electric systems dynamics and stability with artificial intelligence applications part-1
TRANSCRIPT
Electric Systems, Dynamics,
and Stability with Artificial
Intelligence Applications
Jarnes A. Mornoh
Howard University Washington,D.C.
Mohamed E. El-Hawary
Dalhousie University Hali&r, Nova Scotia, Canada
M A R C E L
m MARCEL INC. NEWYORKDEKKER, BASEL
D E K K E R
ISBN: 0-8247-0233-6
This book is printed on acid-free paper.
Headquarters
Marcel Dehher, Inc.
770 Madison A\enue, N ~ NYorh, N Y 10016
tel: 2 12-696-9000: fay : 2 12-685-35-10
Eastern Hemisphere Distribution
Marcel Dekker AG
Hutga\w -1. Postfach 8 I ? . CH--1001 Bawl. Switzerland
tel: -11-61-261-8482; fax: 3 1-61-261-8896
World Wide Web
ht t p : / / w m ~ ~.dekker.coiii
The publisher offers discounts on this book when ordered i n bulk quantities. For more
information. writc to Special Sales/Profehsional Marketing at the headquarters d d r i : ~ ~
;I bove.
Copyright 0 2000 by Marcel Dekker, Inc. All Rights Reserved.
Neither thih book n o r any part may be reproduced or transmitted i n any form or by i111>'
iiieans, electronic or mechanical, including photocopying, microfilming. and recording,
or by any information storage and retriet.al system, without permission in ~ ~ i t i n g f r o i n
the publisher.
Current printing (last digit ):
1 0 9 8 7 6 5 4 3 2 I
PRINTED IN THE UNITED STATES OF AMERICA
To Felicia, for the amazing grace, and Omeiza,
Ojirese, Eshovo, and Toyin, joy of life. -JAM
To Ferial, thanks for your company.
Bob, Ron, and Betty, you make me happy. -MEE
This page intentionally left blank
Series Introduction
Power engineering is the oldest and most traditional of the various areas within
electrical engineering, yet no other facet of modern technology is currently un- dergoing a more dramatic revolution in both technology and industry structure.
Among the most exciting changes are those where new solutions are being ap-
plied to classical problem areas.
System dynamics and stability engineering have been crucial elements of
power system engineering since early in the twentieth century. Smooth, continu-
ous operation of modern power supply systems depends greatly on the accurate
anticipation of interconnected equipment, dynamic behavior, and correct identi-
fication of the system’s operating limits. Proper engineering requires precise
methods that can manage that knowledge and direct i t to the design of economi-
cal and secure power systems. Artificial intelligence offers an exciting new basis
for performing dynamic modeling and stability analysis, one that can provide
considerable value and new insight to these often difficult aspects of power
system performance.
Electric Systems, Dynamics, and Stabiliv with Artificial Intelligence Appli-
cations is an exceedingly comprehensive and practical guide to both power sys-
tem dynamics and stability concepts, and to the use of artificial intelligence in
their modeling and engineering. Drs. Momoh and El-Hawary provide a compre-
hensive introduction to power system dynamics and stability, along with a thor-
ough discussion of recently developed concepts such as transient energy func-
tions. Their book is rich in its appreciation of the intricate operating constraints
and issues that real-world power system engineers and operators must face every
day. But wtiat sets this book apart is its application of artificial intelligence to
these long-recognized power system engineering challenges. Chapters 7-9 ex-
pI a i n ho 11ei i ra I tiet uwrk s, espert sy s t e111s us i in g know I edge-based fra in eu'ork s.
and t ' u~zylogic can be applied to the solution of sotlie of the thorniest problems
in power system dynamics.
Like all books in Marcel Dekker's Power Engineering series, Elecstr-ic. SJ.Y-
tettr.v, ~ > j w t t i i c * s , App/ic.trliotis present?t i t i d Sttihilitjv Il'itli A rlificitrl Itit~~ligutic*ci
tnodern power technology i n ;I context of proven. practical applications: iiset'u
;is a reference book iis well iis for self-study and classroom use. Marcel Dekker':,
Pourer Engineering series bill e\~entually include books co\w-ing the entire field
of ponw engineering, in all ot' its specialties and sub-genres, all ainned at provid -
ing practicing pouw engineers u,ith the knou~ledgeand techniyiies they need to
tneet the electric industrlf's challenges in the twenty-first century.
Preface
The intention of this book is to offer the reader a firm foundation for understand-
ing and analyzing power system dynamics and stability problems as \yell as the application of artificial intelligence technology to these problems. Issues in this
area are extremely important not only for real-time operational considerations,
but also in planning, design, and operational scheduling. The significance of
dynamics and stability studies grows as interconnected systems evolve to meet
the requirements of a competitive and deregulated operational environment. The
complexities introduced give rise to new types of control strategies based on
advances in modeling and simulation of the power system.
The material presented in this book combines the experience of the authors
in teaching and research at a number of schools and professional developtnent
venues. The work reported here draws on experience gained in conducting re-
search sponsored by the Electric Power Research Institute. the National Science
Foundation, the Department of Energy, and NASA for Dr. Momoh. Dr. EI-
Hawary's work was supported by the Natural Sciences and Engineering Rc-
search Council of Canada and Canadian Utilities Funding.
This book is intended to meet the needs of practicing engineers invol\.ed i n
the electric power utility business, as well as graduate students and researchers.
I t provides necessary fundamentals, by explaining the practical aspects of artifi-
cial intelligence applications and offering an integrated treatment of the evolu-
tion of modeling techniques and analytical tools.
Chapter I discusses the structure of interconnected power systems, founda-
tions of system dynamics, and definitions for stability and security assessment.
Chapter 2 deals with static electric network models and synchronous machine
representation and its dynamics. Limits for operations of a synchronous machine
and static load models are discussed as well. Chapter 3 deals with dynamic
models of the electric network including the excitation. and prime mover and
governing system models. The chapter concludes with a discussion of dynamic
load models.
Chapter 4 covers concepts of dynamic security assessment based on tran-
sient stability evaluation. This chapter includes both conventional and extended
formulations of the problem. Chapter 5, a complement to Chapter 4, treats the
more recent approach of angle stability assessment via the transient energy func-
tion idea. Chapter 6 introduces the idea of voltage stability and discusses tech-
niques for its assessment.
Chapters 7 through 9 are devoted to an expose of artificial intelligence
technology and its application to problems of system stability, from both the
angle and the voltage sides. In Chapter 7, we introduce basic concepts of artifi-
cial neural networks, knowledge-based systems, and fuzzy logic. In Chapter 8.
we deal with the application of artificial intelligence to angle stability problems.
and the extension to voltage stability is presented in Chapter 9. Chapter I C
offers conclusions and directions for future work in this field.
In developing this book. we have benefited from input from many of 0111'
students. colleagues, and associates. While they are too many to count, we wish
to tnention specifically encouragement by H. Lee Willis, the editor of the Powei-
Engineering Series for Marcel Dekker, Inc. The continual counsel and prodding
of B. J. Clark was extremely helpful. We acknowledge the able administrative
support of Linda Schonberg and the assistance of our respective deans.
We are grateful to Dr. Chieh for the great inspiration and generous contribu -
tions, and to many others, whose names are not included, in the development
ot' this volume. Our students, both present and former. contributed their time
and many valuable suggestions. Many thanks to them and especially to the
young research assistants at the Center for Energy Systems and Control for
putting up with the burdensome challenge of producing this book just in time.
Finally. the book would not have been published without the help of our
Creator and the support of our families.
Contents
Series 1titi.odiic.tiotz H. Lee Willis
Prefiice
1 Introduction
1 . 1 Historical Background
1.2 Structure at a Generic Electric Power System
I .3 Power System Security Assessment
2 Static Electric Network Models 10
2.1 Complex Power Concepts 1 1
2.2 Three-Phase Systems 14
2.3 Synchronous Machine Modeling 21
2.4 Reactive Capability Limits 31
2.5 Static Load Models 32
Introduction 10
Conclusions 35
3 Dynamic Electric Network Models 36
Introduction 36
3.1 Excitation System Model 36
3.2 Prime Mover and Governing System Models 40
3.3 Modeling of Loads
COn c1LI s ions
4 Philosophy of Security Assessment
Introduction
4.1 The Swing Equation
4.2 Some Alternative Forms
4.3 Transient and Subtransient Reactances
4.4 Synchronous Machine Model in Stability Analysis
4.5 SU bt ran s i ent Eq u at i ons
4.6 Machine Models
4.7 Groups of Machines and the Infinite Bus
4.8 Stability Assessment
4.9 Concepts in Transient Stability
4.10 A Method for Stability Assessment
4.1 1 Matheinatical Models and Solution Methods i n Transient
Stabi I ity A sse ssmen t for General Networks
4.12 Integration Techniques
4.13 The Transient Stability Algorithm
Conc 1us ion s
S Assessing Angle Stability via Transient Energy Function
Introduction
5 . I Stability Concepts
5.2 System Model Description
5.3 Stability of a Single-Machine System
5.4 Stability Assessment for ri-Generator System by the
TEF Method
5.5 Application to ;I Practical Power System
5.6 Boundary of the Region of Stability Conclusion
6 Voltage Stability Assessment
Introduction
6. I Worhing Definition o f Voltage Collap\e Study Terms
6.2 Typical Scenario of Voltage Collapse
6.3 Time-Frame Voltage Stability
6.4 Modeling for Voltage Stability Studie\
6.5 Voltage Collapse Prediction Methods
6.6 Clas\ i ficat i on c) f Vo1t age Stab i 1i ty ProbIe111s
6.7 Voltage Stability As\es\ment Techniques
Col1 ter 1f.\ xi
6.8 Analysis Techniques for Steady-State Voltage
Stability Studies 135
6.9 Parameterization IS 1
6.10 The Technique of Modal Analysis 156’
6.1 1 Analysis Techniques for Dynamic Voltage Stability Studies 157
Conclusion 169
Modal Analysis: Worked Example 170
7 Technology of Intelligent Systems 175
Introduction 175
7.1 Fuzzy Logic and Decision Trees 177
7.2 Artificial Neural Networks 177
7.3 Robust Artificial Neural Network 183
7.4 Expert Systems 191
7.5 Fuzzy Sets and Systems 206
7.6 Expert Reasoning and Approximate Reasoning 213
Conclusion 220
8 Application of Artificial Intelligence to Angle Stability Studies 22 1
Introduction 22 I 3 3 3 eh& 8. I ANN Application in Transient Stability Assessment
8.2 A Knowledge-Based System for Direct Stability Analysis 238
Conclusions 257
9 Application of Artificial Intelligence to Voltage Stability
Assessment and Enhancement to Electrical Power Systems 259
Introduction 259
9.1 ANN-Based Voltage Stability Assessment 260
9.2 ANN-Based Voltage Stability Enhancement 265
9.3 A Knowledge-Based Support System for Voltage
Collapse Detection and Prevention 272
9.4 Implementation for KBVCDP 278
9.5 Utility Environment Application 287
Conclusion 287
10 Epilogue and Conclusions 289
298
31 I
332
35 I
This page intentionally left blank
Electric Systems, Dynamics,
and Stability with Artificial
Intelligence Applications
This page intentionally left blank
1
Introduction
1.1 HISTORICAL BACKGROUND
Electric power has shaped and contributed to the progress and technological
advances of humans over the past century. I t is not surprising then that the
growth of electric energy consumption in the world has been nothing but phe-
nomenal. In the United States, for example, electric energy sales have grown to
well over 400 times in the period between the turn of the century and the early
1970s. This growth rate was 50 times as much as the growth rate in all other
energy forms used during the same period.
Edison Electric of New York pioneered the central station electric pouter
generation by opening of the Pearl Street station in 1881. This station had a
capacity of four 250-hp boilers supplying steam to six engine-dynamo sets. Edi-
son's system used a I 10-dc underground distribution network with copper con-
ductors insulated with a jute wrapping. The l o b i t \d tc ige of tlic-,cii-mit limited the
service area of a central station, and consequently central stations proliferated
throughout metropolitan areas.
The invention of the transformer, then known as the "inductorium." made
ac systems possible. The first practical ac distribution system in the United
States was installed by W. Stanley at Great Barrington, Massachusetts. i n 1866
for Westinghouse, who acquired the American rights to the transformer from its
British investors Gaulard and Gibbs. Early ac distribution utilized 1000 V over-
head lines.
By 1895, Philadelphia had about twenty electric companies with distribu-
tion systems operating at 100 V and 500 V two-wire dc and 220 V three-wire
dc: single-phase, two-phase, and three-phase ac; with frequencies of 60, 66. 125,
and 133 cycles per second; and feeders at 1000-1200 V and 2000-2300 V.
The consolidation of electric companies enabled the realization of econo-
mies o f scale in generating facilities. the introduction of a certain degree of
equipment standardization, and the utilization of the load diversity between
areas. Generating uni t sizes of up to 1300 MW are in service, an era that w;i?
started in I973 by the Cumberland Station of the Tennessee Valley Authority.
Underground distribution o f Lwltages up to 5 kV was made possible by thc
de\.elopment of rubber-base insulated cables and paper insulated, lead-co\.erec
cables in the early 1900s. Since that time higher distribution voltages hi1L.e beer
necessitated by load growth that would otherwise overload low-voltage circuit!,
and by the requirement to transmit large blocks of power over great distances.
Coninion distribution voltages in today's systems are in 5 , IS, 25, 35. and 6'1
kV \vltage classes.
The growth in size of power plants and in the higher voltage equipment
u ' a s accompanied by interconnections of the generating facilities. These inter-
connections decreased the probability of service interruptions, made the utiliza-
tion of the most economical units possible, and decreased the total reserve ca-
pacity required to meet equipment-forced outages. This growth was ;11so
accompanied by the use of sophisticated analytical tools. Central control of thc
interconnected systems was introduced for reasons of economy and safety. Th,:
advent of the load dispatcher heralded the dawn of power systems engineering.
\+!hose objective is to provide the best system to meet the load demand reliablq.
safe1y , and economical 1y , U t i 1i zi ng state-of-the-art computer fac i 1i ties.
Extra high voltage (EHV) has become the dominant factor in the transmi>-
sion o f electric power over great distances. By 1896, an 1 1 kV three-phase line
M . ~ Stransmitting 10 MW from Niagara Falls to Buffalo over a distance of 2 3
miles. Today, transmission voltages of 230 kV, 287 kV, 345 kV, 500 kV, 735
kV, and 765 k V are commonplace, with the first I100 kV line scheduled fclr
energization in the early 1990s. One prototype is the I200 kV transmission
tower. The trend is possible. resulting in more efficient use of right-of-wa;!.
lower transmission losses, and reduced environmental impact.
The preference for ac was first challenged in 1954 when the Swedish Stale
Power Board energized the 60-mile, 100 kV dc submarine cable utilizing I!.
Larnm's Mercury Arc FAves at the sending and receiving ends of the world s
first high-voltage direct current (HVDC) l ink connecting the Baltic island o f
Gotland and the Swedish mainland. Today, numerous installations with voltages
up to 800 kV dc have become operational around the globe. Solid-state technol-
ogy advances have also enabled the use of the silicon-controlled rectifiers (SCR)
of thyristor for HVDC applications since the late 1960s. Whenever cable trans-
mission is required (underwater or in a metropolitan area), HVDC is more eco-
nomically attractive than ac.
Protecting isolated systems has been a relatively simple task, which is car-
ried out using overcurrent directional relays with selectivity being obtained by
time grading. High-speed relays have been developed to meet the increased
short-circuit currents due to the larger size units and the complex interconnec-
tions.
For reliable service, an electric power system must remain intact and be
capable of withstanding a wide variety of disturbances. It is essential that the
system be operated so that the more probable contingencies can be sustained
without loss of load (except that connected to the faulted element) and so that
the most adverse possible contingencies do not result in widespread and cascad-
ing power interrupt ions.
The November 1965 blackout in the northeastern part of the United States
and Ontario had a profound impact on the electric utility industry. Many ques-
tions were raised and led to the formation of the National Electric Reliability
Council in 1968. The name was later changed to the North American Electric
Reliability Council (NERC). Its purpose is to augment the reliability and ade-
quacy of bulk power supply in the electricity systems of North America. NERC
is composed of nine regional reliability councils and encompasses \Tirtually all
the power systems in the United States and Canada. Reliability criteria for sys-
tem design and operation have been established by each regional council. Since
differences exist in geography, load pattern, and power sources, criteria for the
various regions differ to some extent.
Design and operating criteria play an essential role in preventing major
system disturbances following severe contingencies. The use of criteria ensures
that, for all frequently occurring contingencies, the system will, at worst, trans-
mit from the normal state to the alert state, rather than to a more severe state
such as the emergency state or the in extreenzis state. When the alert state is
entered following a contingency, operators can take action to return the system
to the normal state.
1.2 STRUCTURE OF A GENERIC ELECTRIC POWER SYSTEM
While no two electric power systems are alike, all share some common funda-
mental characteristics including:
1 . Electric power is generated using synchronous machines that are driven
by turbines (steam, hydraulic, diesel, or internal combustion).
2. Generated power is transmitted from the generating sites o\rer long dis-
tances to load centers that are spread over wide areas.
3. Three phase ac systems comprise the main means of generation. trans-
mission and distribution o f electric power.
