Escolar Documentos
Profissional Documentos
Cultura Documentos
in
THE FACULTY OF GRADUATE STUDIES
(Department of Electrical and Computer Engineering)
We accept this thesis as conforming
to the required standard
October 2001
Abstract
The quality of the electric power delivered to customers by utilities may not be acceptable
for some types of sensitive loads, which are typically power electronics- and computer-based
loads, particularly in the control of industrial processes. There are cases where the increasing use of power electronics to enhance process eciency and controllability creates power
quality problems. The growing application of shunt capacitors for voltage support, power
factor correction, and system loss reduction, as well as the use of series capacitors (xed
or controlled, for line reactance compensation) will increase the potential risk of transient
disturbance amplications and potential electrical and mechanical resonances in the presence
of more and more power electronic devices, and of steam and gas turbines in distributed and
co-generation power plants. As the natural order of the system grows, so does its ability
to oscillate more! At the same time, new power electronic devices also oer the means for
adequate \power conditioning", to meet the special requirements of electric power quality in
a system.
To evaluate the promising solutions oered with the introduction of more and more power
electronic devices in transmission and distribution systems, such as FACTS (Flexible AC
Transmission Systems) Controllers and Custom Power Controllers, as well as to analyze
their interaction and impact on either the load or the network side, computer programs
based on the EMTP (Electromagnetic Transients Program) are becoming more useful. The
development of new EMTP-based models for representation of controls and power electronic
devices has been the main subject of this Ph.D. thesis project. Its main contributions are
summarized as follows:
development of a \simultaneous solution for linear and nonlinear control and electric
power system equations" (SSCPS) in EMTP-based programs, through the compensation
method and the Newton-Raphson iterative algorithm. This solution method eliminates
not only the one time step delay problem at the interface between the solution of power
and control circuits, but also all the internal delays, which may exist in methods based
on the transient analysis of control systems (TACS) since 1977. A \circuit approach"
was proposed in this thesis, as an innovative alternative to the solution presented by
A. E. A. Araujo in 1993;
R
(the UBC version of the EMTP), based
experimental implementation in MicroTran
on SSCPS, of a \simultaneous solution" for: linear and nonlinear current and voltage
dependent sources; independent current and voltage sources, which can also be connected between two ungrounded nodes; hard and soft limiters; transfer functions; mathematical and transcendental FORTRAN functions; special control devices and some
digital logic gates; transformation of variables (such as the abc to 0 transformation
ii
ABSTRACT
iii
interaction with a Brazilian utility company and industries for the realization and analysis of eld measurements of electromagnetic phenomena aecting the quality of power,
such as voltage sags and voltage swells; harmonic current and voltage distortions; transients, etc., with determination of causes, consequences and investigation of possible
solutions for power quality problems, as for example, the application of Custom Power
Controllers;
synthesis of simulation guidelines for the evaluation of the impact of power electronic
devices on the quality of power, based on realistic eld measurements and EMTP time
and frequency domain simulations.
The assessment of electric power quality, with the use of EMTP-based programs, and the
evaluation of the technical impact of power electronic devices on the quality of power, can
hopefully be performed with the models developed in this Ph.D. thesis project.
Contents
Abstract
ii
List of Tables
vi
List of Figures
vii
Acknowledgements
xii
Quote
xiii
1
3
4
6
9
12
13
.
.
.
.
.
.
.
15
18
18
21
32
33
36
CONTENTS
2.3
2.4
2.5
2.6
v
.
.
.
.
.
40
42
48
54
58
64
65
74
78
88
91
104
104
120
121
125
125
133
150
153
Bibliography
160
List of Tables
3.1 Comparison between voltage and current in a diode as a function of its parametric values.
80
4.1 Global harmonic distortion limits for the system voltages recommended in Brazil. . . . . 110
4.2 Comparison between eld measurements and EMTP simulation results for the operating
condition with the harmonic passive lters turned OFF. . . . . . . . . . . . . . . . . 119
4.3 Comparison between eld measurements and EMTP simulation results for the operating
condition with the harmonic passive lters turned ON. . . . . . . . . . . . . . . . . . 120
vi
List of Figures
1.1 Typical Design Goals of Power-Conscious Computer Manufacturers. (Source: IEEE Std.
446-1987, \IEEE Recommended Practice for Emergency and Standby Power Systems for
Industrial and Commercial Applications." ) . . . . . . . . . . . . . . . . . . . . . .
1.2 CBEMA curve revised by the Information Technology Industry Council (ITIC). . . . . .
1.3 (a) Thyristor in an industrial power converter. (b) Thyristors in a high voltage direct
current (HVDC) System. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
2.10
2.11
2.12
2.13
2.14
2.15
2.16
2.17
2.18
2.19
2.20
16
19
20
21
23
24
25
27
29
29
29
30
30
31
31
33
34
35
39
41
vii
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
8
9
LIST OF FIGURES
2.21
2.22
2.23
2.24
2.25
2.26
2.27
2.28
2.29
2.30
2.31
2.32
2.33
2.34
2.35
2.36
2.37
2.38
2.39
2.40
2.41
2.42
2.43
2.44
2.45
Simulation results of circuit with ideal operational amplier (noninverting amplier circuit).
Transfer function. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Observer form block-diagram of transfer function in equation 2.80. . . . . . . . . . . .
Possible computer implementation of the transfer function block-diagram in Fig. 2.23. . .
Block-diagram representation of a rst-order transfer function. . . . . . . . . . . . . .
Observer form block-diagram of rst-order transfer function of Fig. 2.25. . . . . . . . .
Possible computer implementation of rst-order transfer function of Fig. 2.25. . . . . . .
Realistic rst-order lag circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Time domain simulation of rst-order transfer function. . . . . . . . . . . . . . . . .
First-order transfer function with windup (static) limiter. . . . . . . . . . . . . . . .
First-order transfer function with non-windup (dynamic) limiter. . . . . . . . . . . . .
Transient response of a rst-order transfer function with windup and non-windup limiter.
Soft limits. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Zero-order transfer function with soft limits. . . . . . . . . . . . . . . . . . . . . . .
Time domain response for a sinusoidal excitation input u(t) illustrating the eects of hard
and soft limits on the output x(t). . . . . . . . . . . . . . . . . . . . . . . . . . .
Open loop control system with "supplemental devices S 1,S 2 and S 3". . . . . . . . . . .
Nonlinear control block-diagram with a sinusoidal intrinsic FORTRAN function. . . . . .
Circuit implementation for the simultaneous solution of a sinusoidal FORTRAN function.
Transport delay control device. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Circuit implementation for the simultaneous solution of a transport delay control device. .
Transient simulation of a transport delay control device. . . . . . . . . . . . . . . . .
Transient simulation of a pulse delay control device. . . . . . . . . . . . . . . . . . .
Pulse delay control device with arbitrary input signal. . . . . . . . . . . . . . . . . .
Logic gate "NOT". . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Circuit implementation of a logic gate "NOT" for simultaneous solution. . . . . . . . .
viii
41
42
44
45
45
46
46
47
47
49
49
50
52
53
53
54
55
55
58
59
60
61
62
62
63
66
70
71
72
72
73
LIST OF FIGURES
3.7
3.8
3.9
3.10
3.11
3.12
3.13
3.14
3.15
3.16
3.17
3.18
3.19
ix
76
76
77
78
81
82
82
85
85
86
86
87
87
93
94
95
96
97
97
98
99
100
101
102
103
106
107
108
LIST OF FIGURES
4.16 (a) Phase \A" current measured with harmonic passive lters turned on. (b) Phase-to-phase
\A-B" voltage measured with harmonic passive lters turned on. . . . . . . . . . . . . 109
4.17 THD harmonic trend, with harmonic passive lters turned o from 12:00 midnight to 06:00am. 109
4.18 THD harmonic trend, with harmonic passive lters turned on all the time. . . . . . . . 110
4.19 Distribution substation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.20 Current-source, parallel-resonant inverter for induction heating. . . . . . . . . . . . . . 114
4.21 Amplitude of the positive sequence system impedance at the PCC with harmonic lters. . 114
4.22 Phase angle of the positive sequence system impedance at the PCC with harmonic lters. 115
4.23 (a) Phase \A" current simulated with harmonic passive lters turned o. (b) Phase-tophase \A-B" voltage simulated with harmonic passive lters turned o. . . . . . . . . . 116
4.24 (a) Phase \A" current measured with harmonic passive lters turned o. (b) Phase-to-phase
\A-B" voltage measured with harmonic passive lters turned o. . . . . . . . . . . . . 116
4.25 (a) Phase \A" current simulated with harmonic passive lters turned on. (b) Phase-tophase \A-B" voltage simulated with harmonic passive lters turned on. . . . . . . . . . 117
4.26 (a) Phase \A" current measured with harmonic passive lters turned on. (b) Phase-to-phase
\A-B" voltage measured with harmonic passive lters turned on. . . . . . . . . . . . . 117
4.27 Instantaneous ideal compensation current to be \injected" by a shunt active lter. . . . . 118
4.28 Voltage sag measurements (%RMS versus time duration) with an overlay of the CBEMA
curve. For time durations less than 1 cycle the equipment seems to measure peak values. . 122
4.29 (a) Phase-to-phase \A-B" measured voltage sag. (b) Phase-to-phase \A-B" simulated voltage sag. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
4.30 Instantaneous voltage
uctuations causing light
ickering eect. . . . . . . . . . . . . 124
4.31 Modulated voltage and respective amplitude frequency spectrum . . . . . . . . . . . . 124
4.32 Control block diagram of a second order dierential equation with poles on the imaginary
axis of the complex plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.33 Solution of system with bounded resonance oscillations. . . . . . . . . . . . . . . . . 126
4.34 Introduction of a one time step delay in the control block diagram. . . . . . . . . . . . 127
4.35 Solution of system with unstable resonance oscillations caused by the introduction of one
time step delay. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
4.36 Classical linearized \swing equation", used in power system small-signal stability studies of
a single machine connected to an innite bus. . . . . . . . . . . . . . . . . . . . . . 130
4.37 Simulation results of the synchronous machine rotor angle deviation, in the presence of a
positive damping torque coecient. . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.38 Simulation results of the synchronous machine rotor angle deviation, in the presence of
negative damping torque coecient. . . . . . . . . . . . . . . . . . . . . . . . . . . 131
4.39 Canonical second order transfer function representation of the single-machine innite bus
system. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
LIST OF FIGURES
4.40 Circuit for the dynamic control of the ring angle (\") of a thyristor. . . . . . . . . .
4.41 Voltages and currents in a circuit with dynamic control of the ring angle of a thyristor. .
4.42 Circuit for the dynamic control of the ring angles of a three-phase six-pulse thyristor-bridge
rectier. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.43 Voltages and currents with dynamic control of the ring angles of a three-phase six-pulse
thyristor-bridge rectier. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.44 Dynamic control of the ring angles of a three-phase six-pulse thyristor-bridge rectier. .
4.45 Dynamic voltage control signals at the output of the proportional-integral (PI) and the
limiter control blocks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.46 Circuit for the dynamic control of three-phase PWM voltage source inverter (VSI). . . .
4.47 Phase \A" modulation and triangular carrier waveforms for generation of gating signals
through sinusoidal pulse width modulation (PWM). . . . . . . . . . . . . . . . . . .
4.48 Node voltage \vSA " generated by a three-phase PWM voltage source inverter (VSI). . . .
4.49 Voltage across the load \vSA NEUTR " and current supplied to the load by a three-phase
PWM voltage source inverter (VSI). . . . . . . . . . . . . . . . . . . . . . . . . . .
4.50 Load currents supplied by a three-phase PWM voltage source inverter (VSI). . . . . . .
4.51 Line-to-line voltage generated by a three-phase PWM voltage source inverter (VSI). . . .
xi
135
136
138
139
140
141
144
145
146
147
148
149
Acknowledgements
I would like to thank God for the gift of learning. My most sincere thanks to my parents,
Dorival and Isolina, and to all my relatives for their unconditional love and support. To my
wife, Luciana, and my daughters, Alexa and Aline, my love and my heartfelt thanks for their
strong participation in this life project altogether. I dedicate a very special note of thanks
to our special friends Wany, Fernando, Fulvia, Alexandre, Martha, and Richard for their
careful and kindness personal assistance.
I owe a tremendous debt of gratitude to Dr. Hermann W. Dommel, my Ph.D. thesis
supervisor, for all his personal and professional encouragement, share of wisdom and support
for the development of this thesis. (The responsibility for any remaining errors is solely
mine.) I also thank Dr. Dommel for the honor and opportunities of have being his teaching
assistant.
I also thank Dr. William G. Dunford for kindly accepting to be my Ph.D. thesis cosupervisor, with Dr. Dommel becoming a Professor Emeritus at UBC. I have also learned
with Dr. Jose R. Mart, who has excellent teaching skills. Professor Sandoval Carneiro
Jr. from the Federal University of Rio de Janeiro (UFRJ), Brazil, has gently been very
supportive, right from the start of this Ph.D. program in Canada.
I most specially appreciate the help, acceptance and advice of many individuals without
whom this opportunity would never have become fruitful. Professors, sta members, past
and present colleagues and friends at the Department of Electrical and Computer Engineering
of the University of British Columbia (UBC) have been the source of inspiration and support
to pursue scientic and personal growth. I also thank my former Brazilian professors and
colleagues at the State University of Campinas (UNICAMP), and at the Federal School of
Engineering of Itajuba (EFEI), for building and enhancing the foundation of my knowledge
in science and engineering.
I would like to sincerely thank the Fundaca~o Coordenaca~o de Aperfeicoamento de Pessoal
de Nvel Superior (CAPES), Braslia - Brazil, for the nancial support to this Ph.D. thesis
project. Without it, my dream would never come true.
I also thank the Brazilian utility company ELEKTRO - Eletricidade e Servicos S.A., with
a special reference to Francisco Alfredo Fernandes, for providing opportunities for a practical
interaction in power quality analysis, through a professional cooperation with the engineer
Ernesto A. Mertens Jr.. I acknowledge and thank students, professors and sta at the ETE
Prof. Armando Bayeux da Silva, a technical high school of the CEETEPS - Centro Estadual
de Educac~ao Tecnologica Paula Souza, S~ao Paulo, Brazil, for all the teaching experiences I
was able to conduct, which enriched my communication and leadership skills.
Last but not least, I thank and acknowledge the contributions of many people, not
mentioned, not forgotten, who certainly have had an impact and in
uence on my living and
studying at UBC, Vancouver, B.C., Canada, since August 24, 1997.
Vancouver, B.C., Canada
Benedito Donizeti Bonatto
October 09, 2001.
xii
\Engineering:
engineering is the application of mathematical and scientic principles to practical ends, as
in the design construction, and operation of economical and ecient structures, equipment,
and systems. It's art and communication, politics and nance, modeling and simulation,
invention, approximation, measurement and estimation, and more. It's a way to think
about problems."
http:nnwww.eg3.com
xiii
Chapter 1
Electric Power Quality and Power
Electronic Devices: An Overview
HE PURPOSE of this research project was to develop reasonably accurate models for
control systems and power electronic devices to evaluate their impact on the quality
of power. These models and methods were developed for implementation in the Electromagnetic Transients Program (EMTP) [1], [2], or in similar programs. A \circuit approach" is
used for the simultaneous solution of the control and electric power system equations, thus
eliminating any time step delay in the digital time domain simulation. Such time step delays
can cause inaccuracies or numerical instabilities. The advantages of the circuit approach is
its \generality and
exibility" for modelling multi-terminal linear and nonlinear control devices, which are needed in the analysis of electromagnetic phenomena aecting the quality of
power. This chapter presents an introduction to power quality problems and their relation
with power electronics, followed by a description of the motivations for this Ph.D. thesis
research and its relevant contributions.
industry has been proposed as a solution to make the present utility companies more competitive in oering better services and better quality at lower prices to their customers.
However, as in any business, this may require some investments in the infrastructure of the
power system, to cope with the new demands of the modern types of loads (power electronics and microcomputer based). In this scenario, traditional economic analysis, such as pay
back return or rate of interest, might show that these investments are only feasible with
concurrent increases in electricity taris. The paradox of more quality for less money still
remains a topic for discussion in forums such as government regulatory agencies.
With the growing utilization of automation and control based on the use of microprocessors, of power electronic devices, and of modern manufacturing techniques, industries have
been able to produce goods faster and with increasing quality. However, with such modernization new issues have emerged regarding the quality of electricity. Sensitive loads tend to
shut down if there are small variations in the network voltage. Also, harmonic distortions
caused by nonlinear loads may result in wrong operation or may increase the losses in power
system components. Another problem is capacitor switching in the utility system, which
may cause problems for adjustable speed drives (ASD's), which are used more and more by
industry. All these problems point out that more attention must be paid to power quality problems. Economic losses expressed in terms of interrupted production, of damage to
equipment, and of time delays in the processing of goods and the consequent negative impact
on customers have caused a rising number of complaints about power quality problems in
many electric utility companies. Estimating the cost of poor power quality is a dicult task.
Nevertheless, the annual approximate value would be in the order of hundreds of million
of dollars in damage. As an example, the cost per year to U. S. A. industry in lost time
and revenue due to power related problems were estimated in 1993 as US$26,000,000,000 [3].
The costs tend to grow as the sensitivity and use of microprocessor-based devices tend to
increase. The Electric Power Research Institute (EPRI) stated that in the year 2000, 60%
to 70% of total utility power generated within the U. S. A. would be controlled by power
electronics, compared to 30% in 1995.
Power electronic devices, however, are also able to \guarantee" a certain expected level
of electricity quality to a sensitive or special load, and such devices exist today. Flexible AC
Transmission Systems (FACTS) technology, Custom Power Controllers, active lters, among
other power electronics applications, oer promising solutions for improving the quality of
power in transmission and distribution systems.
The introduction of more and more power electronic devices into the network will create
issues of compatibility of operation not only in steady state, but also under transient conditions. New models for innovative equipment, as well as new philosophies for their control
and operation, will then be required. It is not enough to evaluate the electric quality conditions only at the interface of power electronic systems with the electric power system. The
propagation of electromagnetic phenomena into the industrial or utility network must be
evaluated as well [4]. Therefore, software packages such as the ElectroMagnetic Transients
Program (EMTP), have become important and necessary tools to analyze the impact of large
power electronic devices on the quality of electric power. The aim of this research project
was the development of new EMTP-based models for control and power electronic devices,
thus allowing the accurate evaluation of the impact of high power electronics on the quality
of power. As part of the project, eld tests were conducted in cooperation with a Brazilian
utility company, to provide realistic power quality data measurements.
1.1.1
Digital computer simulation of electromagnetic transients in single- and multiphase networks is well established [1]. Since the publication of [1] in 1969, signicant improvements
have been made in models for electrical power system components such as transmission lines
[5], transformers, turbine-generators and cables [6]. Of particular interest for this research
project are the necessary advances in the simulation of power electronic devices within power
systems.
Although steady-state solutions at fundamental and harmonic frequencies have been proposed to analyze power quality problems, the complexity of periodic switching in power electronic devices can only be studied thoroughly through time-domain simulations, e.g., with
EMTP-based programs [4], [7], [8], [9], [10], [11], [12], [13].
EMTP-type simulation is particularly useful for the analysis of the dynamic interaction
of distributed Custom Power Controllers in a power system. Despite international demonstration projects for some Custom Power Controllers already with some years of operating
experience, how these new devices will interact within each particular power system is still an
open question. The performance obtained from some prototypes, or from simulations with
simplied models, may not be sucient for real applications, where dangerous resonance
and other unforeseen problems may occur. These concerns make more realistic electromagnetic transient program based simulations important when detailed models of Custom Power
Controllers become available. This could avoid more expensive corrective actions after installation. Measurements and simulations also become necessary for performance evaluations
under dierent network and load conditions.
All those facts have encouraged the development of new practical models, methods and
guidelines for the appropriate use of EMTP-based programs as potential tools for power
quality studies.
Nevertheless, good engineering judgment in setting up the power quality problem and
representing the physical phenomena with appropriate models, does still represent the major
challenge, despite the impressive accuracy obtained with the models available today.
