mean different possibilities of
connecting the load or other elements such as compensators
tilters, generators, etc.. Thetotal equivalent impedance is then
calculated at each harmonic frequency in star-grounded and
connected to the transmission busbar as a shunt element.
Consequently, there is no alteration of the dimension of the
transmission system matrix, See illustration inFigures 3.3 and
3.4. A composition of different arrangements can be
represented at the same busbar.
..............................................
Load&t
Other
Elemems
Figure 3.3. Distribution system, loadsandothel' elements
3.3 The Modeling of Loads
In this section, the modeling of individual elements is
discussed in detail. Considering that there is some
disagreement regarding which harmonic models are best for
loads, transformers, generators, etc [2], various proposed
models are discussed. Also simpler equivalents for
disUibution and transmission systems at relJlote points of the
area of interest are discussed.
Consumers' loads playa very important part in the harmonic
network characteristic. They constitute not only the main
element of the damping component but may affect the
resonance conditions, particularly at higher frequencies.
Indeed, measurements [3] have shown that maximum plant
conditions resulted in a lowering of the impedance at the
lower frequencies, butcause an increase at higber frequencies.
Mahmoud and Shultz [4] observed in simulations that the
16
addition of load can result in either an increase or decrease in
harmonic flow.
Transmission
System 3-phase
Representation
Distribution system
and other elements
Figure 3.4. Overall systemrepresentation
Consequently, an adequate representation of the system loads
is needed. However, it is very hard to obtain detailed
information about this. Moreover, as Ule general loads consist
of an aggregate number of components, it is difficult to
establish a model based on theoretical analysis.
The necessity of practical measurements on distribution
points, at 13.8kV for example, together with detailed
information of the network: under study, is vital for the
understanding and establishment of a realistic model.
Attempts to deduce a model from measurements have been
made. See Bergea1 et al [5] and Baker [6]. However, more
comprehensive measurements and system data are needed.
Although practical experience is still insufficient to guarantee
the best model, system studieshave to proceedwith whatever
information is available. Thus,load characteristics are looked
at in detail and alternative models developed-in.the following
sections.
A typical composition of consumers' plantmaybe as shown in
Table 3.1. From Table 3.1, it seems evident that there are
basically two sorts of loads - resistive and motive. That
would imply a simple combination of resistances and
inductances. However, the difficulty in obtaining detailed
information about composition, power and variation with time
makes the task very hard. Nevertheless, it is possible to
approach the problem of representing loads for harmonic
studies by using alternative models according to the load
characteristics and infonnation available.
3.3.1 Recommended Models
Loads are generally expressed by their active and reactive
power P and Q, respectively, which are used to calculate the
equivalent impedance from load now study results at
fundamental frequency. Hovwever,- at. harmonic frequencies P
and Q cannot be used straightforwardly because the active
power absorbed by a rotating machine does not exactly
correspond to a damping value and so additional information
is necessary. The following alternative models A-D can be
used according to the load characteristics and information
available:
R
T able 3..
1 L oadeomposition resistive part only
Nature Type of Load Electrical
Characteristics
Domestic Incandescent Lamp Passive Resistive
Compact Non-linear Figure 3.5. Load model A
Fluorescent Passive Inductive
Small Motors Non-linear
Computers Non-linear(*)
Home Electronics
Commercial Incandescent Lamp Passive Resistive
Air Conditioner Passive Inductive R R
Resistive Heater Passive Resistive I
Refrigeration Passive Inductive
Washing Machine Passive Inductive
Fluorescent Lamp Non-linear(*) resistive motive part
(Std) Non-linear(*) part
ASDs Non-linear(*)
Fluorescent Non-linear(*) Figure 3.6. LoadmodelB
(Electronics) Non-linear(*)
Computers B. The equivalent resistance is estimated as above, but
OtherElectronic with an inductance in parallel. This should be
Lo-ads evaluated using an estimation of the number of
Small Fan Passive Inductive motors in service, their installed unitary power (not
industrial Pump Passive Inductive demand), and their negative sequence inductance.
Plants Compressor Passive Inductive However, as precise information on the number of
(Low Resistive Heater Passive Resistive motors, etc. in use at anygiven time is unavailable, a
Voltage) Arc Furnace Non-linear(*) fraction K of the total MW demand must be used to
ASDs Non-linear(*) represent the motorpart. This is thenmultiplied by a
Other Electronic Non-linear(*) factor of, for example, 1.2 in order to consider the
Loads installed power which should be used. To calculate
the equivalent negative sequence inductance, a factor
(*)These loads are harmonic producing. Hence, they do not K h proportional to the severity of the starting
exhibit a constant R. L, or C, ie. they are non-linear and condition should be used. This model is a
therefore cannot be included in an equivalent network of combination of common practices proposed by
impedances. Fortunately, there is every~OD to believe they reference [19]. Therefore, we will have:
have insignificant effect (open circuit) on the harmonic
impedance andmay be neglected. R= ..1!:- L= V2
P(l-K) 1.2 (K+~ K 1 Pm
A. At harmonic frequencies, the reactive power
estimated may havea negligible effect in some cases. where P =totalMW demand
Thus, the P is considered equivalent to a resistance K =motorfraction of the total MW
of value R=V2/P, V being the nominal voltage at Ke =electronic controlled load fraction of
fundamental frequency (see Figure 3.5). 'Ibis totalMW
representation should be used when the motor part K1 =severity of starting condition
is very small, i.e. for commercial and domestic loads CJ) =radian frequency
in which the motive part is so partitioned that the
resistive effectis predominant. K assumes values around 0.80 for industrial loads and around
0.15 for conunercial and domestic loads. K1 assumes values
between 4 and 7. KE can assume valuesaround O. It maywell

17
be that it is sufficiently accurate to ignore the resisuve
component of the motor pan. However, an additional
resistance representing the motor damping can be included as
Rl=LlK.2, where K2 is a fraction of the negative sequence R
inductance or locked-rotor inductance. K2 assumes values 1
around 0.20.

c. When a big induction motor or group of motors are

R L
connected directly at Intermediate voltage levels,
which is the case in industrial plants, the motive part resistive motive part
is better represented by a resistance in series with part \ ---J ',,-_ _.,,~
the negative sequence inductance of the motor (see
Section 3.5.4b on rotating machines representation). Figure 3.7. Load model C
The model can be assessed as follows: - The
equivalent resistance, the resistive part, and the With the approach adopted, a composite load model at the
negative sequence inductance of the motor is transmission system substation can be represented. The
estimated asin B, and the series resistance estimated expected effect is a better representation of the load. Because
by R = CIllJK], where most supplyelectricity companies have Dot thoroughly studied
K3 =effective Q of the motor circuit- 8 their own loads, a comprehensive investigation of the load
CJ) =radian fundamental frequency composition is necessary to enable the engineer to choose a
Alternatively, a seriesinductance LT to represent the better model or composition of models-based on an estimate
equivalent leakage reactance of the distribution of the system load.
transformers at lower voltage connecting the
resistance load can be incorporated (see
Figure 3.7). A value of mLT =O.IRcan be assumed.

D. This model was developed from experiments

performed on medium voltage outputs using audio-
frequency ripple-control generators at EDF [7]. The
circuit suggested was an inductance in series with a
resistance. This branch was connected in parallel
with another inductance. The estimated P and Q are
resistive motive part
used in empirical formulae to calculate the part
equivalent impedances. Thus, R =
V',/P; L1 =
0.073Rim; L2 = R/(6.7tan(Pbi)-O.74)co; tan(phi) = figure 3.8. Load model D
QIP (See Figure 3.8). Although this model was
obtained based on two frequencies only, 175 and 3.3.2 Other Considerations
495Hz, and the information available is not clear
enough [8] on how the equivalent circuit was When the harmonic number increases, it is necessary to use
derived, the parameters do not differ substantially larger values ofR. As noinformation is yetavailable, a factor
from models B and C. L2 seems equivalent to the of h1l2, where h is the harmonic order, seems a reasonable
motor part inductance and RILl to the resistive value as a first approximation. Pesonen et al [2] bave
circuit. suggested a factor of approximately O.6h lJ1

The harmonic impedance of distribution systems and loads

bas acmally been measured at a few sites in the U.K. The
results could not be satisfactorily reproduced digitally until the
downstream system from 33kV andcapacitance at4lSV were
represented [9]. Measurements collected by Balcer [6]
showed that there is a strong indication of an effect of po~u
factor correction capacitance on the harmonic impedance of
11kV, 33kV, and 132kV systems. Therefore, there are
reasons to believe that PFC capacitance should be
represented. The PFC MYAt could be up to half of the ~1W

18
numerically, depending on the local PFC policy and system
conditions, i.e. whether maximum or minimum plant. Hence,
the overall load representation should be as Figure 3.9. The
PFC MYAI should be represented as a fraction of the total
MW estimated.

Figure 3.11. Harmonic impedance variation (p.u.) with

transf. - line fed load Capacitive variation -
constant resistance

T
r
I

~1
linear :2Q' 0
Z C P.F.C.,
load capacitance
3I.U
I. 0.50

Figure 3.9. Overall load representation

Figure 3.12. Harmonic impedancevanatton (p.u.) with
3.3.3 Sensitivity Tests transf.-line fed load Resistive and Capacitive
variation
In order to illustrate the sensitivity of theequivalent harmonic
impedance with the load level and composition, a set of 3.4 Modeling of Other Elements
examples is shown in Figures 3.10 to 3.12. Typical 3.41 Distribution Lines and Cables
parameters for a 69 kV distribution system are used: The
resistive, inductive, and capacitive parts of theload are varied Distribution lines and cables are represented by their exact
andthe equivalent impedance calculated. equivalent pi [11]. An estimated correction factor for skin
effect is applied by increasing the line resistance with
The examples show considerable variation in the equivalent frequency by:
impedance for variations of the resistive and reactive
components of theload. For instance, when the resistive part R= R (I + O.646h2 ) lines
of the load approximates the surge impedance of the line, the 192+O.518h2
resonance effect is significantly reduced (see Figure 3.10). R =R(O.187+O.S32h 1ll ) cables
Conversely, changes in the reactive 'part may affect
considerably theequivalent impedance. 3.42 Transformers
Complete representation of transfonners, including
These examples do show very clearly the importance of an capacitances, is not practical and cannot be justified for
accurate estimation and representation of the distribution hannonic frequencies. Experience has shown that
system and loads. Although the variations imposed seem capacitances start to have some effect at lOkH~ i.e. well
exaggerated, it is very likely that such deviations between the above the common harmonic frequencies present in power
estimated and the actual parameters may occur, as the systems, i.e. 2kHz. Transformer impedance is shown to be
information is Dot easily obtainable. proportional to the lealcage reactance and linear with
frequency. Various impedance representations have been
~..) MJa: 13Ia'f UM
suggested. The following alternative models can be
.:...... ~,.",....+-r~ ......,..,.~....&.~..... represented:
A. A resistance in series with the leakage inductance.
Here a correction factor of hl .IS can be used [10]
SeeR 3.13).
.
:'l:
..
,
. 20.0

I..a

Figure 3.10. Harmonic impedance variation (p.u.) with

transf.-line fed loadResistive variation Figure 3.13. Transformer model A

19
B. The leakage reactance in parallel with a resistance.
This is calculated by multiplying a factor times the
reactance. A factor of 80 is suggested in the CEGB
program[11] (See Figure 3.14).
Figure 3.14. Transformer model B
c. Pesonen et al [2] suggested a resistance Rs in series
with an assembly of inductance L in parallel with a
resistance Rop. Resistances R, and Rp are constant
whatever the frequency and an estimate of their value
can be obtained as provided by expressions:
9O<V2/SRc<110
13<SRplV2<30
with S being therated powerof the transformer (see
Fi e3.15.
Figure 3.15. Transformer model C
3.43 Rotating Machines
(a) Synchronous Generators
(b)
When non-linear currents/voltages appear in the
stator of a synchronous machine, the fundamental
component is responsible for the energy conversion
process and sets up a rotating mmf wave which
reacts with the rotor mmf to produce the resultant
fundament3.l mmf gap flux. Conversely, the
harmonic components set up mmf waves rotating at
different frequencies, but there is no armature
reaction. Therefore, the reaction offered to
harmonics is not related to synchronous parameters
but an equivalent impedance which should be a
function of the leakage path. Also, it may be
assumed that synchronous machines produce no
harmonic voltages and they can be represented by a
shunt equivalent impedance at harmonic frequencies.
20
However, the literature is notin agreement regarding
appropriate impedances at harmonic frequencies.
Westinghouse [12], Williamson [13], and Pesonen et
al [2] suggest a reactance derived from either the
subtransient or negative sequence inductance:
Shilling [14] suggests X =X,", while Campbell and
Murray [15] suggest X =Xes'. Fresl [16] suggests X
= 112(Xd"+X2) , where X2 = 112(Xd"+Xq").
Westinghouse [12] suggests a correction of the
equivalent inductance. This is because when
frequency increases, a smaller amount of flux
penetrates the rotor. The amount is not known
accurately but normally taken as unity for the
fundamental and0.8 at 1000Hz.
When using typical values of synchronous machine
reactance to calculate the equivalent reactance X, it
can be observed that the sabtransient reactance
seems a reasonable value and should be used. A
resistance representing the damping can be
incorporated. Electra 32 [11] suggests a slcin effect
correction factor of hO.96 Regarding the equivalent
ctrcuit, Personen et al [2] suggest a parallel
combination of R and L. Here a series combination
is more appropriate, as the equivalent circuit of a
synchronous generator can be visualized as an
induction motor for harmonic frequency. However.
regarding practical values, the skin effect
representation and the way to combine the
impedances will not cause any significant difference
on the equivalent impedance. IIr the program, a
series or parallel combination can be used. Skin
effect and inductance correction can be represented
as well. A damping resistance based on the losses
can be addedfor both series or parallel combination.
Induction Motor
The well known configuration of an equivalent
circuit of an inducting motor is shown in Figw-e
3.21a. The Slip, s, at harmonic frequencies s(h) is
approximately equal to 1 as
s(b) = h:(1-s(I)/h) - 1, where s(l) -0.02
With Xm negligible, the equivalent circuit in Figw-e
3.21 is a reasonable approximation. Here L is the
locked-rotor inductance, which can be calculated
from the severity of starting condition. R is the
damping resistance which is derived from the motor
losses. For detailed analysis. see Chalmers [17] and
Klingshirn and Jordan [18]. Induction motors are
 generally present as part of the loadandin a group of higher frequencies is unpredictable as small capacitance may
different sizes. have a dominant effect producing resonances. Measurements
[3] have shown that in some cases the minimum impedance at
the higher frequencies occurred at the minimum fault level.

The net effect of increasing the load is to reduce the

R impedance to both fundamental and harmonic frequencies.
The combined effect of increasing generation and load is to
R reduce system resonances by increasing system inductive
elements and increasing the damping by lower resistance paths
L toground.

The effect of the line capacitance is to reduce the harmonic

(a) (b)
Figure 3.16. Synchronous generator representation
(a) series combination
(b) parallel combination
3.5 Supply System Eguivalent
In supergrid studies there is a need for an equivalent of the
transmission system as the representation of the whole
network is not practicable. A simpleinductance based on the
fault level contribution has been used in the past but its
representation is clearly unsatisfactory. The factors
influencing the equivalent impedance are discussed and a
simple butmore realistic modelis adopted.
impedance for higher frequencies. However, the combined
effect with inductances may cause parallel resonances and
thus have the oppositeeffect.
There is no typical equivalent that can be used in any system
without question, but measurements ~3-6] have shown similar
characteristics in the harmonic impedance of a transmission
system. A general point is the prominence of the first two
resonance frequencies, i.e. parallel (COp) and series (IDs). This
can be observed from Figures 3.18 [3] and 3.19 [6].
Inspection of the impedance characteristic shows that a T
circuit can be assumed as in Figure3.20a. This produces an
impedance as in figure 3.20b, which represents the system
impedance more accurately than the fault level inductance
when the necessary data are available. There are reasons to
believe that the inductive part is inversely proportional to the
fault level contribution. That is,

L = L 1 + ~ = (nominal voltage in kV)2

(fault level in MVA)*21tf

Thelinecapacitance is responsible for thecapacitance C in

thecircuitof Figure3.20a. Thus, the impedance is given by:

(a) Z{m)=coL t -[~.(l/coC)]/[mL2 -(l/coC)]=

R L =COLt - [mL,,/( Q)2~ C-l)] (3.1)
I\, VV\.-.rY'.

L= locked rotor
inductance

"
'- (b) ...... E
.2
Figure 3.17. Induction motor representation I
(a)Complete representation
(b) Equivalent harmonic model 1
~
The harmonic impedance of a transmission system is
determined/affected by factors such as fault level, system
J.. I I I ! ! .. ! I I
loads, capacitance of lines and cables, compensations, etc. In
general, an increase in fault level reduces the harmonic
impedance at lower frequencies. However, the behavior for Figure 3.18. Harmonic impedance of 132kV system

21
In order to give more flexibility to the model, the parameters
Lt - L2, and C as a function of COp and Cl)s canbe assessed. That
is because the transmission lines determine fundamentally the
behavior of the a.c. network and thus COp and COs could be
estimated from the knowledge of the lines, which would better
approximate the system impedance.
Resistances in series with the inductances can be incorporated
to provide a damping component to attenuate resonance peak..
In order to illustrate the accuracy of the model, the measured
impedances of Figures 3.18 and 3.19 are reconstructed in
Figures 3.21 and 3.22. respectively.

.........",._--.---
1.11"---
-..;.;- wn.=-,--,,l
~4~C~"Y
~
h- ......
1

t.~ .. '. - - - - - . _.. t ... __.- --1

1-- . . ,. _. _..:...-,1
~i\.~
Figure 3.21. Calculated harmonic impedance 132kV system
corr ondin tomeasurementsofFi e3.18

Figure 3.19. Harmonic impedance of 132kV cablesystem

withoutlocal generation, minimum plant
condition

(&)

Figure3.22. Calculatedharmonic impedance 132 kV system

corresponding to measurements of Figure 3.19

3.6 Distribution System Example

A simple but typical 11 kV distribution system was modeled

.p
( b)
Figure 3.20. Transmission systemequivalent model
L - fault level inductance - Olp. m._frequencies ofresonance
Equation (3.1) given previously is a second-order equation.
Therefore, considering the two frequencies where the parallel
and series resonances occur, the equations can be solved in
termsof Lit ~, and C.
as shown in Figure 3.23. In the diagram the
consumer has a
connected load of 865 kVA at a power factor of 0.8 lagging
and a power factor correction capacitor of 250 kvar. The
consumer wishes to connect an adjustable speed drive using a
6-pulse converter at the IIkV busbar. The system fault level
at the busbar, including the transformer, is 30MVA and the
source impedance may be considered as purely inductive.
The maximum harmonic currents (5th, 7th, 11th, and 13th)
injected are specified and typical for 6 pulse drives.

1. Z( COp) = infinite, therefore mp2L'1.C= I

C =1/0>pL2 =I/CDp2(L-Ll)

2. Z(COs) = 0, therefore msLl = a>sJ-,J(COs'1.L'1.C-1)

= =
L1 L21( C1>l'1.~C-I ), COs'1.L2C 1+L~l = l.JL1
C =(lJLl)x(1/COs2~)
Comparing C from (I) and (2), it will can be seen that LI =
L( 00p!00s)2 andthe equivalent circuit is completely determined.

22
 1"10" ~--...-.- ........ ~-,....--..,..--r-o'!'-...-.......-...,....--.--..,....,.~.....,
..
....... _. )-(',;';..... . ::eo
. 1,:.: u (l.c~~~ . (.~~'1: 1. ,. I

./ "',
.. P.J' ~~.OOQ' . . -' . ::OIA1(1l) 10

'. ... : .. :. ~. : . ... 1.. '" .-'.-...-..-....,,-, ...

~.

~:~~~~.:~~~ : ~4 :L.:..: .,.. ,. .: _~.:.: ., :

t
:
:
O.l

1 ) ) 4 S, 1; 9181112111'1jUI111J9~21"23.2$
..................-....I.-....."""--I.........1.,,;,;I
J Ol-....................I..ootIo.--'---.~--....

.. .:... .. .. 'ji :"lc.r~~d:cnC'! j .. .. Ii

t~~ : ., .. . .. ::: :. ~.!: ~: . Q.~'.~
: 1 Iii;
":"-: /~.-.:: ,-t:," ,tl13 ... ;- _t . .' . Figure 3.24. Equivalent 11 kVFrequency Response
~.(: ;'r'>'}'; ';r :::1 ..": :::::: :::::::::: ..
Impedance - LoadRepresented by Series Model 1

.. ::~. .: ~1:\ O'~Fr~~~l.: : :'.: :~.: ., ,."..., -..............

-_
r:. tl: ::~l:L~Rl~~;m~r, :.
. . . . .. .. 0.. . . .. .. ..
..........-.---..... .. .....-.........
~

...: ~1: I
i
I
~ 3 4 , , 7 i 9 10 U 11 13 It 1$ 14 11 J 19 JI :1 J3 D ~2S

1.
Figure 3.25. Equivalent II kV Frequency Response
Figure 3.23- System Schematic Diagram Impedance - LoadRepresented by Parallel Model 2
In order to illustrate the influence of the load modeling in
harmonic studies, Figures 3.24 to 3.26 show the equivalent
harmonic impedance viewed from the 11 kV bus using
different load models. The dominant parallel resonance
harmonic frequency is estimated commonly by

MVA sc OJ
h= .1...............- -'--..........e.....l..................~ ~........--....r......M-........--'--~
MVAr C4p 1 2 J 4 .s , 7 I 9 11 a n 1t is U 11 II 1SJ I 21 ~ :3 1C 15
k
which can be derived by finding the unity power-factor Figure 3.26. Equivalent 11 kV Frequency Response
frequency of the system. This calculation assuming no load Impedance - Load Represented by EdFICIGRE Model 7
indicates that the resonance frequency is around 10.95 times
the fundamental frequency. At the resonance frequency the When the harmonic currents are injected it can be observed
impedance of the 11 kV bus becomes very large as it can be that at the 11til harmonic the resultant voltage obtained with
seen on Figure 3.24, when load model 1 is used. Thus, a parallel representation(model 2) is 66 V or 1.04%,
significant voltage distortion may result at the 11 th whereas with the series representation (model I) the 11th
harmonic. The high harmonic voltages will also result in harmonic voltage on the l1kV bus was 332 V or 3.23%.
high harmonic currents both in the capacitor bank and the Thus, near the parallel resonance frequency the impact of
system reactance. A more detailed agalysis, however, the load representation can be very significant. Using an
reveals that I) the resonance frequency varies with the alternative serieslparallelload representation (EdF model 7),
resistance of the system and the amplitude of equivalent the frequency response of the equivalent impedance is
harmonic impedance, 2) the outputvoltage is not necessarily shown in Figure3.26. Two important facts can be noted.
maximum at the resonant frequency, and 3) is a function of FU'St the resultant VOltage on the llkV bus is now 48 V or
the damping (resistance) of the circuit. However because 0.69%. Second the resonant frequency shifted slightly
current is only injected at the 11 th harmonic one does not higher (from 11th harmonic to near the 13th harmonic).
need to consider other frequencies, but rather remember the Table3.2 shows a summary of the casessimulated where the
sensitivity of the system harmonic impedance (around the model and load andcomposition were varied.
resonance frequencies) to par~eters variations.

23
 2 Load ModeI'Ing and C onditions simulated
T abl e 3..
Case Linear Load Model
Case 1 NoLoad Representation
Case 2 P, Q- Basic Load Flow OV~("-)

Case 3 P, Q - Basic Load Row .V1'(~)

25
Case 4 50% Induction Motor
Case 5 25 % Induction Motor
Case 6 50% Induction Motor
Case 7 75% Induction Motor
c-.
Case 8 90% Induction Motor JCIII
4 c-.
S c..
Case 9 25% Ind. Motor + Skin Effect 4; :-
: c..

Case 10 75% Ind. Motor + Skin Effect ~ CIII YSC'I Vtt~

to

Skin effect was included in cases 9 and 10 to account for the
impact on the system impedance of the frequency
dependence of the resistive component of the load. Figure
3.28 illustrates the amplitude of the 5th and 11 til harmonic
voltage (%) at the II kV busfor all models used.
CJV5(r.)
.v,t(,..
Figure 3.29. Harmonic Voltage (%) at the l1kV Bus for
Different Load Models
Modeling loads using just the economic model (P and Q
only) is inadequate for harmonic -studies. No load (case 1)
representation should not be used for harmonic studies. The
load models (2 suggested in the litesature can not be used
indiscriminately without a comprehensive check of the
actual load characteristics and composition. The
appropriate representation is particularly crucial near the
parallel resonant frequencies of the system. exactly where an
accurate estimanon of the system behavior is most
necessary. Frequency response of the system impedance is
sensitive to both the methodology (modelingltopology) and
the actual load composition. A comprehensive list of linear
load models for harmonic studies found in the literature is
illustrated in Annex 1

3.7 The Need for a Complete Load Representation

Figure 3.28. Harmonic Voltage (%) forDifferent Load
Model
Figure 3.29 demonstrates more clearly how much the
resultant voltage can vary depending on the model and load
composition used. When comparing to standards such as
the IEEE 519, it becomes clear that the violation of the
standard may depend on the load model used for the
calculation of theresultant distortion.
General loads in a transmission or distribution system are
generally expressed by theiractive and reactive power P and
Q, respectively, which are used to calculate the equivalent
impedance for load flow studies at the fundamental
frequency, assuming the system voltage is known.
However, at harmonic frequencies P and Q cannot be used
directly because, for example, the power absorbed by
rotating machines does not exactly correspond to a damping
value, neither does the motor equivalent inductance bear
any direct or simple relationship to the reactive power
estimated at the fundamental frequency. In addition a
measurable percentage of any general load nowadays is
electronically controlled and needs to be properly
represented. Electronic loads are harmonic producing and
consequently do not exhibita constant R, L or C. Therefore,
they cannot be included as part of the passive component
of the equivalent impedance. They should be represented
by a harmonic source at all frequencies of importance.
With the proliferation of the utilization of power electronics
a progressive conversion of traditionally linear loads to
electronically controlled will happen.