4. Voltage and frequency levels are required to remain within tight toler-
ance lekrels to assure a high quality product.
The basic elements of ii generic electric poww system are displayed i n
Figure I . 1 . Electric power is produced at generating stations (GS) and transmit-
ted to consumers through an intricate network of apparatus including transmis-
sion lines, transformers. and switching devices.
Transmission network is classified as the folloufing:
I . Transmission system
2. Si i btransmi ss i on system
3. Distribution system
,s~~.strinThe t i . i ~ i i ~ s i ~ i i . s . s i ~ ) i i i n t erconnec t s a I 1 major genera t i ng s t ;i t i c) n s ;I nd main load centers in the system. I t forms the backbone of the integrated p o \ \ ~ r
system and operates at the highest voltage levels (typically, 230 kV and ab0F.e 1 .
The generator \vltages are usually i n the range of 11-35 kV. These are steppej
up to the transmission cultage level, and power is transmitted to transmission
hubstations where the voltages are stepped down to the subtransmission I t . \ , c . l
(typically, 69 kV to I38 kV). The generation and transmission subsystems ;it e
often referred to as the hi i lk po\r.ei. systerii.
The .viihtr.iiri.siiii ,s.siori sy.stew transmits power at a lower voltage and in
s mii 1ler qitan t i t i es froi n the t ran s iii iss ion s i i bst at i on to the d i s t ri bi i t i on subst ;I-
tions. Large industrial customers are commonly supplied directly from the s u b
transmission system. I n some systems, there expansion and higher \,ohage l e \ ~ l s
becoming necessary for transmission, the older transmission lines are otten re1 2-
gated to s i i bt ra n smi ss ion function . The di.stri/mtioii .sj:steiri is the final stage in the transfer of' pourer to tlie
individual customers. The primary distribution voltage is typically betwwn 4.0
kV and 34.5 kV. Small industrial customers are supplied by primary feedcrs
at this \.oltage level. The secondary distribution feeders supply residential and
commercial customers at I20/240 V.
The function o f an electric power system is to con\.ert energy from one of
the naturally available forms to the electrical form and to transport i t to the
points of consumption. Energy is seldom consumed i n electrical form but is
rather converted to other forms such as heat, light, and mechanical energy. The
rid\rantage of the electrical form of energy is that i t can be transported and
controlled with relative ease and with a high degree of efficiency and reliability.
A properly designed and operated power system should. therefore. meet lie
fol low i ng fitnda men ta I reqii ire men t s:
--
--
5
22 kV
- 500 kV 500kV 230kV
-20 kV
Tie line to neigh-boring system
I I 1 1 I Trans-
I Transmission system
I
mission (230 kV) Tie line
distribution
Subtransmission
Residential Commercial
Figure 1.1 Basic elements of H power system.
1. The system must be able to meet the continually changing load demand
for active and reactive power. Unlike other types of energy. electricity
cannot be conveniently stored in sufficient quantities. Therefore, ade-
quate “spinning” reserve of active and reactive power should be main-
tained and appropriately controlled at all times.
2. The system should supply energy at minimum cost and with minimum
ecological impact.
3. The “quality” of power supply must meet certain minimum standards
6
with regard to the following factors: ( I ) constancy of frequency; (l )
constancy of voltage; and (3) level of reliability.
Several levels of controls it~vol\~inga complex array of devices are used to
meet the above requirements. These are depicted in Figure I .2, which identifit*s
the various subsystems of a power system and the associated controls. In this
overall structure, there are controllers operating directly on individual system
elements. In a generating unit these consist of prime mover controls and excih-
tion controls. The prime mover controls are concerned with speed regulaticm and control of energy supply system variables such as boiler pressures, temperi-
tures, and tlows. The function of the excitation control is to regulate generatlir
bultage and reactive power output. The desired MW outputs of the individual
generating units are determined by the system-generation control.
The primary purpose of the system-generation control is to balance the total
system generation against system load and losses so that the desired frequency
and power interchange with neighboring systems (tie flows) is maintained.
The transmission controls include power and bdtage control devices. sw:h
as static viir compensators, synchronous condensers, switched capacitors ;I id
Frequency Tie Generator
flows power
Supplementary
, -- - ------- - ------------------ ------ --- -,Cmd-. ---------I
I I I
I I It
I I Excitation I I I
I I I
+ System and control
4-
-
II
I I
I
I I I I I II
II
I
I
I
1I
I I1
I
W Rime mover
i
Voltage
Shaft
Power
-Generator
power speed/
-
Power
; i I *
i Transmission Controls
; I
; I
i ’Reactive power arid Voltage control, INDC
associated contro s
transmission and
Figure 1.2 Subsystems of a power system and associated controls.
I n troductio ri 7
reactors, tap-changing transformers, phase-shifting transformers, and HVDC
transmission controls.
These controls described above contribute to the satisfactory operation of
the power system by maintaining system voltages and frequency and other sys-
tem variables within their acceptable limits. They also have a profound effect
on the dynamic performance of the power system and on its ability to cope with
disturbances.
The control objectives are dependent on the operating state of the power
system. Under normal conditions, the control objective is to operate as effi-
ciently as possible with voltages and frequency close to nominal values. When
an abnormal condition develops, new objectives must be met to restore the
system to normal operation.
Major system failures are rarely the result of a single catastrophic disturb-
ance causing collapse of an apparently secure system. Such failures are usually
brought about by a combination of circumstances that stress the network beyond
its capability. Severe natural disturbances (such as a tornado, severe storm, or
freezing rain), equipment malfunction, human error, and inadequate design com-
bine to weaken the power system and eventually lead to its breakdown. This
may result in cascading outages that must be contained within a small part of
the system if a major blackout is to be prevented.
1.3 POWER SYSTEM SECURITY ASSESSMENT
The term Po\ivr. Sysfeni Stability is used to define “the ability of the bulk power
electric power system to withstand sudden disturbances such as electric short
circuits or unanticipated loss of system components.” In terms of the require-
ments for the proper planning and operation of the power system, i t means that
following the occurrence of a sudden disturbance, the power system will:
1 . Survive the ensuing transient and move into an acceptable steady-state
condition, and
2. In this new steady state condition, all power system components are
operating within established limits.
Electric utilities require security analysis to ensure that, for a defined set of
contingencies, the above two requirements are met. The analysis required to
survive a transient is complex, because of increased system size, greater depen-
dence on controls, and more interconnections. Additional complicating factors
include the operation of the interconnected system with greater interdependence
among its member systems, heavier transmission loadings, and the concentration
of the generation among few large units at light loads.
The second requirement is verified using steady state analysis in what is
Power Svstern Stablllty 1 . 4
T+ + + Strwrural Volruge (Slow) Oher Tvpes (SSR, Lo 6
I I Frequencb Oscillatiw s, etc
I 4 + +
Small Small Static Disturbance Disturbance (Power
(Contingency) (Quasi-Static)+-Cntical Sensitir it y Bifurcation Bifurcation Eigenvaluc AMIYSIS I Theory (Hopf. Analyv\ e tc . )
Bifurcation T a p Changer Singular Value Decompoiition
Techniques Frequency M Y Energy D o m i n Functions Analysis Angle
Stability Linear Selective Rogramrmng
Vdtage M a N m m Jacobian A M I y si s Stability Lnadability
~~~~-~~~~~~~~
4 Unlned Use of the Energy Function and h e Bifurcalion Analytical Studies Approach I
Linear Rogrammng for DecompositionOpimal Nonlinear Ragramming Germation Rescheduling dnd Power H o w (OPR wth (NLP)with quadraticEnhancement
(Prevcatlve Load Control Transieru Energy Margin and objccti~cfunctions and
Coatrol) Vdcagc Energy Margin ds Linear Rogramrmng (LP) fu
Generation and Load Control
Control using Exper( System Control Selections. output Arttndal Neural Networb IFS) Pnmiizatim Classifier and Analyzer (ANN), and/ a Expert
intelligent Systems fa solution outputSupport System
re l'erred to :is "s t ;i t i c sec i i r i ty iisse ss men t ." The t'i rs t reqUi renit' 11t i s the hii bjec t
deii 1t \v i t h i 11 "d y n amic sec i i r i ty assessme11t ." Dy nam ic sec11 r i ty st i id ies ;I re
broadly classified ;is being either "angle stability studies" or "voltage stabil; t),''
;is depicted i n Figure 1.3.
In "angle stability studies." problems are classified as either "large distitrb-
mce.' tix transient e\'aluiitioii, or "small disturbance" for steady state stability
t'\,:ilurition. A similar classification for \soltage probleiix is indicated i n Figure
1 ..3. Solution techniqites tor transient unglc stability e\,aluation incluck:
Ti me domai n s i mii 1at ion
Direct methods
Hybrid methods
Probabi I i st ic nie t hods. pat te1-11 recog n i t ion
COnn pit t ;it iona1 i n te 1I igence
Time domain simulation techniques invol\Ve judicious use of integration
methods such as Runge-Kutta. trapezoidal rule of integration, and ~ ~ a \ . e t c ~ r m
re 1axat ion. These met hods are part ic41ar1y usefu 1 for off- 1i ne transit‘ 11t s t abi 1i t y
ana I y s i s.
On the other hand, in “voltage stability studies.” the problems are classified
as “large disturbance” in some contingency cases, “small disturbance” i n quasi-
static cases. and “static,“ which requires solutions to the general pouw tlo\v
(algebraic) equations only. I t is as a result of this classification that the solution
techniques and requirements are derived. The bifurcation theory. Linear Pro-
gramming applications. and the use of the energy fiinction are but ii few such
tech n i qLies. Agai n, the t i me domai n s i m u I at i ons t h ii t ;ire i 11\YII \red take ad \,;in t age
of various numerical integration methods mentioned earlier. (The machine dj8-
namics lead to differential equations that are inherently nonlinear.)
The unified approach as indicated on Figure 1.3 is aimed at encompassing
the similarities and differences that distinguishes the Lrarious techniques used in
assessing the stability of the electric power system. This is \4,hether or not the
problems are a result of voltage or angle instability. The resulting enliancenient
that is brought forth by this approach, measurable ;is a benefit-to-cost index. lies
in the de\~lopment and use of more robust tools for solLting present ~ind long-
range problems. I n this light, various programming and optimization schemes
that are applicable include decomposition Optiiutrl Po\t*or Flo\tp (OPF). Liiiorrr
Pi .oSr . c r i i i i i i i r i ,~(LP), and Qir(rcltntic. Pi.oRt .cr i i i i i i i i iS (QP), with the necessary and
su f fici en t sy stem and net work constr ai n ts. Finally. this book introduces three fundamental types of intelligence support
systems that truly adds the rigor, value, and robustness to the desired enhance-
ment schemes. These support systems include expert systems (ES), fuzzy logic
(FL). and artificial neural network (ANN). Each have their unique characteris-
tics (decision-support, classifiers, learning capabilities, etc.) and are ridaptable
in providing viable solutions to a variety of voltage/angle instability problems
associated with the electric power system. The discussion on this area of artifi-
cial intelligence applications to power system stability rind dynamics is pre-
sented i n the final few chapters of this book.
&
Static Electric Network Models
INTRODUCTION
The power industry in the United States has engaged in a changing busincss
environment for some time, by moving away from a centrally planned sys tm to one in which players operate in a decentralized fashion with little knowleclge
of the full-state of the network, and where decision-making is likely to be mar-
ket driven rather than through technical considerations alone. This new environ-
ment is quite different from the one in which the system operated in the past.
This leads to the requirement of new techniques and analysis methods for func-
tions of system operation, operational planning, and long-term planning.
Electrical power systems vary in size, topography and structural compo-
nents. However, what is consistent is that the overall system can be diLiried
into three subsystems, namely, the generation, transmission, and distribution
subsystems. System behavior is affected by the characteristics of each of the
major elements of the system. The representation of these elements by means
of appropriate mathematical models is critical to successful analysis of sys ;em
hehavior. Due to computational efficiency considerations for each diffe -ent
problem, the system is modeled in a different way. This chapter describes some
system models for analysis purposes.
We begin in Section 2.1 by introducing concepts of power expressed as
active, reactive, and apparent. This is followed in Section 2.2 by a brief reLiew
of three phase systems. Section 2.3 deals with modeling the synchronous ma-
chine from an electric network standpoint. Reactive capability curves are exam-
ined in Section 2.4. Static and dynamic load models are discussed in Section
2.5 to conclude the chapter.
2.1 COMPLEX POWER CONCEPTS
In electrical power systems one is mainly concerned with the flow of electrical
power in the circuit rather than the currents. As the power into an element is
basically the product of the voltage across and current through it, it is reasonable
to exchange the current for power without losing any information. I n treating
sinusoidal steady-state behavior of an electric circuit. some further definitions
are necessary. To illustrate, we use a cosine representation of the sinusoidal
waveforms involved.
Consider an impedance element 2=ZL$. For a sinusoidal voltage, v ( r ) is
given by
v(t)= tlCOSOt
The instantaneous current in the circuit shown in Fig. 2.1 is
i(r)= /,,, cos(or - Q)
where the current magnitude is:
The instantaneous power is given by
p(r)= i r ( t ) i ( t )= XI I,,,[cos(or)cos(or- $)]
Using the trigonometric identity
1coscl cosp = -[cos(a - p) + cos(cl + p)]
2
we can write the instantaneous power as
+
V Z
-
Figure 2.1 Instantaneous current in a circuit.
The average power p ( , is seen to be
Since through 1 cycle. the a \ erage of co\( 2ot - @) i \ zero. this term contri b-
ute\ nothing to the acerage of / I .
I t i \ more convenient to use the effectiLre ( r i m ) calue\ of boltage and curt-cnt
than the iiiiixiinuni ~ a l u e s .Sub\tituting xi = v% Ci,,,,). and I , , , = -\/'?(/,,,,,). \+e get
Thus the p o ~ wentering any network is the product of the effectilre v:iIies
of' terminal iaoltage and current and the cosine ol' the phase angle betLveen the
\,oltage and ciirrent which is called the i ~ o ~ ~ i ~ ~ ) . , t ~ / ~ . t o ) .(PF). This applies to siniis-
oidal Lroltages and currents only. For ii purely resistive load. cos$ = 1. and I he
current in the circuit is fully engaged in con \q ing power from the soiirce to
the load resistance. When reactance (inductive or capacitilte) as t+,ell ;is m i s -
[;incc are present, ii component o f the current in the circuit is engaged i n con\~r:~-
ing energy that is periodically stored in and discharged from the reactance. This
stored energy, being shuttled into and out of the rnagnelic field of ;it1 inductaiice
or the electric field of a capacitance. adds to the magnitude of the current in the
circuit but does not add to the a\.erage power.
The a\wage power i n ii circuit is called i1ctiL.e power. and loosely speaking
the po\+rer that supplies the stored energy i n reacti\re elements is called reacliite
po\i.er. Acti1.c pou.cr is denoted bjf P, and the rci1ctii.e pourer. is designattx ;i\
Q.Thej' are expressed as
I n both equations. 1' and 1 are rim calues of terminal Lfoltage and current.
and Q, is the phase angle by mrhich the current lags the \,oltage.
Both P and Q arc of the same dimension, that is in (Joules/s) Watts. Hou-
eb'er, to emphasize the fact that Q represents the nonactive power, i t is measiired
i n reiictiLpe k~oltampere units (\'at-).Larger and tiiore practical units arc k i 1 o r . m
and megavars. related to the basic unit by
As\urnt. that 1'. 1' cos$, and 11 sin@,a11 shown in Fig. 2.2, are each multi-
plied bqr /. the r i m cralue of the current. When the components of iultage 1' , :OS@
vvI 3Static' Electric N e t u v r k Models
Isin $
I
Figure 2.2 Phasor diagrams leading to power triangles.
and V sin$ are multiplied by current, they become P and Q respectively. Sinii- larly, if I, I cos$, and I sin$ are each multiplied by V, they become V I , P. and Q respectively. This defines a power triangle.
We define a quantity called the complex or apparent poufer, designated S. of which P and Q are orthogonal components. By definition,
S = P +j Q = i/*
= V/ cos@+ j V 1 sin@
= V/ (cos@+ , j sin@)
Using Euler's identity, we thus have
s = VIC"*
or
s=VIL$
If we introduce the conjugate current defined by the asterisk (* ' )
I* = l l ( L @
i t becomes obvious that an equivalent definition of complex 01- apparent POW-
er is
s = VI" ( 2 . 5 )
We can write the complex power in two alternati\,e fo rm by using the relationships v=Z7 and 7 = Y v
This leads to
s = ZI I" = Z J I J ? (2.6)
or
S=VY*V"= Y" ,V I " (2.7)
V
I Figure 2.3 Series circuit of n impedances.
Consider the series circuit shown in Fig. 2.3. Here the applied Lvltage i:,
cqual to the sum of the voltage drops:
\'=I(& +z,+ .. . +Z,:)
Multiplying both sides of. this relation by /* results in
s = \'I* = I / * ( Z ,+ 22 + . . . + Z , ! )
or
uith
s,= ,II?z,
being the individual element's complex power. Equation (2.8) is knourn as tke
sunimiition rule for complex powers. The summation rule also applies to paral1i:l
circuits. The use of the summation rule and concepts of complex pourer ai.e
ai\rantageous i n solving problems of power system analysis.