1.1.2
In recent years, tremendous improvements have also been made in digital measuring instruments, because of the use of microprocessors and digital signal processing techniques
[14], [15]. This made it possible to conduct power quality surveys in many countries around
the world. The Electric Power Research Institute (EPRI) commissioned an extensive survey
of distribution system power quality in the U. S. A. [16]. In Canada, the Canadian Electricity Association (CEA) developed a guideline for the power quality that utility customers
experience, in a three-year project [17]. A quantitative measure of the deterioration of the
ideal sinusoidal waveform from the growing utilization of power electronic devices resulted
from these surveys. In many of the real-world problems, momentary voltage variations have
been the main cause for shutting down microprocessor controlled industrial processes. The
diversity of problems to cope with, and their complex interaction, has created a need for
more research on power quality issues [18]. Moreover, with the present deregulation process
in the electricity industry, power quality has become a key factor for utilities and customers
in a competitive market.
The experience in monitoring power quality phenomena has increased in the latest years,
as reported by the related literature [15], [19], [20], [21], [22], [23], [24], [25], [26], [27]. Well
known power quality problems have been summarized and at the same time new problems
have been discussed in [28], [29], [30], [31], [32], [33], [34], [35], [3], [36], [37]. The need
for more detailed information on power disturbances, and the search for new techniques to
process the amount of available measured data, have motivated research into applications of
modern theories in the power quality area, as discussed in [38], [39], [40], [41], [42].
According to the author's experience as a \power quality engineer" in a Brazilian utility
company, there are many dierent reasons why electricity customers become dissatised with
the quality of the electric power delivered by utilities, and why they complain. This happen
partly because there may indeed be technical problems. Partly, the motivation comes from
the need to reduce electricity costs in the industrial production process, looking for better
taris and better contracts, as everybody tries to survive in a competitive and aggressive
business environment.
On the technical side, the most common power quality problems are caused by a fatal
combination of sensitive electronic-based loads and a high incidence of voltage sag phenomena [43], [44], [45], [46]. Most of these voltage sags are due to faults in transmission and
distribution systems, caused by lightning phenomena, which cannot be easily avoided or
minimized. It is also common to nd poor voltage regulation within the industry electric
system [45], [46]. This aggravates the impact of voltage sags, causing frequent process malfunction or interruption with nancial losses, which are rarely presented explicitly by the
industry personnel, unless any kind of nancial compensation is legally required. There is a
wide range of alternatives for possible solutions to technical problems in the quality of the
electric power supply. Usually, the immediate most cost-eective measure is to minimize the
cause or eects of the problem close to its origin, depending on the type of electromagnetic
phenomena involved. Typically, voltage sag problems can be minimized by proper adjustments in the sensitivities of the load or load control, whenever this is technically feasible.
However, in some cases compensation through the use of power electronic devices might be
a promising alternative, either for an individual sensitive load or for an entire industrial process. Usually, the utilities comply with the standards of supply set by the regulatory agency,
but the customer process is much more sensitive and some kind of electronic compensation
would be necessary. The potential con
ict is sometimes created when a possible technical
solution requires high nancial investments. A cost versus benet analysis usually leads to
a cheap compromise solution; in the absence of clear regulations one needs \to live with the
problem!"
1.1.3
Power quality has become an important issue because of the increasing use of power
electronic devices. The Institute of Electrical and Electronics Engineers, Inc. (IEEE) has
therefore developed standards to address power quality problems, which are brie
y discussed
here.
The problems related to the quality of electricity are not new, since there was never an
ideal sinusoidal waveshape, with frequency and voltage exactly at their rated values. However, with the changes in the type and sensitivity of the loads in recent years, harmonic
current and voltage distortions, short- and long-duration voltage variations, impulse and oscillatory transients, voltage
uctuations (causing visual
icker), power frequency deviations,
voltage unbalance among the phases in a three-phase system, and other electromagnetic
phenomena are increasingly causing power quality problems.
IEEE Std. 1159-95 [47] denes and characterizes electromagnetic phenomena which may
cause power quality problems. It also provides recommended practices for monitoring electric
power quality.
Most utility regulations dealing with harmonics are based on IEEE Std 519-1992, [48],
[49], [50]. This standard presents recommended practices and requirements for harmonic
control in electric power systems. It addresses most of the issues of harmonic generation
by power electronic converters, arc furnaces, static VAR compensators and power electronic
controlled drives. It also discusses:
eects of harmonics;
harmonic measurements;
In the later task, time-domain simulation can be particularly useful to predict equipment
and power system behaviour. It thus can help engineers to provide some answers in detecting
harmonic related or other power quality problems. The IEEE Std. 519-1992 [48] is currently
been revised to account for interharmonics in power systems and the possible application of
probabilistic approaches in harmonics evaluation.
IEEE Std. 446-1987 [51] covers the recommended practice for emergency and standby
power systems for industrial and commercial applications. A computer \voltage tolerance envelope", shown in Fig. 1.1, also known as the CBEMA curve (Computer Business Equipment
Manufacturing Association curve) is presented in this standard.
The Information Technology Industry Council (ITIC) revised the CBEMA curve, which is
presented in Fig. 1.2. It shows that computer- and power electronics-based loads, properly
designed by the manufacturers, should be able to withstand a complete interruption of
voltage supply for up to 20ms, a voltage sag of 30 percent for 0:5s, 20 percent for 10s or
10 percent in steady state. It also denes the upper limits in the input voltage that should
be tolerated. The CBEMA (ITIC) curve has been widely used as an important \reference"
400%
PERCENT VOLTAGE
300%
200%
VOLTAGE
BREAKDOWN CONCERN
115%
106%
COMPUTER VOLTAGE
TOLERANCE ENVELOPE
100%
87%
LACK OF STORED ENERGY IN
SOME MANUFACTURERS'
EQUIPMENT
0%
0.001
0.01
0.1
30%
0.5 1.0
10.0
100.0
1000.0
2s
Figure 1.1: Typical Design Goals of Power-Conscious Computer Manufacturers. (Source: IEEE Std.
446-1987, \IEEE Recommended Practice for Emergency and Standby Power Systems for Industrial and
Commercial Applications." )
for the susceptibility level of computer- and power electronics-based loads. However, due
to the great variety of products and processes, and their response to transient variations in
the supply voltage, there are cases where the load sensitivity is much more strict than the
CBEMA (ITIC) curve, which has to be determined then case-by-case for an adequate power
quality assessment and proposal of solutions.
IEEE Std 1100-1992 [52] presents the recommended practice for powering and grounding
sensitive electronic equipment. It addresses the multidisciplinary area of power quality,
giving practical guidelines on load and source compatibility concerns.
Voltage
uctuations causing visual
icker are being studied by the Task Force IEEE
P1453 on Light Flicker, which is considering the adoption of existing standards and practices
600
500
400
300
Applicable to SinglePhase
120V Equipment
200
140
120
100
80
70
40
0
Voltage Tolerance
Envelope
10
1us
10
110
90
10
10
10
10
10
1ms 3ms
20ms
0.5s
Duration in Cycles (c) and in seconds (s)
10
10s
10
Steady
State
Figure 1.2: CBEMA curve revised by the Information Technology Industry Council (ITIC).
of the IEC (International Electrotechnical Commission) and UIE (International Union for
Electroheat) for measuring such types of disturbances. This task force is also reviewing other
IEEE standards and recommendations on this issue. Other IEEE Standards within the IEEE
Color Series Books (http://www.ieee.org) provide useful recommendations about complex
issues on topics associated with the quality of power in utility, industrial and commercial
installations.
1.1.4
10
High voltage direct current (HVDC) and
exible AC transmission systems (FACTS technology) have been used for some time to extend power transfer capability, to improve power
system stability, and for other reasons. Dr. Narain G. Hingorani introduced the acronym
FACTS (Flexible AC Transmission System) for high power electronics applications in transmission systems [53], [54], [55]. HVDC, static Var compensator (SVC), thyristor controlled
series compensations (TCSC), static synchronous compensator (STATCOM), static synchronous series compensation (SSSC) and unied power
ow controller (UPFC) are examples of the so called FACTS devices. Reference [56] provides an annotated bibliography of
HVDC and FACTS devices. It also includes a list of FACTS installations, with data on
manufacturers, utility companies, countries, etc. It shows that, despite the high costs of
these high power electronic devices, they are gaining in acceptability around the world.
The term \Custom Power" was also introduced by Hingorani, to represent power electronics applications designed to mitigate power quality problems in industrial and distribution systems [57], [58], [59]. The distribution static condenser (D-STATCOM), the voltage
sag compensator (also known as DVR - dynamic voltage restorer), the solid-state breaker
(SSB), the solid-state transfer switch (SSTS), among others, are examples of such Custom
Power Controllers. Various manufacturers have proposed shunt, series, or shunt/series dynamic compensation schemes, with dierent acronyms, as solutions to specic power quality
problems. \The D-STATCOM, although based on the STATCOM, has a wider range of
applications. In fact, the D-STATCOM can be designed for reactive power control, or for
voltage control of the fundamental frequency, but it may also include higher frequencies as
in shunt active power lters." The integration of series- and shunt active lters, referred
to as unied power quality conditioner (UPQC) [60], [61], is promising to be the denite
solution for the majority of power quality problems. \However, its high cost may make it
useful only in some special cases. On the other hand, the shunt or series devices such as the
D-STATCOM or the voltage sag compensator will probably play a signicant role in future
distribution systems" 1 .
The ongoing deregulation process in many countries is also fostering competition in the
1 From personal communication with Dr.-Ing. Maurcio Aredes, COPPE/UFRJ, Rio de Janeiro, RJ,
Brazil.
11
electric power industry, which accelerates the application of new technologies in the transmission and distribution system. For example, there are applications being developed for
superconducting magnetic energy storage devices (SMES) for low voltage distribution systems, which will provide voltage support for a few seconds to sensitive processing equipment
during times of voltage sags.
Case studies with practical applications of Custom Power Controllers can also be downloaded directly from the web sites of some manufacturers, as for example:
(a)
12
(b)
Figure 1.3: (a) Thyristor in an industrial power converter. (b) Thyristors in a high voltage direct current
(HVDC) System.
devices used. Although the semiconductor power devices act as switches, they are not ideal
switches and many physical limitations do apply.
To analyze the in
uence and interaction of dierent new power electronic devices on
the quality of power is important for electric utilities and their customers. Moreover,
with a worldwide deregulation process in the electricity industry, power quality analysis
13
will rise in importance and urgency. Therefore, as part of this project eld tests were
developed in a Brazilian electric utility company, where realistic power quality cases
were analyzed and simulations were performed;
The opportunity to conduct practical eld tests in cooperation with an electric utility
company was a valuable experience, and necessary for the validation of digital computer
models.
With the addition of ideal operational ampliers, transfer functions can be implemented with a \circuit approach", where the circuit elements R, L, C are solved by
the main code of the EMTP. If integration methods are changed in the EMTP, for
example from trapezoidal rule to backward Euler as done in some versions at instants
of discontinuities with the CDA technique, no extra coding is needed. Operational
ampliers are not aected by integration rule changes. Moreover, if ideal operational
ampliers are implemented in steady-state solution, the frequency response of linear
14
control systems could be easily calculated in EMTP-based programs by just using the
frequency scan option.
This Ph.D. thesis is organized as follows: Chapter 2 presents the simultaneous solution
method for control and electric power system equations (SSCPS) in EMTP-based programs.
Chapter 3 discusses the developments made for power electronics models in EMTP-based
simulations. Chapter 4 presents simulation cases of power quality assessment with the use
of the existing features of MicroTran, the UBC version of the EMTP. SSCPS simulation
cases with the new models of Chapter 2 and the developments for the dynamic control of
power semiconductors presented in Chapter 3 are illustrated in practical power electronics
controllers. Simulation guidelines for the evaluation of the impact of power electronic devices
on the quality of power are summarized in Chapter 4. Finally, Chapter 5 presents the main
conclusions and contributions made in this Ph.D. thesis, and also points out the author's
recommendations for future work.
Chapter 2
Simultaneous Solution of Control and
Electric Power System Equations
(SSCPS) in EMTP-based Programs
2.1 Previous Developments on Transient Analysis of
Control Systems (TACS)
The computer subroutine TACS (acronym for \Transient Analysis of Control Systems")
was developed in 1977 [64] for the simulation of control systems in the EMTP (acronym
for \Electromagnetic Transients Program"). The general philosophy of the solution method
adopted at that time required a one-time-step delay at the interface between TACS and the
electric network solution 1 . This non-simultaneous approach was probably used because it
was easier to write a code separated from the main program, with a simple interface. The
main program passed information to the TACS program, which then returned information
to the main program for use one time step later, as illustrated in Fig. 2.1. Moreover, control
system equation matrices in TACS are usually unsymmetric, whereas the network elements
in the EMTP result in symmetric matrices. By separating the solution into two parts, the
code for symmetric matrices in the EMTP could be maintained.
The solution in two parts, with a time delay of one t between them, was an expedient
way to implement control system equations, but it proved to be the cause of critical numerical
1 Many other software programs, such as PSIM [65], also require a one-time-step delay between the solution
of control and power systems equations, which makes the solution method non-simultaneous.
15
16
Electric Network
Solution
( EMTP )
Time Delay
1 t
Control System
Solution
( TACS )
Figure 2.1: EMTP and TACS interface with 1 time step delay.
instabilities and inaccuracies in some cases in the time domain simulation of electric and
power electronic system transients [66], [67], [68], [63]. In cases where the EMTP and TACS
elements form a closed loop (or feedback system according to control theory), the eect of
the interface delay cannot always be eliminated by using a small step size t, as stated in
[69].
Besides the time delay between TACS and EMTP, even more time-step delays were
introduced by the internal solution algorithm of TACS, in order to deal with nonlinearities
in feedback control loops. The TACS internal solution is therefore non-simultaneous for
some control cases, and also sequential for its implemented devices. Improvements have
been made through the years in the TACS subroutine of some versions of the EMTP, such
as better ordering of its variables to minimize the number of delays inside TACS [70], using
the compensation method to eliminate the one-time-step delay in the EMTP-TACS interface
[71], development of a new TACS program \MODELS" [72] and its possible applications for
simultaneous solution of power electronics systems equations [73].
In 1993 A.E.A. Araujo proposed a simultaneous solution of both sets of equations, electric
network equations and control systems equations, as a way to eliminate the one-time-step
17
delay problem at the interface, as well as the internal control delays [67], [68], [63]. The augmented matrix with the control equations becomes unsymmetric due to the structure of the
control equations. Most of the equations of both the electric network and the control systems
are usually linear, while some are nonlinear. A proper partition of the system of equations
would allow the solution to be separated into two subsystems, one linear and another nonlinear. A.E.A. Araujo chose to solve the system of linear equations inside the EMTP, and
the system of nonlinear equations (including nonlinearities from the electric network and
from the control system) with the compensation method in an iterative Newton-Raphson
algorithm as in [74]. The control equations, both linear and nonlinear, were developed inside
the subroutine \CONNEC", which is a user-dened subroutine in the MicroTran version of
the EMTP of the University of British Columbia. \Similarly to TACS, the trapezoidal rule
of integration was used to numerically integrate the rst-order dierential equations inside
CONNEC, for example, in the implementation of transfer functions. The code was written
to prove the ideas, but as far as the author knows, was not implemented in a production
version of the EMTP."
In this research project, the simultaneous solution of the electric network and control
equations in EMTP-based programs is achieved with a \circuit implementation" of the control system. With this novel approach for EMTP-based programs, elements of the control
circuit which already exist in the EMTP, such as resistances and capacitances, are solved by
the EMTP proper, while elements missing inside the EMTP, such as ideal operational ampliers 2 and current and voltage dependent sources, are solved in the subroutine CONNEC
with the compensation method. This circuit approach is an alternative to the mathematical
representation adopted by Araujo, and gives some important advantages, such as \generality
and
exibility" for control modelling in EMTP-based programs.
The compensation method with an iterative Newton-Raphson procedure is used for the
solution of the added linear and nonlinear control system elements, such as dependent
sources, dierent types of limiters, as well as intrinsic FORTRAN functions and some special
control devices, as explained in the following sections. \Among the added elements, the de2 The author acknowledges the help of Mr. Jesus Calvi~no-Fraga for indicating in 1998 in his technical
report for a graduate course, the need for modeling operational ampliers in MicroTran [75].
18
pendent sources are the most important ones for control system modelling." The FORTRAN
code for the added elements in subroutine CONNEC has approximately 5,000 lines of code,
compared to 15,000 lines of code in the main part of the MicroTran version of the EMTP.
Compensation Method
The compensation method has long been used in EMTP-based programs for solving
the equations of nonlinear elements with the Newton-Raphson iterative method [74]. If
the nonlinear elements are not too numerous, this approach connes the iterations to a
19
relatively small system of equations, compared to the nodal equations for the entire system.
This approach is used here for solving the equations of dependent sources as a special case
of nonlinear elements. Without limiters, the equations are linear, but with an unsymmetric
matrix.
When there are M nonlinear elements in a circuit, the following system of equations 2.1
to 2.6, allows the simultaneous solution of the nonlinear equations with the rest of the linear
network [2],[78], which is then represented by its M-phase Thevenin equivalent circuit, as
illustrated in Fig. 2.2:
[ vOPEN ]
[ rTHEV ]
[i]
[ v]
i1
i2
i3
i4
...
vOPEN _ 1
iM
vOPEN _ M
ZM
...
vM
...
v4
vs(M)
Z4
v3
vs(4)
Z3
v2
vs(3)
Z2
v1
vs(2)
Z1
vs(1)
2
6
[vOP EN ] = 6
6
4
vOP EN
vOP EN
..
.
vOP ENM
1
2
(2.1)
3
7
7
7
5
(2.2)
[rT HEV ] =
6
6
6
4
[i] = 6
6
4
[v ] =
6
6
6
4
i1
i2
..
.
iM
v1
v2
..
.
vM
20
3
7
7
7
5
(2.3)
3
7
7
7
5
(2.4)
3
7
7
7
5
(2.5)
(2.6)
If the branch equations in 2.6 are linear, as in the case of dependent sources, they can be
represented in the form of a voltage source behind an impedance, as illustrated in Fig. 2.3,
or in the form of a current source in parallel with an impedance, as shown in Fig. 2.4. It is
assumed here that the branch impedances are not coupled, and that they are resistive (Rk ).
For other types of impedances, the equations would have to be modied.
ck
ik
Rk
vk
vsource(k)
dk
Figure 2.3: Representation of branch equation k as a voltage source in series with a resistance.
21
ik
ck
Rk
vk
isource(k)
dk
isource(k) = vsource(k) / Rk
Figure 2.4: Representation of branch equation k as a current source in parallel with a resistance.
After the two systems of equations 2.1 and 2.6 have been solved in subroutine CONNEC,
the currents [i] of 2.4 are returned to the main program, which adds the eect of the M nonlinear branches to the previously calculated open-circuit solution for all nodes with unknown
voltages,
[e] = [eOP EN ]
[zT ] [i]
(2.7)
where:
[e] is a column vector with the nal solution for the N node voltages;
[eOP EN ] is a column vector with the previously calculated open circuit solution for all the N
nodes with unknown voltages;
[zT ] is a rectangular matrix with N rows and M columns (N = number of nodes with
unknown voltages and M = number of branches solved with the compensation method) 3 ;
[i] = column vector with the M compensating branch currents.
2.2.2
Dependent Sources
This section presents the necessary equations for implementing current and voltage dependent sources in EMTP-based programs by using the compensation method. The following
important assumptions are made:
3 For further details about the compensation method, and the calculation of matrix [zT ], please, see
reference [74].
22
Proper precautions are taken to handle extremely large numbers and zero values.
The following models are derived: Current-Controlled Voltage Source (CCVS), CurrentControlled Current Source (CCCS), Voltage-Controlled Voltage Source (VCVS) and VoltageControlled Current Source (VCCS). In all cases, the equations from the Thevenin equivalent
circuit are the same, namely, for the controlling branch j
(2.8)
(2.9)
where:
vOP ENk = voltage vk for [i] = 0 (open circuit).
rkk = Thevenin resistance (self resistance of branch k).
rkj = Thevenin resistance (coupling or mutual resistance between branches k and j ).
vj = va
vb
ij = iab
(2.10)
(2.11)
and that the dependent source, CCVS, is connected between nodes c and d with branch
voltage
vk = vc
vd
(2.12)
23
ik = icd
(2.13)
ij
vOPEN
[ rTHEV j ]
ik
Rin
Rout
vj
vk
[ rTHEV k ]
ij
vOPEN k
vj = Rin ij
(2.14)
vk =
ij + Rout ik
(2.15)
where:
Rin = Input resistance of branch j .