24
 Another component normally overlooked in harmonic R2 = XL2
studies is the power factor correction (PFC) capacitance of
distribution systems. Measurements have shown that there
K2
where
is a strong correlation of the effect of power factor
correction capacitance on the harmonic impedance of
distribution systems. The PFC Mvar could be up to half of
v = System vottage
XLl = Transformer Reactance
. the MW numerically, depending on the local PFC policy
and system conditions.
Cl =Estimated Capacitance of the Load
/1 = Estimated Harmonic Current Source
where
Therefore, the proposed model shown in Figure 3.30 makes
P =Total Active Power
an attempt to incorporate all these aspects in the general K =Fraction of Induction Motors
modeling of loads for harmonic studies and consequently
KE =Fraction of Electronic Loads
make the so called general, normal or "linear" load
Rl =Equivalent resistance representing thepurely resistive
representation for harmonic studies much more realistic.
component Of the load
Detailed research should be carried out to assess the actual
~ = factorfor skin effect correction
load composition and determine the proper representation
and parameters of each load or aggregate of loads.
XU = Equivalent inductance representing the induction
motors
The equivalent model should consist of a combination of
R2 =Damping factorfor the induction motor
seriesand parallel combination of resistances, inductances, representation
Kl =Severity of Starting Condition
capacitances, and harmonic currentsources as indicatedin
Figure3.30. Km =Installed Motor Factor
K2 =Fraction a/the locked-rotor (ornegative sequence)
inductance
h =Harmonic order
. . . . r~. ~:~~~ ..~~ . :~ . : : . . XLI = Leakage inductance of transformers at lower
~. oJ 1h~ lSj"!'ttl!"ft
. . . . .. . . . .. . . . voltages connecting the resistive load
.. .. : !" .................._
....................... :
: .
. /1 =Ideal harmonic current source (use typical values
f'!e:'.r~i=: . . ?rC. Cat"A]dlenc~
l.oed . . :
according to type Of toadfeeder).

The resistance RI is estimated from the actualresistive load

connected to the bus, that is, discounting the induction
motor and electronic load part. The skin effect can be
incorporated in the equivalent resistance ..by choosing an
appropriate factor as indicated. The inductance of the
induction motors should be evaluated using an estimation of
the fraction of the total load that represents induction motors
and their installed unitary power (not the demand). Also a
Figure3.30. General Load Representation forHarmonic factor Kl representing the severity of the starting condition
Studies should be used to calculate the equivalent inductance. R2
represents the damping component of the equivalent
Because the reactive power of the load estimated at the induction motor impedance. Also background distortion
fundamental frequency has little to do with the equivalent should not be neglected. Harmonic simulation studies will
impedance of the load at harmonic ffequencies, it is have to include background distortion if they are to be
suggested that Q (estimated for the load without any PFC) become more accurate. Background distortion can increase
should be totally disregarded for the estimation of the or decrease the resultant distortion .depending on phase
equivalent harmonic impedance of the load. Thus, starting relationship. A harmonic current or voltage source
with the total active power P and additional information representing the harmonic contribution of the non-linear
about the load composition the following procedure is component of the load must be modeled.
suggested for calculating parameters forharmonic studies.
3.8 Conclusions
XL2
v2 This chapterdemonstrates thatthe representation of thepower
KmKKlph
system loads and extended networks can be improved by
using alternative models. The distribution system, loads, other

25
elements and equivalents of extended networks have been Model 4
considered in detail.The models developed allow a more Parallel
realistic representation of the system and, consequently, a Combination
more accurate assessment of the harmonic currents and
voltages throughout the transmission network. Guidance has From P and
been provided on modeling of individual loads and on typical fraction of
load composition. System tests are necessary to provide motors
verification of the modeling methodology developed, as well
as adding to the knowledge of system load characteristics.

This chapter demonstrates that the representation 0 f linear ModelS

elements is very important for harmonic studies and should SerieslParallei ::~r::-:r:::
not be neglected or represented without full consideration of . 0.:__ " . 0 I " . 0
Combination 1 C " . "
:1 : : .c
0
'" 1 : :- " 0

theload characteristics and composition. (CIGREIEclF

....... t2" ~
:" .
Model) :0"1" ~.
Consequences regarding system predictions include wrong
indication of excessively high harmonic distortion. In
""1"'
H ...
..
..
."
.

addition, utilization of sophisticated harmonic penetration Model 6

programs with inaccurate basic tnformanon, and/or
inadequate modeling is a waste of money, and the
SeriesJParallel
Combination 2
Lnr:{~~J"rT:
consequences of the interpretation of the results might cost " . I "..:-".::'- " . .
" ..~ ...--1:2":..... ..
even more. In conclusion, remember that the accuracy of ~"'".
any calculation cannot be better than the data on which it is . .01" 0""
Fl"'l"
.
.
. .. .
0 . . .}

.
.. .

..
"

based.
Model 7
Annex 3.1. Linear LoadModels SeriesJParallel
Combination 2
Model Equivalent Circuit
Description FromP,
Modell inductanceof
Series distributioon
Combination transformers
(Common and induction
Practice) motors

From P andQ
Model 2
Parallel References
Combination 1
l____---r___-...
.
..
..
.
[1] ROLLS,T.B.,"PowerDistribution in Industrial
(Common Installations," lEE Monograph Series 10, 1972.
Practice) ~P;P ~ l~ ~" .01.
~l.. ~ ~ ~ ~0.~ [2] P~ONEN, M.A.,"Harmonics, Characteristic
..
_ .

...
a . '

.. Parameters, Methods of Study, Estimates of Existing Values

From P andQ . .. in the Network."Electra, Vol.77, pp. 35-54, 1981.
plus skin effect [3] HUDDART, K.W., and BREWER, OL., "Factors
Influencing the Harmonic Impedance of a Power System."
Model 3 Conference on High Voltage DCTransmission., lEE No. 22,
Parallel pp. 4SG-4S2, 1966.
Combination 2 [4] MAHMOUD, A.A. and SHULTZ, RD., "A Method for
Analyzing Harmonic Distribution in a.c. Power Systems,"
FromPand IEEETrans., PAS-IOI, No.6, pp. 1815-1824, 1982.
fraction of [5] BERGEAL, J. and MOLLER, t; "Influence des
induction Chargessur la propagation des perturbations de type
motors harmoniques principales consequences," Internal Repon
E.D.F., HRl221034, 31.12.80.

26
[6] BAKER, W.P., "Measured Impedances of Power
Systems," International Conferenceon Harmonics in Power
Systems, UMIST, 1981.
[7] BERGEAL, J. and MOLLER L., "Influence of Load
Characteristic on the Propagation of Disturbances," CIRED
1981.
[8] MEYNAUD, P., E.D.F., Direction des Etudes et
Recherches, Privatecommunication, 25 Fevrier, 1983.
[9] HOWROYD, D.C.,CEBG Technology Planning and
Research Division, Privatecommunication 29.12.83.
[10] Electra, Vol. 32
[11] HARPO 3 - Harmonic Impedanceand Penetration
Program. CEGB Report CS/CIP300.
[12] Westinghouse Electric Corporation,"Electrical
Transmission and Distribution Refrence Book," 1950.
[13] WILLIAMSON,A.C., 'The Effects of System
Harmonics upon Machines," International Conferenceon
Harmonics in PowerSystems, UMIST, 1981.
[14] SHILLING, WJ., "Exciter armature reaction and
excitation requirements in a brushless rotating-rectifier
aircraftalternator," TransAm.Inst.Elect.Eng. 1960, 79, pt.
II.
[IS] CAMPBELL, L.C. and MURRAY,N.S., "Harmonic
Penetration into-Power Systems," 5th Universities Power
Engineering Conferences, Swansea,Wales, 1970
[16] FRESL, V.: "Sistermiuzbude generatoraHE 'Derap,"
Inf. RadeKONCAR, 1974,60-63, pp. 67-77.
[17]CHALMERS, Bl., "Induction-motorlossesdue to
non-sinusidal supply waveforms," Proc.1EE, Vol. 115, No.
12, 1968.
[18] KLINGHRIRN, E.A and JORDON, H.E.,"Polyphase
induction motorperformance and losses on nonsinusoidal
voltagesources,"IEEETrans., 1968, PAS-87.

27
 Chapter 4

MODELING OF HARMONIC SOURCES

POWER ELECTRONIC CONVERTERS

Gary W.K. Chang WilsunXu

Siemens Power T&0 University of Alberta
BrooklynPark, MN, USA Edmonton, Alberta, Canada

4.1 Introduction The device can be operated as a six-pulse converter, as

shown in Figure 4.1, or configured in parallel arrangements
Harmonic problems are not new to electric utility and for higher pulse operation. Theoretically, a static power
industrial power systems. In the past, most harmonic- converter load draws currents from the source system that
relatedproblems werecaused by largenonlinear loads such consist of positive and negative currents which are equally
as arc furnaces. These types of problems have been separated. The pulse number refers to the number of
effectively mitigated. However, due to the widespread "humps" on the de output voltage that are produced during
proliferation of power electronic controlled devices every ac cycle.
nowadays, the problems caused by harmonics are of
increasing importance. Power electronic loads offer a
number of advantages in controlling power flow and in
efficiency, but they perform this by chopping, flatting, or
shaping sinusoidal voltages and currents. Harmonics are
produced in the process.

Among today's power electronic applications, most of

the harmonic problems are caused by the static power
converters. The static converters are used in many types of
industrial applications. The purpose of this chapter is to
present modeling and simulation techniques for power
electronic devices, focusing on the harmonic modeling for
static power converters. First, we briefly review the Figure 4.1. Six-Pulse Line Commutated Converter
commonly seen power electronic type harmonic sources.
Next, a number of major converter harmonic models used In Figure 4.1, each pair of thyristors is~~ggered (firing
in harmonic simulation will be described. Two converter angle) and conduct until they are reverse-biased. If a
models used as harmonic simulation examples are then thyristor is triggered at zero firing angle. it acts exactly like a
presented. diode. 1be term line commutated converter refers to the fact
that the load actually turns thyristors off, rather than them
4.2 Review of Power Electronic Hannonic Sources being turned off by external control circuits. The ideal ac
current waveform for a six-pulse converter is on for 120
D~ to the advanced technologies in power electronicsdegrees and off for another 60 degrees. During the on
development overthe pastdecade, the application of power period, the de load current is assumed constant in the ideal
electronics has been widely spread to all types of case due to the assumed existence of a large series dc
industries. Commonly observed examples are.. inductor. Assuming no commutation overlap and balanced
three-phase operation, it can be shown that the phase a
Line Commutated Converters current is

The introduction of economic and reliable line
commutated converters has caused a significant increase in
harmonic-generating loads, and they have dispersed over
the entire power system. In most cases, line commutated
converters are the cause of hannonic problems in power
distribution systems. Thesedevices areworkhorse circuits
for acJdc power conversion. The common application of
static power converters is in adjustable speed drives for
motorcontrol. Another application is in HVDC tenninals.
28
ia(t) =1:.!!.Sin(hlJt + Ok). (4.1)
h h
where h = 1, S, 7. 11, 13, .... We see that the ac harmonic
currents generated by a six-pulse converter include all odd
harmonics except triplens. Harmonics generated by
converters of any pulse number can be expressed by
h =pn 1, where n is any integer and p is the pulse number
of the converter. For the ideal case. converter harmonic
current magnitudes decrease according to Ilh rule. Table 4.1
gives the fllz-rule and typical harmonic currents (in per unit 4.3 Review of Static Power Converter Models for
of the fundamental component) for six-pulse converters Harmonic Simulation
[1].
In order to simulate the propagation of harmonics
Table 4.1. Theoretical and Typical Harmonic Currents for throughout a network, adequate models for harmonic-
Six-Pulse Converters. generating loads as well as system components must be
developed. In general, the power electronic devices that
generate harmonic currents can be modeled by using simple
current source models or complicated device-level models.
The harmonic simulation can be in frequency domain, in
Pulse-Width Modulated Converters time domain, or in both. In this section, an overview of
common harmonic modeling techniques for static power
PWM converters use power electronic devices that can converters for simulation studies will be described.
be turnedoff and turned on. Therefore, voltage and current
waveforms can be shaped more desirably. The switching Power electronic converters for harmonic analysis can be
simply represented by a harmonic current source or a model
components can be thyristors that are forced off by external
control circuits, or they can be GTOs or power transistors.that takes into account the interaction between ac system
The latter devices are usually used because of their fast network and the converter dc system. When the latter
switching characteristics are neededfor effective PWM. situation is considered, a more sophisticated converter
analysis for the resulting harmonic currents as a function of
In a PWM converter, the switching devices are system reactance, delay angle, and commutation angle is
controlled to switch on and off to produce a series of required. The accuracy of converter model needs also to be
pulses. These pulses are to be variedin width to produce a considered to guarantee the convergence of the simulation.
pulsed three-phase voltage wave for the load. Due to their At present, there are several techniques that have been
low efficiencies, PWMconverters are limited to low power developed for modeling of power.e1earonic converters in
applications in the several hundredkWor hp ranges. harmonic simulation. These techniques can be categorized
as:
Cycloconverters
1. Current injection model.
Thecycloconverter is a device that converts ac power at
one frequency into ae power at a lower frequency. 2. Frequency- or time-domain Norton equivalent circuit
Cycloconverters are usually used in low speed and large model.
horsepower applications. The harmonic frequencies
generated by a cycloconverter depend on the output 3. Harmonic coupling matrix model.
frequency, which is varied in operation to control motor 4. Time- or frequency-domain device model used with
speed. The output frequency of a cycloconverter can be frequency-domain network model.
controlled by precisely timing the firing pulses at its
thyristor gates through computer control. 5. Time-domain model.
Static V AR Compensator (SVC) 1be following sections give a brief overview on the
aforementioned converter models for harmonic simulations.
The static var compensator is used as a voltage For the details of these models, please refer to the
controller in the power system. 1bis device controls corresponding references.
network voltage by adjusting the amount of reactive power
supplied to or absorbed from the power system. The Current Injection Model
applications of the SVC are usually for local compensation
of reactive power to industrial loads and for regulation of The mostcommon techniquefor harmonic simulation is
utility network voltages to improve transfer. capabilities to treat static power converters as known sources of
across the transmission system. Typical configuration of an harmonic currents with or without including phase angle
SVC consists of shunt capacitors with a thyristor- information. This is due to the fact that the converter acts
controlled reactor (TCR) connected in parallel. as an injection current source to the system in many
operational conditions. Generally, the steady-state
Other Power Electronic Devices
condition is assumed. The following frequency-domain
Other power electronic devices which may generate matrix equations for each harmonic are used to compute
harmonics in the power system include static phase the network harmonic voltages:
shifters, isolation switches, load transfer switches, and
energystorageand instantaneous backup powersystems as (4.2)
well as those devices covered under the subjects of or
Aexible AC Transmission System (FACTS) and Custom (4.3)
PowerSystems (CPS) [2].
29
Then. superposition is applied to convert the solved
values of each Vh into (he time domain for each network
bus k as follows:
H magnitude of the harmonic voltages on each network bus is
Vk(I)=LV:sin(h"11+~), (4.4)
11=1
where H is the highest harmonic order under
consideration.
In the current injection model, the magnitudes of
harmonic currents can be determined simply following
the l/h rule as stated in (4.1) or represented by
measurements, as shown in Table 1. The phase angles of
the current sources are functions of the supply voltage
phase angle [3] and can be expressed as . known [7]. The switching function is used to detennine
6h =8h-spec + h( 61 - 61- spec ) , (4.5)
where 61 is the phase angle obtained from the load flow
solution for fundamental frequency current component,
and 6h-spec is the typical phase angle of the harmonic
source current spectrum. Many times, especially for
studies involving one converter, the phase angles are
ignored and only the magnitudes are used in the harmonic
simulation. Once .the hannonic voltages are known,
harmonic currents through network elements are
determined, and they can be convened to time-domain
currents.
The advantages of the current injection method are
that the solution can always be obtained directly (non-
iterative) and it is computationally efficient Ideally, this
method is able to handle several harmonic sources
simultaneously. The drawback of this method is that
typical harmonic spectra are often used to represent the
harmonic currents generated by the converter which
ignores the interaction between the network and the
converter. This prevents an adequate assessment of cases
involving non-typical operating modes, such as partial
loading, excessive harmonic voltage distortions and
unbalanced network conditions. Reference [4] suggests
that the current injection model should be used carefully
(if at all) when the converter source voltage 1HD is on
the order of 10% or more. More information on the
current injection method and associated models can be
found in [4-6].
The aforementioned drawbacks can be overcome by
using more advanced converter models and harmonic
analyses desaibed in the following sections. The models
generally include the effects of harmonic voltages on the
converter current waveform. Therefore, these advanced In (4.6), t; and Voc can be expanded to include both positive
methods couple the converter with the system admittance and negative sequence components, and Ie can be expanded
matrix, such as shown in (4.3), or some other more to include the firing anglecontrols [13].
complicated expression of the power system. Given an
initial estimate of harmonic current injections at the
converter, the network bus harmonic voltages are
30
determined. A new estimate of the harmonic injection
currents is then obtained from the computed harmonic
voltages. This process is repeated until convergence in the
reached.
Frequency- or Time-Domain Norton Equivalent Circuit
Model
In this model, the converter is represented by a Norton
equivalent circuit, where the Norton admittance represents
an approximation of the converter response to variation in
i~ terminal voltage harmonics. A common approach for
thi~ m<:>del to h8:ve the converter switching represented by a
SWItching function whose frequency-domain expression is
the ac side harmonic phasors directly from the dc side
~armonic phasors. TIlls model is then iteratively improved
In a frequency-domain networksolution process. nus type
of solution process in harmonic power flow analysis is
usually called iterative harmonic analysis (IRA) [8].
Reference [9] also presents a similar model in time domain
with an iterative simulation technique. More information
on this model can be found in [10] and.[.ll].
The advantages of the Norton equivalent model are that
the solution process has better convergence characteristics
and that a direct solution for the interaction between the
converter and the network can be obtained. However, [12]
reports that the problem with this model is that the
converter is indeed an interface between the ac and dc
systems, with only the ac system represented in the entire
iterative solution process. If the converter controller needs
to be modeled, a separate iterative process is required for
solving the converter interaction with the dc system at each
iteration.
Hannonic Coupling Matrix Model
In [13] and [14], the authors proposed an efficient
technique by the linearization of the interaction between
the converter de system and the ac network. Then, the
entire system is solved via the harmonic coupling matrix
equation to account for the interaction between the
converter dc system and the ac terminal voltage. Figure
4.2 shows a simplified single- or three-phase converter
model, where the harmonic coupling matrix equation can
expressed as
[Vltlc]=[A
C D Ide
BIVac]. (4.6)
de
 (de -----
Vac
---
lac
Figure 4.2. Simplified Converter Model
The harmonic coupling matrix provides a relationship
between the harmonic components of ac side voltage/
current and the de side voltage/current of the converter.
Each element in the matrix is a sub-matrix and is a
function of the converter states and commutating
inductance. This model can be used either in .the time or
the frequency domain with the incorporation of the
iterative approach, and it has been developed for both
single-phase and three-phase converters while ignoring
the effects of converter controls, commutation variations,
and resistance in ac networkimpedance [12].
Time- or Frequency-Domain Device Model used with
Frequency-Domain Network Model
In this model, the converter is described in terms of
the actual time-domain differential equations that govern
its performance. Then, convener currents are solved in
the time-domain and converted intothe frequency-domain
by the use of Fouries analyses. Next, the harmonic
currents are injected into the network model and the
harmonic voltages at each network bus are calculated.
The computed voltages are then used to recalculate the
converter currents in the time domain. In Newton-
Raphson or Gauss-Seidel types of harmonic power flow
analysis, this procedure iterates until convergence criteria
are met The HARMFLO and HAR:MFLO+ computer
programs are well-known products that use the
combinations of time- and frequency-domain solutions.
More details aboutthis model can be found in [15-17].
Reference [18] also presents a frequency-domain
model which formulates a general set of non-linear
equations to describe the converter in steady state. The
formulation convolutes periodic sampled quantities in the
frequency domain with square pulsesampling functions.
The use of sampling functions in this manner-is similar to
other work using the switching function [7]. The non-
linear equations are then solved using Newton's method
in conjunction with the frequency-domain network
model.
Time-Domain Model
In the time-domain model, the solution method used is
a time simulation of the entire system (both the converter
and the ac network). These solution methods are the
most mature of harmonic simulations. The programs
such as EMTP~ ATP, and EMTDC can be used to obtain
31
a complete time-domain solution. The actual periods of
operation within each cycle of converter operation are
described by differential equations. No attempt is made to
convert to the frequency domain. Both balanced and
unbalanced conditions can be handled, and the converter
model can be as detailed as necessary. However, the
solution time and engineering effort increase significantly.
References [19] and [20] also provide other insights for the
time-domain model.
4.4 Case Study
In this section, we evaluate two converter models used
in harmonic simulation. The two models are current
injection model and harmonic Norton equivalent circuit
model. A commonly seen PWM type adjustable speed
drive (ASD) is chosen for evaluation. An ASD mainly
consists of a converter (rectifier or front-end), a dc link, a
controller, and an inverter. Generally, the harmonics
produced in the inverter part are negligible as seen from
the converter ac side because of the harmonic current path
formed by the dc link capacitor. Therefore, the converter is
modeled as the only part that injects harmonic currents into
the power system for the PWM type. ASD. Figure 4.3
shows the converter circuit of the AS}), where the inverter
and the motor load are modeled as a direct current source.
R
t
c
1
Figure 4.3. Converter Circuit Model of thePWM ASD
Norton Equivalent Circuit Model
The converter circuit shown in Figure 4.3 is solvedwith
the harmonic analysis technique described in [7]. The end
result is a delta connected Norton equivalent circuit. The
circi .. is then interfaced with the supply network in an
iterative fashion, as described in [11], to determine the
harmonic current injections from the ASO. Parameters
needed to run the model are 1) the firing angle of the
converter thyristors, a; 2) the direct current flowing into
the inverter, ld; and 3) the de link R, L and C component
t
values.
The firing angle of the PWM type ASD is almost zero
because of the use of diodes as the front end. The direct
current flows into the inverter can be estimated from the
motor load as
 = 2. 34VP cosa ,
:Z50 .
ld (4.7) 200
g
150

where P is the motor load including the losses, and Vg is 100

the line to ground voltage of the supply system.

In a typical harmonic study, ex and ld need to be 0.004 Off6 ~ 0 O'~ 0.012 0014 0.016 0 G18

varied for investigation of the various ASD-motor

operating conditions. Also, representation of the dc link
100

150
l\\\~pr.~f 1\
~.,.~
III
ts essential for the correct harmonic simulation. If the de
link parameters are not available, a simplified model such
200 tJ ~
250 ...:
as the current injection model may be proposed. Time (Me.)

Figure4.4. Typical Current Waveforms of PWMType

Current Injection Model ASDs
An ASD may be represented as a harmonic current
source. Table 4.3 gives the typical harmonic magnitude Model Verification
and phase spectra that can be used to model an ASD. The
corresponding waveforms are shown in Figure 4.4. Lab tests are conducted to verify the three-phase ASD
These data are obtained from the Norton equivalent model developed in this study. The tests are performed on a
model that simulates an actual PWM ASD and are 30 hp PWM-typeASD serving a 20 hp induction motorload.
verified by lab tests. Extensive analytical and numerical The lab setup is shown in Figure 4.5. The ASD supply
studies indicate that the data is suitable for modeling voltage (line to line), Vs, is 600 V. The motor mechanical
PWM type ASDs.
load is varied to simulate different-operating conditions. 13
Table 4.3. Typical Harmonic Spectra of PWM Type operating conditions are recorded. For each operating
ASDs condition, waveforms of 15 voltage and current quantities
are measured when the system is in steady-state. These
waveform snapshots, including source side voltages and
Load Level 100% 75% 50%
currents, motor side voltages and currents, de link voltage
h-order Mag. Angle Mag. Angle Mag. Angle and inverter current, are synchronised and sampled at a rate
1 100.00 0 100.00 0 100.00 0 of 100kHz. In addition to the above operating tests, the dc
3 0.35 -159 0.59 -44 0.54 -96
5 60.82 -175 69.75 -174 75.09 -174
link is measured at various frequencies in a standstill test to
7 33.42 -172 47.03 -171 54.61 -171 determine its component parameters.
9 0.50 158 0.32 -96 0.24 -102
11
13
15
17
19
3.84
7.74
0.41
1.27
1.54
166
-177
135
32
179
6.86
4.52
0.37
7.56
3.81
17
-178
-124
9
9
14.65
1.95
0.32
9.61
7.66
16
71
28
10
16
- -I
"be,Itlt
abc
8
21 0.32 110 0.43 -163 0.43 95
23 1.08 38 259 11 0.94 -8
25 0.16 49 3.70 10 3.78 7
Figure 4.5. Lab Test Setup for ASDHarmonic
Measurement

To use this model, the ASD is first represented as a The ASD equivalent circuit model can be verified by
constantpower load at the fundamental frequency. The real comparing the measured ASD currents into the supply
power load is equal to the ASD/motor load and the reactive system against the calculatedones. Wavefonncomparison is
power load is zero. The network is then solved at the more desirable than harmonic spectrumcomparison because
fundamental frequency. The ASD current magnitude and
t the former ensures that both the harmonic magnitudes and
phase angle aredetermined as 160 and 960, respectively. The phase angles are checked. A network model representing the
harmonic current source representing the ASD is calculated test conditions is constructed. The up-stream supply system
by scaling up the magnitude column of Table 4.3 by 160 and is modelled as mown three-phase harmonic voltage sources
by shifting the phase angle column by h960 It can be seen determinedfrom the measured data.
that the current source model is easy to use and needs less
input effort. The disadvantages are that the model cannot Figure 4.6 gives the comparison for the three-phase ASD
simulate a wide variety of ASDs and ASD/system currents between measured and calculated values under one
operating conditions. representative operating condition. In the figure, the
measured waveforms are plotted as solid lines while the
32
 4.5 Summary
: ' \(J

~i The common models of power electronic converters used

f)
in harmonic analysis are presented in this chapter. These
~ models are built either in time-domain or frequency-domain
SO and in conjunction with the proper power system network
100
model. Once built. the model is ready for use in a harmonic
0 1l.()") 5 0.01 0.015 0.02 0.025 0.03 0.035 simulation technique such as non-iterative or iterative
analysis.
Testl OiRaled S ed 100% Load : Pnase-B CUmnt 10System
100

<:

0.005 0.01 0.015 0.02 oms 0.03

References

1. R. P. Stratford, "Analysis and Control of Harmonic

I OO'-----'--~---'----"---'-----'-----'
o 0.005 0.01 0.015 0.02
Time (Seconds)
Test I Raled S ed 100% Load : PllaseA CUmnt S
IS . ...... .. . . . .. . . . ... . -:. . ... . . . .- -......~ ---- . - . . .... .. . .
~ Applications," IEEE Trans. on. Power Delivery, Vol.
<:
IS 20
Frequency (Hannonic Order)
Figure 4.6. Comparison of Measured and Calculated
Waveforms
computed waveforms are plotted as dashed lines. Some
discrepancies between the measured and calculated
waveforms are observed. A few factors may have
contributed to the discrepancies. The first factor is that the
size <?f ASD is relatively small. This makes the variations
of thyristor electric characteristics more noticeable in the
waveforms. The high frequency components of de link
current generated by the PWM scheme and inaccuracy of
dc link parameters may.also contribute to discrepancies.
Tests showed that the waveforms obtained with an inductor
added between the drive and the motor are in a closer
agreement with the calculated ones. In addition, it is
noticed that harmonics in the supply voltage have some
impact on the calculated waveforms. Unfortunately,
including more harmonics does not enhance the agreement.
Overall, the agreement between the measured and the
calculated waveforms show that the proposed models are
acceptable.
Current in Systems with Static ~qw.er Converters," IEEE
0.025 0.03 oms Trans. on Industry Applications, Vol. lA-I?, No.1,
JanuarylFebruary 1981, pp. 71-78.
m
2. IEEE Power Electronics Modeling Task Force & Digital
Simulation Working Group, "Guidelines for Modeling
Power Electronics in Electric Power Engineering
12, No.1, January 1997, pp. 505-514.
3. Task Force on Harmonics Modeling and Simulation,
2S 30 35 'The Modeling and Simulation of the Propagation of
Harmonics in Electric Power Networks Part I: Concepts,
Models and Simulation Techniques," IEEE Trans. on
Power Delivery. Vol. 11, No.1, January 1996, pp. 452-
465.
4. M. F. McGranaghan, R C. Dugan, and W. L. Sponsler,
"Digital Simulation of Distribution System Frequency-
Response Characteristics," IEEE Trans. on Power
Apparatus and Systems. Vol. PAS-100, No.3, March
1981, pp. 1362-1369.
5. D. J. Pileggi, N. H. Chandra, and A. E. Emanuel,
''Prediction of Harmonic Voltages in Distribution
Systems:' IEEE Trans. on Power Apparatus and
Systems. Vol. PAS-loo, No.3, March 1981, pp. 1307-
1315.
6. T. H. Ortmeyer, "Harmonic Analysis Methodology,"
IEEE PES Tutorial Course, Course Text 84 EH0221-2-
PWR, February, 1984, pp. 74-84.
7. W. Xu, J. E. Drakos, Y. Mansour, and A. Chang. "A
Three-Phase Converter Model for Harmonic Analysis of
HVDC Systems," IEEE Trans. on Power Delivery, Vol.
9, No.3, July 1994, pp. 1724-1731.
8. J. Arrillaga and C. 0: Callaghan, ''Three Phase AC-DC
Load and Harmonic Flows,' IEEE Trans. on Power
Delivery, Vol. 6, No.1, January 1991, pp. 238-244.