The phasor diagrams shown in Fig. 2.2 can be converted into compltmx
pourer diagrams by simply following the definitions relating complex power .o
tzoltage and current. Consider the situation with an inductive circuit i n urhic-h
the current lags the voltage by the angle $. The complex conjugate of the current
M i l l be in the first quadrant in the complex plane as shown in Fig. 2.3(a).
Multiplying the phasors by V, we obtain the complex power diagram in
Fig. 2 . 4 b). Inspection of the diagram as well as previous development lexis to
ii relation for the power factor of the circuit:
Pcos$ = __
1s I
2.2 THREE-PHASE SYSTEMS
A significant portion of all the electric power presently used is generated. trails-
mitted. and distributed using balanced three-phase bultage systems. The single-
Static Electric N e t w w k Models 1.5
Figure 2.4 Complex power diagram showing the relationship among voltage. current,
and power components.
phase voltage sources referred to in the preceding section originate in many
instances as part of the three-phase system. Three-phase operation is preferable
to single-phase because a three-phase winding makes more efficient use of gen-
erator copper and iron. Power flow in single-phase circuits is known to be pul-
sating. This drawback is not present in a three-phase system as will be shown
later. Also, three-phase motors start more conveniently and, having constant
torque, run more satisfactorily than single-phase motors. However, the compli-
cations of additional phases are not compensated for by the slight increase of
operating efficiency when polyphase systems of order higher than three-phase
are used.
A balanced three-phase voltage system consists of three single-phase volt-
ages having the same magnitude and frequency but time-displaced from one
another by 120". Figure 2.5(a) shows a schematic representation where the sin-
\ 120°
van120Pnce120°
Figure 2.5 (a) A Y-connected three-phase system and (b) the corresponding phasor
diagram.
gle-phase \,()Itage sources appear in ii wye o r Y-connection; a delta o r A configu-
ration is also possible, as discussed later. A phasor diagram shouring each
the phase voltages is also given in Figure 2.S(b). As the phasors rotate at the
iiiigular t'reqiiency cu with respect to the refkrence line in the counterclockc~,isc
(ciesignated as positive) direction, the positive maximum value first occurs t ' c r
phase ( I and then in siiccession for phases h and (*. Stated i n ;I different ~t'aq'.to
;in ohser\~erin the phasor space. the voltage of phase ( I arri\,es t i n t follo\zred 1'4, that o f h and then that of (*. For this reitson the three-phase Lroltage of Fig. 2 5
i h said to have the phase sequcncc t r b c . (order, phase sequence. or rotation a11
mean the ss;itiie thing). This is important tor certain applications. For es~irnple.
in three-phase induction motors. the phase seqitence dcterniines u,hether tlic
i i io tor rotates clock\i.ise o r coiinterclock~ise.
2.2.1 Current and Voltage Relations
Balanced three-pha+e \y\tem\ c;in be \tudied u\ing technique\ de\reloped t o r
$ingle-phaw circuit\. The arrangement o f the three \ingle-phaw Ioltagc\ i n t o ;I
Y o r ;i A configuration require\ \ome modification\ in dealing U ith the o\er.iIl
\>\tern.
2.2.2 Y-Connec tion
With reference to Fig. 2.6. the cotiimon terminal I I I \ called the neutral o r <tar
(k') point. The ~ol tagesappearing between any t u o of the line terminal\ 1 1 . h,
m d ( h a b e different relation\hips in magnitude and phaw to the \oltage\ apptar-
ing bet\$een any one line terminal and the neutral point 1 1 . The \et o f colta;e\
\', . 1; , ,itid itre called the line coltage\, and the \et of coltage+ li,. and \'\'(
c i ~ creferred to ii\ the phaw c oltage\. Con\ideration o f p h a w r diagram\ pro\ ide\
t he rcq ii I red re I ;it ionsh i p\.
The etfectike kaliie\ o f the phase \wlt:ige\ ;ire \hohn III Fig. 2.6 i i \ \,',, \;,.
and l'!.Each ha\ the wile magnitude, and each 15 di\placed 120" from the olher
tuo ph;i\or\. To obtain the magnitude and phare angle o f the line \oltnge froin
( I to h (i.e . \ ' , , I . L+C apply Kirchhoff'\ voltage l a ~ :
This equatioii state\ that the ~o l t ageexisting from ( I to h i h equal to the
coltage from ( i to 11 (i.e.. \,',) piu\ the voltage f rom I I to h. Thus Eq. (2 .10 ) can
be re~irittenas
(.!.I I )
I7
Qn = vp L -120O
Figure 2.6 Illustration of the phase and magnitude relation\ beiueen thc phaw and
line coltage of' a Y-connection.
Since for a balanced system, each phase Lroltage has the same magnitude.
let UI set
I K,! I = ' &,I I = I Yll I = 1; (2 .12)
where 1;: denotes the effective magnitude of the phase voltage. Accordinglj u e
may write
K,, = y,o0 (2.13)
Kll= y - - 120" (2.14)
yl,= y L- 230" = y L120" (2.1s)
Substituting Eqs. (2.13) and (2.14) in Eq. (2.1 I ) yields
K,?= y ( 1 - 1 i - 120")
= f i p 3 0 0 (2.16)
S i ni i 1arly we ob t ai n
X( = f iy- - 90" (3.17)
(2.18)
The expressions obtained above for the line boltages shou that they consti- tute a balanced three-phase voltage system M’hose magnituctes are v‘? time\
thow of the phase voltages. Thus we write
1.: = fiy (2.111)
A current tlowing out of a line terminal ci (or 11 or i \ the same a\ thi\t( a )
tlon ing through the phase source Lroltage appearing between terminal\ 11 and ( I
(or 1 1 and h or I I and c) . We can thus conclude that for a Y-connected three phase wurce, the line current equals the phase current. Thu\
l, = f,, (2 .2 ) )
In the aboce equation, l, denotes the effective \ d u e of the line current aid /,, denotes the effectike tfalue for the pha\e current.
2.2.3 A-Connection
We now consider the case when the three single-phase sources are rearrangd to forrn ;I three-phase A-connection as shown in Fig. 2.7. I t is clear from in5ptc- tion of the circuit shown that the line and phase \oltage\ haw the wine magiii-tude:
The phase and line currents. howec er are not identical, and the relation\liip betkveen them can be obtained by using Kirchhoff’s current lau at one of ‘he 1i ne term i nals.
In a manner similar to that adopted for the Y-connected source, let U\ con-\icier the pha\or diagram shoun i n Fig. 2.8. Assume the phaw currents to tw
/ n \b bb’ ob’
C Ibc
I,,’ C’
Figure 2.7 A A-connected three-phase s w r w .
I
Stntic Elect t-ic NetUY)t-k Mode Is
Figure 2.8 Illustration of the relation between phase arid line currents in a A-connection.
/ < I / , = I/ ,
I/,[ = I,, i - 1 20"
I ,,, = I,, i 120"
The current that flows in the line joining [ I to ci' is denoted by l$,,,,and is
given by
As a result, we have
I,,,,, = I,,[ I 1 20" - 1 i 01L!
which simplifies to
I,,,,, = &,L I 50"
Similarly.
Note that a set of balanced three-phase currents yields a corresponding set
of balanced line currents whose magnitudes are fi times the rnagnitudes of the
phase values:
20
\+hxl, denotes the m:ignitude 01' any o f the three line currents.
2.2.4 Power Relationships
Arrume that ii three-phare generator is supplq ing a balanced loud \+,ith thc t h r w rintiroidal phase voltagcs:
\' ( f 1 = \ '7 1;: \ l I l ( l ) t
\ ' , ( f ) = \ 21,: \111((1)/ - 120- )
\' ( I ) = \ 21,: \ l l l ( ( I ) f + 120 )
~chcre@ is the phase angle bt.t\swn thc current iind voltage i n c d i ph;iw. ?'he total po\ser i n the load is
This t ims out to be expanded US:
L:sing ;I trigonometric identit,~, ive get
Yote that the last three terms in the abo\.e equation add tip to ~ r o .l'hus n'c obtain
When referring to the voltage level of' a three-phase s\'stem. by con\'ention. one in\.ariably tinderstands the line l-oltages. From the iibo\.e cliscussion thc relationship bet\\,t.cn the linc and phase \.oltages i n ;i Y-connectetl s~~s tem s
The power equation thus reads in t e r m of line quantities:
p 1 0 = f i , l f l 11, l C O 4
We note that the total instantaneous power is constant. hairing a magnitude
of three times the real power per phase.
We may be tempted to assume that the reactiLre power is of no importance
in a three-phase system since the Q terms cancel out. Houte\zer, this situation is
analogous to the summation of balanced three-phase currents and \dtages that
also cancel out. Although the sum cancels out, these quantities are still \very
much in evidence within each phase.
We extend the concept of complex or apparent power ( S ) to three-phase
systems by defining
SIO= 3y1;
where the L1ctik.e and reactijre powers are obtained from
In terms of line Lralues, we can assert that
sl,= fiv1:
and
2.3 SYNCHRONOUS MACHINE MODELING
I n po\s'er system stability analysis. there are several types of models used for
representing the dynamic behavior of the synchronous machine. These models
are deduced by using some approximations to the basic machine equations. This
section gives a brief introduction to synchronous machine equations.
2.3.1 Stator and Rotor Voltage Equations
In de\.eloping performance equations of a synchronous machine, the follou,ing
assumptions are made:
I . The stator urindings are sinusoidally distributed along the air-gap its far
a s the mutual inductance effects with the rotor are concerned.
2. The stator slots cause no appreciable variation of the rotor inductances
Lvith rotor position.
3. Magnetic hysteresis is negligible.
1. Magnetic saturation effects are negligible.
Based on these assumptions, a synchronous machine can be represented b j
six Lvindings as shown in Fig. 2.9. The stator circuit consists of three-phast-
;irniatiire uindings carrying alternating currents. The rotor circuit consists o .
field and iiniortisseur windings. The positive direction of ;I stittor winding cur.
rent is rissumed to be into the machine.
The Lwltage equations of the three-phase armature uindings are:
(),, = - N , , i , , t l t
Rotation
0,elec. rad/s
‘ c
Figure 2.9 Stator mcf rotor circuits of ;I synchronous rnachinc. ci. h. c: Stator ph: w
windings: jil: field cvinding: k d : d-axis armature circuit; k q : q-axis armature circuit; 1 =
I . 2, . . . . r l ; I I = number of xinature circuits; 0 = angle by Lvhich cl-axi5 leads the iiiag-
rietic axis of phase winding. electrical radiaii~;: ( I CO,= rotor angular \‘elocitj. clectri1:al
rad/x.
Static Electric Net\t*ork Models 23
where
L,,,,.L,,,,,L,, are self-inductance of nhc windings.
L,,/,. L,,,, LI,,,, LI,,, L,,,,LIIlare mutual inductances between two stator winding (cih, hc, NC')
L,,,,/,L/,,,/,L,,,/ are mutual inductances between stator and fields windings
LlIAll,L,lL,I, LlL,/are mutual inductances between stator winding and d-axis
armature circuit
L,,A,,,LhAll,L,A,lare mutual inductances between stator windings and q-axis
armature circuit.
The rotor circuit voltage equations are given by:
(2.24)
The rotor circuit flux linkages are given by
v/d= Lff,/;/,/+ L[?.diA[/- Ldd;/>- LhAili/i- L,kqi ,
vkd = Llldild + LALdilrl - &iAdi , i - LRAdili - L,A,/;(
vAq = LAAqikt, - LiiAyio - Lldqih - LcAq;c ( 2 . 2 5 )
Equations (2.22) and (2.23) associated with the stator circuits together with
equations (2.24) and (2.25) associated with the rotor circuits completely describe
the electrical performance of a synchronous machine.
The fact that mutual and self inductances of the stator circuits vary with
rotor position q which in turn varies with time, complicates the synchronous
machine Eqs. (2.22) to (2.25). The variations in inductances are caused by the
variations in the permeance of the magnetic flux path due to nonuniform air
gap. This is pronounced in a salient pole machine in which the permeances
along the two axes are significantly different. Even in a round rotor machine
there are differences between the two axes due mostly to the large number of
slots associated with the field winding.
The self and mutual inductances of the stator circuits are given by
I,, = I, ( , + 1, ! CO\ 78
1, = I. ,, + + 3 '
[ :]=:
4!!- K ,;,, r.1.28, tit
where
All the inductances expressed as dqO components are seen to be constant, i.e..
they are independent of the rotor position. It is interesting to note that i,,does not
appear i n the rotor f lux linkage equation. This is because the zero sequence com-
ponents of armature current do not produce net mmf :icross the air-gap.
While the dqO transformation has resulted in constant inductances i n Eqs.
(2.28) to (2.30). the mutual inductances between stator and rotor quantities are
not. For example, the mutual inductance associated with the f lux linking the field ufinding due to current i,, flowing in the d-axis stator winding from equa-
tion (2.30) is (3/2) L,,,,,. whereas from equation (2.29) the mutual inductance
associated with flux in the d-axis stator winding due to field current is L,,,,,.This
difficulty is overcome by an appropriate choice of the per un i t sqrstcm for the
rotor qu an t i t i es.
2.3.3 Per Unit Representation
I t is usually conivenient to use a per uni t system to normalize system irariables.
to offer computational simplicity by eliminating units and expressing sqrstem
quantities as dimensionless ratios. Thus
A we11-chosen per unit sy steni can m i n i 111 ize comp i t ;i t i on a I effort, x imp1i f y
e\'aluation. rind facilitate understanding of system characteristics. Some baw
quantities may be chosen independently and quite arbitrarily, Lvhile others t'ol-
low automatically depending on the fundamental relationships between system
\.ariables. Normally, the base values are chosen so that the principal ttariable?
mrill be equal to one per unit under rated operating conditions.
I n the case of a synchronous machine, the per unit system inay be used t c I
remoF.e arbitrary constants and simplify mathematical equations s o that they rnq
be expressed in terms of equivalent circuits. The basis for selecting the per uni
system for the stator is straightforward, while it requires careful consideration for
the rotor. The L,,,,-base reciprocal per unit system will be discussed here.
The following base quantities for the stator are chosen (denoted by sub-
scripts)
f?,h,l,r. = peak ~(alue of rated line-to-line voltage, v
= peak value of rated line-to-line current, A
,f\.,,,< = rated frequency, Hz
The base \ d u e of each o f the remaining quantities ;ire automatically st't
and depend on the above :is follows:
CO,,,~,,= 271 .fh,l,c. electrical rad/s
2 o,,,,,.,,, mechanical rad/s = \clh.l,, -
TFf ' ~ \ t 7 . l W
Z+lre= -:-, ohms 1\h,i\r.
The stator voltage equations expressed in per unit notations are g i \ m b j
_ _
27
The corresponding flux linkage equations may be written as _ _ - - - --
w,/ = --&/id + LJ/</+ L<lL,/~L/
- _ - - -
wq = -L& + Ld,/i</ -WO = -MO (2.32)
The rotor circuit base quantities will be chosen so as to make the flux
linkage equations simple by satisfying the following:
I . The per unit mutual inductances between different windings are to be
reciprocal. This will allow the synchronous machine model to be represented by
simple equivalent circuits.
2. All per uni t mutual inductances between stator and rotor circuits i n
each axis are to be equal.
3. The following base quantities for the rotor are chosen, i n view of the
L,,,/-base per unit system choose,
(2 .33)
The per unit rotor flux linkage equations are given by
(2.34)
Since all quantities in Eqs. (2.31) to (2.34) are in per unit, we drop the overbar
notation in subsequent discussion.
If the frequency of the stator quantities is equal o the base frequency, the per unit reactance of a winding reactance is numerically equal to the per unit
inductance. For example:
xl/= 2zf Ll/ (Q)
= 2nfhd,rL\hd\r.,Dividing by Z\hd\c if f = f b d , c , then the per unit values of X, / and L,,
are equal.
2.3.4 Classical Representation of the Synchronous Machine
The per unit equations completely describe the electrical and dynamic perfor-
mance of a synchronous machine. However, except for the analysis of very
\mall \ j \tern\, these equation\ cannot be used directly for \}stem \tability \tud-
I C \ . Some \implification\ and approxitiiations are required to reprewit the \ j n-chronoii\ machine i n stabilitj \tudie\. For large \ y stem\. i t i \ nece\wrq to neg-
1ec.t the tr:in\tornier voltage term\ \ir<,arid \ir, atid the et'f'ttct o f \peed \ ariation\.