Rout = Output resistance of the dependent source in branch k.
= Gain over the controlling or measured current, applied as voltage dependent source at
branch k.
Inserting equation 2.14 into 2.8 and equation 2.15 into 2.9, results in:
(2.16)
(2.17)
24
Using the two equations 2.16 and 2.17 is preferable to using the four equations 2.8,
2.9, 2.14 and 2.15, because it reduces the number of equations which have to be solved in
subroutine CONNEC from 4 to 2. Whenever possible, the voltages should be eliminated in
this reduction from 4 to 2 equations, because the solution will then produce the currents,
which are the variables that have to be passed back to the main program.
For an ideal current-controlled voltage source, Rin = 0 and Rout = 0, from which results:
(2.18)
(2.19)
If expressed in matrix form, one can see that the matrix becomes unsymmetric, since matrix
element j k is no longer equal to matrix element k j .
ij
vOPEN
[ rTHEV j ]
ik
Rin
vj
vk
ij
Rout
[ rTHEV k ]
vOPEN k
vj = Rin ij
(2.20)
(2.21)
25
where:
B = Gain over the controlling or measured current, applied as dependent current source at
branch k.
By inserting equation 2.20 into 2.8 and equation 2.21 into 2.9 one can also obtain, respectively, the following equations:
(2.22)
+ 1 ik + ::: + RrkM
out iM = 0
(2.23)
Observe that the division by Rout as done in equation 2.23, allows the use of very large
numbers for Rout without numerical diculties.
For an ideal current-controlled current source, Rin = 0 and Rout ! 1, resulting in:
(2.24)
Bij + ik = 0
(2.25)
ij
vOPEN
[ rTHEV j ]
ik
Rin
Rout
vj
vk
[ rTHEV k ]
vj
vOPEN k
26
vj = Rin ij
(2.26)
(2.27)
where:
A = Gain over the controlling or measured voltage, applied as dependent voltage source at
branch k.
By inserting equation 2.26 into 2.8 and dividing the resulting equation by Rin to avoid
numerical diculties, results in equation 2.28. In order to eliminate the voltages and keep
only the currents as variables, and also to allow the use of very large numbers for the gain
A, the following calculations are done: (equation 2.26 inserted into 2.8) minus the result of
[(equation 2.27 inserted into 2.9) and divided by the gain A]. This procedure eliminates Rin
in the resulting equation 2.29:
vOPENj
rj 1
+ Rin i1 + :::
R
in
rjj +Rin
::: + Rin ij + Rrjkin ik + ::: + rRjMin iM
k + r
vOP ENj + vOPEN
j1
rkj A
::: + rjj A ij + rjk
::: + rjM rkM
iM = 0
A
=0
rk1 i + :::
1
A
rkk +Rout i + :::
k
A
(2.28)
(2.29)
Based on the equations 2.28 and 2.29 for a voltage-controlled voltage source, if
A ! 1,
Rin ! 1, and
Rout ! 0,
then equations 2.30 and 2.31 are obtained, which can be used to model \ideal operational
ampliers". Note that the use of equation 2.31 only makes sense if there are feedback paths
modelled in the network part, which create the \rjk " coupling resistance. Then equation 2.31
27
will produce the correct current ik , which is returned to the main program for the calculation
of voltages by compensation. Note also that the equation 2.31 is exactly stating that vj = 0.
(Please, see equation 2.8.)
ij = 0
(2.30)
(2.31)
28
\In this thesis, if not otherwise clearly indicated, the assumption is made that all operational ampliers are ideal"!.
There are many variations and combinations of OP AMP circuits. The two basic ones
are the inverting amplier (Fig. 2.9) and the non-inverting amplier circuit (Fig. 2.10), with
the transfer functions are given by equations 2.32 and 2.33 respectively. Fig. 2.11 illustrates
an adder, a special case of the inverting amplier, where the output is a linear sum of the
input voltages, with the transfer function given by equation 2.34. Fig. 2.12 shows an ideal
integrator, with the transfer function as of equation 2.35.
E o ( s)
Ei (s)
Eo (s)
Ei (s)
Eo (s)
R4
R2
R1
= 1 + RR21
Ei1 (s)
R1
(2.32)
Ei2 (s)
R2
(2.33)
Ei3 (s)
R3
(2.34)
29
R2
R1
ei ( t )
eo ( t )
eo ( t )
R2
ei ( t )
R1
R4
R2
ei1 ( t )
R3
eo ( t )
ei2 ( t )
ei3 ( t )
Eo (s)
Ei (s)
RCs
(2.35)
Fig. 2.13 presents a generalization of the inverting amplier circuit, which is very useful
to obtain Laplace transfer functions by using the impedance approach [81]. With the ideal
operational amplier, a \virtual ground" potential appears at the inverting input terminal,
30
C
R
ei ( t )
eo ( t )
Z2(s)
Z1(s)
Eo ( s )
Ei ( s )
Z2 (s)
Z1 (s) :
(2.36)
For example, in the circuit shown in Fig. 2.14, the transfer function is derived with ideal
operational amplier using the impedance approach.
The complex impedances Z1 (s) and Z2 (s) for this circuit are:
Z1 (s) = R1
Z2 (s) =
Cs+ R12
(2.37)
R2
R2 Cs+1
(2.38)
31
R2
C
R1
eo ( t )
ei ( t )
Figure 2.14: First-order lag circuit using ideal operational amplier.
The transfer function is therefore obtained as
Eo (s)
Ei (s)
R2
1
R1 R2 Cs+1 :
(2.39)
ij
vOPEN
[ rTHEV j ]
ik
Rin
vj
vk
vj
Rout
[ rTHEV k ]
vOPEN k
vj = Rin ij
(2.40)
vk = Rout vj + Rout ik
(2.41)
32
where:
= Gain over the controlling or measured voltage, applied as dependent current source at
branch k.
By inserting equation 2.40 into 2.8 and dividing the resulting equation by Rin to avoid
numerical diculties, results in equation 2.42. In order to eliminate the voltages and keep
only the currents as variables, and also to allow the use of very large numbers for Rout , the
following calculations are done: times (equation 2.40 inserted into 2.8) minus the result
of [(equation 2.41 inserted into 2.9) and divided by Rout ]. This procedure eliminates Rin in
the resulting equation 2.43:
vOPENj
rj 1
+
R
Rin i1 + :::
in
::: + rjjR+inRin ij + Rrjkin ik + ::: + rRjMin iM
k +
rj 1
vOP ENj + vOPEN
Rout
kj i +
::: + rjj Rrout
rjk
j
::: + rjM RrkM
out iM = 0
=0
rk1 i + :::
Rout 1
rkk +Rout i + :::
k
Rout
(2.42)
(2.43)
For an ideal voltage-controlled current source, Rin ! 1 and Rout ! 1, resulting in:
2.2.3
ij = 0
(2.44)
(2.45)
Ideal Transformers
Even though an ideal transformer model has already been implemented in most EMTPbased programs, with the equations described in [2] , or with similar equations, an ideal
transformer can also be implemented as a special dependent source. The necessary equations
for the implementation of the ideal transformer as illustrated in Fig. 2.16 are 2.8 and 2.9,
as well as:
33
1 :n
ij
ik
vj
vk
vj 1
=
vk n
(2.46)
ij
= n
ik
(2.47)
where:
nj
1
n = nk = turns ratio of the ideal transformer.
From the equations above and from 2.8 and 2.9, one can easily obtain:
ij + nik = 0
k + r
vOP ENj + vOPEN
j1
rkj n
::: + rjj n ij + rjk
::: + rjM rkM
n iM = 0
(2.48)
rk1 i + :::
1
rkkn i + :::
k
n
(2.49)
Equations 2.48 and 2.49 can be used to model an ideal transformer. It is important to
mention that, normally, better models for electric transformers, which may include saturation
eects, are used in EMTP-based simulations. For further details, please see, for example,
references [2] and [6].
2.2.4
Independent Sources
It may be useful in a circuit or device model to have an independent current or independent voltage source connected between two ungrounded nodes. This can be accomplished
34
with the same technique used for dependent sources, but using only one equation in this
case.
vk = vc
vd
(2.50)
ik = icd
(2.51)
then the necessary equations for the implementation of an independent current source as
illustrated in Fig. 2.17 are:
ik
vk
isource
[ rTHEV k ]
Rout
vOPEN k
where:
(2.52)
(2.53)
35
(2.54)
+ isource = 0
ik + isource = 0
(2.55)
Of course, there is a much easier way to represent an independent current source between
nodes c and d directly in the nodal equations of the EMTP: inject the current source into
node c and with a negative sign into node d [2].
ik
Rout
vk
[ rTHEV k ]
vOPEN k
vsource
(2.56)
vk = vsource + Rout ik
(2.57)
36
where:
vsource = independent voltage source of branch k, which can be constant or a function of
time.
From the equations above, one can also obtain the following equation:
(2.58)
(2.59)
Another approach for voltage sources between ungrounded nodes frequently used in
EMTP-based programs is the insertion of an ideal transformer between the two ungrounded
nodes, with a voltage source to ground on the other side.
2.2.5
Newton-Raphson Algorithm
The equations for current and voltage dependent sources have been presented in the
previous sections, as well as the equations of independent sources which may be connected
between two ungrounded nodes. The solution of these equations is based on the compensation
method, which is already being used to solve nonlinear equations associated with nonlinear
elements in electric or electronic circuits with Newton-Raphson (N-R) iteration schemes. The
Newton-Raphson algorithm is well known, widely used and has quadratic convergence if the
initial estimate is close to the solution. For completeness, the Newton-Raphson algorithm is
presented in this section as it is in [77] and the reader is referred to mathematical books or
numerical analysis books or network solutions books, if more detailed information is needed.
In the scalar case the N-R iteration to solve
f (x) = 0
(2.60)
is given by
xk+1 = xk + xk = xk
(2.61)
37
f1 (x1 ; x2 ; : : : ; xM ) = 0
f2 (x1 ; x2 ; : : : ; xM ) = 0
..
.
fM (x1 ; x2 ; : : : ; xM ) = 0
(2.62)
Denote, for easy notation, the vector of variables by [x] and the vector of functions by
[f (x)]. Then 2.62 has a compact form:
[f (x)] = 0
(2.63)
Assume that the system has a solution; denote it by [x ] and expand each function in a
Taylor series about [x]:
@f (x x ) + @f (x x ) + + @f (x
f1 (x ) = f1 (x) + @x
xM ) +
1
2
1
2
@x
@xM M
@f
@f
@f
f2 (x ) = f2 (x) + @x (x1 x1 ) + @x (x2 x2 ) + + @xM (xM xM ) +
(2.64)
..
.
@fM (x
fM (x ) = fM (x) + @f@xM (x1 x1 ) + @f@xM (x2 x2 ) + + @x
xM ) +
M M
1
1
2
2
2
Assuming that x is close to x , higher order terms may be neglected and the system may
be written in linearized form:
[f (x )] [f (x)] + [J ] ([x ]
where
[J ] jx =
2 @f1
@x1
6 @f2
6 @x1
6 .
4 ..
@fM
@x1
[x])
@f1
@x2
@f2
@x2
@f1
@xM
@f2
@xM
@fM
@x2
@fM
@xM
..
.
...
..
.
(2.65)
3
7
7
7
5
(2.66)
jx
is the Jacobian matrix of the function [f (x)], which has to be calculated at each iteration
step. If equation 2.65 is set to zero and solved, the result will not be the vector [x ] (because
the higher-order terms have been neglected) but some new value for [x]. Using superscripts
to indicate the iteration count results in:
f xk + [J ] xk+1
k
x =
(2.67)
38
[J ]
f xk
(2.68)
xk = xk+1
k
x :
(2.69)
Then
[J ] xk =
f xk
(2.70)
(2.71)
Formulae 2.70 and 2.71 represent the Newton-Raphson algorithm for systems of equations, which reduce the error norm iteratively so that
f
xk+1
f xk
(2.72)
This iterations scheme is repeated until the errors are lower than a specied tolerance.
For the case of a system of linear equations, as in the case of linear dependent sources,
convergence to the solution is achieved with just one iteration step. More iteration steps
are required for the solution of a system of nonlinear equations depending on how close the
initial guess is to the nal solution. For highly nonlinear functions, the standard application
of the iteration scheme of the Newton-Raphson algorithm may cause numerical problems
involving computer over
ows.
The solution algorithm experimentally implemented in the MicroTran version of the
EMTP at The University of British Columbia is presented in Fig. 2.19. This method
presents \generality and
exibility properties", thus looking promising for future work in
detailed modelling of circuits and devices, as will be explained later in this thesis.
INITIAL GUESS
FOR CURRENTS
CALCULATION OF
BRANCH VOLTAGES
[from Eq. (2.6) or (2.1)]
SET VARIABLE
TO ITS LIMIT
CALCULATION OF
RIGHT HAND SIDE
[negative values
of Eq. (2.1)]
IS IT
ACCURATE
ENOUGH?
Yes
Yes
CHECK FOR
LIMITS
Yes
STOP
IS THERE
ANY LIMIT
VIOLATION?
No
MAX. NO. OF
ITERATIONS WAS
EXCEEDED?
No
No
UPDATE
CURRENTS
39
40
Possible Applications
The methodology presented in the previous sections for the implementation of dependent
sources in EMTP-based programs permits the computational development of many practical
applications, such as:
1. Current and voltage sensors;
2. Operational ampliers;
3. User-dened coupled branches in a circuit;
4. Modelling of electronic components, where the physical behavior would need to be
represented by nonlinear equations;
5. User-dened linear and nonlinear functions;
6. User-dened modelling of linear and nonlinear devices, limited only by the creativity
and ingenuity of the user.
Figure 2.20 and Fig. 2.21 illustrate the solution method with an example of a noninverting
amplier circuit, as commonly used in practical analog electronics. It consists of a sinusoidal
voltage source, an ideal operational amplier and 2 resistors (Rf and Rg ). The ideal operational amplier was modelled using equations 2.30 and 2.31, whereas the sinusoidal voltage
source and the resistors are part of the network, represented through a Thevenin equivalent
circuit. If Rf = 2Rg , then voutput = 3vinput , as shown in Fig. 2.21. Alternatively, to get
an amplication of 3 in circuit simulation, one could just use a voltage-controlled voltage
source, with equations 2.28 and 2.29 with the gain A set to a value of 3. In this case a
load resistor should be connected in the output of the dependent source, to avoid numerical
oating subnetwork problems.
Indeed, in theory this noninverting amplier circuit should result in:
voutput
R
= 1+ f
vinput
Rg
(2.73)
41
Rf
voutput
vinput
vinput=1.0 0o [V]
f=60 [Hz]
Rg
voutput
Voltage ( V )
vinput
4
0
10
15
20
25
30
35
40
45
50
Time ( ms )
Figure 2.21: Simulation results of circuit with ideal operational amplier (noninverting amplier circuit).
The next sections will present a technique for the simulation of transfer functions in
EMTP-based programs, some improvements already made for the implementation of saturation or limits for the elements or sources presented in this work, as well as the implementation
of some other special control devices.
42
H(s)
X(s)
X (s)
b sm + b sm 1 + ::: + b1 s1 + b0
B (s)
H (s) =
=k m n m 1n 1
=k
1
U (s)
an s + an 1 s + ::: + a1 s + a0
A(s)
(2.74)
with n m and an 6= 0.
It is possible to reorganize the terms of equation 2.74 as follows:
(an sn + an 1 sn 1 + ::: + a1 s1 + a0 ) X (s) =
k (bm
sm + b
sm
+ ::: + b1 s + b0 ) U (s)
bm sm n +
an
X (s) = k
+ ab1n s1 n + ab0n s n
bm 1 m
an s
U (s)
n + a0 s n
an
n + ::: + b1 s1 n
an
bm sm n + bm 1 sm
an
an
(2.75)
an 1
an s
+ ::: +
X (s) =
+ abn s
0
(2.76)
U (s)
+ :::
a1 1 n
an s
a0 n
an s
X (s)
(2.77)
43
X (s) = k
+ ab1n s1 n + ab0n s n
X (s) =
+s1
U (s)
k abnn U (s) + s
n k b1 U (s)
an
::: + s
k abn U (s)
1
+ bnan s 1 + :::
bn
an
an 1
an s
+ ::: +
n
+s
a1 1 n
an s
n k b0 U (s)
an
a1
an X (s)
+s
X (s)
+ :::
i
a0
an X (s)
an 1
X
(
s
)
+ :::
an
an 1
an X (s)
k bnan U (s)
a0 n
an s
k bnan 1 U (s)
i
a1 X (s)
an
k abn U (s)
0
a0
an X (s)
io
(2.78)
(2.79)
(2.80)
44
U(s)
...
kb0
___
an
kb1
___
an
1
___
s
a0
___
an
+
+
kbn-1
___
an
1
___
s
...
a1
___
an
kbn
___
an
1
___
s
+
+
X(s)
an-1
___
an
...
45
u(t)
...
an
___ M
kb0 -1 F
an
___ M
kb1
-1 F
1M
an
_ ___ M
a0
an
___ M
kbn-1
an
___ M
kbn
-1 F
1M
...
-1M
1M
an
_ ___M
a1
x(t)
1M
an
_ ___ M
an-1
...
Figure 2.24: Possible computer implementation of the transfer function block-diagram in Fig. 2.23.
10
__________
U(s)
X(s)
0.01 s + 1
X (s) = s
1
b
k 0 U (s)
a1
a0
X (s)
a1
(2.81)
46
required in the \more economic" computer implementation. There may be cases where the
realistic implementation is needed, which the proposed method can handle as well without
any restrictions. Fig. 2.29 presents the time domain transient response x(t) of the rst order
transfer function implemented as in Fig. 2.27 and as in Fig. 2.28, for a pulse u(t) of 1V
with a duration of 25ms.
U(s)
kb0
___
a1
X(s)
1
___
s
a0
___
a1
Figure 2.26: Observer form block-diagram of rst-order transfer function of Fig. 2.25.
u(t)
a1
___ M
kb0
- 1 F
- 10 k
- 1 F
x(t)
u(t)
1 k
x(t)
a1
_ ___ M
a0
Figure 2.27: Possible computer implementation of rst-order transfer function of Fig. 2.25.
the circuit of Fig. 2.28, which validated the simulation results presented in Fig. 2.29.
47
10 k
1 F
u(t)
10 k
- x( t )
1 k
x(t)
10 k
x(t)
Voltage ( V )
u(t)
1
10
15
20
25
30
35
40
45
50
Time ( ms )
48
with the modied nodal analysis (MNA), the network and control system equation matrix
becomes unsymmetric, and zero diagonal elements may appear, which requires pivoting techniques. Also, as the number of added branch equations increases, requiring extra columns
and extra rows, the dimension of the matrix may become very large, which may eventually
decrease the computational eciency, if proper techniques are not used. Possibily, if the simulation time becomes a very important issue, as in the case of digital real-time simulators,
a combined solution could be investigated, such that all linear control and system equations
could be solved using the MNA with appropriate techniques, and the remaining nonlinear
equations could be solved with the compensation method. The main motivation here in this
thesis for using the compensation method with a Newton-Raphson iterative algorithm is its
\generality and
exibility" to model nonlinear (and linear as a special case) control devices
in EMTP-based programs, particularly because the branch Thevenin equivalent circuit is
readily available, as in the CONNEC subroutine of MicroTran.
The implementation of limits associated with rst order transfer functions, as well as
special devices and intrinsic FORTRAN functions, are presented in the following sections.
49
[2]. Reference [66] presents an appropriate model for a proportional-integral (PI) controller
(which can be represented as a transfer function with one zero) with a non-windup limiter.
In a lead-lag control function block, for example, the way in which a non-windup limiter can
be realized is not unique; the interpretation of the limiting action should therefore be based
on the electronic implementation of the physical device [86].