33
9. J. Usaola and J. G. Mayordomo, --Fast Steady-State 21. W. Xu, 1. R. Marti, and H. W. Dommel, "Harmonic
Techniques for Harmonic Analysis," Proceedings of Analysis of Systems with Static Compensators," IEEE
the 4th International Conference on Harmonics ill Trans. on Power Delivery, Vol. 6, No.1, February 1991,
Power Systems (ICHPS IV), Budapest, Hungary, pp. 183-190.
October 1990, pp. 336-342.
22. L. J. Bohmann and R. H. Lasseter, "Harmonic
10. B. C. Smith, N. R. Watson, A. R. Wood, and J. Interactions in Thyristor Controlled Reactor Circuits,"
Arrillaga, "A Solution for the Steady-State Interaction IEEE Trans. on Power Delivery, Vol. 4, No.3, July
of the AC/DC Converter with Weak AC and DC 1989,pp.1919-1926.
Systems," Proceedings of the 7th International 23. J. Vittek and M. Y. Najjar, "Common Methodology for
Conference on Harmonics and Quality of Power Steady State Harmonic Analysis of Inverters," IEEE
(ICHQP VII), Las Vegas, NV, October 1996. Trans. on Power Delivery, Vol. 10, No.3, July 1995, pp.
11. W. Xu, J. R. Jose and H. W. Donunel, 'A Multiphase 1628-1634.
Harmonic Load Row Solution Technique," IEEE 24. J. J. Rico, E. Acha, and T. J. E. Miller, "Harmonic
Trans. on Power Systems, Vol. PS-6, February 1991, Domain Modeling of Three Phase Thyristor-Controlled
pp. 174-182. Reactors by Means of Switching Vectors and Discrete
12. B. C. Smith, J. Arrillaga, A. R. Wood, and N. R. Convolutions," IEEE Trans. on Power Delivery, Vol.
Watson, "A Review of Iterative Harmonic Analysis for 11, No.3, July 1996, pp. 1678-1684.
AC-DC Power Systems," Proceedings of the 7th
International Conference on Harmonics and Quality of
Power(ICHQP VII), Las Vegas, NY, October 1996.
13. E. V. Larsen, D. H. Baker, and J. C. McIver, "Low-
Order Harmonic Interactions on AClDe Systems,"
IEEETrans. on Power Delivery, Vol. 4, No.1, January
1989, pp. 493-501.
14. S. G. Jalali and R. H. Lasseter, "A Study of Nonlinear
Harmonic Interaction Between a Single Phase Line-
Commutated Converter and a Power System," IEEE
Trans. on Power Delivery, Vol. 9, No.3, July 1994,
pp. 1616-1624.
15. D. Xia and G. T. Heydt, "Harmonic Power. Row
Studies, Part I - Formulation and Solution, Part II -
Implementation and Practical Application", IEEE
Trans. on Power Apparatus and Systems, Vol. PAS-
101, June 1982, pp. 1257-1270.
16. V. Sharma, R. J. Fleming, and L~ Niekamp, "An
Iterative Approach for Analysis of Harmonic
Penetration in Power Transmission Networks," IEEE
Trans. on Power Delivery, Vol. 6, No.4, October
1991,pp.1698-1706.
17. M. Valcarcel and J. G. Mayordomo, "Harmonic Power
Flow for Unbalanced Systems," IEEE Trans. on Power
Delivery, Vol. 8, No.4, October 1993, pp. 2052-2059.
18. B. C. Smith, N. R. Watson, A R Wood, and J.
Arrillaga, "A Newton Solution for the Harmonic
Phasor Analysis of AC/'OC, Converters," iEEE Trans.
on Power Delivery, .Vol, 11, No.2, April 1996, pp.
965-971.
19. B. K. Perkins and M. R. Iravani, "Novel Calculation of
HVDC Converter Harmonics by Linearization in the
Time-Domain," IEEE Trans. on Power Delivery, Vol.
12, No.2, April 1997, pp. 867-873.
20. M. Grotzbach and Ch. Ried, "Investigation of ACIDC
Converter Harmonics by an Analytical Based Time-
Discrete Approach," IEEE Trans. on Power Delivery,
Vol. 12, No.2, Apri11997, pp. 874-880.

34

Modeling of Harmonic Sources - Magnetic Core Saturation
5.1 Summary
This section introduces the modeling of harmonic sources
due to magnetic core saturation and several casestudies.
Different transformer models have been developed in the
past for steady state and transient analysis of power
systems. Some of these models have nonlinear components
to take into account the magnetic core saturation
characteristics so that harmonic generation can be
simulated. Case studies based on these models are
presented to demonstrate the harmonic generation
behaviors of transformers under different saturation
conditions.
5.2 Introduction
Magnetic core saturation of power transformers and
rotating machines can- generate harmonics. Figure 5.1
illustrates the principle of harmonic generation from
magnetic core saturation [1]. In order to maintain a
sinusoidal voltage, sinusoidal flux must be produced by the
magnetizing current. When the amplitude of the voltage (or
flux) is large enough to enter the nonlinear region of the B-
-_..-_ __.
H curve, the magnetizing current needed will be greatly
distorted fromsinusoidal and containharmonics.
.....
. . . . . . . . . . fII . . . . . . . . . . .-..-..._~~
- 1':'1)
:-:
IB
...
w.........;"
Figure 5.1. Principleof harmonic generation from
magnetic core saturation
Before converter loads were widely used, one of the
principal harmonic sources in the power system was the
excitation current of power transformers. Although modern
transformers do not generate significant harmonics under
Chapter 5
YiluLiu and Zhenyuan Wang
Department of Electrical Engineering
Virginia Polytechnic Institute andStateUniversity
Blacksburg, VA 24061-0111 , USA
Yll..U.LIU@VT.EDU
normal steady state operating conditions [2] they can
considerably increase their harmonic contribution under
abnormal conditions when their magnetic cores are
saturated.
5.3 Examples of magnetic core saturation
There are many situations which contribute to magnetic
core saturation. The following are some common
examples.
Normal excitation
Even under normal excitation conditions, the transformer
core may have entered (slightly) the saturation region and
begun to generate some harmonics in the excitation
current. The degree of the saturation depends on the
transformer design.
Overexcitation
Overexcitation is basically caused by overvoltage. This
problem is particularly troublesome in the case of a
transformer connected to large rectifier plant following
load rejection [2].
As in Figure 5.2, overvoltage drives the Reak operation
point of the transformer excitation characteristics up to the
saturation regionso that more harmonics are generated. In
.-...
....
thiscase the magnetizing current is often symmetrical.
OpcnIiClll point ttlOYiD&
mononlinearp\
c
\
~resulted
operatioIl pom
Noaml openbClII pom
H/i
Figure 5.2. Principle of overexcitation resulting in
transformer saturation
Converter load
Converter loads maydraw DC and low frequency currents
from supplying transformers. The transformer cores are
biased by theseload currents anddriven to saturation.
For example, a cycloconverter with single phase load as in
Figure 5.3 will draw DC currents from source transformer
35
when its output frequency fo and input frequency fi have the
relationship of f j = 2nfo ' where n is an integer [3].
f. f.
I 1
Three phase
transformer eye loconverter
Figure 5.3. A cycloconverter with single phase load
Geomagnetically Induced Currents (GIC)
Geomagnetically induced currents (Ole) flow on the
earth's surface due to geomagnetic disturbances (GMDs).
They typically have a frequency of 0.001 to 0.1 Hz and
could reach peak values as high as 200A. As in Figure 5.4,
they can enter transformer windings by way of grounded
wye connections and bias the transformer cores to cause
half cycle saturation [4-10].
where N is the number of transformer terminals. Vi (i= 1.N)
denotes the voltage of terminal i, ij U= I ,N) denotes the
current flowing into terminal j, Rij and L ij (ie l.N; j=l.N)
denote the resistance and inductance between terminals i
and j, respectively [14]. This model is the type of
Load
transformer model used in all versions of electromagnetic
transient programs.
A simple Tee model
Shown in Figure 5.5 is an equivalent circuit model of a
two-winding transformer. It can be used for teaching
concepts, investigating and demonstrating simple
phenomena, and simulating single phase or three phase
transformer banks. The Rm can be represented by a
piecewise linear V-I curve [16,19] or a constant value
resistance [18,21,22] . The Lm is often modeled by a two-
slope linear inductance [14,16] when the saturation B-H
curve has a sharply defined knee, which is usually the case
for grain-oriented steel cores [15], or more precisely by a
multi-slope piecewise curve [15,17,21-23]. The
characteristics of Rm and Lm are usually found from 00-
load tests [23].
Ip Rp Lp Ls __=.. Rs

Figure 5.4. GIC entering the transformer windings

5.4 Modeling of magnetic core saturation

A large amount of work bas been documented in literature
on modeling of transformer core nonlinearity. Being the Rp, Lp: primary winding resistance and leakage inductance.
Rs,Ls: secondary winding resistance and leakage
predominant factor of power transformer nonlinearity,
inductance..
saturation is the major issue relative to hysteresis and eddy Rm: core losses (hysteresis loss and eddy current loss).
currents. Hysteresis modeling is important in a transient Lm: nonlinear excitation inductance.
study such as switching or fault condition simulations of
Figure 5.5. A simple Tee model for two-winding
transformers [11-13] but is often neglected in harmonic
transformers
analysis [14-17].
Duality based models
There are different approaches for transformer modeling
and analysis: the matrix models [12-16] use an impedance Duality based models are often used to represent three
or admittance formulation relating terminal voltages and phase transformers [20-22]. This may be due to the fact
currents; the equivalent circuit models [11, 17-19] often that the complexcore topology of three phase transformers
use a simplified Tee circuit whose element values are can Dot be represented sufficiently by an equivalent circuit
derived from test data; the duality based models [20-22] model or conveniently by a matrix model. Here nonlinear
account for core topology and the connection between inductances are used to model core saturation [21,22] and
electric and magnetic circuits. Although the latter two types the modeling circuits are derived based on the principle of
of models can also be presented in matrix format, they are duality betweenmagnetic and electriccircuits..
easier to understand from a circuit point. Due to space
Figure 5.6 shows four types of duality based three phase
limitations, only a few model examples will be discussed in
transformer modeling circuits. They have been connected
this chapter. as WyelWye (YfY), DeltalWye (DIY), WyelZigzag caz:
A matrix model and DeltafZigzag (DfZ). The models can be used for
harmonic analysis and low frequency transient analysis.
A matrix modelis written as:
'LIl
GIC saturation models
VI ill R12 RIN i l L l2 LIN i1
R21 Rn R2N i 2 L 21 L n L2N d i2 For a transformer under a severe ole bias which causes
v2
+ . -dt heavy half cycle saturation, it becomes necessary to
vN RN1 RN2 RSN iN L N1 L N2 LNN iN account for the flux paths in and between core,tankand air
36
gaps. A detailed model based on 3D finite element
calculation may be necessary [9]. Shown in Figure 5.7 is
the equivalent magnetic circuit model of a two-winding
shell type transformer. The model can be used for
harmonic studies of a balanced three phase shell type
transformer bank under GIC bias. The circuit model has
four branches:
Branch 1: Represents the sum of core and air flux all
within the excitation windings, this total flux is
the sum of both DC(GIC) and AC flux.
Branch 2: Represents the flux path in the yoke segment.
Branch 3: Represents the sum of flux entering the side leg.
Part a f this flux will leave the side leg and enter
the tank.
Branch 4: Represents the flux leaving the core from the
center leg. Part of this flux loops back in the air
and the rest flows through the air gaps and the
tank.
x"'---- -.,
An iterative program can be used to solve the circuitry of
Figure 5.7 so that nonlinear components are considered
properly. Also, the harmonic balance method can be used
to solve the nonlinear time domain circuit and the
frequency dependent linear circuit iteratively [24].
5.5 Case studies
Cases #1 to #3 are based on the system shown in Figure
5.8. The transformer can be either a Y/Y, or a DIY, YIZ,
DIZ connection. DC bias (if any) are injected into
secondary windings by current sources Idea, Ideb and Idee.
Primary winding currents are Iwa, IWb and Iwe Power system
linecurrents are I sa, I Sb and I se
Transformer
.--------_..-- Resistive
load
~
Primary

x. --.
A :::;J

y.
~ ---,
y. -..
i _

Figure 5.8. Schematic diagram of a three phase

B ~ transformer with resistive load

z
~ Case #1: Harmonics during normal excitation
Transformers may generate harmonics under rated
operating conditions (rated voltage, no DC bias). Shown in
Figure 5.9 are the typical excitation current waveform and
spectrum of phase A of a three phase DIY connected
(a) YIY and DN models (b) yrz and Dflmodels transformer. It can be seen that, except for fundamental
component, 3rd and 5th harmonics dominate thecurrent
t; Lbx, Lx" I.m: duality-derived inductances
~B,C,X, Y,Z: primary terminals
a, b,c: secondary terminals
~ ..
1...0: leakage fluxpaths outside windings A
0.'

I..m: leakage flux paths between outer and

innerfmtermediate windings -G.
OJ

. leakage flux paths between intermediate 4)

I
2ClO 100 soc, _
0.02 vUl o.ta \/.. ~., ,~ ~

and inner windings T.-(I) 81

major flux paths via transformer cores (a) Waveform (b) Spectrum

Figure 5.6. Duality models for three-phase transformers Figure 5.9. Phase A excitation current of a DIY connected
threephase transformer underrated operating condition
Rc2
The generated harmonics are different in frequency and
amplitude with different transformer connections. As
Rei
shown in Figure 5.10, YIY and YIZ connections have less
harmonics generated than DIY and Dfl connections. (The
connection type is indicated beforethe current indicator in
AC
the figure. For example YIY_Iwa means phase A primary
winding current of a YIY connected transfonner.)
DC
Case #2: Harmonics due to overexeitation

Rc2
Under overvoltage conditions; harmonics increase
amplitudes with respect to excitation voltage. However, the
Figure 5.7. The equivalent magnetic circuit model of a harmonic spectrum pattern is unchanged (compare Figure
single phase shell type transformer 5.1l(b) with Figure5.9(b) ).
37
 r,
'" r-. .-

A rs

,:.!

" :. :1 A .. .oS :<.. I~~. :- :.; . w.. . ;-:. :.... o

~IITC ~ t) N,

(a) Waveform (b) Spectrum

Figure 5.13. Phase A excitation currentof a DIY
Figure 5.10. Current harmonics of three phase connected three phase transformer under 10% unbalanced
transformers under rated operation condition DC bias
,-,',--- - - - - - --.
"
A
.. A

.:.(,11 ~<t

T_I.'
(0,(4 ., I
~. ~
7".,.(1)
~ .ct
M.
(a) Waveform (b) Spectrum (a) Waveform (b) Spectrum
Figure S.l1. Phase A excitation current of a DIY connection Figure 5.14. Phase A excitation currentof a DIY
three phase transformer under 110% overvoltage condition
connected three phase transformer under 50% unbalanced
Again, the generated harmonics are different in frequency DC bias
and amplitude with different transformer connections, YN
and YIZ connection have less harmonics generated than
DIY and DIZ connection (Figure 5.12).
;;:il
l,...
DOC

!D'"
ioloO
~ ~ 3 a~ a~ a~ ~ ~ ~ 5 a~ a~ I. .
.....
or>Cl
D'" $ ~ ~ a
0-

0'"
.oc
or>Cl
Figure 5.15. Current harmonics of threephase
...... transformers under 10% unbalancedDC bias

Figure 5.12. Harmonics of three phase transformers under ~"'''''' DC''''.''''''

110% overvoltagecondition

Case #3: Harmonics due to unbalanced DC bias

Under unbalanced DC bias, harmonics become

significantly higher comparing to the same balanced DC
bias level. For the given DC bias levels (Phase A has a
......
positive DC bias X%, Phases B and C have equal negative
DC bias -O.5X%), most of the harmonic amplitudes Figure 5.16. Current harmonics of threephase
increase along with the DC bias levels but only a few transformers under 50% unbalanced DC bias
decrease (see Figures 5.13 and 5.14). This may be due to
the fact that the excitation point has already entered the Case #4: Harmonic generation and cancellation of a
heavysaturation region (see section 5.3 for details). transformer-converter-motor system

Underunbalanced DC bias, the YIY connectedtransformer A transformer-converter-motor system is very common in

seems to have less TIID in the source line currents than
other three types when DC bias becomes larger, but the
difference is not significant(see Figures 5.15 and 5.16).
modern industry. It could be a large harmonic source to the
power system and it is important to know its harmonic
generation behavior.
The block diagram of such. a system is shown in Figure
5.17. The transformers are modeled by circuits of Figure
5.6. Four types of transformer connections and three motor

38
speeds are studied. Results are listed in Table 5.1 and
shown graphically in Figure 5.18.

I -l I .1
Hz .. '* .,.

(e) V/Y. system. 15Hz (f) sn: system, 15Hz

A A
Figure 5.17. Block diagram of a transformer-convener-
motor system
Table 5.1. THO of transformer primary winding currents 2M ... ~ .. .~. 'HI 1M .. Hz .. '. 'HI

and power system linecurrents, DC bias in secondary

(g) DIY. secondary. 15Hz (h) DfZ. secondary. 15Hz
winding currents
eM MS TIlD (%) IDe
(pIs) (Hz) Isa Isb Isc Iwa Iwb Iwc (%)
15 18.2 18.1 18.1 56.0 56.1 50.9 11.2 A A
YIY 5 18.0 18.1 17.8 53.7 53.2 53.0 2.33
2.5 24.0 23.9 23.9 58.8 51.8 51.9 0.68
15 17.6 17.6 17.6 55.4 56.3 51.4 11.2 '~II ,I
Hz ...- ... Hz .. .- ".
Y!l 5 18.1 17.8 18.0 54.4 53.4 53.6 2.03
2.5 24.0 24.1 24.0 52.0 52.3 51.9 0.29 (i) ON. primary. 15Hz G) on; primary, 15Hz
15 18.4 18.4 18.4 59.9 52.5 58.5 11.0
MY 5 18.3 18.4 18.4 53.4 53.9 54.0 1.78
2.5
15
23.9
18.6
23.9
18.6
23.9
18.6
52.0
52.9
52.3. 52.2
54.0 59.5
0.26
12.2
...
.
...
III

A A-
sn. 5 18.2 18.0 18.4 53.4 53.6 54.0 2.38
2.5 24.0 23.9 24.0 51.9 51.9 52.0 0.26 .. .
I
, .1 I _I
Hz ....- '.. Hz ...- ...
(k) DIY. system. 15Hz (1) OfZ. system. 15Hz

A A
"
" ..
A. A.

Me -. iii .. ,... ,.. - - iii .. ,... ,..

(a) YIY. secondary. 15Hz (b) YfZ, secondary, 15Hz
Hz ...
I1 1

itt -
(m) DfZ. secondary, 5Hz (n) OIL. secondary. 2.SHz

A A

',.. ..
.
A A.

'--Hz"''''''' --jiz"'''''' .
(c) YIY, primary. 15Hz (d) Yrl,primary, 15Hz 1 .I.
Hz ......... Hz - .....
(p) DIL primary, 5Hz (q) Oil, primary. 25Hz

39
 1800
1\ "
1600
,I \\, ,," ,,
\
A " 1400 I f
,,
,, ,,
A "
" ,I
,
\
\
1200
\ ,
, ,, ,,,
\
I ,
I
I I I
1000 \ \
it - .- ~ ,, I \,
, ,
'a- A-t 800
, ,,
(r) DIl. system. 5Hz. (s) DIZ, system, 2.5Hz ,,
. 600 I , ,I
, I
\

,,,
I

400 , ,,
J \
,
Figure 5.18. Frequency spectrum of currents in a 200
\\
,
I
\
\ I
\
\
\
transformer-convener-motor system ' ... _--
0
, I
,------' I
'"
As can be seen from Table 5.1, at the same motor speed, .200 '-----"---....------.....---,.......- - ' - - _ . 6 . - - J
o 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
there is no significant THD difference in primary winding Time (s)
currents among the four types of transformers. When motor
speed is half of the rated value (2.5Hz), THD in the power Solid line - normal condition
system line currents are much higher than those at rated Dashed line - ole condition
speed (5Hz) and above. Figure 5.19. Excitation current waveform of a two-
winding transformer under GIC
As can be seen from Figure 5.22, current harmonics in
primary and secondary windings of different transformer
connections are nearly the same (considering the current
-s 107
ratio), which include those generated by the cycloconverter C
~
t:
-----.- ....
::s 10'
and those generated by the transformer due to DC bias. u

The majority of the transformer winding current harmonics

= 10'
0
-=:!
.tj - l p . u.
are generated by cycloconverters. When motor speed -e
~ 10
4
- - .7p.u.

decreases, these harmonics shrink to oddharmonics except

triplens of power frequency and increase in amplitude so
e
I~ l~ 10 lrJ
that the THDs of the system line currents become higher. Ole (A-t)

Harmonic cancellation is obvious in the power system line Figure 5.20. Excitation current rms value of a two-
currents, which means that most harmonics find their Winding transformer vs. ole at 1 p.u. and 0.7 p.u. AC
circulating path through the other two phases of the same voltages
transformer or through the other two transformers rather
than entering the source (see Figure 5.17 and Figure 1.5

5.18(e), (f), (k)9 (1)). These circulating barmonics will c:

increase the loss of transformers" but will decrease the i
u
c 1.0
.S!
.harmonics entering into the power system. S
.~
-1,.11.
Case #5: Harmonics due to GIC .5o.s - - .7'....

GIC may cause extremely large harmonic currents to flow o.o ..................................-.................-....~ ............,~ .........
10 101 103 10 10'
from a transformer into the power system and it is essential Gte (A-t)
to know the size of these currents under different Gle
levels in order to analyzepower system responses [9]. Figure 5.21. THO in excitation currentof a two-winding
transformer vs. ole at 1 p.u. and 0.7 p.u. AC voltages
By applying different levels of DC bias to the models
shown in Figure 5.7, the excitation current waveforms are
obtained and two of them are shown in Figure 5.19. The
rms value and THD of excitation current are shown in
Figures 5.20 and 5.21, respectively. Excitation current
harmonicsare plotted in Figure 5.22 againstDCbias level.