Therefore, the rnachine equation tie\cribed bjf Eys. (2.33)and (2 .33 )become
( 2 . 3 6 ,
By defining the follou~ing variables
the riiachine eyuationx becomes
( 2 . 3 7 )
M here Et: i \ the q-a\ii\ c-omponent of the \ oltage behind transient rcactanc\e
\'. I" , , i \ the open-circuit tran\ient time con\tant. E, i \ the \folt:ige proportiond
to I , , and olo i \ the ~o l t ageproportional to E,<,.Sincc per u n i t \ = I_ , l'rom
l!c~u~ltlon(2.37) m e hace
For studies i n \i.hich the period of analj'sis is sinall i n comparison c+,itli I-,:, thc riiachine inodel i x often simplified bj' assuming that E'; is constant throiigh-
o i i t the st lidy period. Th i s ;issiinipt ion e 1imi nates t he on 1j , d i ftere n t i a1 ey i i ;it ion
asxociiited M i t h the electrical chxiicteristics o f the machine. A further approui-
rriation is to ignore transient saliency by assuniing that .v(; = .v(; and to :issiiriie
that the t'lux linkage alw remains constant. With these assuinptions. the \,oIt;ige
behind the transient impedance R,,+jx$ has a constant magnitude. The equi\.a-
lent circuit is shown in Fig. 2.10. The machine terminal voltage phasor is repre-
sented by
q=E’i 6 - (R(,+.j.~;)i,
The machine dynamic model is represented by
T, @4! = M,,,- M ,(It
P, = R , ( V , I )
v,= E’J - (R,,+ jx ; ) i, (2.39)
where V , is the machine terminal voltage phasor and can be calculated from
power flow considerations. Then ’ can be calculated. The machine scl?ing equa-
tion can then be solved.
Equation (2.39) is the so-called classical model of the synchronous machine
and is widely used in power system stability studies. This classical model is
often used for three different time frames: subtransient, transient. and steady-
state. Figure 2.1 I surnmarizes these three simple synchronous machine models.
The subtransient and transient assume constant rotor flux linkages. and the
steady-state model assumes constant field current. These models neglect sali-
ency effects and stator resistance and offer considerable structural and computa-
t i onal s i mpl i ci t y .
Figure 2.10 Eqiii\~alcntcircuit synchronous urith x,; = xi:.
t t
(a) Subtransient model
E’L6
(b) Transient model
EqL6
(c )Steady-statemodel
Figure 2.11 Simple hjmchronous machine model.
Stcrtic Electric Netwwrk Models
2.4 REACTIVE CAPABILITY LIMITS
It is important in voltage stability and long-term stability studies to consider the
reactive capability limits of synchronous machines. Synchronous machines are
rated in terms of maximum MVA output at specified voltage and power factor
(usually 0.85 or 0.9 lagging) which they can carry continuously without over-
heating. The active power output is limited by the prime mover capability to a
value within the MVA rating. The continuous reactive power output capability
is limited by three considerations: armature current limit, field current limit, and
end region heating limit.
Figure 2.12 demonstrates a family of reactive capability areas for three
different values of hydrogen coolant pressure. Note that the higher the pressure.
the larger the capability curve. In Fig. 2.12, the region AB is the field current
limited while the region BC is due to armature heating constraints.
t 0.6 p.f. lag t
Figure 2.12 Reactive capability curves of a hydrogen-cooled generator at rated \dtage.
2.5 STATIC LOAD MODELS
Con\re11 t ionit 1 transient -stab i 1i ty s t L i d i es were i n I v ed main 1y w it h gene rat o r s t ii .
bilit),. and little importance wits attached to loads. Recently, significant atteiitioii
has been g i \ w to load modeling. Much ot' the domestic load and some industrial
load consist o f heating and lighting. particularly i n the winter. and i n earl)' loati
111ode 1s t hese \+'erecons i dered iis constant i mpedances . Rot ii t i ng eclLI i pine n t
often inodeled ;IS ii siiiiple form o f synchronous machine and coniposite load
u'ere simulated by a mixture of these two types of load. A lot of uurk hits gon:
i n t o the development ot' more accurate load models. These include some corn-
ples niodels of specific large loads. Most loads. howe\ter. consist o f ;I I x g c
c l i i a nt i t y of cli \wse eqiii pmen t of v;iry i ng I c \ ~ ls atid coinposi t ion and some
cq11 i va I en Imodc1 is iiecessary .
A static l o d model expresses the churacteristic o f the load at m y instarit
ot' tirric in terms o f algebraic functions o f the bus \.()Itage niagnitude and t'w-
qi1cnc.y at that instant. The active power component P and the reacti\,c po\f.crr
c0111 pone 11t Q are c011 s i dered separate I y . ,A general load chxacteristic may be adopted such that the M V A loadii ~g
Lit :Iparticulx bus is ;I function of' lroltage m d fi-eclucncj~:
Q = K,/\'" f
$1here A',, atid A'(, are constant\ \+hich depend upon the nominal alue o f t ie
\ :triable\ P and Q. For constant frequencq operation. u e write:
~+hmP and Q are acti\,e and reacti\re coinponetits of the load \+(hen the hiis
\'oltage niagnitude is 1'. The siibscript 0 identifies the Fdiics o f the rcspccri\~
\ , x iiib Ics ;it the i ii i t i 11 I or iiom i nal c i perat i tig cond i t ion.
Static loads are reliiti\Jeljfiiniiffected by frecliicncj' change\, i.e.. ni, = 1 1 , =
0. and with constant impedance loads i i i , = 1 1 , = 2. The iniportance of' acciiiate
load models has been deriionstrated f o r \coltage sensitive loads. Figure 2. I3 dtbiii-
oiistratcs the po~verand current characteristics 01' constant poLver, constant c iir-
rent, and constant inipedance loads.
Manq reseiirchers identified the characteristic load parameters f:x \,iirioiis
homogeneous loads. iypical \,alues are shoLvii i n Table 2.1 . These charac*terihtic*\
t
IVI
Nominal voltage Nominal voltage
(a) (b)
Figure 2.13 Characteristic4 of different load models. ( a ) ActiLte and rcacti\c po\\er
\ersus voltage. (b ) Current \er\u\ \oltage.
inay be combined to @\re the overall load characteristic at a bus. For esample,
a group of homogeneous loads, each with a characteristic of N. j , and a nominal
power of P, inay be combined to give an overall characteristic of:
,=I
Table 2.1 Typical Values of Characteristic Load Parameters
Filament lamp I .6 0 0 0
FI Uoresc t' n t I amp 1.2 3.o - I .o -2.8
Heater 2.0 0 0 0
Induction motor half load 0.2 1.6 I .s -0.3
Induction motor fu l l load 0.I 0.6 2.8 I .8
Reduction furnace 1.9 2. I -03 0
A 1iiini nu m plant I .8 -.-3 3 -0.3 0.6
An iilternatii,e model which has been widely used to represent the Lroltagc:
dependence of loads is the polynomial model:
This nwdel is corninonly referred to as the ZIP model. since i t is composed
of constant impedance ( Z ) ,constant current (11, and constant power ( P )conipo-
nents. The parameters of the model are the coefficients P , to P , and Q , to
which define the proportion of each component.
When the load parameters m , and 1 1 , are less than or equal t o unity, a
problem can occur when the voltage drops to a low value. As the voltage magni-
tilde decreases, the current magnitude does not decrease. In the limiting c;i!,e
with zero cultage magnitude, ii load current flows m*hich is clearly irration: 1.
gikren the nondynamic nature of the load model. From a purely practical poi i t
of Lieut. then the load characteristics are only valid for a srnall Ldtage deviation
from nominal. Further, i f the voltage is sinall. small errors in magnitude a i d
phase produce large errors in current rnagnitiide and phase. This results in lo.;s
of accuracy and with iterative solution methods of poor convergence. The ie
effects ciin be overcome by using ii constant impedance characteristic to rep-
resent loads where the voltage is below some predefined ~ ~ a l u e , for example
0 . 8 p11. The parameters o f this model are exponents ci and h. With these exponerrts
equal to 0, 1 . or 2, the model represents constant power. constant current, or
constant i In pe dan ce char ;ic te ri s t i cs , respect i c'e1y . For composite I oads , thei r v i 1-
lies depend on the aggregate characteristics of load components.
The frequency dependence of load characteristics is represented by n iu l t i -
plying the exponential model or the polynomial niodel by ;I f'actor as follow>:
or
where Af is the frequency derivation (.f-.f;,).
CONCLUSIONS
I n this chapter we offered a review of power concepts for single and three phase
systems. We also treated the fundamentals of synchronous machine models for
stability evaluation including the idea of a reactive capability curk’e and static
load models. The per unit system was reviewed and extended to quantities not
frequently encountered, such as time and frequency. Also. some modeling as-
pects of static loads, such as frequency dependent loads in typical electric po\tw
systems, were also discussed.
The reader is referred to the bibliography section for references dealing
with materials in the chapter and contributions made by many other pioneers in
the field. Please note that an annotated glossary of terms is given to summarize
the key definitions and terminology employed in this chapter.
3 Dynamic Electric Network Models
INTRODUCTION
Chapter 2 focused on steady-state models to represent power system elemcr'is
for the s o ~.alledstatic analysis studies. We nou turn O L I ~ attention to inodt Is
for t ransiciit or dy niiiii ic operation a1 st l idies. Th i s cha pter describes soiiie sy s te n i
niodclh for analytical purposes. A model of an excitation system is studied i n
Section 3.1 and Section 3.2 gikres ii discussion of a model of' the prime nio\er
and go\~eriiorsystem. Dyii;iinic load models arc discussed i n Section 3.3 to
conclude the chapter.
3.1 EXCITATION SYSTEM MODEL
The basic function of an excitation system is t o probride ;i direct current to lit.
synchronous machine field Lvinding. I n addition, the excitation system performs
control and protecti\,e functions essential t o the secure operation of the s>rstm
by controlling the field Lmoltage and hence the field current 10 be within Licccpt-
;ibIe le\.e1s 11nder d i fferent operat i rig conditions.
The control tunction~ include the control o f iroltage and reiictiire po'i'er
tlow , thereby enhanc i ng pourer sy ste111 stabi I i ty . The protect i ve functions e n s i i re
t h ;i t t he cii pi1b i 1 i t y 1i 111its o t the sy iichronou s iii iich i ne , e sc itat i on s ~ ' st e111, In d
other ecliiipmcnt are not exceeded.
37
i
Power Power i i Power source
(regulator) source
(regulator) ; i
i j
'iource (exciter)
w 1 r i i Excitation 4 - 1
Reamplitier Power i : power
Amplifier i SWrCe 7 machine
I + i (exciter) - 1 I : I : I
I I I L - I
I
III
I
I I I
I
I I I
I
I II
I -I I II : I I
I
~ , ~ 74 Regulator i M a n u a L E ~ S + S y n c h r o n o u s ~ ~'iConlrol: ; machine 4 Excitation *
system Excitation
control system *:
Figure 3.1 Functional block diagram of a synchronous excitation control sy\tein.
Figure 3.1 shows the functional block diagram of a tj'pical excitation con-
trol system for ii synchronous generator. The following is ;i brief description of'
the various subsysteins identified in the figure. The exciter pro\sides dc p o ~ ' e r
to the synchronous machine field winding and constitutes the pouw stage of'
the excitation system. Usually an exciter is modeled by the first-order system
as shown in Fig. 3.2. The effect of saturation is considered by introducing S,
and K,,, and thus the exciter model is given by
1 (3.1 1E, =
1 + S,,+ K, + T,sv,
The \.ohage regulator processes and amplifies the input control \ign;il\ to ii
leiel and form appropriate for the control of the exciter. This include\ both
regulating and excitation system stabilizing function (rate feedbach or lead-lag
compensation). Nornially. the regulator is inodeled by a fir\t-order \ j stem a\
sho\+n i n Fig. 3.3. The regulator model is given by
Figure 3.2 Block diagram model of exciter
The terminal c oltage transducer senses the generator terminal voltage. recti-
fie\ and filters it t o dc, and compares i t with a reference which repre\ent\ the
h i r e d terminal c oltage. Figure 3.3 shows the terminal voltage tran\duc-er
model given by:
The power system stabilizer provides an additional input signal to the r e p -
lator to further damp power system oscillations. Some commonly used i n w t
signals iire rotor speed deviation, accelerating power, and frequency debiatirn.
A pouw system stabilizer is modeled as shown in Fig. 3.5. The PSS model
is gikwi by
A V = K,,G'(s)Au+ Ki,G(.\)At;' (.$.4)
U'here
Figure 3.3 Block diagram of the voltage regulator.
V S v F
Figure 3.4 Block diagram of a terminal voltage transducer model.
Limiters and protective circuits ensure that the capability limits of excitor and
synchronous generator are not exceeded.
The full excitation system is modeled as shown in Fig. 3.6. The system is
represented by the following set of linear differential equations:
T , & V - I j dt
?;C'VX=-V, -K(I : : , -y+l(+y)dt
T = V, - (S,+ K , + I ) E,dr
dVT,--t = -v+ &(& - (S , + K , + 1 ) E,)
d r T
X 2 1+T1S '"s -KPSS I + T5S l+T 2s 1+T2S
Figure 3.5 Block diagram of a typical Power System Stabiliter (PSS) model.
40
3.2 PRIME MOVER AND GOVERNING SYSTEM MODELS
The prinie soiirces of electrical energy supplied by utilities itre the kinetic enctg>'
o f U ater and the thermul energ~' deri\wl frorn fossil fuels m d nucleiir fission.
The prirnc rnoi.ers cot1Lw-t these soiirces of energj' into niechanicd entrgj ' tl- at.
i n tu rn . is con\,erted to electrical form by the synchronous generator. The pri iiie
m0 i .er go1.ern i ng sy ste111 prov i des a means of c011 t 1-0 ng pow er and t'rccl iicIIL'1'.I 1i
The !.Utict i oti;i 1 re I iit ion sh i p betwee11 the biis i c eIcme11t s ;issoc i iit ed wi t 11 PO\Ic I
gcner:itioii and control i x shokkrn in Fig. 3.7. This section introduces the i i i o c cls
I'oi- 111 draiilic turbines and go\,crning systems ;is well iis steam turbines atid tlicir
go\~crnitig s>~stcms.
3.2.1 Hydraulic Turbines and Governing System Model
The h>draulic turbine model describes the characteristics of' gate opening p :incl
o11 t p i t rnccha 11i cal power . I 11 power system dynam i c ;I 11a 1y s i s . the h >,drii LI I i c t LI r-
biiic is ~isuallymodeled by an ideal lossless turbine along U ith the consideration
ot' "wiitcr hiitiinicr" effect caused by the uwer inertia. is gic,en bl,
3.h)
M here T , )i \ \+raterstarting time.
BCC;ILIWo f the "LI. ater h;itiiiiier" effect, ;I change in gate position produce\
L
Energy Supply
System t
Speed Valve 1_1) Turbine Generator
Governor or
~
Speed
Figure 3.7 Functional block diagram of pobver generation and control.
an initial turbine pouer change which is opposite that \shich I \ de\ired. For
stable control performance, a large transient (temporari\ ) droop M it11 a long rc-
setting time I \ therefore required. This is accomplisht.d by introducing a tran-
sient gain reduction compensation in the governing \i\ \tern. The coiiipen\ation
retard\ o r limit\ the gate niovement unti l the mater flou and pouer output h a ~ c
time to catch up. The gmerning system model i \ \ho\+n in Fig. 3.8.
Pilot Valve max gateand
Figure 3.8 Bloch dingram of gojwning system for ;I hjdraulic rurbinc.
4 2
The goirerning system model is given by
7(-,4! = -p + x, t f t
( 3 . 7 )
= pilot v a l \ ~and ser\mi(itor time constant
K , = servo gain
7;7 = main servo time constant
R,,= permanent droop
R, = temporary droop
& = reset time
ql,,,,, ,,,, = maximum gate opening rate
q,ll,i,il,,rC = maximum gate closing rate p = gate position
3.2.2 Steam Turbines and Governing System Model
A \team turbine convert\ stored energy of high preswre and high teniperatiire
jteani into rotating energy. The input of the steam turbine i \ control \ ,al\e pvsi-
tion (Ayl).nthile its output is torque (AT,). In power \tability analj\is. a
order rnodel is used for \team turbine, i.e.,
ufhere r,,= time constant
Comparing the turbine models for hydraulic turbine and \team turbine, I t I \
clear that the re\ponse of a \teain turbine has no peculiarity wch a\ that extiib-
1tt.d by ;I hydraulic turbine d~ ie to water inertia. The governing requiremen('t o f
\team turbine\, in this re\pect, are more straightforward. There i \ no need for
t riin 4 i e11t droop coin pe n \a t ion.
The go\ erning \ j \tern model i \ given by
- - max gate
band 1r x l -
' @ KS - l + s T G
S
Figure 3.9 Typical block diagram of a steam turbine
cx= K , (U,, - U, - R,,X,) (3 .9 )lit
A typical governing model for steam turbine is shown i n Fig. 3.9.
3.3 MODELING OF LOADS
Load models are traditionally classified into two broad categories: static models
and dynamic models. Earlier, we considered the static load models (Chap. 2) .
In this section, the dynamic load model is discussed.
3.3.1 Dynamic Load Models
Typically, motors consume 60 to 70% of the total energy supplied by a power
system. Therefore, the dynamic effects due to motors are usually the most sig-
nificant aspects of dynamic characteristics of system loads. Modeling of motors
is discussed in this section.
An induction motor can be represented by the equivalent circuit shomm i n
Fig. 3.10. which accounts for quantities in one phase.
In the equivalent circuit all quantities have been referred to the stator side.
The directions of current shown are positive when operating as a motor, i n
which case the slip s is positive. The rotor equation of motion is given by
(3.10)
The torque (r)is slip-dependent,
44
Figure 3.10 f:qiiiwlc.iit circuit ol' ;I ihrw-pli;isc. iiiductioii in;icliiiic..
(3.1 I I
where /: is the nuinher o f poles atid s is the slip. def'incd ;IS:
,< = !'C ! I !