The main dierence between windup and non-windup limiters is the way in which the
limited variable comes o its limit. To illustrate that, the rst-order transfer function presented earlier in Fig. 2.25 is assumed to have a windup limiter as in Fig. 2.30, and a
non-windup limiter as in Fig. 2.31. The time domain simulation of the output variable x(t)
for both cases, for a pulse input excitation u(t) of 1V, is presented in Fig. 2.32.
+ 5
slope=1
10
__________
U(s)
X(s)
0.01 s + 1
U(s)
10
__________
X(s)
0.01 s + 1
50
10
Voltage ( V )
x(t) with
windup limiter
(static)
x(t) with
nonwindup limiter
(dynamic)
u(t)
0
10
15
20
25
30
35
40
45
50
Time ( ms )
Figure 2.32: Transient response of a rst-order transfer function with windup and non-windup limiter.
The implemented solution for limiters uses the methodology proposed in Section 2.2.1.
As indicated in the algorithm illustrated in Fig. 2.19 of Section 2.2.5, a simultaneous system
solution is rst found without considering any of the limits. Then, each limit violation is
veried in the sequence of the input data given by the user. If a particular limit has been
reached, all previous indications of limit violations are cleared, and a solution is found for this
particular limiter and all of its consequences on the other limiters. This cause-consequence
iterative process has been found to be a very \robust method" in all cases tested, and has
given the correct solution for all limiters, \independent of the ordering of the input data given
by the user". Also, because the compensation method is properly applied 8 , the solution for
limiters is simultaneous without time delays.
The maximum (xmax ) and minimum (xmin ) limiting values are part of the input data. For
51
example, in the case of the rst order transfer function with a non-windup limiter illustrated
in Fig. 2.31, with a computer model as in Fig. 2.27, it is possible to represent the non-windup
hard limiting action with a simple change in the equations for the ideal operational amplier,
such that equations 2.30 and 2.31 are replaced by equations 2.82 and 2.83, respectively:
limit
= xmax , or vk
limit
limit
(2.82)
=0
(2.83)
= xmin .
By using these equations it becomes easy to observe the limits accurately. In practice,
in a realistic rst order lag circuit, as in Fig. 2.28, the clamping action is done with the use
of Zener diodes connected in parallel with the capacitor in the feedback loop of the rst OP
AMP for a non-windup (dynamic) limiter, or with Zener diodes connected in parallel with
the resistor in the feedback loop of the \second" OP AMP for a windup (static) limiter 9 .
Another example is the simple limiter control block. In this case, one could use the
equations for an \ideal voltage-controlled voltage source" including the limiting values in
the output voltage, as follows:
ij = 0
vOP ENk + rk1 i1 + :::
::: + rkj ij + rkk ik + ::: + rkM iM + vk
where vk
limit
= xmax , or vk
limit
(2.84)
limit
=0
(2.85)
= xmin .
52
Hard limits are then just a special case of soft limits when the slopes are set to zero. The
equations for soft limits, with the notation from Fig. 2.33, are:
8
if xmin < Ku(t) < xmax ,
< Ku(t);
x(t) = xmin + Kmin[u(t) umin]; if Ku(t) xmin,
(2.86)
:
xmax + Kmax[u(t) umax]; if Ku(t) xmax .
x(t)
slope=Kmax
x max
u min
u max
slope=Kmin
u( t)
x min
Consider, for example, the zero-order transfer function (constant gain) in Fig. 2.34. The
time domain response for a sinusoidal excitation input u(t) of 1V is presented in Fig. 2.35,
illustrating the eects of hard and soft limits on the output x(t).
In this thesis project, soft limits (and hard limits as a special case) have been implemented
for all the current and voltage dependent sources presented in Section 2.2. Limits can also
be easily implemented for all the FORTRAN functions and special devices which will be
discussed in the following sections.
+ 2
K=4
U(s)
53
X(s)
- 2
K min= + 0.1
( 10% )
Voltage ( V )
u(t)
x(t) with
hard limit
10
15
20
25
30
35
40
45
50
Time ( ms )
Figure 2.35: Time domain response for a sinusoidal excitation input u(t) illustrating the eects of hard
and soft limits on the output x(t).
54
S1
G1 ( s )
G2 ( s )
S2
G3 ( s )
S3
Figure 2.36: Open loop control system with "supplemental devices S 1,S 2 and S 3".
In this research project a \truly simultaneous solution" is achieved for intrinsic FORTRAN functions, and no special ordering is necessary, i.e., the input of data by the user is
arbitrary. The solution technique applies the compensation method in a similar way as done
in Section 2.2 for the implementation of current and voltage dependent sources. Assume, for
55
example, the control block-diagram of Fig. 2.37, with a nonlinear relationship between the
output voltage vk (t) and the input voltage vj (t).
vj
vk
K2 SIN ( K1 vj )
Figure 2.37: Nonlinear control block-diagram with a sinusoidal intrinsic FORTRAN function.
A simultaneous solution can be obtained for this nonlinear function by representing the
block-diagram of Fig. 2.37 in the form of an electric circuit as shown in Fig. 2.38. The necessary equations are 2.87 and 2.88 and the branch equations 2.89 and 2.90. These equations
resemble those of a voltage-controlled voltage source presented in Section 2.2.
ij
vOPEN
[ rTHEV j ]
ik
Rin
Rout
vj
vk
[ rTHEV k ]
vOPEN k
K2 sin ( K1 vj )
Figure 2.38: Circuit implementation for the simultaneous solution of a sinusoidal FORTRAN function.
For the controlling branch, the equation is
vOP ENj + rj 1 i1 + :::
(2.87)
::: + rjj ij + rjk ik + ::: + rjM iM + vj = 0
and for the dependent source branch it is
vOP ENk + rk1 i1 + :::
(2.88)
::: + rkj ij + rkk ik + ::: + rkM iM + vk = 0
where:
vOP ENk = voltage vk for [i] = 0 (open circuit).
rkk = Thevenin resistance (self resistance of branch k).
rkj = Thevenin resistance (coupling or mutual resistance between branches k and j ).
vj = Rin ij
(2.89)
56
(2.90)
where:
K1 = Gain over the controlling or measured voltage.
K2 = Gain applied to the dependent source in branch k.
From the equations above, one can also obtain the following equations:
vOPENj
rj 1
+
R
Rin i1 + :::
in
::: + rjjR+inRin ij + Rrjkin ik + ::: + rRjMin iM
=0
(2.91)
(2.92)
If a large number is used for Rin 10 , then the solution for the current ij will be very small
but not exactly equal to zero. This allows the convergence to a non trivial solution for the
current ik , and of course, for the input and output voltages of this nonlinear control block
of a sine function.
These equations can then be solved with the implemented Newton-Raphson algorithm
illustrated in Fig. 2.19, of Section 2.2. Note that for the proper application of the compensation method there must always be possible a Thevenin equivalent circuit. Therefore, in
cases where there is a
oating subnetwork, i.e., a node without connection to ground (as for
example if the output of an intrinsic FORTRAN function has no circuit elements connected
to ground), then the insertion of a big resistance between this node and ground easily overcomes this problem, as is usually done in some versions of the EMTP for the solution of
nonlinear elements [2].
Applying this technique, the following nonlinear intrinsic FORTRAN functions were implemented,
SIN
10 In theory Rin
! 1, making ij = 0, which would result in a trivial solution for equation 2.92. Rout is
obviously assumed to be equal zero.
COS
TAN
COTAN
SINH
COSH
TANH
ASIN
ACOS
ATAN
EXP
LOG
LOG10
SQRT,
57
multiplication ()
division (=)
exponentiation ().
Proper precautions were taken to handle mathematical and computational problems such
as division by zero, square root of negative number, exponentiation of negative number
with non-integer exponent, logarithm of zero or of negative values, etc.. The evaluation of
trigonometric functions accepts the argument in degrees, which is then converted to radians
internally. The inverse of trigonometric functions gives the answer already converted to
degrees.
11 Addition (+) and subtraction ( ) can be implemented with the use of just one ideal operational amplier.
58
Transport Delay
Assume, for example, the time delay control block-diagram of Fig. 2.39, (also called
"transport delay" in TACS), where the input voltage vj (t) only aects the output after the
elapsed time t + , or conversely, the output voltage vk (t) only depends on the past history
value of the input voltage, i.e., vj (t ).
vj
DELAY
vk
(2.93)
ij
vOPEN
ik
Rin
[ rTHEV j ]
59
Rout
vj
vk
[ rTHEV k ]
vOPEN k
K2 vj ( t - )
Figure 2.40: Circuit implementation for the simultaneous solution of a transport delay control device.
where:
vOP ENk = voltage vk for [i] = 0 (open circuit).
rkk = Thevenin resistance (self resistance of branch k).
rkj = Thevenin resistance (coupling or mutual resistance between branches k and j ).
vj = Rin ij
(2.95)
vk = K2 vj (t ) + Rout ik
(2.96)
where:
K2 = Gain applied to the controlling or measured past history voltage vj (t
an independent source at time t, in branch k.
), to create
From the equations above, one can also obtain the following equations:
vOPENJ
rj 1
+
Rin
Rin i1 + :::
::: + rjjR+inRin ij + Rrjkin ik + ::: + rRjMin iM
=0
(2.97)
(2.98)
ij = 0
(2.99)
(2.100)
60
which can be solved with the implemented Newton-Raphson algorithm illustrated in Fig.
2.19, of Section 2.2.
Considering that the delay time is not usually an integer multiple of the simulation
time step t, some type of interpolation must be used. Linear interpolation has been chosen
for that purpose, in a way similar to the transient time domain simulation of a transmission
line model [1], [2]. Fig. 2.41 illustrates the time response of a transport delay control device
where = 4:1667ms, t = 166:6667s, K 2 = 1; the input voltage signal vj (t) is a sinusoidal
source of 1V. Note in Fig. 2.41 that the output voltage signal vk (t) is actually equal to the
input voltage signal vj (t), but delayed in time in 4.1667ms.
2
delayed signal
1.5
Voltage ( V )
0.5
delay
4.1667ms
0.5
1.5
10
15
20
25
30
35
40
45
50
Time ( ms )
Pulse Delay
Applying the same technique used for the implementation for a transport delay, it is
possible to develop a model for a pulse delay control device. In a pulse delay, the negative to
61
positive and positive to negative zero crossings of the input signal are detected and a pulse is
created with the specied delay and with the width of respective time between the two zero
crossings of the input signal. This way there is no need to store all the past history values of
the input signal, just the respective times of zero crossing, which is a \more computational
economic" delay if the output signal will always have to be a pulse, irrespective of the shape
of the input signal. Fig. 2.42 illustrates the time response of a pulse delay control device
where = 20ms, K 2 = 1, and the input voltage signal vj (t) is a 1V pulse source. Fig. 2.43
shows the time response of a pulse delay control device where = 20ms, K 2 = 1, and the
input voltage signal vj (t) is an arbitrary signal source.
2
delayed pulse
1.5
Voltage ( V )
0.5
20ms delay
0.5
1.5
10
15
20
25
30
35
40
45
Time ( ms )
50
62
delayed pulse
1.5
Voltage ( V )
0.5
20ms delay
0.5
1.5
10
15
20
25
30
35
40
45
50
Time ( ms )
Figure 2.43: Pulse delay control device with arbitrary input signal.
vk = K2 (1 vj ) + Rout ik
(2.101)
where:
K2 = 1 is the gain over the controlling or measured voltage (with either vj = 0 or vj = 1),
being applied as a dependent source in branch k.
vj
vk
Figure 2.44: Logic gate "NOT".
ij
vOPEN
[ rTHEV j ]
63
ik
Rin
Rout
vj
[ rTHEV k ]
vk
vOPEN k
K2 ( 1 - vj )
Figure 2.45: Circuit implementation of a logic gate "NOT" for simultaneous solution.
From equations 2.93, 2.94, 2.95 and equation 2.101, one can also obtain the following
equations:
vOPENJ
rj 1
Rin +
Rin i1 + :::
::: + rjjR+inRin ij + Rrjkin ik + ::: + rRjMin iM
=0
vOPENK
r
k
1
vOP ENj
+ 1 + rj 1 + K2 i1 + :::
K
2
::: + rjj + rKkj2 ij + rjk + rkk +KR2 out ik + :::
::: + rjM + rKkM2 iM = 0
(2.102)
(2.103)
ij = 0
(2.104)
(2.105)
Other logic gates, such as "AND", "NAND", "OR", "NOR", etc. can be implemented
in a similar way.
Chapter 3
Power Electronics Modelling in
EMTP-based Simulations
\The application of semiconductor devices in the electric power eld has been steadily
increasing, and a study of power electronics (as it is commonly called) is now a feature
of most electrical and electronics engineering courses. The power semiconductor devices,
such as the diode, thyristor, triac, and power transistor, are used in power applications as
switching devices. The development of theory and application relies heavily on waveforms
and transient responses, which distinguishes the subject of power electronics from many
other engineering studies" [88].
\Computer simulation can greatly aid in the analysis, design and education of Power
Electronics. However, simulation of power electronics systems is made challenging by the
following factors: 1) extreme nonlinearity presented by the switches, 2) time constants within
the system may dier by several orders of magnitude, and 3) a lack of models. Therefore, it is
important that the objective of the computer analysis be evaluated carefully and appropriate
simulation packages be chosen" [89].
\In system level investigation, it is often adequate to represent semiconductor switches
within converters by ideal switches. This is important in order to minimize the overall
simulation time. But, it is very desirable if the same simulation package has the detailed
device models to design snubbers and gate drives. The simulation package should also be
able to represent the controller portion of the converter system by its functional features
64
65
DIODE
THYRISTOR
Cathode
Cathode
Cathode
Gate
( turn-on &
turn-off )
Gate
( turn-on )
Anode
66
Anode
Anode
MTO
ETO
MCT
Cathode
Cathode
Cathode
Turn-off Gate
Turn-on
Gate
Turn-off
Gate
Gate
( turn-on &
turn-off )
Turn-on Gate
Anode
Anode
Anode
TRANSISTOR
IGBT
MOSFET
Emitter
Emitter
Source
Base
Collector
Gate
Collector
Gate
Drain
67
Guidelines for modeling power electronics in electric power engineering applications, especially for use in EMTP-based programs, can be found in [91], [92]. These guidelines can
also be useful when using other digital simulation tools. The approximately 60 test cases in
the computer exercise collection of Dr. N. Mohan [93], [94], [89] which are available for both
R
EMTP and PSpice
simulations, are also useful for power electronics digital simulations.
In particular, the dierences between existing EMTP and PSpice models may give hints for
the improvement of EMTP models. Dr. Ned Mohan also presents in [89] a power electronics
library with special sub-circuits, which are used to represent switching electronic devices
and some control devices. According to [89] \in case of extreme nonlinearity, PSpice uses
extremely small time steps and is also prone to problems of voltage convergence. To avoid
this problem of extreme nonlinearity such as that associated with diodes, R-C snubbers are
connected across them. The values of R and C in these snubbers are not optimized, rather
these are based on speeding up the simulation without distorting the system voltage and
current waveforms signicantly."
As stated in [91], [92], power electronics modelling depends on the objectives of the
study. Depending on the type of study, dierent software tools and solution techniques can
be applied. In steady-state or harmonic analysis, the main concern is the injection and
propagation of harmonic currents into the transmission and distribution system, which may
cause unacceptable voltage distortions and dangerous resonances. The power electronic subsystem is then modelled as \known" harmonic current sources. These currents are assumed
to be independent of voltage variations at the point of common coupling (PCC), where the
power electronics load is connected. However, in many applications such as adjustable speed
drives, active power conditioning, FACTS and Custom Power Controllers, etc., the operation
of a power electronics device closely depends on and can aect the dynamics and the electric
transient behavior of the connected system. Variations of the system parameters, such as
voltage and current amplitude, frequency, phase-angle displacement among the phases in
a three-phase system, instantaneous or average power, etc., need to be used by the power
electronics control, to properly adjust the ring time of the semiconductors, which in turn
might have a feedback eect on the system. Therefore, for transient analysis, a more complex
and detailed representation of the power electronics devices as well as of the supplying power
68
system is required. Some model simplication and system reduction [92] might be necessary
for practical reasons. This can be acceptable, provided that the equivalent model is validated
against practical measurements, and that it ts the study investigation needs.
The representation of semiconductor switching devices is commonly simplied in power
level application studies. Therefore, the nonlinear characteristic of a diode is usually represented in a simplied form as a two-terminal uncontrollable unidirectional current
owing
switch, or in some programs as a voltage-controlled switch. Series on-state and parallel ostate resistances can be added to represent the semiconductor losses. In some EMTP-based
programs, such as MicroTran, series resistances must be included to allow multiple switch
connections on the same circuit node. Parallel resistances can provide a resistive connection
between the DC sides of rectiers and inverters and the AC local ground, thus avoiding
oating sub-network problems [95].
The use of simplied switch models for power electronics devices may be justied to
speed up the simulation time for system level studies, but it may also give wrong and
misleading results, especially related to semiconductor commutation phenomena. Also, the
EMTP solution at xed discrete time intervals t may result in inaccurate turn-on or turno switching times, causing unrealistic high frequency transients in the simulation of power
electronic devices. Backtracking techniques [67] and/or resynchronization techniques ([96]
pages 185, 204, 207), or even the Clock Synchronized Structure Changing Concept (CSSC)
[97] can be used to minimize the problem. Interpolation and/or extrapolation as well as
resynchronization techniques seem to be more and more applied even in the EMTP-based
solution of modern control for power electronics systems, as in the software PSCAD/EMTDC
[98], [99]. Therefore, for EMTP-based simulations of power electronics, it is much more
important to use such techniques than to reduce the time step size.
Numerical oscillations caused by the trapezoidal rule of integration in solving the system
of equations may also be a problem for EMTP-based simulations. The use of techniques
such as CDA (\Critical Damping Adjustment" [84], [85]) is eective in the elimination of
numerical oscillations. MicroTran has CDA implemented, but other EMTP versions may
not, or may use dierent approaches.
69
If the gating circuit is not considered in the study, three-terminal, controllable, unidirectional current
owing semiconductor devices can be represented by simplied switches
with gate turn-on and turn-o controls. Dierent ring controls can be applied to represent
thyristors, GTO's, IGBT's, etc. However, \in many actual power electronics applications, in
order to provide a continuous current
ow path for an inductive load, a reversal diode (free
wheeling diode) is used in parallel with a controllable switching device to form the basic
power electronic unit" [91]. The implementation of a \basic power electronic unit" in digital programs requires special care with respect to \instantaneous commutation phenomena"
[95], [96], [88], [97].
The UBC version MicroTran of the EMTP, up to now, only allows pre-dened timing
for the closing and opening of switches representing semiconductor devices. This is done
through the denition of a modied ring angle (\ ") and of a \switching frequency" for
each semiconductor. The frequency is needed to calculate the switching period and the time
of switching from , which assumes a xed reference at time t = 0. This implementation with
xed opening and closing times has limitations, because it does not use the real ring angle
(\") dened from a zero crossing detection related to the voltages at the semiconductor
terminals, and because it ignores the dynamics of the control circuits. MicroTran users could
write their own ring control subroutine based on an available ALPHA subroutine, but only
a few users have used this option. Also, sensing voltages and currents from the main program
would introduce a one time step delay with this approach, and users must be aware of that.
For PWM control techniques, the auxiliary program \PWM" can be useful to calculate and
dene the closing and opening times, which would then be read in as a switching table. The
dynamics of the control circuit would be ignored, which limits the application of \PWM" to
steady-state behavior.
As part of this thesis project, a subroutine \GATE" was developed to simulate power
electronics dynamic control schemes with more accuracy and
exibility. As its name indicates, the subroutine GATE allows a simplied gate ring control of a semiconductor, i.e,
it can control its turn-on time, and, for some devices, also its turn-o time. The control
signal is assumed to be a gate voltage signal, dened for simplicity, between the gate node
and ground. This subroutine was derived from the subroutine ALPHA (which can be user
70
vgate
VoltageControlled
Bidirectional
Switch
With the subroutine GATE, most of the three-terminal controllable power semiconductor
devices can be represented. Fig. 3.3 presents some test cases for the transient simulation of:
Figs. 3.4, 3.5 and 3.6 illustrate the controlling properties of the bidirectional switch, thyristor and GTO, respectively. Since the solution for switches follows the algorithm already
implemented in most EMTP-based programs [1], including MicroTran, the change of switch
position (\status" on or o) only happens one time step after the enabling gate signal [100].