40
 Engineering Subjects". American Power Conference.
10: a) Fundamental component Vo1.49. pp487-492
10':
- Ipu. [2J 1. Arrillaga, D. A. Bradley. P. S. Bodger, Power System
10~ 07 P u Harmonics, pp94-98. John Wiley & Sons. 1985.
lO~
..-.-..-_ _ ~
~ .~ .=._::'....
~._.~ :': [3] Brian R. Pelly, Thristor Phase-Controlled Converters and
10~~-----------_..-_---"---
Cycloconverters. Operation. Control and Perfonnance.
1 ~ e6 b) 2nd harmonic pp361. John Wiley & Sons. 1971

lOe6 -lp.u. _ _-_ _.._ "

[4] L. Bolduc, 1. Aubin. "Effects of Direct Currents in Power
._..... 07 p.u. Transformers. Part I. A General Approach. Part II.
$ 5.0c5 Simplified Calculations for Large, transformers", Electric
Power System Research. 1. 1978
SOe.S c) 3rd harmonic [5] D. H. Boteler. R. M. Shier, T. Watanabe. R. E. Horita,
"Effects of Geomagnetically Induced Currents in the B.C.
2.~eS
---- 0.7
1 p.u.
p.u.
Hydro 500kV System", IEEE Trans. on Power Delivery.
Vo1.4. No.l, Jan 1989

0 [6] J. G. Kappenman, V. D. Albertson. "Bracing for

2.'se.S d) 4th harm"nic
1.2ScS
-lp.u.
- - 07p.u.
0
1..~e.S e) Sthharmonic
i.ces
- lp.u.
S.O~
0
10') 10 1
Gle (A-t)
Figure 5.22. Excitation currentharmonics of a two-
winding transformer vs. GIC at 1 p.u.and 0.7 p.u. AC
voltages
5.6 Future work
The nonlinear magnetizing characteristics of most models
did not account for core loss (hysterssis loss and eddy
current loss) precisely since it uses a constant resistor to
represent the loss. This is acceptable in some situations
where the transformer serves not as a key element of the
simulated system such as in a transformer-converter-motor
system, but notin others where it playsthemajor role such
as in the inrush current calculations. The values of the
model elements in the duality based models are estimated
from special testdata. It maybe desirable to calculate them
from physical dimensions and material characteristics that
can be obtained from the manufacturer. Also, most models
available are for core type transformers and a limited
number of three phase connections has been modeled.
There is not yet a clear guide on how to model a three
phase transformer with arbitrary connection and core type.
If possible, future works should address these subjects.
References
(1] Jason Hess.Chris Richard. Herb Brown. David Smith, Yilu
Liu, "Computer Animations in Teaching Power
Geomagnetic Storms", IEEE Spectrum. March 1990.
pp.27-33
.................
[7] J-C. u, V-Po we, "FFf Algrithms for the Harmonic
Analysis of Three Phase Transformer Banks with Magnetic
Saturation", IEEE Trans. on Power Delivery, Vo1.6, No.1.
Jan 1991
[8] M. A. S. Masoum, E. F. Fuchs, D. 1. Roesler, "Large
Signal Nonlinear Model of Anisotropic Transformers for
Nonsinusoidal Operation. Part Il: Magnetizing and Core
Loss Currents", IEEE Trans. on Power Delivery. Vo1.6.
No.4, Oct 1991
IOJ
[9] Shu Lu. Yilu ue, Jaime De La Ree, "Harmonics Generated
from a DC Biased Transformer", IEEE trans. on Power
Delivery, veis, No.2, April 1993. pp725-731
[10] W. Xu. T. G. Martinich, 1. H. Sawada, Y. Mansour.
"Harmonics from SVC Transformer Saturation with' Direct
Current Offset", 93 SM PWRD 404-4
[11] E. P. Dick, W. Watson. "Transformer Models .forTransient
Studies Based on Field Measurement", IEEE Trans ..
Vol.PAS-l00, No.1, Jan 1981. pp.409-419
[12] D. N. Ewart, "Digital Computer Simulation Model of a
Steel-Core Transformer", IEEE Trans., Vol.PWRD-I.
No.3, July 1986, pp.l74-182
[13] D. Dolinar, J. Pihler, B. Grear, "Dynamic Model of a
Three-Phase Power Transformer", IEEE Trans.
Vol.PWRD-8, No.4, Oct 1993, pp.1811-1819
[14] V. Brandwajn, H. W. Dommel, I. I. Dommel, "Matrix
Representation of Three-Phase N-Winding Transformers
for Steady-State and Transient Studies", IEEE Trans.
Vol.PAS-lOl, No.6, June 1982. pp.1369-1378
[IS] H. W. Dommel, A Van. Sbi Wei, "Harmonics from
Transformer Saturation", IEEE Trans.Vol.PWRD-l.No.Z,
Apr 1986, pp.209-215
[16] A Medina. J. Arrillaga, "Generalised Modeling of Power
Transformers in the Harmonic Domain", 91 SM 406-9
PWRD
[17] Ahsan H. Chowdhury. W. 'Mack Grady, Ewald F. Fuchs.
UAn Investigation of the Harmonic Characteristics of
Transformer Excitation Current under Nonsinusoidal
Supply Voltage", 96 SM 433-3 PWRD

41
[ 18 J 1. David. Charles .~. Gross. "Nonlinear Modeling of
transformers", IEEE Trans. on Industry Applications,
Vo1.24, No.3. May 1988, pp.434-438
(19] Y. Baghzouz, X. D. Gong, "Voltage-Dependent Model for
Teaching Transformer Core Nonlinearity", IEEE Trans. on
Power Systems. Vo1.8" No.2. May 1993, pp.746-752
[20) Arun Narang, Russell H. Brierley, "Topology Based
Magnetic Model for Steady-state and Transient Studies for
Three-Phase core type transformers", IEEE Trans. on
Power Systems" Vo1.9. No.3" Aug 1994, pp.l337-1349
[21] Xusheng Chen, S. S. Venkata, "A Three-Phase Three-
Winding Core-Type Transformer Model for Low-
Frequency Transient Studies", 96 SM 410-1 PWRD
[22] Zhenyuan Wang" Yilu Liu, "Harmonic Analysis of
Transformers under Converter Load with DC and Low
Frequency Bias", Proceedings of the American Power
Conference, Vo1.59" pp449-454
[23] Washington L. A. Neves. Hermann W. Dommel, "On
Modeling Iron Core Nonlinearities", 92 WM 176-8 PWRS
[24] Shu Lu and Yilu Liu, "Harmonics from DC Biased Three-
Phase Transformer Banks", International Journal of Power
and Energy Systems, Vol.17. No.1, 1997

42
 Chapter 6

Harmonic Modeling of Networks

Thomas H. Ortmeyer M. Fayyaz Akrarn Takashi Hiyama

ECE Dept. Dept. of Elec. Eng. Dept. of Electrical
Clarkson University Univ.ofEngineering and ComputerEng.
Potsdam, NY 13699- and Technology Kumamoto
5720 Lahore, Pakistan University
Kumamoto, Japan

Abstract: Harmonic modeling for power systems
involves the incorporation of device models into a
power system model. The development of accurate
system models for harmonic studies involves the
selection of thedevices to include in the model as well
as the selection of device models which achieve a
balance between complexity and accuracy for the study
in question.
electricity consumer. Primary distribution generally
ranges from 4k:v through 36kv, and is typically utility
owned.
There are two reasons to undertake a distribution
system harmonic study: first, to study-rae impact of a
large new harmonic source and secondly, to examine a
harmonic problem on an existing system.

6.1 Introduction
Harmonic propagation studies begin with the
development of a system model which is the group of
device models to be included in the study. In nearly
every study, it is not feasible to include detailed models
of every component of the system. Interconnected
power systems typically include hundreds of generators
andtransmission lines, and even more distribution lines
and customers. Even stand alone systems have more
loads than can be modeled individually.
Every harmonic study must begin with a determination
of the frequency range of interest and the selection of
the system components which will be modeled. This
chapter describes procedures which are used to develop
system models for harmonic studies. The chapter is
divided into two major sections which describe system
development for distribution systems and for
transmission systems, respectively.
6.2 Distribution System Modeling
While distribution systems and transmission systems
have the same components- lines, transformers, and
machines- there are significant practical differences in
developing systems models for the two types of
systems. Distribution systems, in fact, are divided into
two distinct levels- primary distribution and secondary
distribution. Secondary distribution is most often
below 600 volts? and is typically owned by the
Three phase or single phase modeling The first decision
to make in any distribution system harmonic study is
whether a three phase model is required or if a single
phase model will be sufficient. The three phase model
is required when:
a combination of wye-wye and/or delta-wye
transformers leads to harmonic cancellation
single phase or unbalanced capacitors are
present
ground or residual currents are important in
thestudy
significant unbalanced loading is present
As one or more of these cases is present on many
distribution systems, it is often recommended to
implement a three phase model on any distribution
system study. There have, however, been numerous
successful studies which have been single phase in
nature. The typical instances where a single phase
model may be sufficient are:
a single large three phase harmonic source is
thecauseof the study
theremaining system is well balanced
ground currents arenotan issue
The single phase model of thesystem can be attractive
as it is one third the size of the three phase model and
theresults can be more compact and easier to interpret
'Still, it should be employed only when it is clear that it
will be sufficient forthe study being undertaken.

43

Supplying transmission line
Tnmsm1sslon Tl'1IIISIIIission
S0111"Ce soarce
) For a study of harmonic propagation on the primary
( system, the components of the system should be
) modeled as appropriate for the frequency range of the
... : to be modeled with care. The increase in line
Load!
Harmonic
Source
Figure 6.1. The one line diagram of a typical
primary distribution system.
The extent of the system model Most distribution
systems aretied into the interconnected power network.
The exceptions are certain stand-alone systems such as
are encountered on oil-drilling platforms. In stand-
alone systems, it can be feasible to model the entire
power system. In other cases, however, the system is
too large to fully model. A decision must be made as to
which components to model in detail, and which areas
of the system can be modeled with a network
equivalent.
A common primary distribution system is shown in
Figure 6.1. The system is radial, with energy supplied
to the system through a stepdown transformer from the
transmission netwqrk. In many cases, it is sufficiently
accurate to represent the transmission network by it's
60 Hz short circuit equivalent resistance and
inductance. There should be consideration of the fact
that the short circuit strength of the supplying system
will change depending on the system configuration. A
more detailed model is needed when power factor
correction capacitors are present on the transmission
system nearthe step down transformer.
44
The final modeling aspect to be aware of is that the
transmission system can be a major source of
harmonics for the distribution system. The only way to
determine if this is the case is through measurements on
the transformer secondary. In most cases, these
measurements can be made on the existing current and
voltage transformers. The measurements should be
made with an instrument that will provide phase angle
information as well as magnitude information. The
length of time over which the measurements are needed
must be determined on a case by case basis.
study. Typically, the capacitance of overhead lines and
transformers does not need to be included in studies
involving the lower order harmonics. The interaction
of power factor capacitors and the line and source
inductances is the primary driver of the harmonic
impedance at the lower order harmo~i~.-so these need
resistance due to the skin effect provides increased
damping at the systemresonance points and should be
incorporated into the model.
The final component of the distribution systemmodel is
the load. Load modeling is difficult, as it is not
possibleor desirableto identifyexactly what the load is
at any given point in time. The load model is therefore
somewhat empirical in nature, and different methods of
determiningload models are in use.
system loads
As indicated in Figure 6.1, most if not all
generate harmonic currents to some degree. Every
systemwill have some background harmonic level even
when no large harmonic sources are present. A
generic load model therefore includes both a harmonic
source and a harmonic impedance. An example load
model is shown in Figure 6.2. The series impedance is
often taken to include the effect of the distribution
transformer. Several different methods have been used
to select the shunt impedance values for a given
load[1,2]. One method is to model motor load
separately from the other load. The passive loadMYA
is converted to an equivalent parallel R-L impedance.
The motor load is modeled as a single lumped
induction motor with appropriate leakage reactances
and stator and rotor resistances.
Another issue in model development is load
aggregation, as it is not necessary or feasible to model
each load individually. Load "ggregapon was studied
in [3], where it was shown that feeder loads can be
aggregated into fairly large groups without excessive
loss of accuracy. For the feeders involved in that study,
the load was grouped into approximately 10 equivalents
per feeder.
Zseries
Utility
source
oll-S ite
geJl.eratioll

Figure 6.2. Generic per phase load model

T1IJIed
filter
Many harmonic studies involve a small number of large
harmonic sources. In these studies, the background
harmonic level is often ignored in the study, and
considered separately as a source of error. The
background harmonic level, however, is important tor
several reasons, including in the design of harmonic LUt.
filters. At present, the only way of determining Load
background levels on a given system is through
measurement. An-efficient measurement and modeling
procedure is described in [4]. A procedure which
combines measurement and modeling is generally
required in performing harmonic studies on primary ASD
distribution systems.

Secondary distribution systems Studies of secondary
distribution systems involve studies of a single plant or
commercial installation. Many of the modeling
characteristics of primary distribution systems also hold
for secondary systems. The one line diagram for a
Figure 6.3. Industrial plant one line diagram.
Table 6.1. Summary of typical dataneeded for a
distribution harmonic studLYe
:beVice~T:::. :bata~Deed.edH:~~jj~~~~: H~:::~:: :::::(:::::: ",

typical industrial system is shown in Figure 6.3. The transfonner Actual turns ratio,
plant model is likely to include several different voltage connection diagram, shortcircuit
levels and likely will have fewer capacitors, but more impedance
of the capacitors will be installed with tuning coils for overhead phase and neutral conductor size,
filtering purposes. Line and transformer capacitances lines, cables layout,length, or short circuit
are typically negligible, as is also the case for short impedances; capacitance (when
cable runs. In many cases, measurements -are easier to needed)
perform on secondary systems, and load data may be capacitorbank voltagerating, var rating, configuration
more readily available. !( wye, grd wye, or delta)
tuned filter tuned frequency,
Data preparation At present, most studies are
volt, var rating, configuration
performed with one of the several commercial
generatorl subtransient impedance, configuration
harmonics analysis software packages which are
large motor
available. The device data entry and model synthesis
using these packages is convenient once the data has load, linear watts, power factor, composition,
been gathered. Typical data which is needed for balance
studies is summarized in Table 6.1. load, expected level of harmonic current
nonlinear injection, maznitude andphase angle

45
 6.3 Transmission System Modeling
Transmission system modeling is somewhat different
than distribution system modeling for a number of
reasons. The level of capacitance of lines and, in some
cases, transformers, is such that these capacitances
must be included in the models. Transmission systems
have higher X/R ratios than distribution systems, so
that harmonics can propagate for much longer
distances- and distant components can have a
significant impact on harmonic propagation. The third
aspect of transmission system modeling is that the
interconnected system of generators and transmission
lines will typically take on a much wider range of
operating configurations than is observed in most
distribution systems.
A representative transmission system is shown in
Figure 6.4. For convenience, only a single source and
critical bus are shown, although there maybe more than
one of either of these. A transmission system harmonic
study begins with the identification of a local system
which must be modeled in detail- Le., each component
modeled individually with an accuracy appropriate for
the study. Due to size and time constraints, distant
portions of the system must be represented as lumped
equivalents. There is a third, intermediate area where
systemrepresentation is needed for accuracy. The key
to transmission system model development lies in
accurately and efficiently selecting the boundaries of
the intermediate system and selecting appropriate
lumped models for the remote system representations.
This can be a difficult task, and there exist several
differentmethods for making theseselections.
remote systems
local
network
critical
bus harmonic
source
Figure 6.4. Representative transmission system for
a harmonic propagation study.
46
Model size selection. The sizing of the system to be
modeled in detail- the local system- has been
approached in three basic ways.
1. Engineering experience. The engineer
conducting the study decides what parts of the
system are to be modeled based on previous
experience [5~6]. This experience is based on
previous studies- preferably harmonic studies- and
identification of key components such as capacitor
banks, large generators, etc.. This method can
work well, but also can fail when the study is
outside the range 0 f the experience 0 f the engineer
doing the study.
2. Distance methods. Distance from the source
bus is often used as a modeling criteria-geographic
distance, series line impedance, and number of
buses distant from the source have each been used
[7]. In order to get sufficient accuracy when strictly
applied, these methods can result in modeling
unimportant system segments which are a similar
distance (from the source bus) as Important system
components.
3. Sensitivity methods. More rigorous
approaches to system model development involve
the use of one or more network sensitivity tooIs.[S-
10] Two examples of sensitivity methods are as
follows.
Remote system eguivalencing. There are 2 basic
equivalencing methods. The first and simplest is the
use of the fundamental frequency short circuit
impedance (here, short circuit impedance is taken to
mean R+jIDL, where Rand L are constant as frequency
changes)[6]. This approach has the advantage of
simplicity. Some studies have used open circuits or
short circuits in place of the short circuit impedance-
and have used the sensitivity of the results to these two
models to judge the feasibility of using a lumped model
at that bus.
The second approach involves the use of a frequency
response curve which represents the changes in
impedance of the remote system with respect to
frequency variations [11]. This method is moreflexible
than the previous method, and has the capability of
yielding accurate results with a smaller system model.
It is particularly useful in time domain studies where
system size can be more severely limited. It is,
however, difficult to perform switching studies
involving components in the equivalenced network. In
these cases, it is preferable to build a larger system
model which includes component models of all
equipment which will be involved in a switching study.
A second weakness of most frequency response
methods is the inability to model coupling between
remote networks- the systems 1-5 of Figure 6.4.
Sensitivity Analysis Method: The adjoint network
analysis can be used to efficiently determine the
sensitivity of the system response to component
parameter variations [10]. The transmission network
N and its adjoint network N can be used to
determine the sensitivity of the system response to
parameter variations. The transfer impedance T is
defined as the harmonic voltage at the bus of primary
interest divided by the injected harmonic current. The
network N is excited by a unit current "source at the
harmonic source bus to get network branch currents
II .,12 ,-.. , In The adjoint network N which has
9 regions, a simple equivalent can be used. In the case
the sametopology as the original network, is excited by
a unit current source from the output to get adjoint
network branch currents ii ,..
,~ ,Tn. The
sensitivity of a transfer impedance T with respect to
any parameter x (R, L, or C), at frequency OJ,
denoted by Sr. is dermed as
sI = ~(;)
Thesesensitivities can be calculated using
aT -
-_ = I(x)[ (x)
ox
where l(x) and I (x) are the x element branch
currents from the analysis of N and N respectively.
The calculation of the transfer impedance sensitivity is
efficient Its effectiveness is limited to small parameter
variations as it involves partial differentiation.
Bilinear Theorem: The large variations which can
occur in an external system impedance cannot be
reliably predictedby usingsmallsignal analysis. Large
changes in the traDSfer impedance of a network to
cbanges in an element Z (in this case the remote
system equivalent impedance) can be assessed by
pulling that element Z out of the networi. effectively
forming a three port network [9]. For the transfer
impedance V%. the following general equation is
obtained:
T__V2 _ _Z.....
xin....._T_(O_}+_Z__T_<00_)
- I} - Z+Z.xin
47
where T(O) is the transfer impedance when branch
impedance Z = 0 while T( 00) corresponds to the
=
transfer impedance when Z 00. Zxin is the input
impedance looking into the network from the nodes of
Z. In order to assess the modeling accuracy at a tie
bus at some harmonic frequency, T(O), T( 00) and
Zxin are determined through three respective network
solutions. The bilinear formula can then be used to
determine the transfer impedance T for any value of
Z . A typical result of the bilinear analysis is shown in
Figure 6.5, which shows the impedance regions where
largeerrorswill occurin a given transfer impedance. If
the actual system impedance will not enter these
shown in Figure 6.5, the positive error will never
exceed 5% as this sensitivity does not occur in the
positive resistance region.
jX
region where or
exceeds +5%
~
region where error
... ,exceeds -5%
..,," J~
R
Fagure 5. RX diagram showingsensitiveregions
for remote equivalent impeclaDces.
Summary
This chapter desaibes methods which can be used to
develop system models for harmonic studies. The
chapter covers modeling techniques for both
distribution and transmission level barmonic studies.
Theo development of an effective system model depends
onthe accurate determination of which system elements
to model in detail. The second aspect of system
modeling is to use the appropriate device model which
will provide accurateresults witboutundue complexity.
 References IEEE Trans. On Power Delivery, Vol. 3, No.3,
(JUly, 1988).
1. R. C. Dugan, M. F. McGranaghan, and H. W.
Beaty. Electrical Power Systems Quality,
McGraw-Hill, New Yor~ 1996.
2. "Modeling and Simulation of the Propagation of
Harmonics in Electric Power Networks, Part I:
Concepts, Models, and Simulation Techniques"
(Task Force Report), IEEE Transactions on Power
Delivery, Vol. 11,No.1 (Jan., 1996), pp. 452-465.
3. T. Hiyama, M. S. A. A. Hammam and T. H.
Ortmeyer. "Distribution System Modeling with
DistributedHarmonic Sources. IEEE Transactions
on Power Delivery, Vol. 4, No.2 (April, 1989),
pp. 1297-1304.
4. T. H. Ortmeyer and T. Hiyama. "Distribution
System Harmonic Filter Planning," IEEE
Transactions on Power Delivery, Vol. 11, No.4
(Oct. 1996), pp. 2005-2012.
5. B. R Shperling and L. Mennemenlis-Hopkins,
"Generation and Distribution of Current
Harmonics on EHV Systems," Proc. of the 3M lnt,
Cont. on Harmonics in Power Systems, Nashville,
IN, Sepl 1988. pp. 149-154.
6. o. Bergault and G. Morin. "Analysis of a
Harmonic Overvoltage due to Transformer
Saturation following Load Shedding on Hydro-
Quebec-NYPA 765 lev Interconnecuon," IEEE
Trans. On Power Delivery, Vol. S., No.1 (Jan,
1990). pp. 397-405.
7. R. S. lballam, "Harmonic Propagation and
Amplification in a 115 lev network," Proc. Of the
411. Int. Cont. On Harmonics in Power Systems.,
Budapest, Hungary, oe, 1990. pp. 148-153.
8. M. G. Wiclcramasekara and D. Lubkeman,
"Application of Sensitivity Factors for the
Harmonic Analysis of Distribution System
Rec()ntiguratioD and CapacitorProblems," Proc. of
the 3M Int Cont. on Harmonics in Power Systems,
Nashville, IN, Sept. 1988. pp. 141-148.
9. M. F. ~ T. H. Ortmeyer, and J. A.
Svoboda. "External System Modeling for Power
Transmission System Harmonic Analysis using the
Bilinear lbeorem." Electric Machines and Power
Systems, Vol. 23(1995). pp. 141-147.
10. M.'F. Alaam, T. H. Ortmeyel, J. A. Svoboda.
..An Improved Harmonic Modeling Technique for
Transmission NetWork," IEEE Trans. on Power
Delivery, Vol. 9, No. 3 (July, 1994), pp. IS10-
1516.
11. N. R. Watson and J. Arillaga, "Frequency-
Dependent AC System Equivalents for Harmonic
Studies and Transient Converter Simulations."

48
 Chapter 7

Frequency-Domain Harmonic Analysis Methods

s. Mark Halpin" Paulo F. Ribeiro2, and J.J. Dai3

'Mississippi State University, Starkville.. MS
2awx Technologies, Inc., Lynchburg, VA
30peration Technology Inc., Irvine, CA

Abstract: There are a large number offrequency-domain II Yu Yli Ylj YIN VI

analysis methods that are in widespread use. The most
popular of these are frequency scans, harmonic
penetration, and harmonic power flow. Each of these
y.1 Yil Yii Yij YiN V,
= (7.1)
techniques can be employed on a "per-phase" (positive or y.J Yjl Yji Yjj YjN VJ
zero sequence) or "multi-phase" basis and each technique
utilizes an admittance matrix system model developed
from individual component-level models connected
IN YNI YNi YNj _YNN VN
according to system topology. The development of this
admittance matrix system model and the frequency- The complex admittances (b~.-on a knO\\l1
domain harmonic analysis algorithms that make use of it frequency) in (7.1) can be determined from the
are the subjects of this chapter. definitions of admittance parametersas shown in (7.2).

7.1. Introduction (7.2)

The methodology employed in the development of

admittance matrix models is based on multi-port network
theory. Positive sequence admittance matrix models are
developed from device-level two-port admittance (y)
parameters. Multi-phase models are similarly developed
from multi-port admittance parameters. Figure 7.1
shows a general multi-port model with appropriate
terminal parameter definitions.
i1
For cases where a common node voltage reference point
is available (as represented by the dashed line through
the network in Figure 7.1), an advantageous
simplification to (7.2) can be developed. The rules I and
2 as follows can be used to define lOy. complex
admittance Yij in (7.1).
L (all admittances connectedbetween i and j )
1: Yij = -
IN 2: Y = L(all admittancesconnectedto i )
ii

-I . N-port -
IJ These well-known proceduresform the backbone of most
VI +
I
network VN admittance matrix formulation algorithms because the)"
~

v v.J can be applied directly to a complete system. Fonnation

'-------+_ - - - - - - - - t ' - -_ _- J
of component-level admittance models is not required
[1,2J.
In some admittance matrix building algorithms. an
Figure 7.1. N-port Network
incidence matrix [Q] is used to represent network
connectivity. In this approach, a triple matrix product is
The arbitrary network of Figure 7.1 should, for
used to. form the system admittance matrix model as
simplicity, be limited to passive elements. It is possible,
shown in (7.3) where the complex matrix [Ypnm] consists
however, to include certain dependent sources provided
of component-level admittance matrix models formed
that the dependence can be cast in an admittance
according to (7.2) or the simplifying rules I and 2.
relationship. Equation 7.1 describes the general network
of Figure 7.1.
(7.3)