11,
bv i th 1 1 , being the synchronous speed of the machine. where P is the number of' pdes. aiicl,/'is the frequency. I n Eq. (3.I I ) we have
(K, +j .U,)U ,+,; x, ' .U,,,= J -K. +.; IX. + x,..,
I t is noted that Ecl. (3.10)represents ;I stendy state pcrformaiice nioclcl 0 1 the intluction motor. with all qiiantities referrecl to the stator siclc. There ;ire inodcls that represent the trnnsicnt performance of the niotor th ; i t are bnsccl on I l u x liiikngcs. volt;iges. and torque variutions.
CONCLUSIONS
In this brief chapter w e concentrated on models ibr power ~ystciii stahiliiy i n the tiiiie domain. We discussed electric excitation model as well as prime iiIover and governor system models. We concluded with a brief introduction to dy-namic load model. For fiirther reading on the topic. ii reference list ard an annotated glossary of teriiis is provided ;it the hack o f the book.
Philosophy of Security Assessment
INTRODUCTI 0N
We are concerned with the implications of il major network disturbance such as
a short circuit on a transmission line, the opening of a line or the switching 011
of a major load to name just a few. Here, we will consider the behaivior of the
system immediately following such a disturbance. Studies of this nature are
called transient stability analysis. The tendency of a power system to react to
disturbances in such a manner as to maintain its equilibrium (stay i n synchro-nism) is referred to as stability. One way of classifying disturbances is through
the categorization of small versus large. A disturbance is assumed to be small
if the behavior of the system can be adequately represented through a linsariza-
tion of the nonlinear system of dynamic equations of the system.
Stability considerations have been recognized to be among the essential
tools in electric power system planning. The possible consequences of instabilitj,
in an electric power system were dramatized by the northeast pobrer failure of
1965. This is an example of a situation that arises when il seLFere disturbance is
not cleared away quickly enough. The blackout began with a loss of ;I transmis-
sion corridor. which isolated a significant amount of generation from its lo:id.
More recently. a transmission tower in the Consolidated Edison sj~stem u ' a s hit
by a severe lightning stroke in July 1977. The events that follo\s.ed led to the
4 5
shutdown of power in New York City. Both events dramatized the consequences
o f iiii instability in an interconnected electric power system.
Our intention is to give ;in introduction to transient stability in electric
power sjrstems. We treat the ciise of a single machine operating to supply iir
infinite bus. The analysis of the tiiore complex problem of large electric po~vei'
riet\\.orks 1t.ith the i titerconnections taken into consideration is treated ;is istell.
4.1 THE SWING EQUATION
I n the p o ~ ser \j'stem\ engineer'\ terminology, the dynamic equation relating t h t s
inertial toryue to the net accelerating torque o f the s j nchrotious machine rotor
i \ called the s~ ing equation. Thi\ \imply \tate\
(1% , --
T',, i4.1 t tlt -
The left-hand-side i \ the inertial torque uhich is the product of the inerti,i
( i n hg In') o f all rotating tii;i\se\ attached to the rotor shaft arid the angul: r
:icccler:ition. The accelcrating torque T, , is in Nc\\ ton-meter\ and can be c'\ -
pres\ect as:
I n the abole , T,,,i \ the dritirig mechanical torque and i \ the retarding or
load electrical torque. The angular po4tion of the rotor 0 may be eupre\\ed ii\
the f o l l o ~ing win of angle\
The angle cc is a constant Nhich is needed if the angle 0 is mea\ured from
;in different from the iingular reference. The angle w,?t I \ the re\ult of the
rotor angular motion at rated \peed. The angle 8 i \ time \ a r j ing and repre\erts
de\ iation\ from the rated angular displacement\. This gi\ es the ba\i\ f-or our
1s~ i e I-c1;it ion
We find i t iiiore conLtenient to substitute the dot notation
Therefore we have
Jii' = T,,,- T,, (4.5)
4.2 SOME ALTERNATIVE FORMS
Some useful alternative forms of Eq. (4.5) have been developed. The first is the
power form which is obtained by multiplying both sides of by 03 and recalling
that the product of the torque T and angular velocity is the shaft power. This
results in
Jog = P ,,, - P,,
The quantity J o ~is called the inertia constant and is truly an angular mo-
mentum denoted by M (Jdrad.):
Thus the power form is:
M 8 = P,,,- P , ( 3 . 7 )
A normalized form of the swing equation can be obtained by dividing Eq.
(4.5) by the rated torque THto obtain the dimensionless equation.
The left-hand side of the above equation can be further manipulated to yield
a form frequently used. Recall the definition of the kinetic energy of a rotating
body. This gives the kinetic energy at rated speed as
1Wi = - J o ~
2
then
L 2 W ,-
TH o?,T,
We know further that the rated power is
-I8
Thus
Consequently we huvc
A constant which has proved very useful is c-noted by H. which is equ.11 to thc kinetic energy itt rated speed divided by the rntcd power PK
The units of H ;ire in sec. As il result we write the per unit or iiortiia1izi:d suing equation its
Obscrving thiit .,,, = P,,,,, wc cat1 then write
ZH 6 = p,,,- p , (4. 0) On
\vlicrc thc cquution is in p i i .
4.2.1 Machine Inertia Constants
Thc iltigttlitr iiionicntuiii inertia constant M as defined by Eq. (1.6) cu11be oh-tainctl from mnnufacturcr supplicd machine datil. The machine kinetic ~11crgy. ,V. iiiay be written i n ter111sof M i1S follows:
where wK is the angular speed in electrical degrees per second. This i n tttrtl is rclutccl to the frequency by
We c m therefore conclude that
The value of N is obtained from the moment of inertia of the machine
Lisually denoted by WR‘ and traditionally given i n Ib-f’. The con\-ersion formu-
la is:
The relation between H and M can be obtained using Eq. ( 3 . 8 )re\s<rittenas
H = -N (4.14)
G
Here G is the machine rating. Thus
(4.15)
The quantity H does not vary greatly with the rated power and speed of the
machine, but instead has a characteristic value or set of \ due \ for each cla\s of
machine. In the absence of definite information typical \ alues o f H may be used.
The cur\,es in Figs. 3. I . 3.2. and 4.3 give the general characteristic \ ariation of
H for exi\ting and future large turbo generators.
5 - -
#S Figure 4.1 Inertia conatantb for large turbo generator rated 500 M V A and below.
C'htrpter I
4 -
3.5 - -
A
CrJ
.i 3 - -
s 3 2.5 - -E I cn 2 2 -- -3600 RPM FOSSIL c
cn c 0 -C-1800 RPM NUCLEAR
1.5 - -2 -C
1 - -
0.5 - -
, 1 1 1 , , -0 I I
Figure 4.2 Expected inertia constants for liitiire large turbo generators.
In sy stern studies U here several machines ha\ing different ratings are wed,
the H constant for each machine, given to a base of the machine rating. rriu\t
be conberted to the coiiimon system base by multiplying H in Eq. (4.14)b j the
ratio (niachine base MVA/system base MVA).
4.3 TRANSIENT A N D SUBTRANSIENT REACTANCES
In order to understand the concept of transient and \ubtran\ient reactance\ of a
\ynchronous generator, let U\ consider the transient behavior during a balariced
t'ault. The dependence of the talue of the short circuit current in the ele:tric
p o ~ e rs j \ t e m o n the in\tant i n the cycle at which the \hart circuit occur\ can
be verified using ;I simple model. The inodel is ;I generaror uith wrie\
re\istance R and inductance L \ h o w n in Fig. 4.4. The xroltage o f the gene -ator
I \ a\\uriied to Lary a\:
5 I
4 - -I
3 3.5 - -F
$ 3 - -3 z I 2.5 - -v)c. c
tj 2 - - -A 450-514rpm0 0
.E 1.5 - - +B 2Ul-400rpm
r +C 133-180rpm -t
1 - - +D 80-lXIrpm
0.5 - -
. . . . . . . . . . . . . . . . . . . . 0 -
Figure 4.3 Inertia constants of large water wheel generator\.
r ( t )= E,,,sin(cot + 00
With a balanced fault placed on the generator terminals at t = 0. then we
can show that a dc term will in general exist. Its magnitude at t = O may be
equal to the magnitude of the steady-state current term. The transient current i(r) is given by
where
The worst possible case occurs for the value of a g i \ m by:
j w L R
In t h i \ case the ciirrent magnitude ill approach titice the \tcadq -\talc II a \ -
i n i i i i i i \ alue iinmcdiiitely after the \hart circuit. The tran\ient current I \ pi\ -
Thus:
E,,,i ( t )= -1z
I t is clear that the maximum of i (r) is twice that of E,,,/Z.This n~a\~efortii is
shown in Fig. 4%
E,,,For the case tana = or we have i ( t )= - sinot. This a.a\.eform is shourn i nz
Fig. 3.Sb.
I t is clear from inspection of either the expression for the short circuit cur-
rent or the response u'ajseform given in Fig. 4.6 that the reactance of the ma-
chine appears to be time varying. This is so if we assume a fixed Lfoltagesource
E. For our power system purposes we let the reactance vary i n a step-wise.
fashion X:, Xg;and X , , as shown in Fig. 3.6.
The current history i ( t ) can be approximated in three time zones by three
different expressions. In the first, denoted the subtransient inten-al. lasting up
to 2 cycles. the current is I". This defines the direct axis subtransient reactancc
The second denoted the transient gives rise to
Figure 4.5 Short circuit current waLCeforms.
/
Ex; = I'
kchere I' is the transient current and X:, is the direct axis tran\ient reactance. Tie
transient intenfa1last5 for about 30 cycles. The steady-state condition gi\ses t 76'
di rect ii x i \ synchronous re;ict ance.
Note that the subtran\ient reactance can be as low a s 7% o f t h e \ynchronc U\
re;ict;i nce.
4.4 S Y N C H R O N O U S M A C H I N E M O D E L IN STAB1LlTY ANALYSIS
A brief outline of equations to account for flux changes in a synchronous ma-
chine is given to define various electrical quantities and to construct phasor
diagrams. The following approximations are involved in the models discussed:
I . The rotor speed is sufficiently near 1.0 pi( and may be considered a
constant.
2. All inductances are independent of current. The effects due to satura-
tion of iron are not considered.
3. Machine winding inductances can be represented as constants plus si-
nusoidal harmonics of the rotor angle.
4. Distributed windings may be represented as concentrated windings.
5. The machine may be represented by a voltage behind an impedance.
6. There are no hysteresis losses in the iron, and eddy currents are only
accounted for by equivalent windings on the rotor.
7. Leakage reactance only exists in the stator.
Under these assumptions, classical theory allows constructing a model of
the synchronous machine in the steady-state, transient, and subtransient states.
The per unit system adopted is normalized, although the term ”propor-
tional” should be used instead of “equal” when comparing quantities. Note that one p i i field voltage produces 1 .O p i field current and 1 .O p i open-circuit ternii-
nal voltage at rated speed.
4.4.1 Steady-State Equations
Figure 4.7 shows the flux and voltage phasor diagram for a cylindrical rotor
synchronous machine ignoring all saturation effects.
The following comments explain the construction:
I . The flux @, is proportional to the field current 1, and the applied field
voltage and acts in the direct axis of the machine.
2. The stator open-circuit terminal voltage E, is proportional to @, which
is located on the quadrature axis.
3. The voltage E, is proportional to the applied field voltage and may be
referred to as E/.
4. When the synchronous machine is loaded, a flux Q> proportional to and
in phase with the stator current I and when added vectorially to the
field flux @, gives an effective flux a,.. 5. The effective internal stator voltage E, is due to @, and lags i t by 90”.
6. The terminal voltage V is found from the voltage E, by considering the
cf,
Direct Axis
t I I I
I! -- b
Quadrature Axis
\ oltage d r o p due to the Ieahage reiictance X and iiriiiatiirc rt" i+-
tatice K . 7 . B) +iniilaritj 01' triangle+. the dit'terencc betbeen E, and E i \ i n ph i\e
U ith the / X \ oltage drop and i \ proportional to /. Theret im the \ o l ~ i g e
difference n i a j bu treated ;i\ ;i koltage drop ;icro++;in iirniatiire rcict-
ance x, .
The 'riitii o f X,,and X i is termed the synchronous reactance, x>. For the salient pole synchronous machine. the phasor diagram is more coin-
pies. Bec~iii'rethe rotor is syninietrical about both the d and c/ ;ixes i t is cot- \.u-
nient to rusol\~enianj' phasor quantities into coniponents i n these ;ises. 'The
stator currunt niajc be trtxited i n this manner. Although a,,ujill bc proportional
to I,:iind a,,will be proportional to /,,, because the iron paths i n the two :,xes
;ire diftcrent. the total iirniatiire reaction tlus 0 \+. i l l not be proportional t o / iior
necessarily be i n phase \+rith i t . Retaining our earlier normalizing ;tsutiipti )ns.
i t inay be assumed that the proportionality between !*, and @</,is i i n i t y but tlic
proportionality between and @(, is less than uni ty and is a function of the
saliency.
4.4.2 Salient Pole Synchronous Machine
Figure 4.8 shows the phasor diagram of the salient pole synchronous machine. The
d and q axes armature reactance4 are developed a4 in the c) lindrical rotor caw. Di-
rect and quadrature 4ynchronous reactance4 X,, and X,, can be establi\hed. i.e..
From the phasor diagram we have:
where V+ and V,. are the axial components of the terminal voltage V. In steady-state conditions i t is reasonable to use the field voltage E, or tht!
voltage equivalent to field current E, behind the synchronous reactances as thc machine model. In this case the rotor position (quiidrature axis) with respect t o the synchronously rotating frame of reference is given by the angular position of E,.
As the cylindrical rotor model may be regarded as ii special case ofn salient machine ( X , ,= X, , ) , we will consider only the salient pole machine.
4.4.3 Transient Equations
For faster changes in the conditions external to the synchronous machine. the steady-state model is no longer appropriate. Due to the inertia of the f lux l i n l -ages these changes cannot be introduced throughout the whole of the modcl inimedintely. I t is essential to establish new fictitious voltages E,; and E,;. reprc-senting thc t lux linkages of thc rotor windings. These transient voltages call t-e shown to exist hehind the transient reactances X,; and X,;,
E:,= v,,+ R,,I,,- la,x<; E*;= V,,+ RJcl+ l , ,X;
The voltage E, is now considered as the sum of two voltages. E,,and E,, aid is the voltage behind synchronous reactance. In steady-state. current flows only in the field winding and hence, in that case, E,,= 0 and E, = 0.
Allowing for the rotor t lux linkages change with time requires using t’ie following ordinary differential equations:
-E ,SE,;= -r,:..
The phasor diagram of the machine operating in the transient state is s h o w in Fig. 4.9.
Direct Axis E l
i E ,
Figure 4.9 Phasor diagram of a salient pole synchronous machine in the transient
state.
4.5 SUBTRANSIENT EQUATIONS
Other circuits exist in the rotor, either intentionally, as in the case of damper
windings, or unavoidably. These circuits are taken into account if a more exact
model is required. The reactances and time constants involved are small and
can often be practically disregarded. When required, the development of these
equations is identical to that for transients and yields:
The equations are developed assuming that the transient time constants are
large compared with the subtransient time constants. A phasor diagram of the
synchronous machine operating in the subtransient state is shown in Fig. 4.10.
4.6 MACHINE MODELS
I t is feasible to expand the model even further than the subtransient level but
this is rarely done in multi-machine stability programs. Investigations using a
Direct Axis
generator inodel with up to seven rotor \+~indiiigs.ha\,e shown that using the
\tandard niachine data. the I~IOI-e complex models do not necessu-ily j'ield accti-
ixte results. Ho~.t . \ .er , improved results can be obtained if the data, cspeciall>.
t tie t i nie constan ts. are appropriate 1y modified.
The m o s t contwiient method of treating sq'nchronous machines of differir-g
coniplcxit>' is to allow e x h machine the I ~ I ; I S ~ I I ~ L I I ~possible number o f ec1u.t-
tioritr a i d then let the actual model used be cleterniined autoinatically accordir g
to rhe data presented. Thus, fibre models are possible tot ii four-\s~indingro tor
4.6.1 Model 1
4.6.2 Model 2
4.6.3 Model 3
cl- and q-axis transient effects requiring two differential equations (SE:.and ,SE). The follou~ing equations are used. A block diagram is shown in Fig. 3.11 .
E:,= v,,+ R,,I,,- l(/x:/
E:,= \'(, + RJ, , + 1,,x,;
U
Figure 4.1 1 Block diagram representation for model 3.
4.6.4 Model 4
ti- and q-axis subtransient effects requiring three differential equation\ (
$E::,and \E:: ). The following equations are used.
4.6.5 Model 5
t I- a id (1- axi s s i i b t ra11s i e11t effec t s reqi i i ri n g f o ur cl i fferen t i a1 ey ii ;I t ioris ( .;E;. .SE,:. and .SE:;). The follouing equations are used..SE:;,
The following mechanical equations need to be solFred for all these models
s6 = O - OH
4.7 GROUPS OF MACHINES A N D THE INFINITE BUS
Groups of synchronous machines or parts of the system may be represented by
a single synchronous machine model. An infinite busbar, representing a large
stiff system, may be similarly modeled as a single machine represented by
model 1 , with the simplification that the mechanical equations are not required.