If it becomes necessary to avoid this delay problem for certain types of simulations, an
alternative implementation for a \simultaneous" solution for voltage-controlled switches is
presented in the next section of this chapter.
71
i
VoltageControlled
Bidirectional
Switch
vsource
vgate
i
THYRISTOR
vsource
vgate
i
GTO
vsource
vgate
vsource = 10 sin ( t ) [ V ]
vgate = 1 [ V ]
f = 60[ Hz ]
Figure 3.3: Test cases for transient simulation of voltage-controlled, bipolar in voltage and bidirectional
current
owing switch, thyristor and GTO.
Switch Current
10
Current ( A )
Source Voltage
Gate Voltage
Voltage ( V )
10
10
12
14
16
Time ( ms )
Source Voltage
Thyristor Current
Current ( A )
10
Gate Pulse
Voltage ( V )
10
10
12
14
Time ( ms )
16
72
Source Voltage
GTO Current
Current ( A )
10
Gate Pulse
Voltage ( V )
10
10
12
14
Time ( ms )
16
73
74
(3.1)
(3.2)
where:
vOP ENk = voltage vk for [i] = 0 (open circuit).
rkk = Thevenin resistance (self resistance of branch k).
rkj = Thevenin resistance (coupling or mutual resistance between branches k and j ).
It is possible to assume as branch equations:
vj = Rin ij
(3.3)
and
vk = [vj (Ron
Roff ) + Roff )] ik
where:
Ron = On-state resistance, and
Roff = O-state resistance, for the voltage-controlled switch connected at branch k.
From equation 3.4 it is easy to verify that:
if vj = 1, then vk = Ronik ;
(3.4)
75
Moreover,
if Ron = 0, then vk = 0;
else if Roff
! 1, then ik = 0.
In order to sense the gate control voltage, assume Rin ! 1, which results in
ij = 0
(3.5)
(3.6)
which then can be solved with the implemented Newton-Raphson algorithm illustrated in
Fig. 2.19. Observe, however, that the voltage signal vj is calculated using equation 3.1 (i.e.,
vj = vOP ENj rj 1 i1 ::: rjj ij rjk ik ::: rjM iM ) in the solution algorithm, which is
always correct and less prone to numerical problems.
Assume, for example, the simple circuit with a simultaneous solution for a voltagecontrolled switch presented in Fig. 3.7. Applying an enabling control voltage signal (1V) at
the node gate, the switch turns on at the same time the enabling signal is received. When
the gating signal becomes zero (0V), the switch turns o at the corresponding same time,
as illustrated in Fig. 3.8. In the case of a conventional switch, as it is implemented in most
EMTP-based programs, the turn-on and turn-o would occur 1 time step later, as illustrated
in Fig. 3.9.
A similar approach has been proposed in [100] and [73] for the simultaneous solution
(also called "synchronized solution" in [73]) of voltage-controlled switches in EMTP-based
programs. Every time a switch changes its status, special computer techniques, such as, for
example, the Critical Damping Adjustment (CDA) [84], [85], would have to be triggered to
avoid numerical oscillation problems in the simulation of power electronic devices.
"Simultaneous"
VoltageControlled
Bidirectional
Switch
vsource
vgate
vsource = 10 sin( t ) [ V ]
vgate = 1 [ V ]
f = 60[ Hz ]
Source Voltage
Current ( A )
10
Gate Pulse
Voltage ( V )
10
10
12
14
16
Time ( ms )
76
12
Voltage ( V )
Current ( A )
Source Voltage
8
Gate Pulse
2
Time ( ms )
77
78
i (t)
v (t)
CATHODE
qv t
i(t) = Is e( kT )
( )
(3.7)
where:
i(t) is the current through the diode, from anode to cathode,
v (t) is the voltage across the diode, i.e., the potential dierence between the anode and
cathode terminals,
79
Is is a constant which depends on the physical properties of the diode, and is usually in the
range of 10 6 [A] to 10 9[A] 1 ,
q = 1:6022 10 19 C is the charge of an electron,
k = 1:3806 10 23 J= K is the Boltzmann's constant, and
T is the temperature in degrees Kelvin (273:16 K = 0 C ).
When the polarity of v (t) is as shown in Fig. 3.10, the diode is in the conducting region.
At 17 C 290 K , the constant VT kT=q 25mV . For v (t) < 3VT ( 75mV ),
i(t) Is . The value Is is usually referred to as the saturation current. If the diode
is forward biased with v (t) > 4VT (over 100mV ), equation 3.7 may be approximated by
i(t) = Is e(qv(t)=kT ) . Table 3.1 expresses the relationship between v (t)=VT and i(t)=Is , which
are derived from
i(t) =
v(t)
Is e VT
(3.8)
When a constant voltage V0 is applied to the diode, a constant current I0
ows through
it. The pair of values (V0 ; I0 ) is called the operating point of the diode. For each operating
point along the characteristic curve of the diode, one could dene a dynamic resistance of the
diode, which relates increments of the voltage to the increments of the current (dv (t)=di(t)
for v (t) = V0 and i(t) = I0 ). For higher frequencies, additional physical eects come into
play and the diode may no longer be treated as a simple nonlinear resistor. Charges stored
in the semiconductor material will also require the inclusion of \dynamic capacitive eects"
in the nonlinear model of a diode. The value of the capacitance is, in general, a function of
the voltage across the diode. The reader is referred to [77], [101] and other references for
further details. Reference [89], (pp. 71-1 to 71-7), discuss the PN junction diode switching
characteristics, including a sub-circuit for a diode model with reverse recovery.
In this thesis project, the compensation method presented in Section 2.2, with a NewtonRhapson solution algorithm, is used for the solution of the nonlinear model of a diode. Inversion of the diode characteristic curve from equation 3.8 and inclusion of a series resistance
Rout results in branch equation 3.10, which is solved together with the linear network equa1 Actually, Is is temperature dependent and may assume default values of 10 14[A] at 27C [101].
80
Table 3.1: Comparison between voltage and current in a diode as a function of its parametric values.
v(t)
VT
i(t)
Is
..
.
7
6
5
4
3
2
1
0
-1
-2
-3
-4
-5
-6
-7
..
.
..
.
1095.633
402.429
147.413
53.898
19.086
6.389
1.718
0
-0.632
-0.865
-0.950
-0.982
-0.983
-0.998
-0.999
..
.
tion 3.9 to determine the operating point of the diode for a particular network condition, as
illustrated in Fig. 3.11:
h
ik
Is
+1
i
+ Rout ik
(3.9)
(3.10)
where:
vOP ENk = voltage vk for [i] = 0 (open circuit).
rkk = Thevenin resistance (self resistance of branch k).
vk = voltage of branch k, i.e., voltage across the diode (the potential dierence vanode
vcathode ) plus the voltage drop across the series resistance Rout .
ik = current through the diode, i.e., the current
owing from the anode to the cathode
terminal.
Rout = series resistance for the diode model (which may be assumed to be equal zero).
81
vOPEN (t)
Nonlinear Diode Characteristic
Voltage ( V )
iSC (t)
SC
0.5
1.5
2.5
Current ( A )
Figure 3.11: V-I diode characteristic and network Thevenin equivalent circuit equation.
Inserting equation 3.10 into equation 3.9 results in:
(3.11)
82
ik
[ rTHEV k ]
vOPEN k
Rout
vk
vT ln ( ik / Is + 1 )
Figure 3.12: Circuit implementation for the simultaneous solution of a nonlinear diode model.
3
Network Thevenin
equivalent at time "t"
v
OPEN
(t)
VI Diode nonlinear
characteristic
Voltage ( V )
i
1
SC
(t)
3
3
Current ( A )
Figure 3.13: V-I diode characteristic and dierent network Thevenin equivalents.
83
on the contrary, convergence can be very slow for some cases, or numerical problems such as
computer over
ow may even arise. Therefore, it is important to derive a heuristic computer
technique for initial guesses, to accommodate the highly nonlinear exponential characteristic
of a diode, and speed up the convergence of the solution. A robust rule should be general
and insensitive to the network or diode parameters. The detailed development of such a
technique for implementation in EMTP-based programs is presented in the following.
A simple inspection of Fig. 3.11 reveals some \physical" candidates for the initial guess of
the nonlinear diode current, associated with the network conditions, i.e., with the particular
linear network Thevenin equivalent circuit:
ik =
ik =
vOPENk
rkk +Rout )
v
OPENk
Is e VT
(3.12)
ik = Is
(3.13)
(3.14)
Equation 3.12 is the short circuit current ikSC , which becomes the initial guess for the
diode current ik in the conduction mode. Then,
h
ikSC
Is
i
h
ikSC
Is
i
On the other hand, if vOP ENk < 0 and vOP ENk 6:0VT , then again the voltage
across the diode is rst estimated to be equal to vOP ENk , and the initial current is then
calculated with equation 3.13;
If vOP ENk < 0 and vOP ENk < 6:0VT , then the diode current is assumed to be as
calculated with equation 3.14. Alternatively, one could use a linear function instead of
using equation 3.14.
84
There are other possibilities for the initialization of the variables when a Newton-Raphson
method is being used [77]. For example, piecewise linear approximations could be used rst
to determine the initial guess (either for the current or for the voltage across the diode) and
then, the linear characteristic would be replaced by the detailed nonlinear equation. This
approach might be useful for speeding up the solution of networks with multiple nonlinear
elements, such as diodes, transistors, etc.
A two-slope piecewise linear approximation of the diode characteristic is accurate enough
in many cases. Even though the equations are linear in that case it is not known a priori
on which segment the solution ends up. This is particularly true if there are many diodes
to be solved. The production code of MicroTran still uses the Newton-Raphson method for
the two-slope piecewise linear representation, which will then decide, iteratively, the solution
points for the N diodes. Theoretically, there is no guarantee that the iterations will converge
in such cases, but by limiting the new voltages and currents to technically reasonable values,
the method has so far converged in all tested cases, with up to 6 diodes.
The use of linear piecewise or detailed nonlinear models of electronics semiconductors
might also be a requirement for the correct solution of commutation phenomena in power
electronics circuits. \In circuits with gate turn-o thyristors (GTO's), commutation of the
current into a diode must often be instantaneous, without any current interruption. An
example for such situation is the buck-boost (step-down/up dc-dc) converter shown in `Fig.
3.14 [93]' and the half-wave rectier circuit with a freewheeling diode in `Fig. 3.15'. ... If
the diode is represented as a switch which closes when the voltage from anode to cathode
becomes positive (either built into the code as in MicroTran, or controlled in this way
through TACS), then the positive voltage `across the diode' will only be seen in the next
time step immediately following the time step in which the GTO turned o. This is one time
step too late, because the current in the GTO would already have dropped to zero." [95],
[78]. Therefore, EMTP-type simulation of \instantaneous commutation" with simple switch
models might give wrong results in the case of switching converters. User knowledge of the
commutation process can be used to pre-dene the switches which will have simultaneous
commutation [78], but it may not work for all cases. A better option is the compensation
method used in this work, which assures a simultaneous solution of all equations [95].
85
DIODE
v = 155 sin( t ) [ V ]
DIODE 2
- 50 [ V ]
86
1
iDIODE
Vmax = 2 [V]
f = 60 [Hz]
vDIODE
Current ( A )
Diode Current
Diode Voltage
Volltage ( V )
10
12
14
16
Time ( ms )
12
x 10
Is=1.0E12 [A]
Diode Current
8.5
Time ( ms )
1.5
0.5
0.5
1.5
2
7.5
0.5
0.5
1.5
2
0.2
0.2
0.4
0.6
Current ( A )
0.8
1.2
9.5
1.4
Figure 3.19: V-I nonlinear characteristic of the diode resulting from the EMTP simulation.
87
Figure 3.18: Detail of the transient simulation of a nonlinear diode model in an EMTP-based program.
Current ( A )
Voltage ( V )
88
approximations are continuous from segment to segment, the derivatives are discontinuous.
Spline tting techniques can then be used to approximate the device characteristic by separate low-order polynomials between adjacent segments, with the cubic spline being the most
popular one [77].
89
and various non-linearities in the control system also requires appropriate models for a successful computer simulation.
Moreover, with variations in the supply voltage caused by power system disturbances,
the control, ideally, should be able to actively withstand and support its primary regulation
functions without disrupting the electric supply to the controlled load. However, in most
practical power quality cases, typically related to voltage sag problems, power electronicsbased loads (i.e., their control) are either extremely sensitive to momentary voltage variations
causing frequent shut-downs in industrial process operation, or improper designs may become
the cause of many power quality problems.
As part of this thesis project, \basic control devices" were experimentally implemented
in MicroTran. The main advantage of this development, compared to TACS (Transient
Analysis of Control System) used by many other EMTP versions, is that a \true simultaneous solution" is found through the compensation method using a Newton-Raphson iteration
scheme. Therefore, provided that appropriate computer techniques are used to allow convergence in the solution, this method has been shown to be very robust for the cases simulated
up to now, and does not have the 1 time step delay present in the interface of EMTP and
TACS and also does not have any internal delays in the linear and non-linear control solution.
With graphical user-interfaces for MicroTran, it would become easy to dene libraries of
control devices as well as of power components (and sub-circuits, similarly as in [65], [102],
[103]), in such a way that no dierentiation between power and control circuit will then
be necessary, since a unique simultaneous solution approach would be used, based on the
method and algorithm presented in this Ph.D. thesis.
Chapter 4
Evaluation of the Impact of Power
Electronic Devices on the Quality of
Power
HE analysis of the dynamic interaction between power electronic devices and power
systems and the assessment of electric power quality phenomena can be thoroughly
done with EMTP-based programs. The objective of this Ph.D. thesis research project was
to develop reasonably accurate models for EMTP-based programs, with which one could
evaluate the impact of high power electronic devices on the quality of power. To make the
thesis results more valuable to utilities and industries, the following practical tasks were
carried out:
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
91
This chapter presents some simulation cases of power quality assessment with the use of
the existing features of MicroTran, the UBC version of the EMTP. The simultaneous solution of control and electric power system equations (SSCPS), with the new circuit approach
presented in Chapter 2 and with the models developed for the dynamic control of power
semiconductor devices presented in Chapter 3, are illustrated through practical control and
power electronics controllers simulation cases. Important simulation guidelines for the evaluation of the impact of power electronic devices on the quality of power are also summarized
in this Chapter.
4.1 Dynamic Interaction between Power Electronic Devices and Power Systems
The interaction between the utility supply system and power electronics-based loads,
(such as electronic converter controlled electric motor drives), depends on a variety of factors
[104], as for example:
1. the type of \front-end" electronic converter, which converts line-frequency AC into DC
(diode-bridge rectiers (which are unidirectional in power
ow), switch mode converters
(in which, power
ow can be reversed), thyristors converters (which can be made
bidirectional in power
ow);
2. the number of phases (single-phase, three-phase) from the supply system used by the
converter, and the converter conguration (e.g., 6, 12, 24, 48, etc. pulses converters),
which also aects the waveform current distortion;
3. the \strength" (or \stiness") of the utility system, determined by its \short-circuit
power";
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
92
4. the number of power electronics based loads, the point of their connection in the
network, and the electric power of electronic converters, which will in
uence their
impact on the electric supply system and vice-versa;
5. the design and control of the electronic converter. The choice of parameters and components, particularly in cases where low-cost choices are made, can cause a deterioration
on the quality of power of the utility system. On the other hand, the converter operation (i.e., its control) can easily be disrupted by power system disturbances travelling
through the utility network, as is very common in the case of sensitive loads.
The most common lower-power electronics based load uses a single-phase diode-bridge
rectier [94], [104] which draws highly distorted waveshape current from the utility system,
and may cause harmonic associated problems, such as:
risk of resonances with utility or industry power factor correction capacitor banks, etc.
This type of rectier only draws current close to the maximum peak of the \assumed sinusoidal" utility voltage source, in order to recharge the capacitor lter on the DC side of the
rectier. To illustrate that, Fig. 4.1 presents a typical single-phase diode bridge rectier [93],
[94], [95], [104] 3 , where Ls1 represents the system Thevenin equivalent inductance (resistance
ignored here) and Ls2 represents any series inductance added in the AC side of the rectier.
Assume that the diodes are ideal and the circuit has been energized a long time ago, such
that the DC capacitor lter has already been charged (i.e., the transient energization has
gone and \stead state waveforms" are present). Note, from Fig. 4.2 , that the current is
(and consequently id ) only starts to
ow during the positive semicycle of the source voltage,
when vSA is greater than the voltage vdc (i.e., when diodes 1 and 2 are forward biased at
3 Reference [95] presents useful guidelines on power electronics applications using the EMTP.
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
93
time t1 ). Conversely, during the negative subcycle, when vSA is less than vdc , the current is
starts to
ow (i.e., when diodes 3 and 4 are forward biased at time t4 ), with id = is
owing
in the same direction, corresponding then to the electronic current rectication process).
Fig.4.3 presents the harmonic amplitude spectrum of the current drawn from the source by
a single-phase diode-bridge rectier.
id
vSA
Ls1
vPCC
D1
Ls2
is
0.5 mH
D3
vA
0.5 mH
vdc
D4
1000 F
20
D2
-
To better understand the waveform voltage distortions caused by a single-phase diodebridge rectier, one can apply the Kirchho's second law to the circuit of Fig. 4.1, and
derive the following equations,
vSA
vL
vA = 0
(4.1)
where:
vL = voltage across the total inductance L = Ls1 + Ls2 ,
i1 = i2 = is .
(4.2)
200
150
100
50
50
100
150
200
t1
t4
10
SA
15
is
20
25
vdc
Time ( ms )
vA = vSA
Ls1
30
dis
dt
35
40
vA
45
50
94
(4.3)
(4.4)
input, vA , and in the point of common coupling (PCC), vP CC , where many other loads may
4 As a matter of fact, many DC-DC electronic converters rely on the inductor physical properties during
switching transients, to either step voltages up (boost) or step voltages down (buck).
From Fig. 4.4 it can be seen that the inductances in
uence the rate of change (and consequently the waveshape) of the current, such that is starts to grow exponentially, reaching a
maximum peak value when the voltage across the inductance is zero, after which vL changes
its polarity (because the derivative of the current changes its signal 4 ) physically trying to
keep the current
owing until the current nally becomes zero. Following the equations 4.3,
4.4 and Fig. 4.4 one can easily understand the voltage waveforms distortions in the rectier
vP CC = vSA
dis
dt
Figure 4.2: Current drawn from the source by a single-phase diode-bridge rectier.
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
Current ( A )
Voltage ( V )
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
95
30
25
RMS Current ( A )
20
15
10
15
20
25
30
35
40
45
50
Harmonic Order
Figure 4.3: Harmonic amplitude spectrum of the current drawn from the source by a single-phase diodebridge rectier.
be connected and may be aected by the waveform voltage distortion. Fig.4.5 presents the
harmonic amplitude spectrum of the voltage at the PCC.
In a four-wire, three-phase system with \well balanced" loads, composed by single-phase
diode bridge rectiers, as illustrated in Fig. 4.6, the current
owing through the neutral
conductor has mainly third harmonic components (i.e, 180Hz, for a 60Hz fundamental) as
shown in Fig. 4.7 and Fig. 4.8. Observe that the third harmonic component of the neutral
current in Fig.4.8 is approximately 3 times the third harmonic component of the current of
a single-phase diode-bridge rectier as illustrated in Fig.4.3. These high currents may cause
overheating in the neutral conductor (usually designed with a smaller cross section than the
phase conductors), thereby creating a potential hazardous risk, such as re.