49
It should be noted that the procedure based on rules I and
2 is appropriate only when the common node voltage
reference point is present. In general, this is true for
either balanced or radial systems. Unbalanced nonradial
systems should incorporate more generalized admittance
matrix modeling techniques as described in [31.
As previously described.. the modeling approach can
be applied to either single-phase or multi-phase systems.
For multi-phase systems, the rules 1 and 2 given
previously must be considered to apply to 3x3 admittance
matrices. The subscripts i and j, therefore, must be
expanded to represent i+O, i+ 1, andi+2 and j+O, j+ I, and
j+2, respectively. The summations given in the roles
apply so as to sum entries in the system admittance
matrix with those in the 3x3 component-level models. If
the approach of (7.3) is employed, the primitive
admittance matrix will consist of 6x6 blocks provided
that two or more components are not mutually coupled.
The incidence matrix [Q] must be modified accordingly.
References [4) and [5] provide implementation details for
balanced or radial systems and more general unbalanced
nonradial systems, respectively.
For harmonic analysis, the admittance matrix must
be fonnulated at each frequency of interest. The matrix
must be re-built from scratch; direct modifications to
convert the system matrix from one frequency to another
are not usually possible. In general, the matrix is re-built
from the component level RLC parameters for circuit
models for lines, transformers, and other power delivery
equipment The actual matrix construction procedures
described in the previous paragraphs apply without
modification.
7.2 Frequency Scan Analysis
Frequency scan analyses are used to characterize the
response of a power delivery system as a function of
frequency. The term "scan" arises from the systematic
variation of frequency from some initial value 10 to some
final value fF The frequency scan analysis is conducted
through repeated solutions of (7.1) with the admittance
matrix formed for each frequency of interest. Equation
(7.4) provides a clear "frequency-dependent" yersion of
(7.1) where h is used to denote the harmonic frequency
(in either Hz, radls, or PQ).
I(h) = [Y(h>JV<h> (7.4)
Two types of frequency scans are commonly
performed. The first type is based on a single "current
injection" into the power delivery system Ybus model
followedby a solution of (7.1).. Assuming this injection
takes place at node i in Figure 7.1, (7.1) can be solved to
determine the voltages that are produced at each system
node. If the current injection is assigned a value of l/Qo
50
(A or pu), the values of the determined voltages represent
the driving point and transfer impedances as seen at node
i. Because the Ybus model contains only linear elements"
linearity can be applied to scale the results obtained for
the l/Qo (A or pu) to any desired value. This scalability
can be applied to estimate the harmonic voltages that will
be produced at any network bus when a load that draws
nonsinusoidal current is connected at the . . injection"
node. This technique is often used when assessing the
potential impacts of new ASO or other harmonic-
producing load and is very useful for identifying series
and parallel resonances.
Varying the frequency used in the implementation
and solution of 7.1 yields a series of impedance
magnitudes and angles covering the range of frequency
fo-fF A plot of this series provides excellent visual
indication of resonance conditions. Parallel resonances.
which are associated with high impedances to current
flows, appear as "peaks" in the plot. Series resonances,
which are associated with low impedances to harmonic
current flows, appear as "valleys" in the plot A sample
frequency scan is shown in Figure7.2. _".
Posi~ive Sequence Drivinq Point Impedance
40.,.--_ _- . . , . . - _ - _ . . . , . . . . . ~ ~ _ ~ - - - ~
)O+------+----~__f__--+-_+_---_4
ZO+------+-----+-l~-~_+_---_f
12 II
Freguency (M pa)
Figure 7.2. TypicalFrequency (Impedance) Scan
The second type of scan is conducted in the same
manner as described previously, except that a l/Qo
voltage 01 or pu) is connected to one node in the
network. Equation (7.1) is then solved for all other
voltages in the network. The resulting voltages represent
the voltage transfer functions to all other nodes in the
system model. While similar in technique to the
previously described frequency scan, the procedure of
applying a 1/20 voltage source is more typically calleda
"voltage transfer function" analysis to allow the term
"frequency scan" to be associated with driving point and
transfer impedances determined by injecting a known
current into a node.
The voltage transfer function analysis is useful for
investigating the effects of background hannonics. The
terminology ~backgroUDd harmonics" refers to the
harmonic voltage distortion that may be present at the
terminals of any network equivalent. As with the
impedance (frequency) scan. a plot of the voltage transfer
function as a function of frequency can be used to reveal
potential problems. Peaks in the plot indicate
frequencies at which voltages will be amplified and
valleys indicate frequencies at which voltages will be
attenuated. An example plot of a voltage transfer
function is shown in Figure 7.3.
.,~
Voltage Transfer Funct.ion (lV@t.he substat.ion)
1 1.
I
I
L.!O
1\ I those in the Fourier series current vectors for the
I 1.2~ /\ i nonlinear loads.
1.00
.'
/ \
O. 7~
\\ A.
/ \
-.
\ :
\\ V admittance matrix model construction) from
o.n
~~
I ,
0.00 I I
I 0
Figure 7.3. Voltage Transfer Function Plot
As previously mentioned, the admittance matrix
system model can be formed based on sequence networks
or phase-variable networks. Multiphase systems can be
handled without modification. Because the frequency
scans described previously (both types) are only
modifications of this admittance model, the procedures
are equally applicable to sequence- and phase-variable
models. It should be noted, however, that the phase
angles of the voltage or current injections are important
In phase variables, a three-phase positive sequence scan
would be conducted by injecting a current vector of [1/Qo
11-120 1/1200 )T (A or pu) into a three-phase bus. A
zero sequence scan would be conducted in the same
mannerexceptall phase angles would. be equal (typically
0). Of course, a single-phase scan would involve only an
injection into one node of a three-phase bus. The
appropriate injections into sequence variable-based
admittance models can be determined by applying the
symmetrical component transformation to the values
given for the phase-variableinjections.
7.3 CurreDt Source Methods
Currentsource (or currentinjection) methods are the
most popular forms of harmonic analyses. The methods
all make use of the admittance model as described in
section 7.1. The analysis procedure is similar to the
frequency scan analysis for current injections in that
nonlinear loads are represented using harmonic current
spectra of known magnitude and phase. More
specifically, nonlinear loads are represented using a
summation of currents where each entry in the sum
corresponds to a termof known frequency in the Fourier
series representation of the load current Taken
collectively, this sum is.often referred to as a "vector."
With these spectral vectors (magnitude and angle at each
harmonic of interest) known for each load, the analysis
approach proceeds along the following series of steps:
Step 1. Formulate the system admittance matrix model
of the power delivery system including
contributions for all sources and linear loads.
The frequency should be consistent with one of
Step 2. Construct the current injection vector in (7.1) by
extracting the tenn of the appropriate frequency
(which must match the frequency used in the
t'--- II each nonlinear load harmonic current vector.
.. I
I
Step 3. Solve (7.1) to determine the voltages at all
network buses. The frequency associated with
these phasor voltages is the same as that used in
the construction of the admittance model.
The steps 1-3 begin at the lowest frequency
represented in any of the load Ct1lTeQt:bannonic vectors
and repeat for each frequency in all of the nonlinear load
models. It should be clear that not all nodes will posses
harmonic load current injections at all frequencies; some
loads inject 5th , 7 th, 11th.. 13th , etc., while others inject 3rd ,
5th, 7th, 9th, etc. For the case where a nonlinear load does
not inject a current at a particular frequency (but another
nonlinear load does), it is a simple matter to force the
injection to a zero value at that loadbus and continue the
solution of (7.1) at the frequency of interest.
The results of an analysis conducted using the
current injection method is a collection (again, often
referred to as a vector) of harmonic voltages lor eachbus
in the system. Due to the linear nature of the problem
(all nonlinearities have been represented as current
injections), superposition can be applied. Theretore, the
terms in each voltage vector correspond to the Fourier
coefficients of the time-domain voltage. These spectra
(magnitude only; phase is usually not shown) are often
shown graphically as in Figure 7.4. If desired. the time-
domain wavefonn can be easily constructed from the
voltage spectrum at each network bus.
f .000 i i I
i I
I
i
1
1
t
4000
zooo
o - .
Figure 7.4. Example Voltage Magnitude Spectrum
51
 The preceding paragraphs have described the
procedure in general terms, In practice, there are a
number of modifications that are used and that can" in
some circumstance, produce markedly improved results.
The most pronounced modification to the general
procedure is the use of phase information in each
nonlinearload harmonic current vector.
In studies where only a single nonlinear load is
present (or a single nonlinear load dominants all others).
the phase angles for each harmonic current phasor are
not important. In systems with multiple nonlinear loads,
however, neglecting harmonic current phase angles in
load models can lead to inaccurate results. For systems
with multiple nonlinear loads, harmonic currents at each
frequency may be additive or subtractive. so voltage
harmonics determined using steps 1-3 could be either
over-or under-conservative for any particular harmonic.
Another enhancement that is often made is to
include the effects of the fundamental frequency terminal
voltage on the hannonic currents generated by a
nonlinear load [6]. As an example, consider the voltage
and current waveforms shown in Figure 7.5. For the
voltage given in (7.5), the load current is as given in
(7.6).
1 point where the load terminal voltage distortion exceeds
'" 0.8
:;
ci. 0.6
i 0.4
~ 0.2
a 0 +--..-..I~--.....---+---,---+-~--+--~
4td ..0.2
5fa -0.4
~ vectors represent very specific load patterns. It is well
..0.6
-0.8
-I
t known that many nonlinear loads, including ASDs.
ADgle(deg)
Figure 7.5. Example Voltage and Nonlinear Load
Current Waveforms
i.(t) = 2./3 co~Q}t)
7t
~ 2..{j
S7t ~
co-' Scot)
(7.6)
voltage angle is multiplied by "n" in the Fourier series of
the current waveform, where n is the harmonic order of
each term.
.
1,(t) = 2./3 J
-CO~(i)t+u
~)
1t
2/3
--co~Soot + SO) (7.8)
51t .
2/3
+-co(7rot + 76)
71C
Note that this same corrective action of n6 is the basis for
the use of 12 pulse (and higher) drives. In these higher-
order drive systems, significant harmonic Cancellation
can be obtainedfor certain harmonics due only to the no
correction.
One limitation of the current source method is the
validity of the hannonic current vector-representation of
the nonlinear loads. Past experience has shown that this
representation is valid for most nonlinear loads up to the
lOOA.. However, ongoing research in this field indicates a
clear impact of terminal voltage distortion on certain
nonlinear load harmonics, especially those produced by
popular ASDs [7].
In addition, the current source method is limited to
"snapshot" scenarios where the harmonic current source
produce markedly different harmonic currents depending
on load level. Figures 7.6 (a) and (b) show the line
current waveforms (and the associated harmonic
magnitude spectra) drawn by a 2S0hp dc drive for (a)
light load (spinning) and (b) fully loaded. It is difficult
to capture the range of harmonics associated with these
two line current conditions (and all loading points in
between) without a large number of simulations using
the current source methods. Further complicating the
problem is the variation in fundamental frequency
termiDal voltage phase angle that accompanies changing
load conditioDS.

2..{j
+-co~76)t)
7'1t

However, there can be significant differences in

fundamental frequency bus voltage pbase angles in actual
systems. Modifying (7.5) to include an arbitrary pbase
angle 6 as shown in (7.7) leads to the modification of
(7.6) as shown in (7.8). Notice that the fundamental

52
 are required for each frequency included in (7.8)" to form
a complete mathematical model of the system. The
Amps entire set of equations is then solved iteratively using
10 2.08 417 6.25 8.34 ~4.S9 either Newton or Gaussian methods. Linear loads may
be represented with a combination of impedances or with
mSec a constantpower (p+jQ) model.

Amps
rms 1f2J (7.9)

Hannonic
Figure7.6(a). Lightly-Loaded DC DriveLineCurrent
200
100
Amps 0 L---~::::::.:=-~-------'j==~ -:-
1+ -100
-200
mSec
Harmonic
Figure 7.6(b). Fully-Loaded DC DriveLine Current
7.4 Harmonic Power Flow
The deficiencies in the current source method can be
partially overcome using a technique that bas come to be
known as "harmonic power flow" or HPF. HPF
algorithms combinethe current source methodology with
a conventional power flow algorithm. There are two
basic variations of HPFs that find widespread use, and
theyare described as follows:
1. A fundamental frequency power flow solution is
executed using a linear model for all power delivery
equipment and loads, and the resultant fundamental
frequency load tenninal voltages are used to "adjust"
nonlinear load harmonic current vectors (as shown in
(7.8 automatically without additional user action. The
harmonic current vector-tis still required to be known for
each load (as was the case for the current source
methods).
2. All (or some) nonlinear load harmonic current
spectra are represented in the fonn of (7.9) where
C),...,CM represent M control variables that are used to
control various load parameters (such as shaft speed in a
motor drive) and the phasor voltages V.,... ,VN represent
the harmonic voltage pbasors at the load tenninals.
Nonlinear load representations of this fonn are used in
conjunction with (7.1)~ where specific instances of (7.1)
S3
The first HPF version is a relatively simple extension
of the current source method. The same limitations
apply, and the only advantage is the automatic
"correction" for fundamental frequency terminal voltage.
Due to this relatively minor improvement, this first
improvement is considered by many notto be a true HPF.
The second HPF version is an extremely complex
and powerful technique. The system model is formed as
described for the current source method, but the loads
can be modeled in an almost arbitrari~mplex manner
depending on the amount of detail required to obtain the
level of accuracy desired. When a closed-form solution
for the nonlinear load current harmonics can be obtained
as a function of voltage (including voltage harmonics)
and control parameters, it is possible to represent the load
harmonics directly in the frequency domain as shown in
(7.9). In many cases, this closed-form solution can not
be obtained and a combination time- and frequency-
domain techniqueis employed.
The so-called hybrid HPFs utilize a power delivery
system model in the fonn of (7.1) for each harmonic, but
represent the nonlinear loads with ~e-domain
differential equations (8]. Given an initial estimate of
the network voltages, the load models are simulated (they
can be decoupled if necessary) until steady-state is
reached. A new harmonic current vector is then created
from the steady-state current for each nonlinear load.
These current injections are then used as described for
the current source method to obtain an updated terminal
voltage (including bannonics). The procedure continues
until the frequency domain network model converges
with all nonlinear loadmodels in steady-state.
The hybrid methods are by far the most powerful" but
they are also the most complex. It is possible to represent
converter controls, for example" in great detail so as to
account for virtually every possible harmonic scenario.
With this capability, however, comes the requirement
that the user must have the data and expertise required.
More often than not, this is not the case. Detailed
harmonic studiesinvolving complex converter controls or
widely-varying load patterns, therefore, are often best
analyzed using complete time-domain models that are
simulated usinga transient analysis program such as the
EMTP.
 7.5 Conclusions

In this chapter. the most popular frequency-domain

harmonic analysis methods have been presented. While
the approach presented here concentrates on admittance
matrix modeling, it is equally possible to utilizeadvance
impedance matrix models in conjunction with each
analysis type presented. Each of the methods has found
significant practical application as illustrated in the case
studies found throughout this tutorial. It is, however,
always up to the analyst to weigh the costs (improved
complexity) and benefits (increased accuracy) of any
given method for the problem at hand.

7.6 References

[1) W.D. Stevenson, Jr., Elements of Power System

Analysis, 4th ed., McGraw Hill, 1982.
[2] C.A. Gross, Power System Analysis, John Wiley &
Sons, 1986.
(3) M. A. Wo~ D. L. Allen, and L. L. Grigsby,
"Techniques for the Steady State Representation of
Unbalanced Power Systems: Part I. A Systematic
Building Block Approach to Network Modeling," IEEE
Transactions on Power Apparatus and Systems, Vol.
PAS-I04, No. 10, October, 1985.
(4) J. Arrillaga, D.A. BracUey, and P.S. Bodger, Power
System Harmonics.. John Wiley 8 Sons, 1985.
(5) S.M. Halpin, A. Mai~ and C.R Dickinson,
"Calculation of Hannonic Current Flows in Single-Phase
Power Distribution Systems.." Proceedings of the 7th
IEEE International Conference on Harmonics and the
Quality of Power. pp. 516-521, October, 1996.
[6] Task Force on Harmonics Modeling and Simulatio~
"The Modeling and Simualtion of the Propagation of
Harmonics in Electric Power Networks Part I: Concepts,
Models, and Simulation Techniques," IEEETransactions
on Power Delivery, Vol. 11, No.1, January 1996.
[7] D.E. Rice, "A Detailed ADalysis of Six-Pulse
.Converter Harmonic Currents," IEEE Transactions on
Industry Applications, March/April, 1994.
(8) B.K. Perkins and MR Irvani, "Novel Calculation of
HVDC Converter Harmonics by Linearization in the
Time Domain," IEEE Transactions on Power 'Delivery,
Vol. 12, No.2, April 1997.

54
 Chapter 8
TIME DOMAIN METHODS FOR THE CALCULATION OF
HARMONIC PROPAGATION AND DISTORTION
By
e.l. Hatziadoniu
Department of Electrical Engineering
Southern illinois University
Carbondale, IL 62901-6603
hatz @ siu.edu
8.1 INTRODUCTION
Numerical methods for the computation of harmonic
propagation and distortion differ in the manner in
which they represent the harmonic sources and the
system impedance. Iterative methods (e.g. harmonic
load flow) use a phasor representation of these
parameters. Time domain methods, on the other hand,
use a time representation of the system elements and
the harmonic sources. Thus, they are generally more
accurate than the iterative methods.
The simplest system modeling for harmonic
calculations considers rigid harmonic sources and
linearsystem impedance [1-3]. A rigid harmonic source
produces harmonics of only characteristic orders with a
pre-detined and constant magnitude and phase. The
linear impedance is the result of mainly transmission
lines and compensation devices. A model which
includes rigid harmonic sources and linear elements can
be solved by an iterative method with the same
accuracy as a time domain simulation. This is because a
linear model has a phasor representation and the
principle of superposition is applicable.
The presence of non-linear and time varying
elements in the system model can significantly change
the manner by which harmonic currents and voltages
propagate through and interact with thenetwork. Some
of the effects that may appear and which are better
t
studiedby time simulation are [3-7):
a) Under ideal conditions, harmonic devices
(converters, transformers, etc.) produce harmonics of
characteristic orders. For example, in symmetric
saturation a transformer produces all odd orders, if its
terminal voltage is near sinusoidal. An ac/dc 6-pulse
converter produces orders such as 1,5,7,11,13.., if its
terminal voltage is nearly sinusoidal and balanced and
its de current is nearly free of ripple. Most of these
55
devices will produce uncharacteristic .orders if their
terminal conditions are not ideal. Examples are the
inrush current in a transformer and converters operating
with unbalanced voltages.
b) The switching function of power converters is
equivalent to a modulation\demodulation between ac
and de quantities [8,9]. This results in interactions
between harmonics of different order. Such interactions
arenot predictable by linear time invariant models. The
study of this phenomenon is important, especially in
systems likely to have significant barmonic distortion
because a path is provided through me converter for
unrelated harmonics to interact.
c) The gate control of power converters may
interact with harmonics on the system through the
synchronizing loop. This interaction in combination
with the modulation\demodulation property of power
converters can provide a feedback loop with significant
gain for the amplification of harmonics. Extreme
pbenomena that can result from these interactions
include limit cycles and harmonic instability.
The following sections summarize the methods for
the simulation in the time domain of non-linear and
time varying systems. The methods discussed are
suitable for widely usedprograms suchas EM'11>.
8.2 REPRESENTATION OF SYSTEM IMPEDANCE
There are two approaches for the representation of the
system harmonic impedance that are used in time
simulations. The first approach requires the detailed
representation of the network devices, which are
primarily responsible for the impedance properties. The
second approach uses the dynamic equivalent of the
impedance.
A Detailed Modeling. Detailed modeling of the
network involves 3-pbase models of the network
devices. The following is a summary of the modeling
approach for various devices. A detailed description of
network modeling is presented in a previous chapter
[2,3].
The classification of transmission line length is
determined by the wave length of the highest harmonic
frequency of interest. Long transmission lines are
represented by distributed parameter models. Medium
lines can be represented by cascaded Pi-equivalents.
Short transmission linesare usually represented by their
sequence impedance using lumped RLC branches. In
some distribution systems the capacitance may be
neglected for the overhead lines. Mutual inductance
may be included for medium and short lines to reflect
coupling between the phases.
Saturation and hysteresis modeling is required for
transformers, if significant overvoltage is anticipated at
the transformer terminals.
Compensation and harmonic filters are modeled by
lumped RLC branches.
System load contributes significantly to the
damping around resonant frequencies. Normally, a
parallel RL representation is used based on the
aggregate power of the load.
B. Dynamic Equivalent Modeling. This approach yields
a network model whichretains onlya selected group of
the buses (those of interest to the study). SUbsequently,
lumped RLC branches are used to represent the driving
pointand transfer impedances of the selected buses.
The driving point impedance is equivalent to the
Thevenin impedance of a system bus. Physically, the
driving point impedance shows the effect of the bus
harmonic current injections on the bus voltage, as in
(8.1).
With reference to (8.2) the transfer impedance
between twobuses, k and m, shows the effect of current
injections into bus m on the voltage of bus k and vice
versa.
v:=z~.~
(8.1)
11
k
II -- Zlcm
m
II 111
(8.2)
damping for the desired range of frequencies. Inclusion
of non-linear elements is done externallyto the model.
c. Simulation of Large Networks. With reference to
Figure 8.1, large networks are modeled by a
combination of the two approaches. A detailed network
model is obtained for the part of the network that is of
immediate interest. The remaining network is
approximated by its dynamic equivalent at the
interconnection buses.
This modeling philosophy effectivelydecreases the
size and length of a time domain simulation. Routines
that assemble the network dynamic equivalent are
available in programs such as EMTP [10].
8.3 REPRESENTATION OF HARMONIC SOURCES
There are three approaches for representing harmonic
sources in a time domain simutanon by voltage or
current injection through a rigid source, by the
switching function of the converter! or by a detailed
model of the converter.
A. Rigid Harmonic Sources. The generation of
harmonics by commercial loads has a distributed
nature. An aggregate modelcan be constructed for such
loads using ideal circuit sources (usually current
sources) [1]. The equivalent source is composed of the
harmonic spectrum of the load and does not change
according to the networkresponse(i.e. a rigidsource).
The value of the rigid harmonic source is given in
uc:
.- CD
E-
as~
c::.-
~::s
CO-
W
c
.2
oCD
(I)
C CD
C .(1)
o :::::s
u.Q
...
~
CD
..5

The driving point and transfer impedances can be

-0_
calculated either frommeasurements or from frequency CD CD
=-c
scans of the complete network model. SUbsequently, .so
the values of equivalent RLC branches between the cCD:E
selected buses are calculated. Usually, several series
RLC branches are connected in parallel to approximate
the multiple resonances of the system impedance. The
resulting model is a linear, tumped, 3-phase circuit and
it contains the system resonant frequencies and their Figure8.1. SimuiatiOll of largeaetworD using thedetailed aDd
dyaamic equivalent approaches

S6
time according to (8.3) for the desired number, N, of to (8.5). The time function of the current the TCR
studied harmonics. The magnitude of the fundamental injects into the systemis found from (8.6).
can be obtained from the aggregate power of the load.
The magnitude of the harmonics can be obtained from
measurements.

N
i(t) = 1\ cas(4l+ 8\ )+
o 2/ h ocos(hiIJ+8h)
2 (8.3)

The phase angle of the fundamental in (8.3) is obtained

from the system load flow and the load power factor.
The phase angles of the harmonics are not critical for
the calculation of distortion in models containing one
harmonic source. In the presence of multiple sources,
harmonics of the same order may be amplified or
attenuated according to theirrelative phases. Therefore,
Figure 8.2. Definition of the switching functionof a TCR
the correct estimation of the harmonic phase angle
should be used in (8.3). If the harmonic source has odd
symmetry with respectto its fundamental, the harmonic
phaseangle can be computed directly from thephase of
the fundamental, according to (84).
It
8h =h81 +(h+l)-
2 (8.4)

An expression in the form of (8.3) is included in the

time step loop of the simulation for each phase. The
expression is evaluated at each integration step and the
corresponding network source is updated. In ATP [12]
this can be done by coding (8.3) into MODELS and
using a controlled source in the network. In
PSCAD\EMTDC the network source of (8.3) can be Figure 8.3. The steady state voltageaaoss the inductor of the TCR.
directly written in FORTRAN as a user defined model
[11]. Modeling is possible using TACS controlled TCR
sources in conjunction with "FORTRAN" statements.
However, this is not an efficient method if a large
number of harmonics is simulated. NETWORK

B. The Switching Function. The terminal characteristics

of many converters can be approximated in the time
domain by the converter switching function [1,8,9].
Two typical applications are presented for the
definition of the switching tuncuon. V(t)
al Thyristor Controlled Reactors. With- reference to
Figure 8.2, the switching function of the TCR (for 1 S(t)
phase) equals 1 when the phase thyristors are
conducting and 0 when the phase thyristors are turned Figure 8.4. Tune domain simulation of the TeR throughits
off. At steady state, the switching function of the TCR switching function
is symmetric around the voltage zero crossing as is Vn::R = V.rys S(I)
evidentin Figure 8.2. Under steady state conditions, the (8.5)
I
voltage across the inductor of the TCR is shown in
Figure8.3. It can be written as the product between the
bus voltage and the TCR s~tching function according
m(t) J
= \TCR
o
0

(8.6)

57
Figure 8.4 depicts the simulation of a TCR by its appropriate switching function. The representation of a
switching function. The device appears as a current converter with its switching function allows interaction
source to thesystem. between the converter input and outputquantities. It is,
bl ACIDC Converter. The switching function of thus, more realistic than the representation by a rigid
the line commutated ac/dc converter is shown for one source. Some of the phenomena that can be studied
phase in Figure 8.5. It is 1, when the de current flows through the switching function and that cannot be
into the phase in the positive direction. It is -lone, studied by the rigid source representation are: the
when the de current flows in the negative direction and modulation\demodulation properties of the converter,
zero otherwise. The switching functions of the three which account for interaction between different
phases are symmetric and balanced. At steady state, harmonic orders; the generation of non-characteristic
they lag the system voltage by the convener delay harmonics; the propagation of de harmonics in the ac
angle. Equations (8.7) describe the ac current output of side; and operation under unbalanced voltage or
the converter. The dc voltage at the converter de current.
terminals is given by (8.8). The switching function is primarily a steady state
concept. Thus, it is assumed that the control system
... .,...----....--------------, I
delivers the ignition pulses at regular intervals and
uninterruptedly. It is also assumed that line
I
I
I commutation is completed successfully. Therefore,
I
I detailedrepresentation of the converter, control system
I
I is not required. Some types of interactions between the
t
t network harmonics and the converter controls (such as
I
~--~-----,----- the onset of harmonic instability) ciD be detected.
I
I
I
however, by introducing some detail in the
I
I
representation of thesynchronizing loop [9].
I Figure 8.7 shows a possible simulation in the
EMTP of the synchronizing loop of an ae/dc converter
in conjunction with its switching function. With
reference to this figure, the switching function is
represented by a user defined point-by-point function.
Figure8.5. DetiaitiOD of the switching functionof a six-pulse ac\dc Its x-axis corresponds to the equivalent of one period.
CODverter A ramp is used to scan the switching function. The
ramp is synchronized with the commutatiDg voltage of
i4 (1) =idc(I) S(J,(I) the corresponding phase. A circuit detects the voltage
i(J,(t) = idc(t) Sb(t) zero crossings and resets the ramp. The phase-tag with
ic(t) =idc(t) Sc(t) (8.7) the line voltage can be obtained by comparing the ramp
with the converterdelayangle.
The model of Figure 8.7 allows for interaction
between voltage harmonics and the gate control. In the
constant current operating mode of the converter, the
Figure 8.6 shows the time simulation of the converter.
use of a voltage controlled oscillator (VeO)
The device appears as a current sourcefromthe ac side
desensitizes the synchronizing loop from the system
and as a voltage source from the de side.
voltage waveform [8]. This model, therefore, could be
ACiX more appropriate for inverter operation, sincethe firing
CONVERTER pulses are then more susceptible to jitter by harmonics
1------------- in the line voltage. Thus, certain phenomena resulting
~ :. --------. AC from this interaction, such as harmonic instability, can
Network be detected without using a complex model of the
"'--_--.. : 8(1) I convener.
1 1
I Another system study that can be conducted with
~~ vlt)_~
the switching function is the harmonic interaction
between two or more converters in near proximity. The
Figure8.6. nme domainsimulatioD of the six-pulseac\dc converter network is represented by its dynamic equivalent This
may include only the buses of. the interacting
Other converters, such as the forced commutated conveners. such as in Figure 8.1. Each converter is then
voltage source inverter. can be represented by an represented by its switching function and its

58
synchronizing loop. A time simulation of the system
can predict with fair accuracy the changes in the system
harmonic propagation and distortion as result of the
interaction among the converters.
a frequencies and may last for a few seconds. The former
1-----' Lou S(t)
[/1 /1 /'
~ obtained. At the end of the run the system steady state
Figure 8.7. Timedomainrepresentation of thesynchronizing loop
in conjunction with the switchingfunction of a converter
c. DetailedSimulation of the Converter. This requires
thedetailed representation of the converter controls and
the converter circuit [1,4-7]. Simulation of a detailed
converter model is time consuming. Theretore, this
model is not recommended for a general purpose
harmonic study. It is more suitable for verifying the
design of the converter control and studying its
response to transient phenomena.
8.4 TIME SOLUTION OF THE SYSTEM MODEL
There are three stages in obtaining a time solution of
thesystemmodel.
A Steady State. Most transient simulation programs,
such as EMTP, cancalculate the system steady state for
the linear part of the network excited by the sinusoidal
sources only. Some non-linear elements in the system
model must be initialized separately. The flux of non-
linear inductors is one of the variables that must be
initialized in this manner, so that symmetric saturation
can be observed. If the nux is not initialized properly,
the ensuing transient from the inrush current may last
for several seconds. In some programs, initialization of
the non-linear inductor is obtained approximately by
involving the element's unsaturated inductance in the
steady statesolution. In this manner, the de offset of the
flux is minimized and the duration of the transient is
reduced.
The part of the system model that involves the
simulation of the -converter synchronizing loop and
switching function requires separate initialization. This
can be done approximately as follows: FIrSt a steady
state solution is obtained by representing the converter
as an independent source at fundamental frequency.
The source can be calculated from the converter's
power and power factor. SUbsequently, the steady state
solution at the converter terminals can be used to
initialize thephase of its switching function along with
the othervariables in the synchronizing loop.
59
B. System Run. Start-up of the simulation is followed
by a transient interval. There are two distinct causes for
this: The excitation of the network natural frequencies
and the interaction of the system current and VOltage
with the converter controls. The latter occurs at low
lasts for a few fundamental cycles. The transient can be
as shortas one or two fundamental cycles, if the system
model is properly initialized.
Usually, a run of ten fundamental cycles is
is verified. Verification of steady state for non-linear
inductors could be done by checking the symmetry of
the current and flux. Verification of steadystate for an
ac\dc converter represented by its switching function
could be done by checking of its average dc current.
The average de current can be observed from a low
pass filter. At steady state, this current is constant.
C. Derivation of Frequency Response. The last stage of
the solution is to obtain the frequency-somponents of
the desired voltages and currents and evaluate their
distortion. This is done for the last fundamental cycle
of the systemresponse obtained fromthe timesolution.
Most programs, including EMTP and
PSCAD\EMTDC, have appropriate utilities for this
purpose [10 yl1].
8.5 SUMMARY
Harmonic load flow and linear time domain simulation
are widely used methods for the preliminary study of
harmonics. They represent the system by rigid
harmonic injection sources. They are successful in
systems with low voltage and current distortion. In
these systems the effects of non-linearity and converter
modulation'demodulation are not pronounced and they
do not significantly alter the harmonic profile of the
currents and voltages.
Weak systems can produce significant harmonic
distortion. Then, a detail representation of system non-
linearity and power converters is required to accurately
assess the harmonic profile of the system. Non-linear
time domain simulation is more appropriate for this
purpose. This can be conducted in programs such as
EM'IP and PSCAD\EMTDC.
Non-linear time domain simulation is time
consuming and it is not, therefore, the most desirable
means .of study. Some savings in complexity can be
gained by these two modeling approaches. The use of
dynamic equivalents of the network can significantly
reducethe systemcomplexity. Therepresentation ofthe
converters by their switching function provides good
accuracy with less complexity compared to the detailed
modeling of the converter.
 8.6 REFERENCES
(I) Power System Harmonics", IEEE tutorial course. 84
EH0221-2-PWR. by the Power Engineering Society.