This sixth model is thus defined as:
4.7.1 Model 0
Infinite machine-constant voltage (phase and magnitude) behind ci-axis transient
reactance (X: , ) . Only the following equations are used.
4.8 STABILITY ASSESSMENT
In this section, we discuss the conventional approach to stability assessment
applicable to a single machine against an infinite bus. The method leads to the
equal area criterion.
We concluded that a simple representation of the salient pole machine is
offered by the model 0 given by:
U'e \+i I 1 make the to1lowi ng add i t io nal a \ s 11111pt ions:
A\ ;I result:
I:' = I' - I X'
0 = \ ' + I X '
The output po\ier of ttic niachine is g i \ w b j :
\ ' = I ' c o d
I ' , = -\ ' , 4iii i5
The phasor diagram is st1ou.n in Fig. 4.12.
The electric P O M ' C ~output of the salient pole machine is thereti)re given b j :
The L ariation o f thc output for- salient pole rnachinc L+ i th the torcliie o r
po\srer angle 6 is \ho\s,n in Fig. 4. 13.
Network
Imaginary Axis Machine
Quadrature Axis
Direct A xis
Real Axis
"d
Figure 4.1 2 Synchronous niachine and network frame\ o f rc!crence tor dcvclopiiig
electric po\ver oii t pii t formu la.
= Pe
Figure 4.1 3 Power angle characteristics for a salient pole machint..
66
output Power, P
Angle, S (radians)
Figure 4.1 4 Power angle characteristics for ;I round rotor machine.
I n the case of ii round rotor machine, u'e have Xf: = X ( ; and hence
E V , .P , = -- sin6
x:,
The ~ariation of the output power for a round rotor machine uith the angle
6, (torquc or poww angle) is shown i n Fig. 3.14.
Ewmple 1
A \ynchronou\ machine is connected to an infinite bus through a transformer
ha\ ing ii reactance of 0.I p i and a double-circuit transmission line uith 0.45 pi' reactance for each circuit. The \ystem is shown in Fig. 4. IS. All reactance\ are giken to a ba\e o f the machine rating. The direct-axi\ transient reactance of t ie
rnachine i \ 0. 15 p i t . Determine the \sariation o f the electrical pomw uith :ink le
6. Assullle v = I .o p.
SoILitio17
An equitralent circuit of the LtboLre system is shown in Fig. 3.16. From this ~ s ' e
ha\-e the following: X c ( ,= 0.475 p i .
.
Figure 4.15 System for example 1 ,
Changes in the network configuration between the tuo sides (sending and
receiving) will alter the value of Xcyand hence the expression for the electric
power transfer. The following example illustrates this point.
Example 2
Assume for the system of Example 1 that only one circuit of the transmission
line is available. Obtain the relation between the transmitted electric powrer and
the angle 6. Assume other variables to remain unchanged.
Solution
The network configuration presently offers an equivalent circuit as shoikw i n
Fig. 4.17.
For the present we have
Figure 4.16 Equiident circuit for example 1
E L 6
Figure 4.17 Ecliii\ dcnt circuit tor cuainplc 2
Thcretore.
0bserl.e that the ii1;ixiiiiiitii alue o f the i i m clin e i \ l o ~ v rthan the OI c
corrc\ponding to the prec ious example.
4.9 CONCEPTS IN TRANSIENT STABILITY
111 order to gain ;in under\t;iiicling of the concept\ in\ ol\cd iri transient \tahility
prediction. u e k v i l l concentrate or1 the \implil'icd netuorh con\isting of' ;i \eri:\
reactance connecting the machine and the int'inite h i \ . LTiider thew condition\
oiit. pou er cxprtxion recliice\ to
For .\iniplicity of notation u'e \ \ ' i l l ;issiiiiie stcadJr-state \.alues. An iiiiport;int
assumption that we adopt is that the electric changes iii\.olc,ed ;ire much faster
th;in the resulting mechanical changes produced b), the gencrator/tirrbine spc ed
coiitrol. Thus \i'e ;Issiiiiie that the rnechanical p o u w is ii constant tor the purpose
o f triinsieiit stabilitj, cdculations. The functions P,,,iiiid P, are plotted i n Fig.
-4.18.
The intersection o f the t L i ' o functions dcfincs tn 'o values for 6. The lo~re r
\~:ilueis denoted by 6,,.Consequently, the higher is n: - 6,,according to the \>m -iiietrj of the ciir\'e. At both points P,,,= P , , that is ti16/dt' = 0 and L\,C s a j ' that
thc s)'stciii is in ecliiilibi.iitiii.
Power, P (P.u.)
?
(radians)
Figure 4.18 Pou er-angle cun'e.
Assume that a change in the operation of the system occurs such that 6 is increased by a small amount A& Now for operation near 6,,. P, > P,,, and
c126/c/t' becomes negatiLre according to the swing equation. Thus 6 is decreased.
and the system responds by returning to its stable operating or equilibrium point.
We refer to this as a stable operating point. On the other hand, operating at n -6,, results in a system response that will increase 6 and mo\re further from n -6,).For this reason, we call x - 6,)an unstable equilibrium point.
I f the system is operating in an equilibrium state supplying an electric
pourer P,,, with the corresponding mechanical power input P,,,,,,then the corre-
sponding rotor angle is 6(,.Suppose the mechanical power P,,,is changed to P,, , ,
at a fast rate, which the angle 6 cannot follow as shown in Fig. 3.19. I n this
case P;,l> P, and acceleration occurs so that 6 increases. This goes on un t i l [he
point 6, where P,,,= P,, and the acceleration is zero. The speed, houtvw, is not
zero at that point, and 6 continues to increase beyond 6, . I n this region P,,,< P,
and rotor retardation takes place.
The rotor will stop at 6 where the speed is zero and retardation uill bring
6 down. This process continues on as oscillations around the new equilibriiini
point 6, . This serves to illustrate what happens when the system is sub-jected to
a sudden change in the power balance of the right-hand side of the suting equa-
70
Power, P (P.u.)
(radians)
Figure 4.19 Po\scr angle curve.
tion. The \ituation described abo\fe will occur for 5udden change\ in P, ;i\ M I : I I The \ js tem discussed in Examples 1 and 2 \er\re\ to illustrate th i \ point. nhich
u e discii\\ further in the next example.
E\i,?niplc. 3
The \j'sterii of example 1 i \ deliirering iin apparent power of 1 . I p i ( at 0.85 PF
lagging with two circuits of the line in ser\ ice. Obtain the \ource voltage (ewi-
t a t i o n koltage) E and the angle 6 under these conditions. With the \econd cir(*uit
open a\ i n Example 2. ;i new equilibrium 1 1 1 angle can be reached. Shetch the
power mgle curves for the t w ' o conditions. Find the angle 6,,and the electric
POMer that can be tran\terred immediately follow ing the circuit opening, ;I\ \ L ell
i i ) 6 , . A\\ume that the excitation \ oltage remain\ unchanged.
S oILItion
The power delivered is P,,= S cos $, P,,= I . I x 0.85 = 0.94 pi 1
Using P = VI COS @, then the current i n the circuit is I = -~ cos ' 0.85 1 . 1
Thus we can Urite
71
E = V + j X l
1 + (1.1i - 31.79")(O.37SL90")
= 1.28 +j0.44 = I .3S i 19.20" p i r
Therefore, E = 1.35 P.u., 6,,= 19.20" The power angle curve for the line with two circuits according to Example
1 , is
P,,,= 2.1053 x 1.3Ssin6 = 2.83sin6
With one circuit open, the new power angle curve is obtained as in Example
2, thus
3 P,, = I .43 x 1.3Ssin6= 1.93sin6
The two power angle curves are shown in Fig. 4.20.
From inspection of the curves, we can deduce that the angle 6,. can be
obtained from
P,,,= P,sinS, (curve B)
0.93 = 1.93sin6,
6, = 29.15"
We can obtain the value of electric power corresponding to with one
line open as
P,.,o= 1.93sin19.2"
= 0.63 pi4
4.10 A METHOD FOR STABILITY ASSESSMENT
To predict whether a particular system is stable after a disturbance i t is necessary
to solve the dynamic equation describing the behavior of the angle 6 immedi-
ately following an imbalance or a disturbance to the system. The system is said
to be unstable if the angle between any two machines tends to increase without
limit. On the other hand if under disturbance effects, the angles between e\.erq'
possible pair reach maximum value and decrease thereafter, the system is
deemed stable.
Angle. 13 (radians)
Assuming as we hwc :ilrcady done that the input is constant, ncglipi 3le doniping iuid coiist;itit source \dtagc behind the tnitnsient reactance. the angle hetwccn two machines either increase indefinitely or oscillates iif'tcr all disturh-;iiicch h;i\.c occurred. Thcrcforc'. in the c;isc of tw:o machines, these \ \ i l l either f d l out of' step on the first swing or never. Here the obscrvution that the iiia-
chineh angular differetices stay constalt can be tiikt.tl ;IS ill1 indication 01' hystctii
\tahility. A simple method lor determining htiibility knowii iis the c'cluiil-; re;i tiicthotl is nvnilable. We will discuss this here.
The h\vitig ccluation for ;I machine connected to ; in infinite bus ci111he writ-tell a s
We obtain an expression for the variation of the ringirlar speed o w i t h ?. by noting the alternative form
P0 tlo, = (16
M
Integrating, assuming cc) = 0 and integrating the abo\,e equation, \IT obtain
or
The above equation gi\ves the relative speed of the machine M i t h r-eclpcct to
a reference frame nim ing at a constant speed (by definition of the angle 6). I f
the \ystem is stable, then the speed must be zero when the acceleration i \ either
zero or i \ opposing the rotor motion. Thus for a rotor n4iich i \ accelerating. the
condition for stability i \ that a value of 6, exists such that
This condition is applied graphically in Fig. 4.2 I where the net area under
the P,,- 6 cirr~~ereaches zero at the angle 6 as s h o w . Obsenve that at 6,,.P, is
negative and conseqirently the system is stable. Observe that the area A equals
A! as indicated.
The accelerating power need not be plotted to assess stabilitj,. Instead. the
same information can be obtained from a plot of electrical and mechanical pout-
ers. The former is the power angle curve and the latter is assumed constant. 111
this case the integral iiiay be interpreted as the area between the P, c i r r \ ~ and the c~rr~re of P,,,both plotted versus 6. The area to be equal to zero. must consist
of a p0sitii.e portion A I.for which an equal and opposite negatiLre portion of ,4
must exist, for which P,,,< f,,. This explains the term equal-area criterion for
stability. This situation is shown in Fig. 4.22. I f the accelerating power reverses sign before the t ~ ’ oareas A and .4? ;ire
equal. synchronism is lost. This situation is illustrated i n Fig. 3.23. The areki A :
is smaller than A , and as 6 increases beyond the value \!,here P(,re\’erxes sign
again, the area A j is added to A , .
71
A
Figure 4.21 S t ;Ihi I i t J condition for accelerating rotor.
Power, P
P m
-0 6,
b Power
Angle, 8 (radians)
Figure 4.22 Equal-area c*r.itcriori tor \tabilit!,.
75
Accelerating
Power, P , I
Figure 4.23 Accelerating powrer as a function of the torque angle.
Example 4
Consider the system of the previous three examples. Determine whether the
system is stable for the fault of an open circuit on the second line. If the system
is stable, determine 6,, the maximum swing.
Solution
From the examples given above we have
6,,= 19.2"
6r = 29.50'
The geometry of the problem is shown in Fig. 4.24. We can calculate the
area A immediately:
?U IS
A I = 0.94 [29.15 - 19.20In - 1.93sin6 t /S
1*0 I0 2 0
Observe that the angles 6, and 6,)are substituted for in radians. The re-
sult is:
A I = 0.0262
The angle 6,. is
76
L
Power, P (P.u.
2.84
1.93
0.94
0
6,= 1x0 - 6 , = 15o.xf;
This clcurly gives
; init the system is stable. The iingl~.6,. i s obtained hq solving tor A I = A, . Here we get:
77
This gives some algebra
I .93 c o d , + 0.0164 6, = 2.1376
The solution is obtained iteratively as 6, = 39.39"
This example shows the application of the equal area criterion to the case
of a generator supplying power to an infinite bus over tu'o parallel transmission
lines. For the loading indicated above the system is stable. The opening of one
of the lines may cause the generator to lose synchimism e\'en though the load
could be supplied over a single line. The following example illustrates this point.
Example 5
Assume that the system in Example 3 is delivering an actiLre power of 1.8 p i 1
using the same source, voltage E, as before. Determine urhether the sqrstem u?ill
remain stable after one circuit of the line is opened.
Solution
We have for the initial angle 6,,
I .8 = 2.84 sin&,
6,,= 39.33"
The angle 6,. is obtained from
I .8 = 1.93 sin6,
6 , = 68.85"
The area A , is thus:
OX x s
A , = I .8(68.85- 39.33jn - 1 1.93sin6(16 = 0.13 I 8o 31) 7 3
The area A_.is obtained as:
ii
It A,= j 3.86\inti (16 - 1.816,- 6,1- = o.o6
180
We note that A I > A:, and the system is therefore unstable.
I f ii three-phase \hart circuit took place at ;i point on the extreme end o .
the line. there i \ some impedance bet\veen the generator bus and the load ( i n f i -
tiite) h i \ . Therefore, \ome p o ~ e ri \ tran\mitted cvhile the fault i \ \ t i l l on. Tht.
\ituation I \ \iiiiiIiir to the one\ anal ) l ed abo\ e and \ie ii\e the follou ing e'iamplt:
to illu\trate the point.
Ewmple 6
.A generator is deli\,ering 25%- o f PI,,,,to an infinite bus through ;i transmission
line. A f a u l t occurs such that the reactance betuwn the generator and the bii4
i \ incrcased to tuv times its prefuult value.
I . Find the 6,,before the faiilt.
2. Shoci, graphicallq, c i , h a t happens when the f i i u l t is sustained.
3. Find the niiisiriiiiin \raliie of 6 s\\ting i n case of ;I sustained l'ault.
Solution
Figure 4.25 illustrates the situation for this example. The amplitude of the po\i t r
anglc c i i r ~ e tttith the f;iiiIt su\tained is hal f o f the original \-alue.
Before the t'itiilt \+e h a t e
At the t 'u i i l t instant, \4fe get
A \ before. the \tability condition yields
Hence
0 . 5 C O b 6 , + --<6, = 0.5473K
7'
Bq trial and error
6,= 46.3"
A
Power, P (P.u.)
0.50
0.25
Torque
0 Angle, S (radians)
Figure 4.25 Power angle curves for example 6.
The following example illustrates the effects of short circuits on the net-
work from a stability point of view.
Example 7
The system of the previous examples delivers a power of 1 .O p i when subjected
to a three-phase short circuit in the middle of one of the transmission circuits.
This fault is cleared by opening the breakers at both ends of the faulted circuit.
If the fault is cleared for 6, =50°,determine whether the system will be stable
or not. Assume the same source voltage E is maintained as before. I f the system
is stable, find the maximum angle of swing.
Solution
The power angle curves have been determined for the prefault network i n Exam-
ple I and for the postfault network in Example 2. In Example 3 we obtained
E = 1.35 plr
j0.225
Figure 4.26 NctLforh coni'iguration ciuring rhe faitlr.
There t 'rm
During the fault the netnvrk offers ;I different cont'i~uratioti.L+ hic.t- i \
\ho\\ 11 i n Fig. 4.26. We \ + i l l need to reduce the netnorh in wch ii u ; i j ;I<, to
obtain ;I clear path l'rotn the +oiirct' to the infinite h i \ . We do thi \ b j u\iiig ;I 1'
- A tr~iii\t'ortnationii\ indicated in Fig. 4.27.
)(0.25)= I ,2x = 0.45 + 0.25 + (0.45 0.235
and fri i i l t pourer angle curve is given by
f = 1, I3 sin6
The three power angle curl'es are shown i n Fig. 1.28.
The initial angle i h gilren by the equation
1 .O = 2.84 sit&
6,,= 20.62"
The clearing angle is 6, = SO"
Power P (P.U.
2.a
/ \ I .93
1.13
I .a
y i i i II I I I , I I . Torque
6 f ~ n g ~ e .S (radians)
Figure 4.28 Pre-fault. during fault, and post-fault poww angle CLII'VL'S t o r t.xii11113lc 7
S2
The area A I can thus be crilculated as:
The maximum area A ? is obtained using the angle 6,
I = 1.93 siii6:
6;'=31.21"
6,= I80 - 6; = 148.79'
NOW
We note that A I > A?,and the system is theretore stable.
T o calculate the angle o f iiiiisiiiiurn swing we ha\fe
Hence
Bjr trial and error
6, = 66.3''
4.1 1 MATHEMATICAL MODELS A N D SOLUTION METHODS IN TRANSIENT STABILITY ASSESSMENT FOR GENERAL NETWORKS
I t is coiiinioii practice to model static equipment in the transmission sj.! ten1
by lumped equicralent pi parameters independent of the changes arising i n the
generat ing and load eq U i p men t , Th i s approach is enip1oyed i n mU 1t i miichi ne
stabilitj programs because the inclusion o f time Lrarying parameters ~voulcl pro-
duce major computational difficulties. Moreover, frequency, the most obtrious
variable in the network, usually varies by only a small amount and thus, the
errors involved are insignificant. Additionally. the rates of change of network
variables are assumed to avoid the introduction of differential equations into the
network solution. The transmission network can thus be represented in the same
manner as in the load-flow or short-circuit programs, that is, by a square com-
plex admittance matrix.