The widespread use of this type of inexpensive rectier, usually as the front end of low
power appliances, such as television sets, computers and compact
uorescent lamps (CFL)
200
150
100
50
50
100
150
200
vSA
vLs1
vA
vPCC
Time ( ms )
10
12
14
5 The help of Mr. Kenneth Wicks in doing this measurement is gratefully acknowledged.
16
96
(which are used more and more because of its higher energy eciency), can have a cumulative
negative eect on the quality of power supplied to residential, commercial and industrial
customers [105] . For example, Fig. 4.9 and Fig. 4.10 present the voltage waveshape
measured with a digital oscilloscope at the outlet of the Power Electronics Laboratory of
the Department of Electrical and Computer Engineering at UBC, Vancouver, B.C., Canada
5
. The waveshape distortion is probably caused by the large number of computers in the
building, as well as in the entire university. It also aects, presently with minor consequences,
all the other loads in the building supplied from the same common bus, and eventually, also
propagates through the BC Hydro electric supply system. Fig. 4.11 and Fig. 4.12 show
the results of a Fourier analysis of the outlet voltage curve (the DC component present in
the harmonic amplitude spectrum at Fig. 4.11 might be caused by inaccuracies or dc osets
in the measuring equipment). Although, according to the present standards [48], [49], such
Figure 4.4: Current through and voltage across the total inductance, and voltage waveform distortion at
the point of common coupling (PCC).
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
Current ( A )
Voltage ( V )
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
97
100
90
Percentage of Fundamental ( % )
80
70
60
50
40
30
20
10
10
15
20
25
30
35
40
45
50
Harmonic Order
Figure 4.5: Harmonic amplitude spectrum of the voltage waveform distortion at the point of common
coupling (PCC).
isa
10 -6
f = 60 [Hz]
isc
diode
bridge
diode
bridge
iNEUTRAL
diode
bridge
isb
vSB
vSA
vSC
Figure 4.6: Four-wire, three-phase system with \balanced" single-phase diode-bridge rectiers.
harmonic voltage distortions are usually within acceptable limits, other sensitive equipment
may be aected, and better alternatives for power conversion are actually available.
200
150
100
50
50
100
150
200
5
10
vSA
15
20
30
Neutral
25
Time ( ms )
35
40
45
50
98
The choice of a power electronic converter is based on its intended application and on the
price. With the introduction of more strict standards for power quality, and with growing
concerns about the dynamic interaction between power electronic devices and the power
system, new technologies with less impact and/or less susceptibility to power disturbances
are gaining market acceptance. Examples are active lters and other dynamic compensating
Single- or three-phase controlled thyristor converters can also adversely aect the quality of power, due to their distorted current waveforms, the notching of the input voltage
waveform caused by the commutation among the thyristors, and the poor power factor.
By using, for example, step-up (boost) DC-DC converters (consisting of IGBT's switching at high frequency with PWM control techniques, free wheeling diodes and inductors),
in connection with diode-bridge rectiers, power factor corrected (PFC) interfaces can be
designed. Such PFC circuits are able to draw almost sinusoidal currents at close to unity
power factor.
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
Current ( A )
Voltage ( V )
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
99
30
RMS Current ( A )
20
15
10
10
15
20
25
30
35
40
45
50
Harmonic Order
Figure 4.8: Harmonic amplitude spectrum of the current
owing through the neutral conductor.
devices.
For a proper identication and solution of power quality problems, transient and steadystate analysis are needed, which include models not only for the power electronics but also
the power system part, because of the dynamic interaction between them. Power electronics
based loads can either be the cause of problems in the power system, or they can be adversely
aected by electromagnetic transient phenomena coming from the power system. Reference
[24] presents the fundamental denition of electromagnetic phenomena aecting the electric
power quality, with realistic cases and practical monitoring results and examples of power
disturbances, such as:
Transient overvoltages;
Harmonics;
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
100
Phase Voltage
200
150
100
Voltage ( V )
50
50
100
150
200
10
12
14
16
time ( ms )
Figure 4.9: Voltage waveshape measured at the outlet of the Power Electronics Laboratory of the Department of Electrical and Computer Engineering at UBC, Vancouver, B.C., Canada.
Voltage sags are by far the most common cause of disruption of operation of power electronics based loads, such as electronically controlled motor drives. Many industrial processes
(e.g. pulp and paper, textile, automotive, etc.) rely on accurate speed and torque control
through the use of power electronics, and thus become more or less vulnerable and susceptible to power quality problems depending on the sensitivity of these devices. This is actually
at the heart of many power quality problems!
It is also common that utility capacitor switching creates high frequency transients,
which may propagate through the distribution system and cause amplied transient voltage
oscillations in low voltage power factor capacitor banks in industry [106]. The commutation
of thyristors in current source inverter (CSI) adjustable speed drives (ASD) in industrial
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
101
Phase Voltage
200
150
100
Voltage ( V )
50
Fundamental Voltage
50
Harmonic Distortion
100
150
200
10
12
14
16
time ( ms )
Figure 4.10: Measured voltage waveshape, its fundamental component and its harmonic distortion.
plants may excite natural resonance modes of weak distribution and other industrial systems,
causing high frequency oscillations in the voltages and, consequently, \nuisance tripping" of
sensitive loads. [107].
Resonances tend to occur more frequently as more power factor and voltage support
capacitor banks are used in the system, mainly to control voltage on transmission or distribution lines. Typically, the 5th order harmonic current commonly \injected" into the power
system by traditional power electronics converters, has the potential to cause problems, such
as capacitors failures. The importance of power quality then depends on its economic impact on the industry, the utility, the society, and the country. Appropriate means to predict
problems in the early design stage or to diagnose and mitigate problems in existing systems
becomes a very important task of \power quality engineers".
EMTP-based programs, because of many available computer models for power systems
and power electronics, have become a necessary engineering tool for the evaluation of the
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
102
Fundamental = 100%
Percent of Fundamental ( % )
1.5
0.5
10
15
20
25
30
35
40
45
50
Harmonic Order
4.1. Dynamic Interaction between Power Electronic Devices and Power Systems
103
100
degrees ( )
50
50
100
150
10
15
20
25
30
35
40
45
50
Harmonic Order
Figure 4.12: Phase-angle of the harmonic components of the outlet waveshape voltage.
104
The problem of harmonic analysis in power systems is usually studied with steady-state
solution techniques, which use linear solutions at the harmonic frequencies. The characteristic harmonic spectra of non-linear loads are assumed to be known, and are modeled
as current sources at the respective harmonic frequency. In reality, the harmonic current
sources are not exactly known, because they depend on the behaviour of the power system
as well. For example, harmonics from transformer saturation clearly depend on the voltage
magnitude and waveform at the transformer terminal. Only time domain simulations of
the EMTP type can address the interaction between the system and the harmonic sources,
which can result in non-characteristic harmonics as well. Time domain simulations can also
be useful to develop other types of power quality studies, such as fault analysis, transient
impulses caused by switching utility capacitor banks, diagnosing the eects of special loads
into the system, troubleshooting the failure of sensitive loads, evaluating the application of
\Custom Power" devices as solutions to power quality problems, etc.
This section presents an application of the EMTP in a distribution system study, where
the harmonics injected into a distribution feeder by induction furnaces were the prime concern for power quality. It is taken from a harmonic problem experienced by ELEKTRO Eletricidade e Servicos S. A. 6 , an electric utility in the southeast of Brazil. The three-phase
power system with a digital model of the induction furnace as a power electronic load was
6 ELEKTRO - Eletricidade e Servicos S. A. , Rua Ary Antenor de Souza, 321 - Jardim Nova America,
CEP 13053-024, Campinas, S. P. , BRAZIL.
105
simulated with the EMTP over a time span which was long enough to reach steady state.
Voltage and current waveforms were then analyzed with a Fourier analysis program to obtain
the harmonic content of the distorted waveforms. Field measurements at the point of common coupling between the utility and the industry are presented as well. Electromagnetic
transients programs are more accurate in representing nonlinear eects of the supply system,
and allow more detailed modelling of power electronic loads and devices, than steady-state
harmonic programs. Power electronics and power quality have such a strong correlation that
can only be fully described and analyzed with the use of time domain simulation techniques.
Induction heating has gained wide acceptance in industry because this type of heating process is considered clean, quick and ecient. On the other hand, the use of power
electronic devices for induction heating introduces harmonic currents, causing voltage distortions in the electric supply network. The dynamic interaction of these harmonics with the
electric system in terms of system conguration, loading and other conditions, may result
in linear resonances, or even in undesirable steady-state conditions, which all can result in
misoperation, failure and life reduction of equipment, with consequent economical losses.
Induction furnaces are power electronics-based loads, where the heat in the electrically
conducting workpiece to be melted is produced by circulating currents through electromagnetic induction. Series or parallel-resonant inverters are typical congurations used to supply
energy to the induction coil, at a selected frequency, which can be in a range varying from
the power system frequency to a few hundred kilohertz [94]. Figs. 4.13 (a) and (b) illustrate
an induction furnace in operation.
The operation of these induction furnaces has produced distortions in the current and
voltage waveforms, and has created incompatibility problems between these special loads and
other sensitive loads connected on the same distribution feeder. Changes in the conguration
of the power supply system as well as application of passive lters, have minimized the
eects, but have not completely eliminated the harmonics power quality problem. Field
measurements have been made for dierent operating conditions to evaluate the eectiveness
of the already installed harmonic passive lters.
(a)
106
(b)
Figure 4.13: (a) Metal melting by an induction furnace. (b) Induction furnace operation.
107
itor bank was connected to a dierent bus bar from the feeder that supplies the induction
furnaces customer.
In August 1995, new measurements were made at the substation, especially on the feeder
supplying the induction furnaces [108]. Fig. 4.14 shows the feeder current at the moment
of maximum harmonic distortion. Sometimes later, the customer installed 4th and 5th order
harmonic tuned passive lters, with an economic motivation imposed by a new Brazilian
power factor legislation, which essentially changed from a minimum monthly average power
factor of 0.85 to a minimum hourly average of 0.92 (inductive limit from 06:00am to 12:00
midnight and capacitive limit from 12:00 midnight to 06:00am). Measurements were made in
1998 to verify the system overall performance on this feeder and in the respective distribution
substation.
LIM_IV
Phase B Current
SS Wave
600
Fund
192.7
RMS
235.6
400
200
CF
Amps
0
-200
Min
-400
-600
0
20
40
60
80
100
120
140
Time (mSeconds)
2.091
-492.6
Max
487.2
THD
70.41
HRMS
135.7
TIF/IT 101821
120
100
80
Amps
60
40
20
0
0
10
15
20
25
30
Harmonic
BMI/Electrotek
Figure 4.14: Current measurements in a distribution feeder supplying induction furnaces at the time of
maximum voltage distortion.
108
channels. With this equipment, it is possible to monitor power quality phenomena such as
voltage sags, voltage swells, outages, cold load pick-up, transient impulses, waveshape faults,
harmonic distortions and frequency deviations [15], [108], [43].
The following gures illustrate some of the harmonic measurements made in March/April
1998. Fig. 4.15 (a) shows the phase \A" measured current and Fig. 4.15 (b) shows the measured voltage between phases A and B, with their respective harmonic amplitude spectrum.
Both measurements were taken at a particular time when the induction furnace operation
was with the 4th and 5th order harmonic passive lters turned o (see Fig. 4.17).
400
300
200
100
0
-100
-200
-300
-400
10
20
30
40
50
Time (mSeconds)
60
70
35
30
25
20
15
10
5
0
20000
10000
0
-10000
-20000
-30000
% Fund
Amps
Fund 198.9
204.8
RMS
1.531
CF
Min -308.5
313.4
Max
18.89
THD
HRMS 37.58
TIF/IT52660
Amps
10
15
20
Harmonic
(a)
25
30
BMI/Electrotek
10
20
30
40
50
Time (mSeconds)
60
70
4
3.5
3
2.5
2
1.5
1
0.5
0
10
15
20
Harmonic
25
Fund 13557
13592
RMS
1.489
CF
Min -20237
Max
19805
THD
7.172
HRMS 972.3
TIF/IT 234.9
30
BMI/Electrotek
(b)
Figure 4.15: (a) Phase \A" current measured with harmonic passive lters turned o. (b) Phase-to-phase
\A-B" voltage measured with harmonic passive lters turned o.
Figs. 4.16 (a) and (b) illustrate the same measurements, but during an operating condition when the 4th and 5th harmonic passive lters were turned on (see Fig. 4.18).
The melting process inside the induction furnaces aects the total harmonic distortion
(THD) of the measured voltage at the point of common coupling very strongly. Fig. 4.17
shows the THD historic trend during one regular day of work, with the 4th and 5th harmonic
passive lters turned o from about 12:00 midnight to 06:00am. This THD trend presents a
characteristic variation along the day, depending on the load cycle at the industry. During
the time when the harmonic lters were turned o (from about 12:00 midnight to 06:00am)
10
20
15
20
Harmonic
30
40
50
Time (mSeconds)
10
60
25
70
Fund 198.7
RMS
199.0
CF
1.515
Min -301.5
Max
294.2
THD
5.585
HRMS 11.10
TIF/IT24637
BMI/Electrotek
30
20000
15000
10000
5000
0
-5000
-10000
-15000
-20000
-25000
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
10
(b)
15
20
Harmonic
30
40
50
Time (mSeconds)
10
20
25
60
Fund 14011
RMS
14037
CF
1.431
Min -20083
Max
19896
THD
2.589
HRMS 362.7
TIF/IT 115.1
109
BMI/Electrotek
30
70
300
200
100
0
-100
-200
-300
-400
7
6
5
4
3
2
1
0
0
(a)
Volts
13:19:03
14:29:04
15:39:04
17:58:59
19:08:59
20:19:00
21:29:01
22:38:53
23:48:54
00:58:55
02:08:56
03:18:48
04:28:49
05:38:50
06:48:51
07:58:51
09:08:45
10:18:46
11:28:46
THD (%)
8
7
6
5
4
3
2
1
0
12:09:10
From these charts it is possible to conclude that passive lters can minimize the harmonic
voltage distortions, but not eliminate them completely. New power electronic devices such
Figure 4.17: THD harmonic trend, with harmonic passive lters turned o from 12:00 midnight to 06:00am.
16:49:05
it is easy to see a substantial increase in the THD values. Fig. 4.18 presents the THD historic
trend for another operating condition, with the 4th and 5th harmonic lters turned on during
all the time.
Figure 4.16: (a) Phase \A" current measured with harmonic passive lters turned on. (b) Phase-to-phase
\A-B" voltage measured with harmonic passive lters turned on.
% Fund
16:40:48
17:50:49
19:00:50
20:10:42
21:20:43
22:30:44
23:40:44
00:50:38
02:00:39
03:10:39
04:20:40
05:30:41
14:20:54
15:30:47
THD (%)
8
7
6
5
4
3
2
1
0
13:10:54
06:40:33
07:50:34
09:00:35
10:10:36
11:20:28
Table 4.1: Global harmonic distortion limits for the system voltages recommended in Brazil.
110
As far as the author knows, new Brazilian legislation about harmonics and other power
quality phenomena are currently under development. Some recommendations have been used
to guide utility system planners and operators in supplying power to special loads. Reference [50] for example, presents some criteria and global voltage harmonic limits (Table 4.1),
based on the experience of Brazilian power utilities, as well as on standards from CIGRE
(Conference Internationale des Grands Reseaux Electriques a Haute Tension - International
Conference on Large High Voltage Electric Systems), IEC (International Electrotechnical
Commission) and IEEE (The Institute of Electrical and Electronics Engineers).
as active lters, might be able to dynamically compensate the distortions and improve the
quality of power at the interface of industrial and power systems.
Figure 4.18: THD harmonic trend, with harmonic passive lters turned on all the time.
12:00:53
111
112
113
138 kV
25/30MVA
138/13.8kV
10.2% (25MVA)
25/30MVA
138/13.8kV
9.72% (25MVA)
Bus I - 13.8kV
Bus II - 13.8kV
9.0MVAr
13.8kV
477.0 ACSR - 2km
PCC
3.0MVA
13.8/0.48kV
5.9%
0.48kV
Harmonic
Passive Filters
Induction
Furnaces
4th
5th
114
Ld
Lc
VA
INVA
VB
Lr
Cr
VC
Rload
INVB
4
INDUCTION
FURNACE
NEG
45
40
35
30
25
20
15
10
5
0
0
50
100
150
200
250
300
350
400
450
500
frequency ( Hz )
Figure 4.21: Amplitude of the positive sequence system impedance at the PCC with harmonic lters.
the comparison between the simulated and the measured results.
Figs. 4.25 (a) and (b) show the current and voltage waveforms, but for induction furnace
115
100
80
60
40
20
0
20
40
60
80
100
0
50
100
150
200
250
300
350
400
450
500
frequency ( Hz )
Figure 4.22: Phase angle of the positive sequence system impedance at the PCC with harmonic lters.
operation with the 4th and 5th order harmonic passive lters turned on. This operating
condition corresponds to the time of minimum total harmonic distortion (THD) measured
at the point of common coupling (PCC), but in this case with the harmonic passive lters
turned on (see Fig. 4.18). Both operating conditions have approximately the same value
of fundamental current (199A RMS at the phase-to-phase RMS voltage of 13.8kV). Again,
Fig. 4.16 is repeated here as Fig. 4.26 to facilitate the comparison between the simulated
and the measured results.
Voltage ( V )
0
200
10
20
30
40
50
60
Peak Amplitude ( A )
Time ( ms )
40
35
30
25
20
15
10
5
0
0
10
15
20
25
30
Harmonic Order
x 10
2
1
0
1
2
3
0
70
Current ( A )
200
400
0
116
10
20
30
40
50
60
70
Time ( ms )
4
3.5
3
2.5
2
1.5
1
0.5
0
0
10
15
20
25
30
Harmonic Order
(a)
(b)
Figure 4.23: (a) Phase \A" current simulated with harmonic passive lters turned o. (b) Phase-to-phase
\A-B" voltage simulated with harmonic passive lters turned o.
35
30
25
20
15
10
5
0
10
20
30
40
50
Time (mSeconds)
60
70
20000
10000
0
-10000
-20000
-30000
% Fund
Fund 198.9
204.8
RMS
1.531
CF
Min -308.5
313.4
Max
18.89
THD
HRMS 37.58
TIF/IT52660
Amps
400
300
200
100
0
-100
-200
-300
-400
Amps
10
15
20
Harmonic
(a)
25
30
BMI/Electrotek
10
20
30
40
50
Time (mSeconds)
60
70
4
3.5
3
2.5
2
1.5
1
0.5
0
10
15
20
Harmonic
25
Fund 13557
13592
RMS
1.489
CF
Min -20237
Max
19805
THD
7.172
HRMS 972.3
TIF/IT 234.9
30
BMI/Electrotek
(b)
Figure 4.24: (a) Phase \A" current measured with harmonic passive lters turned o. (b) Phase-to-phase
\A-B" voltage measured with harmonic passive lters turned o.
Voltage ( V )
100
0
100
200
10
20
30
40
50
60
Peak Amplitude ( A )
Time ( ms )
7
6
5
4
3
2
1
0
0
10
15
20
25
30
Harmonic Order
x 10
2
1
0
1
2
3
0
70
Current ( A )
200
300
0
117
10
20
30
40
50
60
70
Time ( ms )
1.4
1.2
1
0.8
0.6
0.4
0.2
0
0
10
15
20
25
30
Harmonic Order
(a)
(b)
Figure 4.25: (a) Phase \A" current simulated with harmonic passive lters turned on. (b) Phase-to-phase
\A-B" voltage simulated with harmonic passive lters turned on.
Some care must be taken when dening the time step size (which corresponds to the
inverse of the sampling frequency), for the transient simulation. Relatively large step sizes
introduce errors into the results of a time-domain simulation and consequently into the post-
300
200
100
0
-100
-200
-300
-400
10
20
30
40
50
Time (mSeconds)
60
70
7
6
5
4
3
2
1
0
20000
15000
10000
5000
0
-5000
-10000
-15000
-20000
-25000
% Fund
Amps
Fund 198.7
RMS
199.0
CF
1.515
Min -301.5
Max
294.2
THD
5.585
HRMS 11.10
TIF/IT24637
Amps
10
15
20
Harmonic
25
30
10
20
30
40
50
Time (mSeconds)
70
1.4
1.2
1
0.8
0.6
0.4
0.2
0
10
15
20
Harmonic
BMI/Electrotek
(a)
60
25
Fund 14011
RMS
14037
CF
1.431
Min -20083
Max
19896
THD
2.589
HRMS 362.7
TIF/IT 115.1
30
BMI/Electrotek
(b)
Figure 4.26: (a) Phase \A" current measured with harmonic passive lters turned on. (b) Phase-to-phase
\A-B" voltage measured with harmonic passive lters turned on.