(2) Task force 00 Harmooics Modeling and Simulation. The

modeling and simulation of the propagation of harmonics in
electric power networks Part I: Concepts, models and
simulation techniques", IEEE Trans. on Power Delivery,
Vol. 11. No.1. Jan. 1996. pp. 452-465.

(3) Task force on Harmonics Modeling and Simulation. Theto

modeling and simulation of the propagation of harmonics in

electric power networks Part D: Sample systems and
examples", IEEE Trans. on Power Delivery, Vol. 11, No. 1,
Jan. 1996. pp. 466-474.

(4] Hatziadoniu, C., Galanos, G.D. and Milias-Argitis, J., "An

Incremental Transformer Model for Smdy of Harmonic
Overvoltages in WealcACJDC Systems," IEEE Transactions
on Power Systems. Vol. 3, No.3, July 1988.

(5] Milias-Argitis, 1. and Zacharias, Th., HatliadODiu. C.

Galanos, G.D., "Transient Simulation of Integrated ACIDC
Systems, Part I: Converter ModeliJlg and Simulation," IEEE
Transactions on Power Systems, Vol. 3, No.1, February
1988.

(6] HatziadoDiu, C., GalaDos, G.D., Miliu-Argitis, I. and

Zacharias, Th., "Transient SimulatioD of Integrated A~
Systems, Part n: System Modeling and Simulation." IEEE
Trapsactiogs on Power Systems, VoL 3, No 1, February
1988.

(7) Milias-Argitis, J. and Zadwiu, Th., HatDadODiu, C;

GalaDos, G.D., "An Algorithm far Transient Simulation of
Power Electronics Systems," IEEE Transactioas on Circuits
and Systems, VoL CAS-34, No.8.

(8) HatDadoaiu, C. and GalaDos, G.D., "lDteractiou Between

the AC Voltages and DC Cmreat in Weak ACIDC
IntereODDecUoDS," IEEE Transaetiops on Power Delivery,
Vol 3. No.4, October 1988.

(9) G.D. Breuer eL aL "HVDC-AC harlncEc iDtenctioDs Puts

1 and 2", IEEE Traas. PAS-lOt No. 3, March 1982, pp.
709-718.

(10) DommeL "Electromagnetic Traasients Program. Referea.ce

manual". Department of E1ecUical Eqiaecring UDiveaity of
British Colombia. Aug. 1986.

(11) PSCAD\EMTDC Power System Simulation Software user's

maaual. Maaitoba HVDCResearch Conler, 1"4.

(12) A1temaUve TraasielllS Program. User's Maaual. DPA, 1996.

60
 Chapter 9
Analysis of Unbalanced Harmonic Propagation
in Multiphase Power Systems
W.Xu
University of Alberta
Edmonton, Alberta, Canada
9.1 Introduction
Harmonics related concerns wereinitially centered at a few
large harmonic-producing devices until the early 1980s.
Because these devices, such as HVDC and SVC, and their
supplysystems are wellbalanced among threephases, harmonic
analysis based on positive sequence network representations
was generally sufficient The situation has changed significantly
in recent years. More and more harmonic-producing loads are
being connected to systems which are unbalanced. The effects
of single phase harmonic-producing loads are also becoming
important. The need to investigate harmonic propagation in
unbalanced systems with unbalanced harmonic excitations has
emerged.
Unbalanced harmonic analysis seeks to assess the
propagation of harmonics in each phase of a power system.
Sometimes, the harmonic currents in the neutral or ground
conductors may also need to be calculated. In all cases, full
phase representation of a network is required. As such, many
unique component modeling and network solution issues must
be addressed. Thanks to the pioneering work documented in
references [1-4], methodologies for modeling and simulating
unbalanced harmonic propagation have been established. The
purpose .of this paper is to summarize the progress achieved in
this area. The key issues of multiphase harmonic power flow
solutions are discussed.. Samplecases are presented to illustrate
the features of unbalanced harmonic analysis.
9.2 The Need for Multiphase Harmonic Analys.
In this paper; the term multiphase harmonic analysis is
used to describe harmonic simulations which are based on a full
phase representation of a system. The system and its loads can
be balanced (a special case) or unbalanced. The term three-
phase harmonic analysis is not used here because the full phase
representation of a system often requires certain network
components (such as transformers) be treated as multiphase
components. Typical cases that require multiphase harmonic
analysis are summarized as follows:
1. Utilization system harmonic analysis. Sample utilization
systems are utility secondary distribution systems,
commercial building distribution systems, and aircraft
power systems. These systems may contain many single-
phase harmonic sources. The networks are unbalanced as
well. Sample needs of harmonic analysis are the assessment
of harmonic currents in neutral conductors, the evaluation
of harmonic mitigation devices and the de-rating of supply
transformers.
61
SJ. Ranade
New Mexico State University
Las Cruces, NM, USA
2. Distribution system harmonic analysis. Most utility
distribution systems are unbalanced in both network
structures and connected loads. Even if the harmonic
sources of interest are three-phase, unbalanced harmonic
analysis is often required. The main interest of such analysis
includes the determination of harmonic resonance
conditions, the assessment of harmonic-telephone
interference, and the verification of customer power quality
levels.
3. Analysis of harmonic problems in balanced systems. For
balanced systems, the majority o!. harmonic related
problems can be investigated using 'the one-phase based
methods. However, cases do arise that need unbalanced
analysis. These typically relate to the generation of
unbalanced harmonic currents from specific loads or
operating conditions.
4. Special Cases. The nature of some harmonic caused
problems may warrant multiphase harmonic analysis. For
example, residual (zero sequence) harmonic currents which
are of primary concern for telephone interference need to be
determined using such analysis. The generation of non-
characteristic harmonics is another example.
9.3 Modeling Considerations
9.3.1 Linear Components andNetworks
Linear components are those components that do not
produce harmonic voltages or currents. The generic model of
linear components is a multiphase coupled [Z(h)] or [Y(h)]
matrix, where h is theharmonic number.
Lines and Cables: The basic model is a per-unit length
multiphase series [Z(h)] and shunt [C] matrices that include all
phase and ground conductors of the component A line model
can then be constructed either as a lumped parameter 1t circuit
(for short lines) or a distributed parameter 1t circuit (for long
lines). Inclusion of ground conductors in the model helps to
determine the neutralor grounding currents.
Transformers: The main difficulty in modeling transformers is
the variety of transformer connections and the resultant phase
shift effects. The phase shift effects must be simulated because
they are an important means of harmonic mitigation.
Experience shows that the best approach is to model
transformers as coupled windings that have no pre-determined
connection forms. The coupling is represented by a [Z(h)] or
[Y(h)] matrix with transformation ratios included [3]. A
particular transformer connection is specified in the input data
by renaming the winding terminal nodes in a way similar to the
actual .transforrner connections (Figure 9.1). Thus any harmonic excitations are known, a three-phase harmonic load
transformer configurations can be simulated. Although the model can be constructed using a coupled 3 by 3 [Z(h) matrix
model can include linear magnetizing branches, the effects of as
the branches are insignificant for most harmonic cases.

I[Z,,+2Z, Z,,-Z, Z,,-Z,]

[Z(h)]='3 Z,,-Z, Z,,+2Z, Z,,-Z,
A
a Z,,-Z, Z,,-Z, Z,,+2Z,

where Zp(h) may be determined from the CIGRE load model.

Few works are known that investigated the determination of
B N Zo(h)data. Due to lack of data, it may be assumed that Zo(h) is
b about I to 5 times of Zp(h). The ratio of 1 corresponds to cases
where the load has no mutual coupling. The ratio of 5
corresponds to cases where the load consists of a large percent
of three-phase rotating loads or loads with ungrounded star
c points. The third issue is unsolved and needs considerable
c research. Present practice is to ignore the harmonic currents if
they are small or to represent the entire load as three-phase
harmonic current sources if they are significant. The current
Y-connection .1-connection sources are typically determined from field measurements.

Figure 9.1. Transformer model example: connection of three
two-winding transformers to form a Y-D three-phase
transformer by node renaming
Rotating Machines: These include induction and synchronous
machines. They can be modeled as a three-phase, balanced
fundamental frequency voltage source behind a three-phase
harmonic impedance matrix (Figure 9.2). Again, no motor
connections such as Y or Ii are specified in the model. The
voltage source is determined from the fundamental frequency
load flow solution. The impedance matrix can be determined as
ZO-ZII
z; +2Z"
External Networks: Due to its multiphase modeling capability,
the representation of external networks. _is=easier in multiphase
harmonic analysis than in the one-phase based analysis. The
reason is that each interface between the external and study
networks can be treated as one phase of a multiphase network
equivalence (Figure 9.3). The data can be determined from
frequency scanning of the external network one phase at a time
across all interface phases. The results are a frequency
dependent multiphase [Z(h)] or [Y(h)] matrix in series with
nmltiphase fundamental frequency voltage sources. The
voltage sources are the open circuit voltages of the external
networks:
I-- external

~
external
I-- network 1
ZO-ZII r network 2
where Za=Rn+jhXa is the locked rotor (negative sequence)
impedance of the machine when h=1. Zo is highly dependent on
the machine armature winding pitch design and typical values
are not available.

A
[Y(h)]9x9 t-0-Jh
B 9-phase source
k Figure 9.3. Example of external network equivalence. Since
C there are three three-phase lines connected to the study
network, the equivalence is a 9 phase system
Figure 9.2. Model for three-phase rotating machines
3.2 Nonlinear or Harmonic-Producing Components
General Loads: General loads refer to the aggregate form of
various individual loads. Modeling of general loads must There are no major differences between the multipbase or
address three critical issues. The first issue is the load's one-phase based harmonic analyses in terms of modeling
response to harmonic excitations. The second is its response to nonlinear components. Previous chapters provide a good
unbalanced excitations. The third is that the load may contain summary of the most important aspects of this subject The
harmonic currents. The first issue has not been fully solved. But purpose of this section is to highlight the unique characteristics
some preliminary results, such as the ClORE harmonic load of harmonic source representation under unbalanced conditioDS.
model, are available [5]. Series R and L elements determined
from 60Hz are used as harmonic load models in some cases as 1) Single-phase harmonic sources connected in different
well. Reference [4] analyzed the second issue. The main idea is phases of a system can only interact with each other through
that if the load's responses to the positive and zero sequence the system. Therefore, each source can be modeled
62
 separately. Models of such sources can take the forms of
either harmonic current sources Of detailed iteratively-
improved device models.
2) A three-phase harmonic source, particularly the power
electronics types, can have interactions among three phases
thatcan influence the output of harmonic currents from each
phase. Non-characteristic harmonics which wouldn't exist if
the supply is balanced can be produced. A salient pole
synchronous machine can become a harmonic source when
unbalanced voltages are applied at the terminals as well. If
the effects of non-characteristic harmonics need to be
assessed, thesources must be modeled in detail. The models
based on typical current spectra are nolonger valid.
3) For those sources that have a nonlinear voltage-current
relationship such as magnetizing branches of transformers,
it is also a good practice to model them with a detailed
model because "typical'' harmonic spectra for such devices
do not exist. For multiphase harmonic analysis, the
placement of nonlinear magnetizing branches in an
equivalent circuit depends on thetransformer designs [3].
9.4 Simulation Methods
The unbalanced harmonic propagation in a power system
can be simulated using a multiphase admittance matrix equation
as follows:
where each row of the [Yh] matrix represents one node of the
system. The node can be any phase of a three-phase bus. It can
also be a neutral connection point such as the star point of a Y-
connected transformer. Complicated transformer connections
are included in this matrix through the node-renaming
mechanism [4]. The right hand side is the harmonic current
sources representing theharmonic producing devices.
Floating sub-networks such as delta connected subsystems
and ungrounded motors could be encountered in multiphase
analysis. Because there is no reference voltage for such
networks, part of the [YIII matrix is singular. These structures
can be accommodated by adding appropriate impedances to
ground or by modifying factorization algorithms.
Once this matrix is established, the various network
harmonic solution methods developed for the' one-pbase based
representations can be extended to the multiphase frame. As
described in previous chapters, four types of harmonic analysis
are normally performed:
I. Frequency Scan Analysis: In these studies, the network
frequency response seen at any phase of a bus can be
determined. The positive, negative and zero sequence
frequency responses seen at a bus can also be determined.
In this case, three-phase harmonic currents, in positive,
negative or zero sequences respectively, are injected into
the study bus. Multiphase frequency scans are useful, for
example, to determine harmonic resonance caused by
single-phase capacitor banks. In general, frequency scans
are difficult to use in the multipbase case because of the
largenumber of nodes that mustbeconsidered.
2. Harmonic Analysis Using Simple Current Source Models:
In these studies, the harmonic-producing devices are
63
modeled as simple individual-phase current sources. The
current source magnitudes and angles are determined, for
example, from measured harmonic spectra. It must be
emphasized that the phase angles of a three-phase harmonic
current sources are seldom 1200 apart among three phases.
Even with a slight unbalance at the fundamental frequency,
the phase angle unbalance at harmonic frequencies can be
significant. Therefore, the harmonic spectra should be
determined for each phase.
3. Harmonic Analysis Considering Fundamental Frequency
Power Flow Results: The main problem of current source
based analysis is the lack of fundamental frequency load
flow information. As a result, the magnitudes and phase
angles of the current sources cannot be determined
adequately. In this improved analysis, a multiphase power
flow is first solved. The harmonic currents injected are
determined using the power flow voltages and "typical"
source spectra.
4. Harmonic Power Row Solutions: In this analysis, the
harmonic sources are also represented as current sources.
However, their magnitudes andphases are updated using an
iterative scheme based on detailed (voltage-dependent)
harmonic source models. Inter-phase coupling of the
harmonic-sources can be modeled witH' good accuracy. The
harmonic iteration scheme solves thenetwork one frequency
at a time. The calculated nodal voltages are then used to
update the currentsource model [4]. In theory, simultaneous
solutions of all harmonic orders like those used in the
HARMFLO program [6] can also be developed for the
multiphase analysis, but the algorithm would be extremely
complex.
Due to the availability and widespread use in other
analyses, time-domain simulation tools such as the Alternative
Transients Program(ATP) are also used for studying harmonic
propagation in unbalanced systems. The key problem in such
usage is to identify when steady state conditions have been
achieved.
95 Case Study I
This study illustrates how uncharacteristic harmonics can
arise and require a multiphase analysis in an apparently
balanced system. A ::t150MVar static var compensator is to be
installed at a substation of B.C. Hydro. The SVC is connected
to a 138 kV bus and consists of one thyristor-Switched
capacitor (TSC) and three thyristor-switched reactors (TSR). It
is commonly believed that the TSR-type SVC is harmonic free.
A TSR is essentially a reactor in series with an anti-parallel
thyristor pair. These thyristors are randomly selected. Each
thyristor has a deviation of voltage drop about %.0.07 V when it
is conducting. As a result, one stack of series-connected
thyristors will have a slightly different total forward voltage
drop than the anti-parallel one. This voltage difference is
effectively a direct voltage across the TSR branch. Since the
resistance of the TSR branches and the SVC transformer
windings is generally very small, a small direct voltage can
result in a relatively large direct current This current will
circulate through the secondary winding of the SVC
transformer and can cause a DC-offset saturation of the
transformer (Figure 9.4). As typical power transformers need
little magnetizing current, a small amount of direct current is
sufficient to cause significant saturation and harmonic
generation.
 Statistical analysis showed that the direct current injected g~~""""~""""""""""""
into the SVC transformer has a normal distribution. At the

):t V - V
confidence level of 95%, the expected direct current can be as
high as 19.2 ampere. This is a very large DC current for regular
power transformers. It is therefore considered necessary to
I
analyze the resultant harmonics. The objective is to find the
highest direct current level that can be tolerated from a 0.000 0.005 0.010 0.015 0.020 0.025 0.030 0.035
harmonic distortion point of view [7]. 11m. (S.c.)

Figure 9.5. Harmonics from DC offset saturation of

transformers
sve s
Transformer ....
Q.)
f'-)
9.5.3 Sample Results
~
f'-)
A number of possible SVC operation modes was analyzed.
>
~ Table 9.1 provides harmonic current distortions (worst phase)
00
,..
~ under the condition that the SVC is operating in the fully
inductive mode and with a 1.15pu 138 kV bus voltage. This
mode has the largest DC current generation and the TSCs are
phase-a phase-b phase-c not sinking harmonic currents from the transformer. Current
distortions are listed in both ampere values and percentage
values with respect to the nominal SVC current of 560 A. The
....
~
I ~

~
. ('f")
I
spectra of the currents injected into- the 138 kV system are
shown in Figure 9.6. This figure indicates that only the lower
~ ~ "
Cf.)
Eo-
orderharmonics are of concern.
Table 9.1. Harmonic current distortion at the SVC bus
Max. 12 13 ITIID 12 13 ITH
Figure 9.4. SVC system and the flow of TSRdirect currents DC (A) (A) (A) (A) (%) (%) D
(%)
9.5.1 Modeling ofHanno~ic Sources 14 1.78 1.49 2.76 0.32 0.27 0.49
18 236 1.96 3.65 0.42 035 0.65
The direct current into the SVC transformer is unbalanced. 20 2.66 2.21 4.14 0.48 039 0.74
The worst case is that one phase serves as the return path for 25 3.43 2.86 5.38 0.61 0.51 0.96
other two phases (Figure 9.4). This leads to the generation of 30 4.17 3.45 6.46 0.74 0.62 1.15
unbalanced harmonics which contain positive, negative and 40 5.66 4.63 8.57 1.01 0.83 1.53
zero sequence components. Moreover" because the transformer 50 7.14 5.76 10.7 1.28 1.03 1.92
is saturated with a direct current offse~ both even and odd
harmonics are generated. The grounded primary allows the
3.0
penetration of zero sequence harmonics into thesupply system.
They could interfere with telephone circuits. 2.5 .Phase-A
.Phase-B
c
In this investigation, the TSR branches are represented as
impedances in series with DC voltage sources. The magnitudes =
i
t:
2.0 aPhase-C 1
and polarities of the DC sources were determined from the ~
u 1.5
o
statistical analysis. A harmonic equivalent circuit is used to C
mode} the saturated transformer magnetizing branch. The ~ 1.0
harmonic iteration scheme is used to determine the harmonic :

......- .-
currents generated [8]. Iterations are needed because the strong 0.5 lit
dependency of generated harmonic currents on supply voltage
harmonics. Sample waveforms areshown in Figure 9.5. 0.0
w
I.e
~ ~
I n..... ___
~ G ~ ~ 0 - w ~ ~ G ~ 0 0
_ _ -No. -
Harmonic - - - - - - w

9.5.2 Modeling of Supply System

The nature of the problem requires multiphase modeling of the
supply system. A network model of the system near the SVC
bus was developed (about 300 nodes). It facilitates the
simulation of various network operating conditions. In the
second approach, the supply system is modeled as a 60 Hz
sinusoidal voltage source at the 138kV SVC bus. It effectively
assumes thatthe supply system harmonic impedance is zero.
Figure 9.7. Harmonic currents injected into thesupply system
9.' Case Study n
Harmonics mitigation in a utilization system is investigated
in this case. The system, shown in Figure 9.7, represents a
simplified commercial building distribution system operated at
a 120V level. The singlephase loads contain harmonic sources
such as switcbed mode power supplied devices (PC, printer
etc.) and fluorescent lights. In order to reduce the neutral
current and associated neutral VOltage rise, a zero sequence

64
current trapper would be connected to the receptacle panel. The
current trapper. like a transformer . is made up of six coupled.
equal turn ratio windings. The windings can be configured into
either a zigzag form (Figure 9.8) or a delta-Y form. In the delta-
Y configuration. the star point of the Y windings is connected
to the neutral conductor. Both configurations are capable of
trapping zero sequence fundamental frequency and harmonic
currents. The objective of this study is to determine which
configuration is more effective and what are the winding
loading conditions.
Iph load
Iphload
excitations. The actual trapper configurations are represented
using node renaming in the input data.
Figure 9.9 shows the waveforms of the neutral voltage at
the receptacle panel. The waveform obtained without harmonic
trapper is also displayed in the figure for comparison purposes .
It can be seen from the figure that the neutral voltage can be
quite high (about 20V RMS) if no mitigation measures are
taken. The harmonic trappers can reduce the voltage to as little
as 3V RMS. The neutral voltage is dominated by the 3rd
harmonic component. The results also shown that 6.-Y and
zigzag configurations have almost the same effect in terms of
reducing the neutral voltage (the waveforms are
indistinguishable in Figure 9.9).

Iph load

120V
480V supply

Figure 9.7. Simplified utilization system for case II

A B C N
20

-ee _ ._ _.
\
~ ,-l

.~
,
:

_ .
1l1M1n_
I _~i
_. y..()eftal

_.~ :_~ ..J

._
:
.