The behavior of the network is described by the matrix equation:
I,,, = YV
where I,,, is the vector of injected currents into the neturork due to generators
and loads Y is the admittance matrix of the network, and V is the vector of nodal
vol tages.
Any loads represented by constant impedances inay be directly included in
the network admittance matrix with the injected currents due to these loads set
to zero. Their effect is thus accounted for directly by the network solution.
4.1 1.1 System Representation
Two alternative solution methods are possible. The preferred method uses the
nodal matrix approach. while the alternative is the mesh matrix method. Matrix
reduction techniques can be used if specific network information is not required.
but this gives little advantage as the sparsity of the reduced matrix is usually very much less.
Nodal Matrix Method
In this method, all network loads are converted into Norton equi\-alents of in-
jected currents in parallel with admittances. The admittances can be included i n
the network admittance matrix to form a modified admittance matrix which is
then inverted, or preferably factorized by some technique so that solution at
each stage is straight forward.
The following solution process applies:
I . For each network load, determine the injected currents into the modi-
fied admittance matrix by solving the relevant differential and algebraic
equations.
2. Determine network voltages from the injected currents using the Z-
matrix or factors.
As the network voltages affect the loads, an iterative process is often re-
quired, although good approximations can be used to avoid this.
With the Nodal Matrix method, bus voltages are available directly and
branch currents can be calculated if necessary.
4.1 1.2 Synchronous Machine Representation in the Network
The equations representing ;I )syiichroiioiis miichine ;ire gi\ren in the form of
The\,cnin voltages behind its impedances. This must be modified to ii currint
source in parullel urith ;in admittmce using Norton's theorem. The x h i i t t m c e
of the machine thus formed may be added to the shunt admittance o f the 11i;i-
chine bus and treated as a netuwk parameter-. The \ ector /,!/ thus contains the
Norton equi ktalent c iirrcii t s of the sj'iichroiious iiiach i nes , The sj'iichronous 11I a-
chine equations are kvritten i n a frame OF reference rotating ivith its o \vn rotor.
Thc real anci imaginary coniponenth o f the iietwork equations, iis giLrt.11 i n Fig.
4.29. art' obrai 11ed !I-( )I I I t h t' fo1I o c i ' i iig t ranstornia t i on
The in\,erse relation is
Network
t imaginary axis
Machine Machine
direct axis
The transformation also applies to currents.
When saliency is accounted for, the subtransient and transient reactances i n
direct and quadrature axes frames are different, and the Norton shunt admittance
will have a different \ d u e in each axis, and when transformed into the netLivrh
frame of reference. will be time varying. To circum\rent this difficult). a con-
stant impedance is used while modifying the injected current.
4.11.3 Load Representation in the Network
To be suitable for representation in the overall solution method, loads niust be
transformed into injected currents into the transmission netM-ork from kvhich the
terminal voltages can be calculated. A Norton equivalent model of each load
niust therefore be introduced. In a similar way to that adopted for synchronous
machines. the Norton admittance may be included directly in the net1iu-k admit-
tance matrix.
A constant impedance load is therefore included in the network admittance
matrix and its injected current is zero. This representation is extremely simple
to implement. causes no computational problems, and impro\.es the accurricy of
the netufork solution by strengthening the diagonal elements in the :idmittatice
matrix. Nonimpedance loads may be treated similarly. I n this case the stead>,-
state values of voltage and complex power obtained from the load tlow are uscd
to obtain a steady-state equivalent admittance (7,))which is included i n the net-
urork admittance matrix [ Y ] . During the stability run, each load is s o l k ~ dse-
quentially along with the generators, etc., to obtain a new admittance (r).i.e.:
The current injected into the network thus represents the de\siation of the
load characteristic from an impedance characteristic.
By converting the load characteristic to that of a constant impedance, M hen
the \toitage drops below some predetermined value ( I/,,,,,,),the irijected current
i \ kept relatikrely small. An example of a load characteristic and its correspond-
ing injected current is shown in Fig. 4.30. In an alternative model the low-voltage impedance is added to the net\sorh
and the injected current compensates for the decriation from the actual character-
istic. In this case, there is a nonzero injected current i n the initial steady-state
operational condition.
86
I I )
power
Figure 4.30 Load and injecled currents for a con4tant type load U ith l o ~ 'iolta;e
adju\tiiient. ( a ) Load current. ( h ) Injected current.
4.1 1.4 System Faults and Switching
I n general most power system disturbances to be studied will be caused h y
changes in the network, normally caused by faults and subsequent switchi ig
action. Occasionally the effect of branch or machine switching will be consid-
ered.
Although faults can occur anywhere in the system, i t is much easier comru-
tationally to apply a fault to a bus. In this case, only the shunt admittance at the
bus need be changed, that is. a modification to the relevant self-admittance o f
the Y matrix. Faults on branches require the construction of ;i dummy bus at the
fau l t location and suitable modification of the branch data unless the distar ce
between the f w l t position and the nearest bus is small enough to be ignored
87 Philosophy of Secirr-ity Assessriierir
The worst case is a three-phase zero-impedance fault and this involves plac-
ing an infinite admittance in parallel with the existing shunt admittance. In prac-
tice, a nonzero but sufficiently low-fault impedance is used so that the bus
voltage is effectively brought to zero. This is necessary to meet the requirements
of the numerical solution method.
The application or removal of a fault at an existing bus does not affect the
topology of the network and where the solution method is based on sparsity
exploiting ordered elimination, the ordering remains unchanged and only the
factors required for the forward and backward substitution need be modified.
Alternatively the factors can remain constant and diakoptical techniques can be
used to account for the network change.
4.1 1.5 Branch Switching
Branch switching can easily be carried out by either modifying the relevant
mutual- and self-admittances of the Y matrix or by using diakoptical techniques.
In either case, the topology of the network can remain unchanged, as an open
branch is merely one with zero admittance. While this does not fully exploit
sparsity, the gain in computation time by not reordering exceeds the loss by
retaining zero elements, in almost all cases.
The only exception is the case of a branch switched into a network where
no interconnections existed prior to that event. In this case, either diakoptical or
reordering techniques become necessary. To avoid this problem, a dummy branch may be included with the steady-state data of sufficiently high imped-
ance that the power flow is negligible under all conditions, or alternatively, the
branch resistance may be set negative to represent an initial open circuit. A
negative branch reactance should not be used as this is a valid parameter where
a branch contains series capacitors.
Where a fault occurs on a branch but very close to a bus, non-unit protec-
tion at that bus will normally operate before that at the remote end. Therefore,
there will be a period when the fault is still being supplied from the remote end.
There are two methods of accounting for this type of fault. The simplest method only requires data manipulation. The fault is initially
assumed to exist at the local bus rather than on the branch. When the specified
time for the protection and local circuit breaker to operate has elapsed, the fault
is removed and the branch on which the fault is assumed to exist is opened.
Simultaneously, the fault is applied at the remote bus, but in this case, with the
fault impedance increased by the faulted branch impedance, similarly the fault
is maintained until the time specified for the protection and remote circuit
breaker to operate has elapsed.
The second method is generally more involved but i t is better when protec-
tion schemes are modeled. I n this case, a dummy bus is located at the fault
position. (even though i t is close to the local bus) and a branch with a \U-:,!
small impedance is inserted between the dummy bus and the local bus. Th:
faulted branch then connects the dummy bus to the remote bus and the branc i
shunt susceptance originally associated with the local bus is transferred to the
dummy bus. This may al l be done computationally at the time when the fault i s
being specified. The two branches can now be controlled independently by sui ,-
able protection systems. An advantage of this scheme is that the fault durativn
need not be specified iis part of the input data. Opening both branches et'!'el:-
ticely isolates the fault, \+rhich can remain permanently attached to the dunin
bus. o r if auto-reclosing is required. i t can be reiiiowd automatically after ;I
su i tabIe dei on i ~at i on period .
I f the network is not being solired by ii direct method, the second method
will probably fail. During the iterati1.e solution of the network, slight volta,;e
errors nrill ciiuse large currents to flow through ii branch with ;I 1w-y smill
impedance. This will slow convergence and in extreme cases uill c~iuse diwr-
gence. With ii direct method, based on ordered elimination, ;in exact solution o f
the bus Lwltages is obtained for the iiijected currents specified at that particuhr
iteration. Thus, pro\.ided that the impedance is not s o small that numerical prob-
lems occur L+,hen calculating the admittance, and the subsequent fwtors for 1 he
forward m c l backward substitution, then convergence of the ow-al l solut ioii
betu.een machines m c l netL+x)rk will be unaffected. The k-aluc. o f the lo\+!-impcd-
ance branch between the cluriiriiy and local bus may be set at II fraction of the
total branch impedance, sub-ject to a minimum value. I t this fraction is u n k r
0.001. the change i n branch impedance is \ w y small compared to the iiccur ic>r
of the netuwk data input and i t is unnecessary to niodify the iiiipedance of' the
branch from the remote to the dummy bus.
4.1 1.6 Machine Switching
Machine su,itching may be considered, either as a network o r machine opera-
tion. I t is ~i netu.ork operation if a duiiiiiiy bus is created to which the machine
is connected. The dummy bus is then connected to the original machine hub b!,
;I I O U * - i mpedance branch.
Alternati\rely. i t may be treated as a machine operation by retaining the
original n s t N v r k topology. When a machine is sw.itched out, it is necessai y to
reniove its iri-jected currcnt froni the network solution. Also a n y shunt admit-
tance included i n the network Y matrix, which is due to the machine mu!,t be
re n10ved. Although a disconnected machine can play no direct part i n system stability,
its response should still be calculated as before, nrith the machine stator current
set to zero. Thus machine speeds, terminal \.oltages. etc.. can be obser\xd ci'eii
when disconnected from the system and in the ek'ent of reconnection, sensible
results are obtained.
When an industrial system is being studied many machines may be discon-
nected and reconnected at different times as the ~ ~ ~ l t a g e le\rel changes. This
process will require many recalculations of the factors in\-ol\,ed i n the forumd
and backward substitution solution method of the netmwk. However, these can
be avoided by using the method adopted earlier to account for synchronous
machine saliency. That is, an appropriate current is iiijected at the relt'\mt
buses, which cancels out the effect of the shunt admittance.
4.12 INTEGRATION TECHNIQUES
Many integration methods have been applied to the power \ysteni tran\ient \ta-
bility problem, and the principal methods are discus\ed nou.
4.12.1 Predictor Corrector Methods
These methods for the solution of the differential equation
.sY = F ( Y . X ) with Y ( 0 )= Y, ,
and X ( 0 )= X , ,
hakre all been detreloped from the general k-step finite difference equation:
Basically the methods consist of a pair of equations, one being explicit
( P I ,= 0) to give a prediction of the solution at t(rl + I ) and the other bein,0 1111-.
plicit ( P I ,f 0) which corrects the predicted value. There are a great Larietj of
method\ aLrailable, wch as hybrid methods, the choice of which being made bq
the requirements of the solution.
I t is usual for simplicity to maintain a constant step length with these meth-
od\ if X > 2. Each application of a corrector method inipro\e\ the accurac~ of
the method by one order, up to a maximum given by the order of acc~iracy of
the corrector. Therefore, if the corrector is not to be iterated. i t i \ common to
use a predictor with an order of accuracy one less than that of the corrector.
The predictor is thus not essential as the value at the preLrious step may be u\ed
as a first crude estimate, but the number of iterations of the corrector may be
1arge.
While for accuracy, there is a fixed number of rele\mt iteration\, i t is
desirable for stability purposes to iterate to some predetermined level of conver-
gence. The characteristic root ( z , )of a predictor or corrector when applied to thc
single variable problem
may be found from
Applying a corrector to the problem defined and rearranging gives:
the solution to the problem becomes direct. The predictor is now not necessaiy
as the solution only requires information of 1' at the previous steps. i.e.. .it
\v(ii = i + 1 ), for i = 1 , 2,. . . . . k . Where the problem contains two variables, one nonintegrable, such that:
then
where
Although and ~ i , , , ~are constant at a particular step, the solution is itera-
tive.
Strictly in this simple case, x ( i i + 1 ) could be eliminated but in the general
multivariable case this is not so. The convergence of the method is now a f u IC-
tion of the nonlinearity of the system. Provided that the step length is suffi-
ciently small. a simple Jacobian form of iteration gives convergence in only a
few iterations. It is also possible to form a Jacobian matrix and obtain a solution
by a Newton iterative process, although the storage necessary is much larger
and the step length must be sufficiently small to ensure convergence.
For a multivariable system, the following two equations are coupled
.sY = F( Y , X ) with Y ( 0 )= Y,)and X ( 0 )= XII
and the solution of the integrable variables is given by the matrix equation.
The elements of the vector L' , ,+~are given by the vector form of
and the elements of the sparse matrix are given by
The iterative solution may be started at any point in the loop, if Jacobian
iterations are used.
4.12.2 The Euler Method
Consider the following ordinary differential equation:
Let x(t) be the state vector of this nonlinear differential equation, which is to be
solved by an appropriate integration technique. The Euler method utilizes a
predictor function based on the Taylor series expansion of .r(t + A[), uthere At is
the step size. As such, by neglecting the higher order terms in the series expan-
sion, we obtain the generalized Euler's formula:
d.r(t)s ( t + A t ) = x ( t )+ x(t).Ar where i ( r ) = __
at
The method is not often used for real-time applications in power systems, as i t
is computationally burdensome. Also, the accuracy of this model for integration is sacrificed by the truncation of O[At)'] and higher order terms done in the
Taylor series expansion. Figure 4.3 I outlines an example of Euler's integration
technique to power system dynamic stability assessment.
4.12.3 The Modified Euler Method
Reconsider the following problem to be solved:
i ( t )= f I ( s ) , t ]
Set Loop Control time: T,, & & Set Integration Step size: At
Flow Calculations and obtain P,(t), the machine injection in p.u. MW
Solve the system of differential equations for all machines:
?f w ( t ) = --(& - 4"
H
I Set Fault Time: tf = t s = o - w R'
' w R
=2nf
Perform 'during fault'Power Flow Calculations
Figure 4.31 The transient stability algorithm
@ Q Using EulerS Method, Solve the machine angles:
6( t )= u(t)-2 q
6(t+ Ar) = & r ) + 8(f)Ar
Yes I
SaveDisplay Final Results:
~~
I ~ ~ ~ ~ ~ ~
Set Fault Clearing Time: & = t I Torque Angle, Machine Speed, Generated
Obtain‘new’state of the machine Power, Times (t,
andor the network perturbations tf. tmarr etc.) and i so on.
Figure 4.31 Continued.
where x ( f ) and.f'(x, t ) are scalar or vector quantities. Expanding . d f ) to the riglit
and left using the Taylor expansion yields:
If the O[(At)'Jand higher order terms are truncated for a small step size, i t c;in
be shown that the modified Euler formula is obtained as follows:
Alternatively, we may bfrite:
Note that the modified Euler inethod is not self*-starting and thus requires an
initial prediction on the state variables in the \rector solution.
4.1 2.4 Trapezoidal Method
The trapezoidal rule is a nonself-starting integration technique that is somewhat
related to the modified Euler method and is based on the geometric interpreta-
tion of the problem. The order of truncation in the Taylor series expansion of
.v( t k Ar) is O[(A?)'].The trapezoidal formula is:
The merits of the trapezoidal method is evident in the accuracy brought forward
iis a result of the truncation of the O[(At)']and higher order terms.
4.1 2.5 Runge-Kutta Methods
Runge-Kutta methods are able to achieve high accuracy while remaining sir gle
step methods. This is obtained by making further evaluation of the functions
within the step. Here we present a class of self-starting prediction formulae,
which are applicable i n the assessment of transient analysis of rotating na-
chines, in which case we are solving the swing equation of the generators.
The numerical integration of the ordinary differential equation given b).:
by a predictor generally involves calculation of xn + 1 as ii function of x,,, Y,, I,
x,,-:, . . . . fll. f,,-l, fl,-., . . . , and tl,. I f the predictor is self-starting, i t must be free
of terms in xn-l ,x,,-?, . . . , f,, fll-l, fIl-? and so forth. That is, the next state solution
sould be of the form: x,,+~= F(x,,, fl,, t"). Now, the equation below summarizes
this requirement and is called the Runge-Kutta predictor formula.
where k ( i )= h f ir ( n ) + c,h, ! ( / I ) + 111
/=I
and the sum of all the weights equal to unity, i.e., w, = 1 I= I
The coefficients are uniquely determined, giving rise to various orders and
associated approximate or accurate models of the Runge-Kutta predictor for-
mula. Table 4.1 summarizes the 2"d,3rd,and 4Ih order Runge-Kutta formulae.
Being single-step these methods are self starting and the step length need
not be constant. If j is restricted so that j < i then the method is explicit and c I
must be zero. When j is permitted to exceed i, then the method is implicit and
an iterative solution is necessary.
Table 4.1 Summary of the RK-2. RK-3, and RK-4 Predictor Forniulae
Runge-Kutta Predictor Formulae Coefficients
RK-2
.Vil+] = .Vll + -h (k"' + k ' ? ' ]
2
RK-3
.Vfl+l = .Vfl + -It [k"' + + k'?']