118
400
200
0
200
400
0.05
40
0.052
0.054
0.056
0.052
0.054
0.056
0.058
0.06
0.062
0.064
0.066
0.058
0.06
0.062
0.064
0.066
Time ( ms )
Ideal Compensation Current for Phase "A"
Current ( A )
30
20
10
0
10
20
30
40
0.05
Time ( ms )
Figure 4.27: Instantaneous ideal compensation current to be \injected" by a shunt active lter.
119
(d) Conclusions
This section presented an application of the electromagnetic transients program (EMTP)
to the analysis of a power quality issue in a distribution system, where the voltage and current
harmonic distortions were produced by induction furnaces. The history of this harmonic resonance problem was described. A digital model of the induction furnace as a current-source,
parallel-resonant inverter load, together with three-phase representations of the supplying
distribution system were used in a time-domain simulation. After reaching a time considered
as steady state, the simulated results were processed through a Fourier analysis program to
obtain the harmonic contents of the voltage and current waveforms at the point of common coupling. Field measurements were also presented. A comparison between the actual
measurements and the EMTP simulations is presented in Table 4.2 and Table 4.3. The dierences in the values of the total harmonic distortion (THD) and the telephone in
uence factor
(TIF) are possibly due to simplications or lack of realistic data in the digital modelling.
The knowledge of detailed manufacturer data of the induction furnaces and their operating conditions, such as the natural resonant frequency, would allow the improvement of the
EMTP simulation. Considering unbalanced conditions could also improve the simulation
results. Both were unfortunately not available.
Table 4.2: Comparison between eld measurements and EMTP simulation results for the operating condition with the harmonic passive lters turned OFF.
Phase \A" Current [A]
Parameters Measured Simulated
Fund
198.9
201.8
RMS
204.8
204.2
CF
1.531
1.491
Min
-308.5
-304.4
Max
313.4
304.4
THD [%]
18.89
15.42
HRMS
37.58
31.11
TIF/IT
52,660
64,416
Error %
1.5
0.3
2.6
1.4
2.9
18.4
17.2
22.3
EMTP-based simulations can be useful tools for harmonic analysis, based on the fact that
very detailed eects can be taken into account. Once the system is modelled for an EMTP-
120
Table 4.3: Comparison between eld measurements and EMTP simulation results for the operating condition with the harmonic passive lters turned ON.
Phase \A" Current [A]
Parameters Measured Simulated
Fund
198.7
190.2
RMS
199.0
190.4
CF
1.515
1.474
Min
-301.5
-279.5
Max
294.2
280.7
THD [%]
5.585
5.110
HRMS
11.10
9.72
TIF/IT
24,637
36,910
Error %
4.3
4.3
2.7
7.3
4.6
8.5
12.4
49.8
based software, any type of studies can be performed. The new models developed in this
thesis project, hopefully will contribute to make EMTP-based programs more valuable tools
for electric utility companies and industrial customers in evaluating power quality problems.
4.2.2
Voltage sags or voltage dips are short duration variations in the supply voltage, caused
by faults in transmission lines or in parallel distribution feeders, or caused by the start-up
of large induction motors or other types of sudden load variations. The majority of power
quality problems are associated with voltage sags, which are very common in today's electricity industry. The ride-through characteristics of modern electronic and computer-controlled
loads are very sensitive to short duration variations in the supply voltage. An entire process
may be shut down when the voltage sags momentarily. The equipment tolerance characteristics to voltage sags vary very much among equipment manufacturers. Moreover, in
most cases the equipment ride-through characteristics are not known. For these reasons,
the CBEMA curve [51] (or the ITIC curve) has been widely used as a rst reference for
power quality studies related to short duration voltage variations. Fig. 4.28 presents actual
measurements of voltage sag phenomena with an overlay of the CBEMA curve 7 .
7 \We typically employ the curve only from 0.1 cycles and higher due to limitations in power quality
monitoring instruments and dierences in opinion over dening the magnitude values in the subcycle time
frame." From [24], pp. 37-38.
121
The power system characteristics at the point where the sensitive load is connected is
another important issue for possible mitigation of voltage sag problems. Power system
protection and operating practices may also aect the success or failure of loads which are
sensitive to voltage disturbances.
Fig. 4.29 (a) shows a voltage sag phenomenon caused by a single-line-to-ground fault in a
distribution feeder, which is in parallel to a feeder supplying a \PVC" pipe (and other plastics
derived products) manufacturer with sensitive loads. The industrial process, controlled by
DC drives, is stopped if the voltage sags at the point of common coupling to less than 90%
of the nominal voltage during a time greater than 18 cycles (300ms).
Fig. 4.29 (b) shows the simulation results using MicroTran. The simulation does not
match the measurements exactly, because the dynamic behaviour of industrial loads (typically induction motors) were not included in the simulation model. Nevertheless, EMTPbased programs have the
exibility to include aggregated load models [109]. More research,
however, seems to be needed for the accurate representation in EMTP-based programs of
the dynamic behavior of loads, which is beyond the scope of this thesis project. Custom
Power Controllers such as the dynamic voltage restorer (DVR), is a promising solution for
the mitigation of voltage sag phenomena.
4.2.3
122
300
Max 106.0
Min
73.50
250
% Volts
200
150
100
50
0
0.001 0.01
0.1
10
100
1000 10000
Time (Cycles)
BMI/Electrotek
Figure 4.28: Voltage sag measurements (%RMS versus time duration) with an overlay of the CBEMA
curve. For time durations less than 1 cycle the equipment seems to measure peak values.
IEC Standard 1000-4-15 provides the specications for a
ickermeter with \lamp-eye-brain"
frequency response to light
ickering eects. Fig. 4.30 shows the simulated instantaneous
voltage for a power electronics controlled welding machine . The operating duty cycle, controlled by the semiconductor ring angle, results in a modulating frequency of approximately
7Hz for this case, as shown in Fig. 4.31, which causes the visual light
icker. Cost-eective
solutions for voltage
uctuation problems are usually related to changes in the load duty
cycle, when this does not aect the industry productivity or the quality of the manufactured product. The application of reactive dynamic compensation through power electronic
devices, such as the distribution static synchronous compensator (D-STATCOM) can eectively mitigate this type of power quality problem [110].
% Volts
110
105
100
95
90
85
80
75
70
123
Trigger
Duration
0.233 Sec
Min 73.50
Ave 89.70
Max 100.8
0
0.1
0.2
0.3
0.4
0.5
Ref Cycle
30664
Time (Seconds)
% Volts
100
75
50
25
0
-25
-50
-75
-100
0
25
50
75
100
125
150
175
200
Time (mSeconds)
Uncalibrated Data
(a)
4
2.5
x 10
2
1.5
Voltage ( Volts )
1
0.5
0
0.5
1
1.5
2
2.5
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
Time ( s )
(b)
Figure 4.29: (a) Phase-to-phase \A-B" measured voltage sag. (b) Phase-to-phase \A-B" simulated voltage
sag.
124
950
Voltage ( Volts )
900
850
800
750
700
650
600
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Time ( s )
1000
500
0
500
1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Time ( s )
Harmonic Amplitude Spectrum
0.9
% Volts
1.5
1
0.5
0
40
45
50
55
60
65
70
75
80
Frequency ( Hz )
125
This section provides more explanations about the theory presented in the previous chapters, with some applications to test cases.
(a) The Eect of a One Time Step Delay in the Solution of EMTP-based Simulations
Reference [66] presents some interesting cases where the one time step delay in the solution
of control and system equations, or the internal delays inside TACS, give wrong or inaccurate
simulation results. For example, the simulation of a transfer function with poles on the
imaginary axis of the complex plane is very sensitive to time delays, which can \move" the
poles to the right half of the complex plane, resulting in instability. To illustrate this, Fig.
4.32 presents a control block diagram of a transfer function with a second order dierential
equation. The simultaneous solution is shown in Fig. 4.33. The intentional introduction of
a one time step delay in the control system, as illustrated in Fig. 4.34, leads to an unstable
resonance condition, as shown in Fig. 4.35. These results emphasize the \importance of a
simultaneous solution approach", which is critical in some cases for the correct simulation
of control and power system equations with EMTP-based programs.
v IN
126
1
_____
s
v OUT
1.0 [V]
1
_____
s
Figure 4.32: Control block diagram of a second order dierential equation with poles on the imaginary
axis of the complex plane.
8
vout
vin
Voltage ( V )
10
20
30
40
50
60
70
Time ( s )
80
v IN
127
1
_____
s
v OUT
1.0 [V]
1
_____
s
delay
1 t
Figure 4.34: Introduction of a one time step delay in the control block diagram.
8
out
vin
Voltage ( V )
10
20
30
40
50
60
70
80
Time ( s )
Figure 4.35: Solution of system with unstable resonance oscillations caused by the introduction of one
time step delay.
128
129
K=
!n =
(4.5)
KS
KS 2!H
(4.6)
(4.7)
130
Tes
KS
Tm
Ta
1
_____
2H s
0
_____
s
- 0.1 p.u.
KD
Ted
Figure 4.36: Classical linearized \swing equation", used in power system small-signal stability studies of
a single machine connected to an innite bus.
0.1
Ta
0.05
Te
5*
0
0.05
Te
Tm
0.1
0.15
0.2
0.25
0
5
Time ( s )
10
Figure 4.37: Simulation results of the synchronous machine rotor angle deviation, in the presence of a
positive damping torque coecient.
131
200
150
100
50
50
100
150
10
Time ( s )
Figure 4.38: Simulation results of the synchronous machine rotor angle deviation, in the presence of
negative damping torque coecient.
Tm
K n2
_________________
s2 + 2 n s + n2
- 0.1 p.u.
Figure 4.39: Canonical second order transfer function representation of the single-machine innite bus
system.
132
(c) Voltage and Current Sensors, Control Devices, Function Blocks and Digital
Logic Gates
The solution method for voltage and current sensors, control devices, function blocks and
digital logic gates follows the procedures for dependent sources presented in Chapter 2. For
example, the simultaneous solution through the compensation method for a generic linear
or nonlinear voltage-controlled voltage source can be used
to sense a voltage signal, with the compensating current at the controlling branch j
set to zero, i.e., ij = 0;
then to dene any linear or nonlinear voltage function at the controlled branch k, i.e.,
vk = f (vj );
and nally to use the Newton-Raphson algorithm presented in Section 2.2.5, (Fig.
2.19) to calculate a solution for the compensating currents ij and ik .
If more than one control voltage signal needs to be sensed, their respective compensating
branch currents are \simply set to zero". On the other hand, if currents have to be sensed,
their respective branch voltages can be set to zero. Since control signals are usually voltage
signals (or current signals converted to voltage signals), a current sensor can be represented
by a current-controlled voltage source.
The implementation of control devices (such as transport or time delay blocks, comparators, zero crossing detection and generation of gate ring signals, Clark transformation,
Park transformation, etc.), FORTRAN function blocks (such as SIN, COS, TAN, ASIN,
ACOS, LOG, EXP, SQRT, MULTIPLICATION, DIVISION, etc.) and digital logic gates
(AND, OR, NOT, NAND, NOR, etc.) uses the same concept of the simultaneous solution
for voltage-controlled voltage sources presented before.
The implementation of a simultaneous solution for limiters requires special attention, as
for example in the case of the widely used PI controllers with non-windup limiters. Reference
[66] clearly explains the problem and the solution for a correct EMTP-based simulation.
133
Special functions, as for example for the evaluation of average values (e.g. \rolling average
power" presented in [66]), root mean square (RMS) values, instantaneous total harmonic
distortion (THD) values, etc. can be easily modelled with control blocks, and with subcircuit implementation.
4.3.2
The cases presented in this section are based on references [65], [93], [89], [94], [104], [95].
zero crossing detection and generation of ring pulses with an instantaneous updating
of the controlling ring angle . The controller is enabled by a voltage signal of \1
volt" and disabled by a voltage signal of \0 volt";
The voltage sensor in Fig. 4.40 is modelled with the equations for an \ideal" voltagecontrolled voltage source (VCVS) presented in Section 2.2, i.e.:
ij = 0
k + r
vOP ENj + vOPEN
j1
rkj A
::: + rjj A ij + rjk
iM = 0
::: + rjM rkM
A
(4.8)
rk1 i + :::
1
rkkA i + :::
k
A
(4.9)
where A = 1. For this case, the synchronizing signal for the gate ring controller could have
been sensed directly from the excitation source vSA , without the need for a voltage sensor,
which was included here for completeness of the test case.
134
ij = 0
(4.10)
ik = 0
(4.11)
il = 0
(4.12)
135
iLOAD
vSA
vLOAD
2.5
5 mH
vGATE
vSYNCHR
Voltage Sensor
vALPHA
f=60Hz
pulse width=10
degreees
45.0 [V]
Firing Angle
(degrees)
venable /
disable
1.0 [V]
tstop = 25.0 [ms]
Figure 4.40: Circuit for the dynamic control of the ring angle (\") of a thyristor.
LOAD
LOAD
25
Time ( ms )
GATE
enable / disable
20
15
10
vsynchr
30
35
40
45
50
136
Figure 4.41: Voltages and currents in a circuit with dynamic control of the ring angle of a thyristor.
Current ( A )
Voltage ( V )
137
(b) Dynamic Control of the Firing Angles of a Three-Phase Six-Pulse ThyristorBridge Rectier
Fig. 4.42 illustrates a phase controlled rectier with a feedback control system based on
the manual of PSIM [65]. \It should be noted that, in PSIM, the power and the control
circuit are solved separately. There is one time step delay between the power and the control
circuit solutions" [65].
The six-pulse ring controller implemented with the models developed in this thesis
project uses a simultaneous solution for the power and control circuit equations. Similarly to
the simple ring controller \" of the previous test case, a multi-terminal voltage-controlled
voltage source is used to model it. The inputs to the six-pulse ring controller are the
synchronizing signal (voltage vAC sensed from the supply system), the dynamic ring angle
(resulting from the ACOS control block, which receives the signal from the limited PI
controller, after comparison with the desired reference voltage for the DC load). From
the gating signal generated to the thyristor with identication number \1", all the other
gating signals are derived sequentially by adding a time delay corresponding to 60 degrees
at the 60Hz frequency. For starting purposes, whenever a ring signal is sent to a particular
thyristor, another \isolated" ring signal is sent to the previous thyristor as recommended
in [88].
Fig. 4.43 presents the EMTP simulated voltages and currents, using a t = 16:6667s
and with a dynamic control of the ring angles of the three-phase six-pulse thyristor-bridge
rectier. Fig. 4.44 shows the dynamic behavior of the control variables, emphasizing the
ring control signal , and Fig. 4.45 illustrates the dynamics of the voltage control signals
at the output of the proportional-integral (PI) control block and the limiter control block.
138
5 mH
G1
G3
iLOAD
G5
f = 60 [Hz]
vDC
vLOAD
vA
vB
400 F
vC
G4
G2
G6
vAC
vSYNCHR
1
vREF
100 [V]
vERR
PI
KP=0.01
KI= 1.0
vSENS= vLOAD
vPI
vLIM
ACOS
-1
1 [V]
vENABLE /
vALPHA
G1
G2
G3
G4
G5
G6
DISABLE
Figure 4.42: Circuit for the dynamic control of the ring angles of a three-phase six-pulse thyristor-bridge
rectier.
150
100
50
50
100
150
vDC
iLOAD
vA
10
Time ( ms )
LOAD
15
vsynchr
20
25
139
Figure 4.43: Voltages and currents with dynamic control of the ring angles of a three-phase six-pulse
thyristor-bridge rectier.
Current ( A )
Voltage ( V )
120
100
80
60
40
20
20
5
10
15
20
ALPHA
25
Time ( ms )
30
35
vREF
ERR
45
vSENS=VLOAD
40
50
140
Figure 4.44: Dynamic control of the ring angles of a three-phase six-pulse thyristor-bridge rectier.
Current ( A )
Voltage ( V )
1.4
1.2
0.8
0.6
0.4
0.2
0
5
PI
10
15
20
vLIM
25
Time ( ms )
30
35
40
45
50
141
Figure 4.45: Dynamic voltage control signals at the output of the proportional-integral (PI) and the limiter
control blocks.
Current ( A )
Voltage ( V )
142
143
to-line voltage generated by the three-phase PWM voltage source inverter (VSI) is shown in
Fig. 4.51.
Most of the advanced Custom Power Controllers [60] (and active lters [111]) apply
this type of converter to synthesize voltages or current waveforms according to the desired
\dynamic reference modulating signal". When a current is to be synthesized, dynamic
hysteresis current-band PWM converters are used [60] 9 . Therefore, the models developed in
this thesis, will hopefully be useful for the accurate EMTP-simulation of a variety of existing
and new power electronic devices, especially those aimed at improving the quality of power
in utility and industrial systems.
9 The Ph.D thesis \Active Power Line Conditioners" of Dr.-Ing. Mauricio Aredes, can be downloaded by
IGBT1
450 [V]
144
IGBT3
IGBT5
iSA-NEUTR
vSA
vSB
vSC
vNEUTR
iSB-NEUTR
iSC-NEUTR
3.87
7.7 mH
10 9
IGBT4
IGBT6
IGBT2
vA
0.8 [V]
vB
NOT
0.8 [V]
vC
0.8 [V]
f = 60 [Hz]
COMPARATOR
vTRI
1 [V]
f = 1500 [Hz]
phase = -180 [degrees]
Figure 4.46: Circuit for the dynamic control of three-phase PWM voltage source inverter (VSI).
1.5
0.5
0.5
1.5
10
TRI
Time ( ms )
15
20
25
145
Figure 4.47: Phase \A" modulation and triangular carrier waveforms for generation of gating signals
through sinusoidal pulse width modulation (PWM).
Current ( A )
Voltage ( V )
600
400
200
200
400
600
5
vSA
10
Time ( ms )
15
20
25
146
Figure 4.48: Node voltage \vSA " generated by a three-phase PWM voltage source inverter (VSI).
Current ( A )
Voltage ( V )
600
400
200
200
400
600
5
vSA vNEUTR
Time ( ms )
5 * iSANEUTR
10
15
20
25
147
Figure 4.49: Voltage across the load \vSA NEUTR " and current supplied to the load by a three-phase
PWM voltage source inverter (VSI).
Current ( A )
Voltage ( V )
60
40
20
20
40
60
5
SANEUTR
10
15
SBNEUTR
Time ( ms )
iSCNEUTR
20
25
Figure 4.50: Load currents supplied by a three-phase PWM voltage source inverter (VSI).
Current ( A )
Voltage ( V )
148
149
600
vSA vSB
400
Voltage ( V )
200
200
400
600
10
15
20
25
Time ( ms )
Figure 4.51: Line-to-line voltage generated by a three-phase PWM voltage source inverter (VSI).