Figure 9.9. Neutral voltage at the receptacle panel

Table 9.2. Harmonic trapper winding currents

and conductor neutral currents CA)
w-i
W-2 W-3 W-4 W-5 W-6
- . II 80.5 93.0 75.8 9.8 9.8 9.8
Figure 9.8. A six winding trapper and zigzag configuration Y-D I boflllS 53.8 53.8 53.6 50.0 50.0 50.0
Isms 96.9 107.4 92.8 50.9 50.9 50.9
A multiphase harmonic analysis is used for this case. The 26.8 37.9 21.6 37.9 21.6 26.7
II
loads are modeled as constant power loads at the fundamental
Zigzag I..ofIIIS 53.6 53.6 53.6 53.6 53.6 53.6
frequencies and current sources at harmonic frequencies. Phase
angles and magnitudes of the current sources are determined I DDS 59.9 65.6 57.8 65.6 57.8 59.9
using the load flow results and harmonic source spectra. The No harmonic source 38.2
single-phase loads are assumed to contain two types of Neutral With a-source but no trapper 187.0
harmonic producing loads. switched mode power supply type current with Y -/1 trapper 21.8
loads. dominated by Pes. and composite type loa~ dominated
(RMS) with Zigzag trapper 23.5
by fluorescent lights. The three-phase loads consist of motors
and adjustable speed drives. The supply system is modeled as
balanced voltage sources behind a system fault impedance. Table 9.2 lists the loading conditions. in the form of
Models for the harmonic trapper which has a core-type design winding currents, associated with both configurations. It can be
should be developed with care. One of the main concerns is that seen from the table that the zigzag configuration results in
the model should be able to correctly simulate the circulation of smaller RMS winding currents. The Y-/1 configuration is
the zero sequence flux. The core-type design forces a large therefore less desirable in terms of losses and device
amount of zero sequence flux to circulate outside the magnetic overheating. The neutral conductor currents in RMS values
core . To simulate these effects. a six-phase coupled [Z] matrix were also calculated for various configurations and are shown
is used with each phase representing one winding. Data of the in the same table. The entry of "No harmonic source" is the
[Z] matrix are calculated from the short and open circuit current obtained with the assumption that loads do not contain
impedances determined with both positive and zero sequence harmonic sources. The results suggest that the increase of
neutral current is mainly due to the harmonic currents from the
loads. Due to space limitations. other important issues such as
65
transformer de-rating and harmonic injections into the supply
system are not discussed. They can be easily investigated using
the same system model and analysis tool.
9.7 Case Study III
This hypothetical study illustrates harmonic propagation in
a utility distribution system. The system, shown in Figure 9.10,
is discussed in detail in the chapter of case studies. Voltage
unbalance in this system ranges from 2-4 %. Harmonic sources
are single-phase sources typical of residential and small
commercial load areas.
46
Figure 9.10. A unbalanced utility distribution system
A case of adding a three-phase 110 HP adjustable speed
drive is studied. The front end of this drive was modeled in
ATP as a 6.-Y transformer feeding a rectifier bridge with a
capacitive tilter and a resistive load. The ATP time simulation
is used to determine steady-state harmoniccurrent injections of
the drive. Figure 9.11 shows the spectra of the current drawn by
the drive and the spectrum under ideal balanced conditions.
Although the changes in characteristic harmonic
magnitudes are small, one can notice that a significant third
harmonic appears in the unbalanced case. As discussed in
Chapter 1, harmonics are not associated with a specific
sequence in unbalanced systems. Thus, for example, third
harmonic currents will flow in ungrounded capacitors and
filters. Thus detailed multiphase modeling is warranted for
utility distribution systems
9.8 Summary
In this paper, important aspects of unbalanced harmonic
analysis formultiphase power systems are reviewed. The need
for multipbase analysis is mainly due to three considetations:
the unbalanced system, the unbalanced sources and the
propagation nature of harmonics. Although considerable
progress has been made in the area of multiphase harmonic
analysis, there are still problems to be solved and improvements
to be made. For example, some of the future work in this areas
includes:
I) Quantification of the effects of various load models on the
propagation of harmonics in a power system with
subsequentrefinement of multiphase harmonic load models.
66
2) Development of analysis methods that can assess the
collective effects of a large number of randomly operating
harmonic sources in a utilization system.
3) Improvement of models and associated solution algorithms
for harmonic-producing devices. The models should be
practical but provide more accurate results.
1i 30.-------------;=====
E
E
as
25
-g 20
~
'0 15
it
i.;2 10
~ 5
as
:E 0 ~......~IIfI.I_II..,...a.&.;_~......~L.f_f~fUII_~-+-I'l~~L.+-.t
...
3 5 7 9 11 13 15 17 19 21 23 25 27 29
Hanmnic#
Figure 9.11. ASD current spectra (fundamental omitted)
9.9 References
[I] I. Densem, P.S. Bodger, and J. Arrillaga, "Three-phase
Transmission System Modeling for Harmonic Penetration
Studies", IEEE Trans. on PAS, vol. PAS-I03, no.2, Feb.
1984,pp. 256-263.
[2] M. Miller, W. Price, M. Lebow, and A. Mahmoud, "A
Computer Program for Multiphase Harmonic Modeling,"
Proc. ICHPS-I, Worcester, 1984.
[3] H.W. Dommel, "Electromagnetic Transients Program
Reference Manual (EMTP Theory Book)", Prepared for
Bonneville Power Administration, Dept. of Electrical
Engineering,Universityof British Columbia, Aug. 1986.
[4] w. X~ J.R. Jose and H.W. Dommel, .."A Multiphase
Harmonic Load Flow Solution Technique", IEEE Trans. on
Power Systems, vol. PS-6, Feb. 1991, pp. 174-182.
[5] ClORE Working Group 36-05, "Harmonics, Characteristic
Parameters,Methods of Study, Estimates of Existing Values
in the Network",Electra, no. 77, July 1981, pp. 35-54.
[6] D. Xia and G.T. Heydt, "Harmonic Power Flow Studies,
Part I - Formulation and Solution, Part n - Implementation
and Practical Application", IEEE Trans. on PAS, vol. PAS-
101,June 1982,pp. 1257-1270.
[7] W. Xu, T.G. Martinicb, J.H. Sawada and Y. Mansour,
"Harmonics from SVC Transformer Saturation with Direct
Current Offset", IEEE Trans. on Power Delivery, Vol. 8,
No.2, July 1994,pp. 1502-1509.
[8] H.W. Dommel, A. Van and S. Wei, "Harmonics from
Transformer Saturanon", IEEE Trans. on Power Systems,
vol. PWRD-l, April 1986,pp. 209-214.
 Chapter I()

Harmonic Limit Compliance Evaluations Using IEEE 519-1992

s. Mark Halpin' and Reuben F. Burch, Iy2

'Mississippi State University, Starkville, MS
2A1abama Power Company, Birmingham, AL

Abstract: The end-use applicationof harmonic simulation Note also that there are a number of appropriate footnotes
studies usually involves an assessment of limit that govern the application of the limit values given. The
compliance. This chapter of the tutorial is devoted to reader is strongly encouraged to consult the Standard for
various limit evaluations using the procedures in IEEE this additional information.
519-1992 and the proposed IEEE 519A, the "application
guide" for 519. The procedures are demonstrated based Table 10.1. CurrentDistortion Limits (in % of I L) for
on a sample system with typical nonlinear load General Distribution Systems (120-69,000 V) [1]
characteristics. It is assumed that the reader is familiar IsdIL <11 11~<17 17Sb<23 ~3~<35 35~ TOD
with the harmonic study method details presented in the <20 4.0 2.0 1.5 0.6 0.3 5.0
previous sections of this tutorial. 20-50 7.0 3.5 2.5 1.0 0.5 8.0
50-100 10.0 4.5 4.0 1.5 0.7 12.0
100-1000 12.0 5.5 5.0 2.0 1.0 15.0
10.1 INTRODUCTION >1000 15.0 7.0 6.0 2.5 1.4 20.0

There are two distinct thought processes that can be

Table 10.2. CurrentDistortion Limits (in % of I L) for
applied to limit the amount of harmonics that are present
General Subtransmission Systems (69,001-161,000 V) [1]
in power systems. The frrst, favored by the International
Electrotechnical Commission (lEe), is a series of limits IsdIL <11 11~<17 17~<23 23~<35 35~ TOD
that is appropriate for application at the terminals of any <20 2.0 1.0 0.75 0.3 0.15 2.5
particular nonlinear load. The second, favored by the 20-50 3.5 1.75 1.25 0.5 0.25 4.0
IEEE and the basis for IEEE 519-1992 [1], is a series of 50-100 5.0 2.25 2.0 0.75 0.35 6.0
limits that is appropriate for application at a single more 100-1000 6.0 2.75 2.5 1.0 0.5 7.5
central point of supplyto multiple nonlinear loads. >1000 7.5 3.5 3.0 1.25 0.7 10.0
The philosophy of the IEC limits is based on the
presumption that limiting harmonic production fromevery Table 10.3. CurrentDistortion Limits (in % of I L ) for
piece of equipment will effectively limit any combined GeneralTransmission Systems (>161,000 V) [I]
effects. While conceptually effective, the assumptions IsdIL <11 11:sJ1<17 17Sh<23 23~<35 35~ TOD
made in developing the actual limits are quite different <50 2.0 1.0 0.75 0.3 0.15 2.5
from those in IEEE 519-1992 and it has been shown that ~O 3.0 1.5 1.15 0.45 0.22 3.75
the IEEE limits are somewhat more restrictive due the use
ofboth voltage and current harmonic limits.
The IEEE limits for voltage and current harmonics Table 10.4. Voltage Distortion Limits (in % of V 1) [1]
shown in Tables 10.1-10.4 are dependent on several PCC Voltage Individual Harmonic THDv (%)
variables and concepts defined as follows: Magnitude (%)
S69kV 3.0 5.0
PCC: Point of common coupling. This pointis defined as
69-161 kV 1.5 2.5
the point in the utility service to a particular
~161 kV 1.0 1.5
customer where another customer could be
connected.
The thought processes behind these tables are that 1)
I sc: Available short circuit current. the customer should be responsible for limiting harmonic
IL: IS or 30 minute(average) maximum demand current. currents in accordance with Tables 10.1-10.3 and 2) the
utility should be responsible for limiting harmonic
TOD: Total demanddistortion. TOO is identical to THD voltages in accordance with Table 10.4. The numerical
except IL (as defined previously) is used instead of values in the voltage and current tables are not
the fundamental current component. independent however. The values in Table 10.4 can be

67
derived from those in Tables 10.1-10.3 by assuming that
the Isell t ratio speci fies an equivalent inductive reactance.
Before beginning the case study examples
demonstrating the use of these tables. it must be stressed
that .~e ~alues are intended as guides and may require
rnodlncauon to fit the particulars of a given situation. In
practice, there are a number of "gray" areas where
discretionis required and concessions from both the utility
and the customer may be necessary.
10.2 GENERAL PROCEDURE FOR APPLYING
HARMONIC LIMITS
The following series of steps outlines a general
procedurefor applyingharmonic limits:
Step I: Choose the pointof common coupling,
Step 2: Characterize the harmonic-producing loads,
Step 3: Assess power factor correction needs,
Step 4: Calculateexpectedharmonics at the PCC,
Step 5: Designand implementsolutions (if needed), and
Step 6: Verifyperformance with measurements.
As the case study examples will show, it is not necessarily
required that a complete and detail harmonic study be
conducted. The concept of "automatic acceptance" is
sufficient for the majority of smaller utility customers and
can often serve to eliminate the need for a comprehensive
simulation. The "automatic acceptance" procedure
requires completion of only the first three steps while a
comple~ simulation often covers the first five steps.
While not a simulation issue, result verification via
measured data is always recommended especially if
equipment additions were required to correct a limit
violation.
10.3 AUTOMATICACCEPrANCE
As previously described, there are many cases
involving smaller customers where a complete harmonic
study is not required. It is possible to evaluate limit
compliance issues at ~e Pee without detailed studies
provided that the short-circuit kVA is much greater than a
weighted sum total of the nonlinear load tVA sUpplied
from the PCC. The following series of four steps outlines
the procedure for "automatic" verification of limit
compliance:
Step 1: Determine Sse (short-circuit kVA) at the PeC.
Step 2: Determine the size and type of nonlinear load
served.
68
Step 3: Evaluate Sow =LSD, .. Wi' where SOi is
i
(kVA) demandof the i dl nonlinear load.
Step4: Acceptautomatically (limits will be met) if
Sow
-S- *100% < 0.1%.
sc
The weighting factors Wi are given in Table 10.5 for
various common nonlinear loads. In general, the weight is
larger for nonlinear loads that produce a more distorted ac
current.
Table 10.5. Weighting Factors for Automatic Acceptance
Procedures
TyPe of Load Wi
Single Phase SMPS 2.5
Semiconverter 2.5
. 6 pulse converter, no choke 2.0
6 pulse converter, >3% choke _ 1.0
6 pulse converter, large series L 0.8
12pulse converter 0.5
10.4 COMPREHENSIVE LIMIT COMPLIANCE
EVALUATIONS
Theprevious chapters of this tutorial haveconsidered
the modeling and analysis requirements for a harmonic
study in detail. In general. the analyst is free to choose
their own analysis procedure, but the following
capabilities areessential for limit compliance evaluations:
I. Modeling of network components such as overhead
lines, cables, transformers, etc. (The analyst has some
flexibility in selecting the level of detail needed;
sensitivity studies should be considered to determine
which degrees of detail can be avoided in any
particular study.)
2. Modeling of various nonlinear loads including pre-
defined models and the capability for user-defined
models based on measured or typical data. (It is left to
the analyst to determine what degree of sophistication
is required, especially if time-domain simulation
techniques are to be used.)
3. Modeling of power factor correction capacitors should
be "built in" to any software being considered. (The
representation of the equivalent capacitance is the most
imponant requirement; other details are somewhat
insignificant for limit compliance evaluations.)
4. Modeling of utility system equivalents should be
straightforward. (In an advanced study, the analyst
should consider the -capability to represent the
frequency dependence of network equivalents.)
Fortunately, there are a great number of computer The nonlinear loads are known to be 50 kVA of small
packages available that easily meet (and exceed) these PWM motor drives (no choke) and 100 kVA of small
requirements. A key point is that the level of accuracy PWM drives (with 3% inductive choke). The harmonic
available in the program should match 1) the level of current spectra for these two loads are shown in Table
accuracy available in the data and 2) the level of accuracy 10.6. Note that it is important to use the provided
required in the results. It is up to the analyst to make this harmonic current phase angles to account for possible
determination for each compliance issue to be harmoniccurrent cancellation at the PCC.
investigated.
The end'result of the detailed harmonic analysis must Table 10.6. Harmonic Current Spectra for Nonlinear
be the voltages and currents at the PCC. (Note that Loads
similar values at other system locations may be of interest 50 kVA Load l00kVALoad
in a given study but they are not relevant to limit (no choke) (3% choke)
compliance evaluation.) Given these values, the system h Ih (%) 8h (0) Ih (%) 8h (0)
strength (as indicated by the available short-circuit at the 1 100.00 -7.40 100.00 -15.50
PCe), and the average maximum demand current at the 5 72.46 -219.03 34.81 -268.72
pce, a simple comparison to the limit values given in 7 51.03 -56.20 10.76 -176.84
Tables 10.1-10.4is all that is required. 11 13.75 -285.10 6.30 -149.36
13 5.02 -172.22 3.24 -118.55
17 5.27 -113.89 2.48 -53.90
10.5 CASE STUDIES 19 3.51 -334.00 1.89 -36.00
23 2.21 -288.60 1.16 -332.30
The following case studies demonstrate limit 25 2.17 -153.11 1.12 -309.20
compliance based on both the automatic acceptance
criteria and the detailed harmonic analysis procedures
Automatic Acceptance Evaluation
described previously. Figure 10.1 shows the single-line
The data given in Figure 10.1 and Tables 10.5 and
diagram of the example system. As is typical, only
10.6 can be used to make quick limit compliance
positive sequence data is used in these studies due to the
evaluations. Based on the descriptions of the load, the
assumed balanced nature of the system. The short circuit
weighting factors in Table 10.6, and the short-circuit
MV A, Sse, and the average maximum demand, SAVG,.MAX,
MVA at the PCC, the following calculation demonstrates
at the PCC are total (three-phase) values.
that automatic acceptance is notrecommended.
46kV
SubtransmissioD SDW = 50*2+ 100*1 *100% =0.71%
Sse 28,000

This result is not a true indicator that harmonic limits

46kV: 12kV will be violated due to the fact that the resultant 0.71 % is
only slightly greater than the 0.1 % required for automatic
acceptance. However, the fact that the 0.71 % is not low
enough for automatic acceptance indicates that a detailed
harmonic study should be conducted to evaluate the true
12kVfeeder nature of the harmonic voltages and currents at the PCC
before a limit violation is declared. Unfortunately, there
is no way to correlate (for the general case) the amount by
Pee which the 0.71 % result exceeds the 0.1 % criteria with
whether or not a detailed study will indicate limit
Ssc=28 MYA @ Pee, violations. Values slightly greater than 0.1 % could lead to
SAVG.MAX=2100 tVA violations if strong system resonances are present at
12kV: 480 V frequencies near those generated by the load(s) in
question.
480 V plant bus ...- - - - - -....~
Detailed Harmonic Study
The data given in Figure 10.1 and Table 10.6 can be
used to conduct a detailed harmonic study using anyone
of a number of available computer programs. Based on an
equivalent impedance of 2.286+j5.151 Qlphase seen
Figure 10.1. Example System for Limit Compliance Case
Studies
looking from the PCC to the subtransmission system,

69
Table 10.7 gives the harmonic current magnitudes
calculated at the pee in both amperes and in percent of
the average maximum demand current (=97.22 A).
The final step in the process is the comparison of the
first row in Table 10.1 (based on the ratio IscIIL=13.3) and
the third column of Table 10.7. This comparison is given
in Table 10.8 and shows that limit compliance is not a
problem for this example case study (the more accurate
calculation approach should overrule the more
approximate "automanc acceptance" criteria).

Table 10.7. PCCCurrents for Limit Compliance

Evaluation
Currents @ PCC
h Ih (A) Ih (% of IL)
1 95.44 98.16
5 3.52 3.62.
7 2.24 2.30
11 0.47 0.49
13 0.43 0.44
17 0.32 0.33
19 0.20 0.21
23 0.12 0.12
25 0.02 0.02

Table 10.8. Evaluation of SimulatedResults with IEEE

519-1992 Limits
Currents @ PCC
h Limit (% of IL ) Ih (% of IL )
5 4n 3~2
7 4~ 230
11 2.0 0.49
13 2.0 0.44
17 1.5 0.33
19 1.5 0.21
23 0.6 0.12
25 0.6 0.02

roD 5.0% 4.44%

The reader should be cautioned that using different

software packages may produce results different from
those shown in Tables 10.7 and 10.8. It is, therefore,
important for the 1alyst to bave a complete
understanding of both the harmonic theory and the
particular techniques used and assumptions made by the
software being used.

10.6 REFERENCES

1. IEEE Standard 519-1992, Recommended Practices

and Requirements for Harmonic Control in Electrical
Power Systems, The Institute of Electrical and Electronics
Engineers, 1993.

70
 Chapter 11

Test Systems for Harmonics Modeling and Simulation

TaskForce on Harmonics Modeling and Simulation*
Transmission & Distribution Committee
IEEE PowerEngineering Society

Abstract This paper presents three harmonic simulation test Because the system has balanced bus loads and the
systems. The purpose is to demonstrate guidelines for the transmission lines are transposed, a balanced harmonic analysis
preparation and analysis of harmonic problems through case is generally sufficient for determining harmonic distortion levels
studies and simulation examples. The systems can also be used in this case. Main harmonic analysis issues to be demonstrated
as benchmark systems for the development of new harmonic by this test system are:
simulation methods and for the evaluation of existing harmonic
analysis software. 1. The need to solve fundamental frequency load flows for
harmonic analysis. The load flow results affect the magnitudes
and phase angles of the harmonic current injected from
11.1 Introduction harmonic sources. Correct representation of the phase angles
are important for systems with multiple harmonic sources [1].
Harmonic studies have become an important aspect of The harmonic filters can have a large impact on the load flow
power system analysis and design in recent years. Harmonic results.
simulations are used to quantify the distortion in voltage and 2. 1be harmonic cancellation effects due to YY and Y-Delta
current waveforms in a power system and to determine the transformer connections (at the HWC terminal) and the
existence and mitigation of resonant conditions. Many digital impactof otherharmonic sources (the SVC). For this purpose,
computer programs are available for harmonic analysis. New the HVDC terminal is modeled as two six-pulse harmonic
analysis techniques are being developed. With a wide variety of sources.
solution methods and" modeling assumptions implemented in 3. The effects of using different line models such as the
many different programs, there is a need for benchmark test distributed-parameter model and the lumped pi-circuit model
systems so that the features and results of the programs can be in harmonic resonance assessment.
evaluated and compared.
......~ .. 13
This paper presents the complete data for three harmonic
simulation test systems. The purpose is to demonstrate guidelines
for the preparation and analysis of harmonic problems through
case studies and simulation examples. Several aspects that can G 1
impact the accuracy of results such as modeling of components
and solution methods are illustrated. The benchmark infoimation
provided in the paper is also useful for the development of new
harmonic simulation methods and for the evaluation of existing
harmonic analysis software. The test systems represent the most
common harmonic study scenarios encountered in industry.
Sample results are provided' in the paper. More information on
the test systems and results can be obtained at
http://www.ee.ualberta.caJpwrsyslharmonics.html.
11.2 Test System No.1: A 14-B05 Balanced Transmission
System
Figure 11.1. Test System 1 - 14BusTransmission System
This test system contains two harmonic sources. One is a
twelve-pulse HVDC terminai at bus 3 andthe otheris a SVC at
bus 8 (Figure 11.1 andFigure11.2).
Task force members and contributors are: R. Abu-hashim, R.
Burch, G. Chang, M. Grady, E. Gunther, M. Halpin, C.
Hatziadoniu, Y. Lin, M. Man, T. Ortmeyer, V. Rajagopal~ s.
Ranade, P. Ribeiro (vicechair), T. Sims, W.Xu (chair, editor).

--
HYDe svc
Figure 11.2. HarmonicSources in TestSystem 1

71
 Complete data for this system are shown in Tables 11.1 to 6. The HVDe terminal is modeled as two six-pulse bridge
I 1.4. Key modeling and simulation features for this case are: rectifiers according to the model of reference [3]. Because
voltage distortion at the HYDe terminal is small, sensitivity
1. All transmission lines are modeled using a distributed- studies showed that the terminal can be modeled as two
parameter line model. Long line effects are included in the harmonic current sources. The source spectra is provided in
model. Figure 11.3 shows the effects of using different line Table 11.4. It must be noted that the magnitudes and phase
models. The curves are the frequency scan results seen at the angles should be scaled and shifted according to the load flow
HVDe bus (bus 3). The results suggest that the long-line results [1]. The HVDC terminal is modeled as a constant
effects should be included for long distance transmission lines. power load in the load flow solution,
Table 11.2: Branch Data for S stem 1 Based on lOOMVA
Branch
Type
~l Left
i Bus'
I Right
Bu, , :
I! R1
lpu) !
~ X1
(pu)
81
lpu)
---Oistributed Parameter Lin. Mod., Xfmr(V-Y) i 4 7 0.OO0 0.20900
Lump.d PI Line Model Xfmr(V-Y) i 4 9 0.OO0 0.55618
Xfmr(V-Y) I 5 6 0.OO0 0.25020
Line ~ 6 11 0.09495 0.19887
Line ~ 6 12 0.12285 0.25575
. ~
. Line ! 6 13 0.06613 0.13024
~
i Xfrnr (V-Delta)
~ Xfmr (V-Y)
I
I 7
7 9
8 0.0oo
0,OO0
0.17615
0.11000
.
ooo.....----.;;;;:.:-----=~-
'I
......
a.
--....-~
,.
. . . .- ----.. . ~
i
Line
Une
I
~
9
9
10
14
0.03181
0.0127'()
0.08448
0.27033
Harmonic Ord.r
! I
Figure 11.3. Effectsof using differentline models
e ..
:
~
Line
Line I 10
12
11
13
0.08203
0.22087
0.19202
0.19985

,
Bus
T abill
j Nominal
i voltage
I
1 8 us D ataand Resul ts tior S.ystem I

Load
P Q
Load
LF
Vonag.
LF
Angl.
THD i
Line
! Capacitor@9 I
Line ~
13
9
1
14
0
2
0.17081
0.0oo
0.01937
0.34795
0.0oo
0.05916
0.0633~
O.0527~
~ (kV) I (KW) (KVaf) (pu) (deg)
(%) ~ Line ~ 1 5 0.05402 0.22300 0.0492~
~ Line 2 3 0,04697 0.19794 0.0438~
0 0 1.0600 0.00
1 230 1.767 ! Line 2 4 0.05810 0.17628 0.0374~
2 230 0 0 1.0450 -5.68 2.177 ~ Une 2 5 0.05693 0.17384 0.03386i
3 230 0 0 1.0427 -15.30 1.516 i Line 3 4 0.06700 0.17099 O.0346oi
301 35.4 59,505 3,363 1.0417 . -16.18 9.169 i Line 4 5 0.01335 0,04209 0.0128~
302 35.4 59.505 3.363 1.0417 -16.18 9.169 I Rltsr08:2nd 8 0 0.52510 8.31233 O.0301~
4 230 47,790 -3,900 1.0282 -11.41 0.755 I Filter@8:5th 8 0 0.52510 1.32635 0.0301Sl
5 230 7,599 1,599 1.0337 -9.82 1.462 I Filter08:7th 8 0 0.52510 0.67307 O.0301~
6 230 0 0 1.0700 -15.87 0.468 I FiterOS:11th 8 0 0.52510 0.27515 O.0301~
7 230 0 0 1.0193 -14.47 I Filter03:11th
! Filter@3:11th I
0.423 3 0 0.00136 0.02772 0.24916i
8 13.8 2 12,900 1.0209 -14.49 0.522 3 0 0.00136 0.b2772 0.2491~
9
10
11
115
115
115
29,499
9,000
3,501
16,599
5,799
',800
1.0147
1.0168
1.0394
-16.09
-16.33
-16.21
0.482
0.421
0.394
I Xfmr(Y-Y)
tJ~fm.t!Y:.Q.!tml
t 1
I
_J..___
3 301 0.0oo
~~...Q:ggm_
0.02800 O.~
._9,QQ2Q9j

12 115 6,099 1,599 1.0528 -16.72 0.391 Table 11.3:Generator Data for S stem 1
13 115 13,500 5,799 1.0458 -16.73 0.376
14 115 14901 5001 1,0154 -17.39 0.343
Voltage P Q Sub-
: ; : setUng gen. gen. transient
2. The generators are modeled aseither slackor PVbuses for the
fundamental frequency load flow solutions and as sub-
transient reactance for the harmonic analys~. The sub-
transient reactances are 0.25 per-unit.
3. Transformers are modeled using short-circuit impedances.
The winding connections ire represented in the model so that
thepbase-shifting effects onharmonic currents are included. If 7. The SVC consists of harmonic filters and a delta-connected
harmonics from transformer saturation are of interest, the TCR. The TCR was modeled using the model of reference [1].
magnetizing branches with saturation characteristics should be The firing angle is about 120 degrees. To facilitate the
modeled. The off-Dominal tap ratios of all transformers are 1.0 solution of thecase using programs witbout a TCR model, the
per-unit in thisparticular case. equivalent load and harmonic spectra of the TCR are listed in
4. The loads are modeled as constant power loads for load flow this paper, With this information, the TCR can be represented
solutions and as impedances for harmonic solutions. The as a constant reactive power load in load flow solution and a
harmonic impedances are determined according to the 3rd harmonic current source in harmonic analysis. Because the
model recommended in reference [2]. SVC is relativelysmall as compared to the HVDC, its impact
5. Harmonic filters are modeled as shunt harmonic impedances. on overall system harmonic distortion is not significant.
All filters are the single-tuned type.