150
Use of statistical and other advanced data analysis methods to produce meaningful
information;
Comparison against national and international power quality standards, taking into
consideration system dierences and similarities;
In the analysis of power quality phenomena through time and frequency domain EMTPbased simulations, special attention must be paid to:
the simulation step size t, which has to be chosen as a function of the maximum
frequency expected (or of concern) in the simulation. Usually, the time step size is
set to a value at least equal to one tenth of the period of the maximum frequency,
which will result in a 3% error with the trapezoidal integration rule [2], [112]. It is also
recommended that the step size be such that the period of the fundamental frequency
is an integer multiple of t, in order to avoid the generation of non-characteristic
harmonics in the post processing Fourier analysis. For example, if the system nominal
frequency is 60:0Hz and the maximum frequency expected in the transient simulation
151
is in the order of fmax = 6kHz , then the time step size can be calculated as t =
1=(10 fmax ) = 16:66666s;
the selection of appropriate models to represent power system and load components,
especially if frequency dependence has to be taken into consideration in the simulated
phenomena [92];
nonlinearities, which are usually disregarded in many simulations. They can aect the
accuracy of the simulation results, particularly in the case of transformer saturation;
the use of simplied switch models for power electronics devices. This may be justied
to speed up the simulation time for system level studies, but it may also give wrong and
misleading results, especially related to semiconductor commutation phenomena. Also,
the EMTP solution at discrete time intervals t may result in inaccurate turn-on or
turn-o switching times, causing unrealistic high frequency transients in the simulation
of power electronic devices. Backtracking techniques [67], [68] and/or resynchronization techniques ([96] pages 185, 204, 207) or even the Clock Synchronized Structure
Changing Concept (CSSC) [97] can be used to minimize the problem. Interpolation
and/or extrapolation as well as resynchronization techniques seem to be more and more
applied even in the EMTP-based solution of modern control for power electronics systems [98], [99]. Therefore, for better accuracy in EMTP-based simulations of power
electronics, it is much more important to use such techniques than to reduce the time
step size;
numerical oscillations caused by the trapezoidal rule of integration in solving the system
of equations. The use of techniques such as CDA (\Critical Damping Adjustment"
[84], [85]) is eective in the elimination of numerical oscillations. MicroTran has CDA
implemented, but other EMTP versions may not, or may use dierent approaches.
the one time step delay at the interface between the control and power systems solution, as well as other internal time step delays, which may exist in TACS (\Transient
Analysis of Control Systems") and in other software packages. The method SSCPS
(\Simultaneous Solution of Control and electric Power System equations") proposed in
152
this thesis overcomes this problem, and the user only introduces time delays if needed
to represent system physical behaviour.
Chapter 5
Conclusions and Recommendations
for Future Work
HE MAIN GOAL of this Ph.D. thesis project was the development of EMTP-based
models for control and power electronic devices for electric power quality assessment.
EMTP simulations can oer theoretical and practical insights into the evaluation of power
quality, both by time-domain simulation techniques and by frequency-domain simulation
techniques.
154
due to the manufacturing of usually inexpensive power electronic converters. It is a challenging environment for engineering, politics, economics, etc. since in modern human society,
electricity has become a basic commodity, which almost everybody and almost everything
depends on.
The quality of the electric power delivered to customers by utilities may not be acceptable
for some types of sensitive loads, which are typically power electronics and computer-based
loads, particularly in the control of industrial processes. There are cases where the increasing use of power electronics to enhance process eciency and controllability creates power
quality problems. The growing application of shunt capacitors for voltage support, power
factor correction, and system loss reduction, as well as the use of series capacitors (xed or
controlled, for line reactance compensation) will increase the potential risk of transient disturbance amplications and potential electrical and mechanical resonances in the presence
of more and more power electronic devices, and of steam and gas turbines in distributed and
co-generation power plants. As the natural order of the system grows, so grows its ability
to oscillate more! At the same time, new power electronic devices also oer the means for
adequate \power conditioning", to meet the special requirements of electric power quality in
a system.
To evaluate the promising solutions oered with the introduction of more and more power
electronic devices in the transmission and distribution systems, as well as to analyze their
interaction and impact on either the load or the network side, computer programs based on
the EMTP (Electromagnetic Transients Program) are becoming more useful.
The development of new EMTP-based models for more accurate representation of controls
and power electronic devices has been the main subject of this thesis project. The assessment
of electric power quality and the technical impact of power electronic devices on the quality
of power, can hopefully be performed with the models developed in this work.
The main contributions of this Ph.D. thesis project are summarized as follows:
development of a \simultaneous solution for linear and nonlinear control and electric
power system equations" (SSCPS) in EMTP-based programs, through the compen-
155
sation method and the Newton-Raphson iterative algorithm. This solution method
eliminates not only the one time step delay problem at the interface between the solution of power and control circuits, but also all the internal delays, which may exist
in methods based on the transient analysis of control systems (TACS) since 1977 [64].
A \circuit approach" was proposed in this thesis, as an innovative alternative to the
solution presented by A. E. A. Araujo in 1993 [67];
{ linear and nonlinear current and voltage dependent sources (which allow, for
example the modelling of operational ampliers, ideal current and voltage sensors,
etc.);
{ independent current and voltage sources, which can also be connected between
two ungrounded nodes;
{ hard and soft limiters (which can be used to represent nonlinear eects such as
saturation);
{ transfer functions (which allows the simulation modelling of all types of analog
lters and classical control blocks);
{ special control devices (such as time delays, comparators, etc.) and some digital
logic gates (NOT);
156
interaction with a Brazilian utility company and industries for the realization and
analysis of eld measurements of electromagnetic phenomena aecting the quality of
power, such as:
synthesis of simulation guidelines for the evaluation of the impact of power electronic
devices on the quality of power, based on realistic eld measurements and EMTP time
and frequency domain simulations.
development and implementation in EMTP-based programs of methods for the \automatic calculation of initial conditions" in the simultaneous solution of control and
power electronic circuits;
development and implementation in EMTP-based programs of techniques for the calculation of the \frequency response of integrated control and power system equations",
allowing for example, the determination of transfer functions between a generic input
and a generic output;
157
development and implementation in EMTP-based programs of techniques for the accurate solution of \digital control systems";
development and implementation in EMTP-based programs of algorithms for the simultaneous solution of \generic nonlinear dependent sources" [101];
Some of the author's publications in areas related to the thesis topic are listed here for
easy reference:
158
B. D. Bonatto, T. Niimura and H. W. Dommel, \A Fuzzy Logic Application to Represent Load Sensitivity to Voltage Sags", in 8th International Conference on Harmonics
and Quality of Power (8th ICHQP), IEEE/PES, Ed., Athens, Greece, Vol. I, pp.
60-64, October 14-16 1998.
159
S. M. Deckmann, and B. D. Bonatto, \Damping Introduced by Control Means Considering Generator Capability Limits", IFAC Control of Power Plants and Power Systems
(SIPOWER095), Cancun, Mexico, 1995.
S. M. Deckmann, and B. D. Bonatto, Laboratory of Electrical Engineeering I, UNICAMP/FEEC, Faculty of Electrical and Computer Engineering, December 1992, reedited in December 1993, and in 1998. Campinas-SP, Brazil.
Bibliography
[1] H. W. Dommel, \Digital Computer Simulation of Electromagnetic Transients in Singleand Multiphase Networks," IEEE Transactions on Power Apparatus and Systems, vol.
PAS-88, no. 4, pp. 388{399, April 1969.
[2] H. W. Dommel, EMTP Theory Book, Second Edition, Microtran Power System Analysis Corporation, Vancouver, British Columbia, Canada, May 1992. Latest update:
April 1996.
[3] T. Mueller and D. Gra, \The Use of Surge Protection Devices in the PetroleumPetrochemical Industry," IEEE Transactions on Industry Applications, vol. 34, no. 6,
pp. 1351{1358, November/December 1998.
[4] H. W. Dommel, \Techniques for Analyzing Electromagnetic Transients," IEEE Computer Applications in Power, vol. 10, no. 3, pp. 18{21, July 1997.
[5] J. R. Mart, \Accurate Modelling of Frequency-Dependent Transmission Lines in
Electromagnetic Transient Simulations," IEEE Transactions on Power Apparatus and
Systems, vol. PAS-101, no. 1, pp. 147{157, January 1982.
[6] J. A. Martinez-Velasco, Computer Analysis of Electric Power System Transients Selected Readings, The Institute of Electrical and Electronics Engineers,Inc., ISBN 0
7803 2318 1, 445 Hoes Lane, PO Box 1331, Piscataway, NJ 08855-1331, USA, 1997.
[7] J. A. Martinez, \Power Quality Analysis Using Electromagnetic Transients Program,"
in 8th International Conference on Harmonics and Quality of Power - ICHQP'98,
IEEE/PES, Ed., Athens-Greece, October 14-16 1998, vol. I, pp. 590{597.
160
BIBLIOGRAPHY
161
BIBLIOGRAPHY
162
[15] Basic Measuring Instruments - BMI, Handbook of Power Signatures, Basic Measuring
Instruments, 1993.
[16] E. W. Gunther and H. Mehta, \A Survey of Distribution System Power Quality Preliminary Results," IEEE Transactions on Power Delivery, vol. 10, no. 1, pp. 322{
329, January 1995.
[17] B. Hughes and J. S. Chan, \CEA National Power Quality Survey - CEA Project
Number 220D 711A," Tech. Rep., Powertech Labs Inc., Surrey, B. C. , Canada, June
1995.
[18] A. Domijan, G. T. Heydt, A. P. S. Meliopoulos, S. S. Venkata, and S. West, \Directions
of Research on Electric Power Quality," IEEE Transactions on Power Delivery, vol.
8, no. 1, pp. 429{436, January 1993.
[19] A. E. Emanuel, J. A. Orr, D. Cyganski, and E. M. Gulachenski, \A Survey of Harmonic
Voltages and Currents at Distribution Substations," IEEE Transactions on Power
Delivery, vol. 6, no. 4, pp. 1883{1890, October 1991.
[20] G. T. Heydt, Electric Power Quality, Stars in a Circle Publications, 1994.
[21] A. E. Emanuel, J. A. Orr, D. Cyganski, and E. M. Gulachenski, \A Survey of Harmonic
Voltages and Currents at the Customer's Bus," IEEE Transactions on Power Delivery,
vol. 8, no. 1, pp. 411{421, January 1993.
[22] K. Price, \Practices for Solving End-User Power Quality Problems," IEEE Transactions on Industry Applications, vol. 29, no. 6, pp. 1164{1169, November/December
1993.
[23] B. M. Hughes, J. S. Chan, and D. O. Koval, \Distribution Customer Power Quality
Experience," IEEE Transactions on Industry Applications, vol. 29, no. 6, pp. 1204{
1211, November/December 1993.
[24] R. C. Dugan, M. F. McGranaghan, and H. W. Beaty, Electrical Power System Quality,
McGraw-Hill, 1996.
BIBLIOGRAPHY
163
[25] The Canadian Electricity Forum, The Canadian Guide to Power Quality and Power
Harmonics - Vol 1 & 2, The Canadian Electricity Forum, 1997.
[26] G. T. Heydt, \Electric Power Quality: A Tutorial Introduction," IEEE Computer
Applications in Power, vol. 11, no. 1, pp. 15{19, January 1998.
[27] C. J. Melhorn, T. D. Davis, and G. E. Beam, \Voltage Sags: Their Impact on the
Utility and Industrial Customers," IEEE Transactions on Industry Applications, vol.
34, no. 3, pp. 549{558, May/June 1998.
[28] J. F. Fuller, E. F. Fucks, and D. J. Roesler, \In
uence of Harmonics on Power Distribution System Protection," IEEE Transactions on Power Delivery, vol. 3, no. 2, pp.
549{557, April 1988.
[29] A. E. Emanuel and M. Yang, \On the Harmonic Compensation in Nonsinusoidal
Systems," IEEE Transactions on Power Delivery, vol. 8, no. 1, pp. 393{399, January
1993.
[30] IEEE Working Group on Voltage Flicker and Service to Critical Loads, \Power Quality:
Two Dierent Perspectives," IEEE Transactions on Power Delivery, vol. 8, no. 1, pp.
429{436, January 1993.
[31] M. H. J. Bollen, \Voltage Sags: Eects, Mitigation and Prediction," IEE Power
Engineering Journal, pp. 129{135, June 1996.
[32] M. H. J. Bollen, \Fast Assessment Methods for Voltage Sags in Distribution Systems," IEEE Transactions on Industry Applications, vol. 32, no. 6, pp. 1414{1423,
November/December 1996.
[33] M. H. J. Bollen, \Method of Critical Distances for Stochastic Assessment of Voltage
Sags," IEE Proceedings-Generation Transmission and Distribution, vol. 145, no. 1, pp.
70{76, 1998.
[34] B. C. Smith, J. Arrilaga, A. R. Wood, and N. R. Watson, \A Review of Iterative Harmonic Analysis for AC-DC Power Systems," IEEE Transactions on Power Delivery,
vol. 13, no. 1, pp. 180{185, January 1998.
BIBLIOGRAPHY
164
BIBLIOGRAPHY
165
BIBLIOGRAPHY
166
[52] IEEE Standard 1100-1992, IEEE Recommended Practice for Powering and Grounding
Sensitive Electronic Equipment, The Institute of Electrical and Electronics Engineers,
Inc., 1992.
[53] N. G. Hingorani, \Flexible AC Transmission," IEEE Spectrum, pp. 40{45, April 1993.
[54] N. G. Hingorani, \High Power Electronics," Scientic American, pp. 2{9, November
1993.
[55] Westinghouse Electric Corporation, FACTS - Technical Papers, Westinghouse Electric
Corporation - Energy Management Division, 1996.
[56] W. H. Litzenberger, R. K. Varma, and J. D. Flanagan Eds., An Annotated Bibliography of HVDC Transmission and FACTS Devices 1996-1997, Bonneville Power
Administration, Western Area Power Administration and U.S. Department of Energy,
1998.
[57] N. G. Hingorani, \Introducing CUSTOM POWER," IEEE Spectrum, pp. 41{48, June
1995.
[58] Westinghouse Electric Corporation, Solid-State Custom Power Products for Distribution Feeder Power Quality Improvement, Westinghouse Electric Corporation - Energy
Management Division, 1995.
[59] Westinghouse Electric Corporation, Custom Power - Technical Papers, Westinghouse
Electric Corporation - Energy Management Division, 1995.
[60] M. Aredes, Active Power Line Conditioners, Ph.D. thesis, Technischen Universitat
Berlin, Berlin, Germany, 1996.
[61] H. Fujita and H. Akagi, \The Unied Power Quality Conditioner: The Integration of
Series- and Shunt-Active Filters," IEEE Transactions on Power Electronics, vol. 13,
no. 2, pp. 315{322, March 1998.
[62] CIGRE Working Group 14.17 - Semiconductor Power Devices, Semiconductor Power
Devices for Use in HVDC and FACTS Controllers, CIGRE WG 14.17, 1997.
BIBLIOGRAPHY
167
BIBLIOGRAPHY
168
BIBLIOGRAPHY
169
[82] A. S. Sedra and K. C. Smith, Microelectronics Circuits, 3rd Edition - Saunders College
Publishing, 1999.
[83] P. M. Anderson and A. A. Fouad, Power System Control and Stability, The IOWA
Sate University Press, AMES, IOWA, U.S.A., 1977.
[84] J. R. Mart and J. Lin, \Supression of Numerical Oscillations in the EMTP," 88 SM
732-0 IEEE/PES 1988 Summer Meeting, July 1988.
[85] J. Lin and J. R. Mart, \Implementation of the CDA Procedure in the EMTP," IEEE
Transactions on Power Systems, vol. 5, no. 2, pp. 394{402, May 1990.
[86] P. Kundur, Power System Stability and Control, McGraw-Hill, Inc., 1994.
[87] But-Chung Chiu, A Converter Model for the Digital Simulation of Transients in
AC/DC Transmission Systems, Ph.D. thesis, The University of British Columbia,
Vancouver, British Columbia, Canada, May 1980.
[88] C. W. Lander, Power Electronics, The McGraw-Hill Companies, ISBN 0 07 707714 8,
Berkshire, England, Third Edition, 1993.
[89] N. Mohan, Power Electronics: Computer Simulation, Analysis, and Education Using
PSpice*, Minnesota Power Electronics Research and Education, Minneapolis, MN
55414 USA, 1992.
[90] N. G. Hingorani and L. Gyugyi, Understanding FACTS: Concepts and Technology of
Flexible AC Transmission Systems, IEEE Press, ISBN 0-7803-3455-8, 2000.
[91] IEEE Power Electronics Modeling Task Force and Digital Simulation Working Group,
\Guidelines for Modeling Power Electronics in Electric Power Engineering Applications," IEEE Transactions on Power Delivery, vol. 12, no. 1, pp. 505{514, January
1997.
[92] IEEE Working Group 15.08.09, Tutorial on Modeling and Analysis of System Transients Using Digital Programs, The Institute of Electrical and Electronics Engineers,Inc., IEEE Catalog Number 99TP133-0, 445 Hoes Lane, PO Box 1331, Piscataway, NJ 08855-1331, USA, August 1999.
BIBLIOGRAPHY
170
[93] N. Mohan, Computer Exercises for Power Electronics Education, N. Mohan, 1990.
[94] N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics: Converters,
Applications and Design, John Wiley & Sons, Inc., 1995.
[95] H. W. Dommel, \Power Electronics Applications Using the EMTP," in Canadian
Electrical Association, CEA, Ed., Vancouver, B.C., Canada, March 1995, pp. 1{9.
[96] J. Arrilaga and B. Smith, AC-DC Power System Analysis, The Institution of Electrical
Engineers, ISBN 0 85296 934 1, Michael Faraday House, Six Hills Way, Stevenage,
Herts. SG1 2AY, United Kingdon, 1999.
[97] K. Strunz and et ali, \Ecient and Accurate Representation of Asynchronous Network
Structure Changing Phenomena in Digital Real Time Simulators," IEEE Transactions
on Power Systems, vol. 15, no. 2, pp. 586{592, May 2000.
[98] G. D. Irwin, D. A. Woodford and A. M. Gole, \Precision Simulation of PWM Controllers," in IPST'2001 International Conference on Power Systems Transients, Federal University of Rio de Janeiro, Ed., Rio de Janeiro - Brazil, June 24-28 2001, vol. I,
pp. 161{165.
[99] A. M. Gole, D. A. Woodford, J. E. Nordstrom and G. D. Irwin, \A Fully Interpolated Controls Library for Electromagnetic Transients Simulation of Power Electronic
Systems," in IPST'2001 International Conference on Power Systems Transients, Federal University of Rio de Janeiro, Ed., Rio de Janeiro - Brazil, June 24-28 2001, vol. II,
pp. 669{674.
[100] S. Lefebvre and J. Mahseredjian, \Improved Control Systems Simulation in the EMTP
through Compensation," IEEE Transactions on Power Delivery, vol. 9, no. 3, pp.
1654{1662, July 1994.
[101] D.O. Pederson T. Quarles, A.R. Newton and A. Sangiovanni-Vicentelli, SPICE3 Version 3f3 User's Manual, Regents of the University of California, Berkeley, California,
USA, May 1993.
BIBLIOGRAPHY
171
[102] PSCAD-EMTDC web site at http://www/hvdc.ca, PSCAD-EMTDC, HVDC Research Center, Manitoba, ON,Canada, 2001.
[103] The Mathworks, Inc. web site at http://www/mathworks.com, MATLAB, MATLAB,
P.O. BOX 5936, Holliston, MA, USA, 2001.
[104] N. Mohan, Electric Drives: An Integrative Approach, Ned Mohan, ISBN 0-9663530-13, 2000.
[105] R. Dwyer et ali, \Evaluation of Harmonic Impacts from Compact Fluorescent Lights
on Distribution Systems," IEEE Transactions on Power Systems, vol. 10, no. 4, pp.
1772{1780, November 1995.
[106] M. F. McGranaghan, R.M. Zavadil, G.Hensley, T. Singh, and M. Samotyj, \Impact
of Utility Switched Capacitors on Customers Systems - Magnication at Low Voltage
Capacitors," IEEE Transactions on Power Systems, pp. 908{914, 1991.
[107] L. Tang et ali, \Voltage Notching Interation Caused by Large Adjustable Speed Drives
on Distribution Systems with Low Short Circuit Capacities," IEEE Transactions on
Power Delivery, vol. 11, no. 3, pp. 1444{1453, July 1996.
[108] L. E. O. Pinheiro, O. S. I. Komukai, B. D. Bonatto, and E. Yoshida, \Measurements
for Power Quality Monitoring in Distribution System," in I Brazilian Seminar on
Power Quality - SBQEE'96, Uberl^andia-MG - Brazil, June 10-13 1996, pp. 92{97 (in
Portuguese).
[109] K. Louie, \EMTP Techniques in Transient Stability Analysis," Tech. Rep., The
University of British Columbia, Vancouver, B.C., Canada, December 1996, (Ph.D.
Thesis Proposal).
[110] G. F. Reed and et ali, \Application of a 5MVA, 4.16kV D-STATCOM System for
Voltage Flicker Compensation at Seattle Iron & Metals," in IEEE Power Engineering
Society Summer Meeting, IEEE/PES, Ed., Seattle - USA, July 2000.
BIBLIOGRAPHY
172