72
8. The harmonic distortion results were obtained using the phase-domain representation. The following items must be
harmonic iteration method described in reference [1) . Because considered in the analysis of unbalanced distribution systems:
the results showed that the voltage distortions at the harmonic
source buses are small and the equivalent harmonic current 1. It is difficult to identify or specify harmonic-producing loads.
injections from the HVDe andSVC are made available in this In general, several loads are served from one point and the
paper, a non-iterative harmonic solution method which models harmonic currents represent the aggregate response of several
harmonic sources as harmonic current injections should give harmonic producing devices.
the same solution results. 2. Many distribution systems tend to contain capacitors.
Frequency scan analysis can be helpful to verify if resonance
conditions exist. Due to a large number of possible harmonic
source locations, however, it is difficult to determine the
H-order frequency scan buses. .
1 3. The commonly assumed properties under balanced conditions
5 such as the zero-sequence nature of triple harmonics no
7 longer hold. Harmonic producing devices at the distribution
level can generate uncharacteristic harmonics.
11
4. Load and transformer connections can have large impacts on
13 harmonic propagation. The subject of load modeling for
17 distribution system harmonic analysis still needs considerable
19 research [1].
23
25 As demonstrated in [1], relatively moderate variations in the
models can have a significant impacton results. The test system
is specified in a way that highlights all of these issues. The
11.2 Test System No.2: A 13-Bus Unbalanced Utility Alternative Transients Program was uses-to calculate harmonic
Distribution System propagation in the system [5,6]. paIiiif results are shown in
Table U .5 and Figure 11.5.
This system is based on the IEEE 13 bus radial distribution
test feeder [4]. The system is unbalanced and serves as a Table 11 .5. VoltazeTHO (Fundamental Freouencv Component)
benchmark system for unbalanced harmonic propagation studies. Node # PhaseA PhaseB PhaseC
The system was used in [1] for illustrative purposes and, with 32 1.96 1.034 1.76 1.038) 1.69 1.007
additional modifications, is proposed here as a harmonics test
33 1.96 1.034 1.76 1.038 1.6 1.007
system.
34 0.96 1.018 0.96 1.030 1.0 1.022
71 3.23 1.010 2.76 1.045 2.8 0.969
75 3.35 1.003 2.82 1.048 2.95 0.967
52 3.30 1.008
911 3.()()(0.965)

-----------~~~- l

i
Figure 11.4. Test System 1- Unbalanced Distribution System i
2 3 4 5 6 7 8 9 10 11 12 13 14 1S 16 17
The feeder, shown in Figure 11.4, contains voltage Hannonic.
regulators, three and single phase line configurations, shunt
capacitors, and spot and distributed loads. Phase-ground and Figure11.5. Harmonic Voltage Distortion Spectrum at Node71
phase-phase connected loads are included. For harmonic studies,
load compositions are specified to include harmonic producing
loads. Complete data are provided in the Appendix A. Current Modeling andsimulation features for this case are:
spectra for the three load types, namely fluorescent light banks,
adjustable speed drives. and composite ("other") residential 1. Conventional loadswere modeled as constant RL impedances
loads. are given for test purposes. The analysis of harmonic obtained from the givenkVA at 60Hz.
propagation in distribution systems must necessarily utilize a

73
2. Harmonic producing loads were modeled as current sources Additional data used to conduct a harmonic analysis of the
with the specified spectra using the 'Models' capability of the example industrial system include the following:
AT? Magnitudes were scaled based on the fundamental
component of load current and phase angles were adjusted 1. System equivalent impedance. For this study, the system
based on the phase angle of the voltage across the load impedance was determined from the fault MVA and XlR ratio
obtained from the fundamental frequency solution. at the utility connection point. These values are 1000 MVA
3. The motor and the capacitor at node 34 were assumed out of and 22.2, respectively. Driving point impedance (as a
service. For harmonic frequencies, the motor should be function of frequency) at the connection point was not
modeled using its sub-transient impedance (or locked rotor available, but should be used wheneverpossible.
impedance). 2. The local (in-plant) generator was represented as a simple
4. The voltage regulator was not modeled. Rather, the Thevenin equivalent. The internal voltage, determined from
substation transformer secondary taps on the three-phases the converged power flow solution, is 13.98/-1.52 kV. The
were set at +15,+10 and +13, respectively. equivalent impedance is the sub-transient impedance which is
5. Lines were modeled as mutually coupled1t branches. O.0366+j 1.36510.
3. The plant power factor correction capacitors are rated at 6000
For the case studied, the voltage distortion levels are low. kvar. As is typicallydone, leakage and seriesresistance of the
This is because several loads are connected phase-phase and bank are neglected in this study.
harmonic phase angles are modeled. As described in reference 4. The displacement power factor for the drive load is 0.97
[I], significantly different results are obtaineddepending on the lagging. This high power factor is typical of drives operated
choice of load models and harmonic currentsource models. It is at or near full load.
noted that in the examples in [1], all loads were assumed to be
connected phase-ground, the motor and capacitor at node 34 are Table 11.6. Per-Unit Line and C~e Impedance Data
in service and harmonic source spectra were different from the (base values: 13.8kV, 10,000kVA)
ones used here. From To R x
100:UTIL-69 01:69-1 l}:OOI39 0.00296
11.4 Test System No.3: A 13-805 Balanced Industrial
Distribution System 03:Mll.L-l 50:GEN-I 0.00122 0.00243
03:Mll..L-l 05:FDR F 0.00075 0.00063
This test case consists of 13 buses and is representative of a 03:Mll..L-l 26:FDR G 0.00157 0.00131
medium-sized industrial plant. The system is extracted from a 03:Mll..L-l 06:FDR H 0.00109 0.00091
common system that is being used in many of the calculations
and examples in the IEEE Color Book series [7]. The plant is Table 11.7. TransformerData
fed from a utility supply at 69 kV and the localplant distribution From To Voltage Tap kVA %R %X
systemoperates at 13.8 kV. The system is shown in Figure 11.6 01:69-1 03:Mll.L-l 69:13.8 69 15000 0.4698 7.9862
and described by the data in Tables 11.6-11.9. Due to the
balanced nature of this example, only positive sequence data is 50:GENl 51:AUX 13.8:0.48 13.45 1500 0.9593 5.6694
provided. Capacitance of the short overhead line and all cables 05:FDRF 49:RECT 13.8:0.48 13.45 1250 0.7398 4.4388
are neglected. 05:FDR F 39:T3 SEC 13.8:4.16 13.11 1725 0.7442 5.9537
26:FDR G 29:Tl1 SEC 13.8:0.48 13.45 1500. 0.8743 5.6831
06:FDR H 11:T4SEC 13.8:0.48 13.8 1500 0.8363 5.4360
06:FDRH 19:17 SEC 13.8:2.4 13.11 3750 0.4568 5.4810
SO:GEN-l Table 11.8. Generation, Load, and Bus Voltage Data
l00:tJTII,69
(from power flow study results)
Bus Vma 0 PIS QRtD PIGId <ltC*l
~ Sl:AUX
01:69-1 (p.u.) (deg)
100:lJTll..,-69 1.000 0.00
kW
7450
kvar
S40
kW kvar

03:MILL-l 01:69-1 0.999 -0.13

03:MIll-l 0.994 -2.40 2240 2000
50:GENl 0.995 -2.39 2000 1910
51:Aux 0.995 -3.53 600 530
OS:FDRF 26:FDRG
05:FDRF 0.994 -2.40
49:RECT 0.980 -4.72 1150 290
39:T3SEC 0.996 -4.85 1310 1130
26:FDRG 0.994 -2.40
49 RECT
o 'J:.. T. sfr 29:T11 T T 06:FDRH 0.994 -2.40
T ~9:T3 T SEC 11:T4 sJ l';9:T7 SEC II:T4 SEC 0.979 -3.08
19: T7 SEC 1.001 -4.69
370 330
ASD 2800 2500
Figure 11.6. Test System3 - A Balanced Industrial System 29:TII SEC 0.981 -4.16 810 800

74
 Table 11.9. Harmonic Source Data Table 3.5: Plant Harmonic Voltage Distortion Summary.
Harmonic # Percent Relative Angle Bus VI v, V, THD
1 100.00 0.00 (V LN ) (VL'l) (VLa'l) v
5 18.24 -55.68 (%)
7 11.90 -84.11 lOO:UTIL-69 39645.70 40.37 104.23 0.28
11 5.73 -143.56 01:69-1 39538.00 52.36 135.14 0.37
13 4.01 -175.58 03:MILL-l 7712.77 53.51 138.13 1.93
17 1.93 111.39 50:GENl 7726.55 51.72 133.51 1.87
19 1.39 68.30 51:Aux 262.74 1.72 4.40 1.81
23 0.94 -24.61 05:FDRF 7709.24 54.07 138.35 1.94
25 0.86 -67.64 49:RECT 269.89 12.79 12.83 8.02
29 0.71 -145.46 39:T3 SEC 2240.05 14.83 37.21 1.80
31 0.62 176.83 26:FDRG 7709.07 53.48 138.04 1.93
35 0.44 97.40 06:FDRH 7703.35 53.43 137.91 1.93
37 0.38 54.36 11:T4SEC 260.40 1.78 4.59 1.90
19:T7 SEC 1302.74 8.58 21.78 1.81
Specific issues related to modeling for harmonic analysis 29:Tl1 SEC 256.29 1.71 4.36 1.84
must also be considered if the results presented here are to be
obtained using different analysis programs. Modeling
considerations applicable to this example include: 11.6 Acknowledgment

1. All loads are modeled as series RL circuits. This approach is
taken instead of parallel RL modeling to more accurately
represent the limited harmonic damping offered by typical
induction motors without resorting to extremely detailed
motor models.
2. Frequency dependence of model resistance is neglected. This
is done mainly because of the significant discrepancies that
exist among various programs available. In addition,
neglecting frequency effects on resistance. leads to over
conservative results (which are often preferred).
3. Transformer magnetizing branch effects are neglected. In
addition, increasing winding losses asa function of frequency
are also neglected. As discussed in 2 previously, this is done
to avoid problems when comparing theresults presented here
with those obtained using other analysis programs.
The results of a harmonic analysis of the system of Figure
~11.6 are given in Table 11.10. Fundamental, fifth, and seventh
voltage harmonic amplitudes and THDv are given foreach of the
system buses. These. results, along with those obtained from a
fundamental frequency power flow study (shown in Table 11.8),
give an accurate descnpnon of the voltage profiles in the plant
11.5 ConclusioDS
The Task Force would like to acknowledge the support of the
IEEEPES Harmonics Working Group chaired by Mr. T. Gentile.
Case 1 was prepared by W. Xu, Case2- by~S.J. Ranade, andCase
3 by M. Halpin. Results were verified by R. Burch, M. Halpin,
CJ. Hatziadoniu, andT.H. Ortmeyer.
11.7 References
1. IEEE Task Force on Harmonics Modeling and Simulation,
"Modeling and Simulation of the Propagation of Harmonics
in Electric Power Networks, Part 1 & 2", IEEE Trans. on
Power DeliVery, Vol. 11, No.1 January 1996,pp. 452-474.
2. ClORE Working Group 36-05, "Harmonics, Characteristic
Parameters, Methods of Study, Estimates of Existing Values
in the Network", Electra, no. 77, July 1981,pp.35-54.
3. W. Xu, I.E. Drakos, Y. Mansour, A. Chang, "A Three-Phase
Converter Model for Harmonic Analysis of HVDC Systems",
IEEE Trans. on Power Delivery, Vol. 9, No.3, July 1994,
pp.1724-1731.
4. IEEE Distribution Planning Working Group Report," Radial
Distribution Test System," IEEE Trans. on Power Systems,
Vol. 6, No.3, Aug.1991, pp.97S-98S.

Complete data for three harmonic test systems has been
presented in thischapter. The systems can be used as benchmark
systems for the development of new harmonic analysis methods
and for the evaluation of existing harmonic software.
Researchers, developers and users of harmonic analysis
programs are encouraged to use these systems to test their
programs andreport their comments to the IEEEPES Harmonics
Modeling and Simulation Task Force.
5. Canadian/American EMTP User's Group "Alternative
Transients Program(ATP) Rule Book" , Portland, OR, 1995.
6. H.W. Dommel, "Electromagnetic Transients Program
Reference Manual (EMTP Theory Book)", Prepared for
Bonneville Power Administration, Dept. of Electrical
Engineering, University of British Columbia, Aug. 1986.
7. IEEE Standard 399-1990, "IEEE Recommended Practice for
Industrial and Commercial Power System Analysis", IEEE,
New York. 1990.

75
 Appendix A: Data for Test System 2 10 505: Phasing C N, conductor 1/0,Neutral 1/0
~ = O.OOOO+jO.oooo O.OOOO+jO.oooo O.OOOO+jO.OOOO
Source System: Node 50. O.OOOO+jO.oooo O.OOOO+jO.oooo
Short circuit MVA 1100 at 82 degrees lagging. Balanced. 1.3395+j1.3295
Yabc = 0.OO 0.OO 0.OO
Substation: Node 50 -31 0.OO 0.OO
Transformer: 5 MVA, 115 kV delta -4.16kV wye grounded j4.6178
Impedance z= 1 + j 8 % at 60 Hz.
ID 508: Three-phase URO concentric netural 250 MCM AI...
Voltage Regulators: Connected at node 31 cables, 6" apart on horizontal plane 40 belowground . Neutral is
II

individual phase control. Wye connected, PT Rati0=20, cr 13 #14 Cu . OD over neutral is 1.28 ..
Rating=700 A, R + jX =3+j9Q, voltage level =122 V Zatx:= O.8506+jO.4037 0.3191+jO.0325 O.3191+jO.0325
O.8597+jO.4458 O.2848+j-o.0145
Transformers: Node33 - Node 34 O.8597+jO.4458
500 KV ~ 4160 delta- 480, wye volts, Z = 1.1 + j 2.0 % Yabe = j94.6212 0.OO 0.OO
j94.6212 0.OO
Linephasing and60 Hz impedance matrices j94.6212
Allconductors ACSR. Line geometry is available from the Internet
site. Upper triangle of phasedomain impedance (Zabc ohms/mile) ID 509: Single-phase URO tape shield; I/O copper tape shielded
and admittance (Yabc J.1S/mile) matrices are shown. For non- conductor with separate I/o copper bare neutral on I spacing; 40"
t1

existent phases , matrices have beenpadded with zeros. deep.

~ = O.9806+jO.5146 O.OOOO+jO.OOCXl O.OOOO+jO.OOOO
ID 501: O.OOOO+jO.OOOO O.OOOO+jO.OOOO
Phasing BACN, Phaseconductor 556,500 26n, Neutra1410 6/1. O.OOOO+jO.OOOO
Zabc= 0.3477+j 1.0141, 0.1565+j 0.4n7 0.1586+j 0.4361 YIbc = jO.3915 0.OO 0.OO
0.3375+jl.0478 0.153S+j 0.3849 0.OO 0.OO
0.3414+j 1.0348 0.OO
Yabc = j6.2450 -jl.7664 -jl.3951
j5.8271 -jO.7461 IDe onnecnvttv ta:
j5.6985 Node-I Node-J LenRtb (ft) ID
ID 502: 32 45 500 S03
Phasing CABN, phaseconductor 4/06/1, neutral 410 6/1 33 32 500 S02
~= 0.7538+jl.1775 0.1586+jO.4361 O~1S65+jO.4777
9 45 46 300 503
0.7475+j 1.1983, O.1535+j 0.3849 31 32 2000 501
O.7436+j 1.2112 52 84 800 S09
Yabc = j5.6587, -jl.1943 -jl.5024 71 32 2000 501
j5.2262 -jO.6626 71 84 800 504
j5.3220 71 150 1000 501
10503: 7S 92 500 508
Phasing CBN,phaseconductor 1/0,neutral 1/0 84 911 300 505
71 92 switch
~= . O.OOOO+j 0.OO O.OOOO+j 0.OO O.OOOO+jO.OOOO
1.3294+jl.3471 0.2066+jO.4591 Shunt canacitors (kVara:
1.3238+jl.3569
Node Connection Ph-A Ph-B Pb-C
Yabc = 0.OO 0.OO 0.OO
y
75 200 200 200
j4.7097 -jO.8999
34 Y 125 125 125
j4.6658
911 0 0 100
ID 504: phasing A C N conductor 110, neutral 110
9

Zuc= 1.3238+jl.3569 OJlOOO+jO.OOOO O.2066+jO.4591 Motorloads: Node 34

O.OOOO+jO.OOOO O.OOOO+jO.OOOO 500 HP three-phase induction motor; running power factor 0.8,
1.3294+jl.3471 efficiency 90%; lockedrotor 3000KVA at 0.4 power factor lag.
Yabc = j4.6658 0.OO -jO.8999
0.OO 0.OO
Loads:
j4.7097 The following model codes are used: D-delta or phase-phase
connection , Y- wye or phase-ground connection. S-constant
kVA fundamental frequency model. Z-constant impedance

76
fundamental frequency model, I-constant current fundamental
frequency model.

Node Model Ph-A Ph-A Ph-B Ph-B Ph-C Ph-C

No. kW levar kw kvar kw kvar
34 Y-PQ 42.63 20.18 0 0 0 0
Harmonic load: None
45 V-PO 0 0 170.53 125.09 0 0
Harmonic load: 60% other (composite) types
46 D-Z 0 0 230.22 131.97 0 0
20% fluerescent, 20% ASD, 20% others.
Note: this load is connected between phase B&C.
52 y-z 127.90 85.79 0 0 0 0
10% fluorescent, 10%ASD, 20% others
71 D-PQ 383.70 219.95 383.70 219.95 383.70 213.95
30% fluorescent, 60% others.
75 Y-PQ 486.02 189.07 68.21 60.55 289.91 212.65
15% fluorescent, 20% ASD, 15% others.
92 D-I 0 0 0 0 170.53 151.38
15% fluorescen~ 20% ASD, 15% others.
Note: theload is connected betweenphases C&A.
911 Y-I 0 0 0 0 170.53 80.74
15% fluorescent, 20% ASD, 15% others.
32 Y-PQ 16.48 9.45 66.40 38.06 116.97 97.05
71 Note: this is a distributed loadbetween nodes 32 & 71

Current spectra of harmonic loads:

Phase angles are with respect to thefundamental frequency
vo1tage In
dezrees.
H Fluorescent ASD Other
order Mag. Phase Mag. Phase Mag. Phase
1 1 -41.2 1 -1.5 1 -35.0
2 0 0 0 0 0 0
3 0.2 273.4 0.542 0.7 0.007 -105.8
4 0 0 0 0 0.095 -167.4
5 0.107 339.0 0.152 110.8 0.002 -275.5
6 0 0 0 0 0.083 -42.6
7 0.021 137.7 0.069 151.9 0 0
8 0 0 0 0 0.005 -247.8
9 0.014 263.2 0.043 -95.0 0 0
10 0 0 0 0 0 0
11 0.009 39.8 0.036 -13.9 0 0
12 0 0 0 0 .0 0
13 0.006 182.4 0.029 95.2 0 0
14 0 0 0 0 0 0
15 0.005 287.0 0.025 -182.7 0 0

77
 Conclusions
This tutorial has covered many of the aspects of power system harmonic analysis. While much
has been accomplished in the field over the last few years, there are still many challenging problems
without solutions. Given the projected increases of nonlinear loads in power systems, these solutions can
not come too soon.

Significant areas of concern related to harmonic modeling and simulation include 1) load
modeling, 2) network equivalents, and 3) hybrid solution algorithms. Significantadvances in these areas
are still needed. It is important that these concepts be developed considering not only the needs and
positions of electric utilities, but those of the end-users of electrical energy as well. Only a closely
coordinated effort will produce the best solutions.

The area of load modeling offers a number of avenues for future research. In particular, the
development of effective aggregation procedures is needed. As pointed out throughout this tutorial, load
modeling is essential for accurate results, but it is often impossible to represent all loads in the detailed
manner needed to obtain accurate results. At this time, aggregation methods that are effective for the
general case are not available.

The area of network equivalents suffers from the same deficiencies as load modeling. At present,
a generic network reduction procedure is not feasible. This lack of capability often forces the-aDalyst into
a compromising position between accuracy and feasibility. The need for network equivalents is clear;
every harmonic study will most likely require one or more. There is, however, much work remaining to
be done to properly address this complex issue.

The conventional harmonic simulation methods (current injections, harmonic power flow.. etc.)
are well-suited for the analysis of a number of harmonic-related problems. As nonlinear loads continue to
grow in size and complexity, however, many more operating conditions are likely to emerge.
Unfortunately, it is very difficult to study advanced nonlinear loads with the more conventional methods.
The hybrid methods offer great promise due to their oatural ability to accept device models in an almost
arbitrary form. Much work remains, however, to take this promising avenue from a research topic to a
commercially available product

In conclusion, I would like to thank each of the contributing authors for their bard work and
dedication that was required to put this material together. Dr. Wilsun X~ who organizedthis tutorial and
assisted in every phase of its development, deserves special recognition. In addition, Dr. M E. EI-
Hawary, Chair of the PES Lifelong Learning Subcommittee, is recognized for providing valuable direction
and advice throughout the development of this document

Mark Halpin

78
 Author Biographies

Reuben F. Burch, IV (M 70) was born in Eastman, GA, on August 5, 1948. He received his BEEdegree
from Auburn University in 1970. He is a Principal Engineer in Enhanced Power Quality at Alabama
PowerCo. in Birmingham. AL where he mainly performs voltage flicker and harmonic analyses. He is a
member of IEEE, PES, and IAS andis a registered professional engineer in Alabama and Georgia.

Gary W. K. Chang, (M'94), received his Electrical Engineering Diploma from National Taipei Institute of
Technology, Taipei, Taiwan, in 1982, and the MSEE and PhD degrees from National Tsing Hua
University, Hsinchu, Taiwan, and the University of Texas at Austin in 1988 and 1994, respectively. Dr.
Chang is currently with Siemens Power Systems Control, a Division of Siemens Power Transmission &
Distribution, LLC. His areas of interest include power systems optimization, harmonics, and power
quality. Dr. Chang is a member of Tau Beta Pi and a registered professional engineer in the state of
Minnesota.

J.J. Dai (S86, M94) received his Ph.D. degree from the University of Toledo, Ohio in 1995, in Electrical
Engineering. Since 1992,he has been associated with Operation Technology, Inc., where he is involved in
the development of computer software for power system modeling and simulation studies, and currently
serves as Vice President of Advanced Technology. His interests and experience include power system
stability study and improvement, power systemharmonic analysis and correction, industrial power system
conceptual design, and computermodeling and simulation of powersystem operation andcontrol. Dr. Dai
is a member of IEEE-PES and IEEE-lAS where he serves as secretary of the Power System Analysis
Subcommittee and the Chair of the Low-Order Harmonics Task Force of the lAS Harmonics Working
Group.

s. Mark Halpin (M 93) received his BEE, MS, and Ph.D. degrees from Auburn University in 1988, 1989,
and 1993, respectively. He is currently an Assistant Professor in the Department of Electrical and
Computer Engineering at Mississippi State University. His teaching interests include power systems,
control systems, and network analysis. His research interests are in the areas of modeling and simulation
techniques for large-scale power systems, power system transients and harmonics, and computer
algorithms. He is active in the IEEE PowerEngineering Societyand Industry Applications Society, where
he serves as chairmanof the IEEE-lAS Working Groupon Harmonics.

C.J. Hatziadoniu (M 97) received his Ph.D. in electrical engineering from West Virginia University in
1987. He is currently an associate professor with the department of electrical engineering at Southern
Illinois University. His interests include power electronics, HVDC, and modeling and simulation of power
systems.

Takashi Hiyama received his B.E., M.S., and Ph.D. degrees all in Electrical Engineering from Kyoto
University, Japan, in 1969, 1971, and 1980,respectively. He joined Kumamoto University in 1971, andhe
currently is a Professor of Electrical Engineering and Computer Science. During the period of June 1985
through September 1986, he was at Clarkson University, and was involved with powersystem harmonics
research. His current interests include intelligent control of electric power systems using fuzzy logic
control and neural networks, measurement and real time control of photovoltaic systems, and harmonic
analysis of distribution systems. He is a senior member of IEEE, and a member of lEE of Japan, SICE of
Japan, and the Japan Solar Energy Society.

Dr. Yilu Liu is and AssistantProfessor of Electrical Engineering at Virginia Polytechnic Institute and State
University. She received the BS degree from Xian Jiatong University, China, in 1978, and MS and PhD
degrees from The Ohio State University in 1985 and 1989, respectively. Dr. Liu is the recipient of the
1993 National Science Foundation Young Investigator Award and the 1994 Presidential Faculty Fellow
Award.

79
Thomas Ortmeyer received his B.S.E.E. in 1972, the M.S.E.E. in 1977 and the Ph.D. in 1980, all from
Iowa State University. From 1972 through 1976, he worked in the Operational Analysis Department,
Commonwealth Edison Company, Chicago, Illinois. Since 1979, he has been at Clarkson University,
where he is Professorof Electrical Engineering. Duringthe 1993-1994 year, he was Guest Professor of the
Advanced Technology of Electrical Engineering Chair, Kumamoto University. His current interests
include power systemharmonics, power electronics, machine control, and power system protection. He is
a senior member of IEEE, and a member of Eta KappaNu, Phi KappaPhi, and Sigma Xi.

Satisb J. Ranade (SM) received his BE degree from the Indian Institute of Technology, the MSEE from
New Mexico State University, andthe Ph.D. from the University of Florida. He is currently a professor at
New Mexico State University. Histeaching andresearch interests are in power systems, machinery, power
electronics, and AI applications in operation and control. He is a member of Eta Kappa Nu and Tau Beta
Pi, and serves as the counselor forthe student branch at NMSU.

Dr. Paulo Ribeiro received a BS in Electrical Engineering from the Universidade Federal de Pernambuco,
Recife, Brazil (1975), completed the Electric Power Systems Engineering Course with Power
Technologies, Inc. (pTI)-1979, and received the Ph.D. from the University of Manchester - UMIST,
England, 1985. He is a registered Professional Engineer (PE) in the USA, a registeredEuropean Engineer
(Eur Ing), and a Chartered Engineer (UK). Dr. Ribeiro is a Senior Member of the IEEE, Member of the
IEE(UK), and CIGRE and active in severalpower qualityrelatedtechnicalworking groups such as Power
Electronics Simulations, IEEE 519a Guidelines for Harmonic Limits Application, Harmonics,
Interharmonics, Probabilistic Harmonics, Power Quality, FACTS and Custom Power. Dr. Ribeiro has
extensive experience in power system studies, design, and system integration. Presently, He is with
Babcock & Wilcox - Product Development Department - Naval Nuclear Fuel Division as the Chief
Engineer for the Superconducting Magnetic Energy Storage (SMES) Program.

Wilsun Xu obtained his Ph.D. from the University of British Columbia, Vancouver, Canada, in 1989.
From 1990to 1996,be was withPowerTech Labs Inc. and B.C. Hydro as an electricalengineer. He joined
the University of Alberta, Edmonton, Canada, as an Associate Professor of Electrical Engineering in
September 1996. Dr. Xu's research interests are in the areas of power quality and harmonics, voltage
stability,and distribution automation.

80