6
RK-3 k' ' = F(.vl,,t1,)
= .vil + -I t (k"' + 2k'!' + 2k'" + kf4'] h6 k'?'= F!.Vl t + - k " h + - i2 2
Explicit Runge-Kutta methods have been used extensibdy in transient sta-
bility studies. They have the adkmtage that a packaged integration method is
iisually available or quite readily constructed and the differential equations at-:
incorporated with the method explicitly. I t has only been with the introduction of no re detailed system component models with very small tiiiie constants, th:,t
the problems of stability has caused interest i n other methods.
Again. the advantage o f the Runge-Kutta techniques is that they are all seli'-
starting. The propagated error from one iteration to the next does not increa>e
rapidly, thus the inethod is said to be stable. The exception is the solution 1 0
"stiff' prohleins, &(here the solution may diverge unless a small step size is used.
(The stiffness of the ordinary differential equations ot' the system can be iiie;isiirL d
iis ii functiuon o f the ration between the smallest to the largest eigeindues of the
linearized system or eigenvector analysis of the Jacobian matrix.) Ne\wtheless.
they are not ~ e r yattractive to pourer system engineers iis k poww flow calculatioris
must be done at every iteration. where k is the order of the Runge-Kutta forinula.
The tollou,ing esnmple demonstrates the RK-4 application.
Exc?niple
Application o f the 1"' er \ J \ tc inorder Runge-Kutta integration techniqiie to p o ~
dynamic \tability asws\iiient. The algorithm is giL en belou.
I . Start the RK-4 Subroutine.
2. Initialize all R K - 3 dependencies.
COllIlt, d = 0: Step Size. At
Time. t = 0: M iix i mum Time, t ,11,1,
Maximum iteration = itniax
3. Sol\e the initial pouer flow and the machine equation\, and obt.iin
the generator torcllie angle (6(0)). terminal voltage\. bu\ in.jection\
P;"'(o).etc.
4. InitialiLe the RK-4 Coefficients: k , = I,= 0 for all i E { l , 4 }
5. Set the e\tirritrfc iirc/c.\ to: i = d + I .
6. Perform RK-4 calculations (obtaining the estimate\).
97
7. Increment count: d = d + I . 8. Increment the estinicitr i1ii1e.vto: i = i + I .
I9. Update the torque angle delta to: 6( t )= & ( I ) + 5 k , and sol\^ the pom'er
flow for I v , l,Py"(t), etc.
10. Compute the new coefficients.
Ik , = ( ~ ( 0 )+ - I , - ) - 2Ff)At
2
I , = w ( f i l i - f : " ' ( t ) )At2H
I I . If d < 3. then go to step 7, otherwise continue.
12. Compute the final value of the power angle and the machine speed at
t = t + A t
16(t + At)= 6(t )+ -(ki + 2k: + 2 k , + k, )
6
ICO(t + At) = CO( t )+ -( 1, + 21, + 21, + 1 4 )
6
13. Reset the count: d = 0
14. Compute the final power flow for this time interkal to get 1 V, 1 . P?( t 1.
etc.
IS. Increment the timer ( t = t + At) and make the following comparisons:
-If ( t< t,,,,,,) and ( N 2 itrncix) then flag the user: "Maximum Number of iteration reached!" and go to step 16.
Elseif ( t< t,,,,,,)and ( N < itmax) then increment the iteration count
to: N = N + 1 and go to step 4.
Else Flag the user: "Run time value of tmax reached!" and con-
t i n Lie,
16. Display all results, power angle, machine speed. generator p o ~ ~ e r .gen-
erator terminal voltage, etc.
17. End of subroutine.
4.12.6 The Sampling Method (James J.Ray; James A. Momoh, et al.)
Another promising development in this area is the sampling method. I t u'as
developed at Howard University with collaboration from the Energy Systems
Network Laboratory (ESNL) in 1989. The sampling method, also known as the
"theta" method. takes advantage of the expanded Taylor Series in its formulatior
and approximations. It was incorporated in a stand-alone program capable oi'
performing stability studies and fault analysis on a wide variety of electric,
power systems. The sampling method has been tested against the classical inte-
gration methods. A 4: 1 speed advantage was observed, without sacrificing thr:
accuracy brought forward in the results. Extended application of this method i r i
power system stability studies is being developed for commercial purposes, both
as a design tool and as a computational support system.
4.13 THE TRANSIENT STABILITY ALGORITHM
An overt'iew of the structure of a transient stability program is given in Fik.
4.32. Only the main parts of the program have been included, and as can be
seen, the same system may have several case studies performed on i t by repear-
edly specifying switching data when no further switching data is available. Cor -
trol returns to the start to see if another system is to be studied. With care, the
program can be divided into packages of subroutines each concerned with only
one aspect of the system. This permits the removal of component models when
not required and the easy addition of new models whenever necessary. Thus for
example, the subroutines associated with the synchronous machine, the AVRs,
speed governors, etc., can be segregated from the network. Figure 4.32 shows ii
more detailed block diagram of the overall structure where this segregation is
indicated. The diagram is subdivided into the five sections indicated in Fi;.
4.33. While the block diagrams are intended to be self-evident several logic
codes need to be explained. These are as follows:
KASE
This is the case study number for a particular system. It is initially set to zero
and increniented by I at the end of the initialization and at the end of each case
study.
KBlFA 7
The sparse vectored inverse of the nodal network matrix is obtained using three
bifactorization subroutines. The first and second subroutines are integer routires
which determine bus ordering and nonzero element location. The code KBIFA 1
is set to unity if i t is necessary to enter these two subroutines, otherwise i t is
set to zero.
r+=? Read in steady-state system data Section 1
Section 2
Section 3
r
Either Solve for machines and network
or recalculate nonintegrable variables
(initially and when switching has occurred) Section 4-
1. Print out results (if necessary)
2. Make power balance check of
initial conditions (if necessary)
NO 1
YES NO
Section 5,
- Store initial Reset initial
conditions conditions
Figure 4.32 The transient stability algorithm.
I 0 0
Synchronou s machine AVR calculations ‘wed governorNetwork calculations calcuyns,~ ,~A ,f c a l c u l j y ~ n s ,
KASE =0
KBlFA I = 0
KBIFA3 = 0
input
YES
NO
V
Figure 4.33a Section I
KBIfA3
The elements of the spar\e-vectored inverse are evaluated in the third bifactor-i-
zation subroutine. The code KBIFA3 is set to unity if i t i \ necessary to en er
thi\ wbroutine, otherwise i t i \ set to Lero. When KBIFA3 i \ unity, i t indica e\
that a netiborh discontinuity ha\ occurred and hence it i \ a lw u\ed for t b i \
purpcx .
Time
The integration time.
H
The integration step length. Like KBIFA3, it is also used t o indicate a disconti-
nuity when i t is set to zero.
4 Switching
Determine bus order
and nonzero element
location for bifurcation
KBIFA3=1 TIME=O
++ Is KASE2IIs KASE2I NO NO
and is thae noand is thae no
switching a t T I M E = O ?switching a t T I M E = O ?
YESYES ** II K B I F G P syncmachine *AVR initialKBIFmSyncmachine
t, conditions.initialinitial
YES conditionsconditions + Speedgov. 1 T initialinitial
conditionsconditions
I 4 numerical part
of bifurcation
Print out initial
conditions
when U S E > 1)
Figure 4.33b Section 2.
I 02
v A -7-TDetermine
switching ops.
if branch change
set KBIFA3 = 1
if new branch
set KBIFAI = 1
I
‘ES -
Determine bus
order and non zero
element location
for bifactorization1 KBIFA3 = 0
part of
KBIFA3 = 1
Figure 4 . 3 3 ~ Section 3.
Philosophy of Secirrity Axsessnierit
I Solve for network
KBIFA3 = 1<>-TIME =
E- PRINT TIME
YES Print out
Print out busbar Print out AVR results
and branch results sync. machine
if required results
4 speed gov.
a results if required
Set flag for
step doubling
KBIFA3 = 0
I
Figure 4.33d Section 4.
I04
A v YES YES
r 1
NO Perform power balance
check to confirm
NO
'End of case '
TIME = 0 KASE = KASE + 1
1 KASE = 1 ?
r-? Store initial
steady - state)
(steady -state)
conditions
Figure 4.33e Section 5 .
PKlNTTlME
The integration time :it which the next printout ot' result\ is required.
MAXTIME
The predet'ined mixiniiiiii integration time for the case study
Start solution Lr'
Calculate constants
nonintegrable variables ,-,-,-,,,,-,,---' +I speed gov. calc
I not usually requireda- ------- - - ---- - - -1
Evaluate integrable Same for cach A
varaible using algebraic b Same for each
form of mpezoidal method 4 speed gov.
I
[HALF= 0
Figure 4.34a Section I
ITMAX
Maxiinuin number ot' iterations per step since las t printout of results.
Note that iiiany data error checks are required in a program of this t j p but
they ha1.e been omitted from the block diagram for clarity.
i
106
Figure 4.34b Section 2
I
107
v A YES
NO
Re -evaluate conditions
at beginning of step
H = H / 2
IHALF = IHALF + I
NO
'Not converging '
TIME = MAXTIME
I
Figure 4 . 3 4 ~ Section 3.
I ox
Strcictiire ot Machine '2nd Netivork Iterative SoI~iti017
The ~ - u c t u r e of this part of the program requires further description. T u o forris
o f \olution are possible depending on whether an integration step is being etralu-
ated or if the nonintegrable Lariables are being recalculated after ii discontinuitq,
A bloch diagram is gi\en in Fig. 4.33. The ridditional logic codes uwd in this
part of the program are:
ERROR
The niiiXiniuin di t'ference betu een an) integrable 1 ariable from one iteration to
;in() t her.
I TK
N u in ber of iterations reqii i red for solution.
IHALF
Number ol' inimediate step halving required for the solution.
TCILERANCE
Specified niiiYifnuni \ alue of ERROR for conc ergence.
I f coil\ erpence ha\ not been achieved after a \pecified number of iterati on\
the ca\e study is tei-ininated. Thi\ is done by \etting the integration time cc ual
to the iiia\iiiiuiii integration time. The latest rewlts are thu \ printed out ;iild ;i
neu c;i\e \tudj i \ atternpted.
CONCLUSION
This chapter dealt urith the philosophy of securit), assess~nent based on fre-
quency domain models arid equal area criterion concepts. In particular. we de-
fine the conventional ingredients for power systeni stability including app ica-
tions of the swing equation and its alternate forms.
Frequency domain models of synchronous rnachines introduced the idca of
subtransient. transient, and steady state reactances. Models for a salient pole and
round wound synchronous machines were discussed. The equal area criteria and
its applications were discussed. The chapter concludes with the treatnieiit of
transient stability for the general network case, including ;I floMxhart of ii tran-
s i eri t s t ii bi 1i t 4' assess men t pro gra ni . Again, the reader is reminded that additional information inay be obti.ined
from the list o f references and the annotated glossary of terms.
5 Assessing Angle Stability via Transient Energy Function
INTRODUCTION
In the actual operation of an electric power system, the parameters and loading
conditions are quite different from those assumed at the planning stage. As a
result. to ensure power system security against possible abnorinal conditions
due to contingencies (disturbances), the system operator needs to simulate con-
tingencies i n advance, assess the results, and then take preventiLre control action
if required. This whole process is called dynamic security assessment (DSA)
and preventive control.
Simulation studies (called transient stability studies) can take up to an hour
for a typical system with detailed modeling for a 500-bus. 100-machine system.
Since i t takes a long time to conduct a transient simulation even for a single
contingency, direct methods of stability assessment such as those based on Lya-
punov or energy functions offer attractive alternatives.
It should be noted that a transient stability study is often more than an
investigation of whether the synchronous generators, following the ~ccurr-ence
of disturbance, will remain in synchronism. It can be a general-purpose transient
analysis. in which the "quality" of the dynamic system behavior is in\,estigated.
The transient period of primary interest is the electromechanical transient. LISLI-
ally lasting up to a few seconds in duration. If growing oscillations are of con-
cern, or if the behavior of special controls is of interest. a longer transient period
may be covered in the study.
For transient stability analysis, a nonlinear system model is used. The
system is described by a set of differential equations and a set of algebraic
eq11at ion s. Genera 11y , the differentia 1 equations are m ilch i ne equations , cont ro
system equations. etc. The algebraic equations are system \,ohage equation:*
involLting the network admittance matrix. The time simulation method and direc
method are often used for transient stability analysis. The former method deter.
mines transient stability by solving the system differential equation step by step.
while the direct method determines the system transient stabiliry without explic.
i t l y solving the system differential equations. This approach is appealing and
has receiLved considerable attention. Energy-based methods are a special case o f
the more general Lyapunokr’s second or direct method. the energy function being
the possible Lyapiinov function.
This chapter deals with transient stability by a specific direct method rnainl;,,
the transient energy function (TEF) method. We begin by co\wing some basic
concepts from the theory of nonlinear system stability.
5.1 STABILITY CONCEPTS
Consider an autonomous system described by the ordinary differential equatior .
where i =.t(t), and F ( A )are n-vector\. F(.t) is generally a nonlinear function ( ~ f
t . Stability in the sense of Lyapunob i \ referred to an equilibrium state of Eq.
( 5 .I ). The equilibrium \tate is defined as the stage i r at which t ( t ) remaii \
unchanged for all f . That is,
The solution for -tr from Eq. (5.2) is a fixed state since F(.t) is not an explicit
tunction of t . For convenience, any nonzero ir is to be translated to the origin
( t = 0). That is. to replace t by .I + i t in Eq. ( 5 .I ) to have
which gives
15.3)
I
*fit /--“’
I.... ................................. ........................ ...*........................ “ ....................... +
6 .
It0
Figure 5.1 Illustration of local stability.
Note that the current s differs from the old one by 4. As can be seen later,
from the definitions, this translation does not affect the stability of the system.
Thus, the origin of Eq. (5.3) is always an equilibrium state. I t should be noted
that t in Eq. ( 5 . 3 )may be any independent variable, including time.
5.1.1 Definitions and a Lemma
Stability (Local)
The origin of the system described by Eq. (5.3) is said to be stable if for any
given E > 0, there exists a 6 5 E such that ll.r,~ll < 6 implies ll.r(t)ll < E for all t
where so is an initial state. The origin is called unstable if i t is not stable.
The concept is illustrated in Fig. 5.1 where the initial state x,) has a magni-
tude less than 6. and the trajectory of s remains within the cylinder of radius E .
Asymptotic Stability (Local)
The origin of the system described by Eq. (5 .3) is said to be asymptotically
stable if it is stable and also if given ~ ~ . x l ~ ~ ~< 6 implies s -+ 0, ( I S t -+00. The
arrow is used to mean “approach,”
Figure 5.2 demonstrates the idea of asymptotic stability where the trajectory
tends to 0 as time tends to infinity.
Figure 5.2 Illustration of asymptotic stability.
--
Glolm11~~Asjmptotic- Stnbil it)/
The origin o f the \ystem described by Eq. ( 5 . 3 ) is said to be globally asymptoti-
call> stable i f i t is \table and also implie\ -+0 cis I + 00, for any .v,, in th?
\v ho I e space.
Posit ive Deti'nite Function
A uniquely defined. scalar and continuous function V(.v) is said to be positi\,c
definite i n ii region K if V(.v) > 0 for .v # 0 and V ( 0 )= 0.
A space surt'ace formed by all .v satisfying W . v ) = 0 is called a contour.
Ob\.iousl>r. contours with different \.alues cannot intersect one another. I t ' the>
do. \ ' ( .v) has tM'o \~al~iesat the intersection. We need the follou.ing lenimii r t b -
qiiirecl for the proof o f ii Ljrapunov theorem.
LcYl117lC?
There exists a sphere defined by l[.vll = N i n which V ( . v ) increases iiionotoiiical'4,
along radical Lwtors emanating from the origin. That is, V(PI,) increases niono-
tonicully L b i t h p i n 0 5 p IN for any unit \rector I I started from the origin.
This ciin be shown using the assumption of positi\.e definiteness. First, coil-
t i nu i ty . V(.Y)> 0 and V ( 0 )= 0, assiiiiie that V(pI1)increases monotonically \bri .h
p i n an interval 0 Ip I pIIand begins to decrease after p = pII.GiLfen ii I I .there
i h ;in associated pIl Lbrhich may be unbounded pll(= 00). Let \ t - among all the U ' S
that has the sniallest pi,. then IIp\, n*II= pill\1t.11 = PI, 5 pII. Since \'( PI,) increas$:\
monotonically nith p i n the inter\,al 0 5 P I pi, 5 p,,. \+'eare able to identify the
positi\,e number to be N = pi,.
L i ' c i p i ~ ~ i o vTheoreni
There are three important theorems on 5tability de\ eloped by Lyapunov. V/e
include thew here to form the theorem gi\ten below. In the theorem, V(i)i \ t i e
total d u i \ atikc of \'( t ) on the trajectory \pecified by Eq. (5.3).That i \ ,
\'( \ ) = tll' = \'\( \ ) \ = 1' ( 1 I / ( t
tlt
Uhere \',( i)i \ the row \ ector formed by the partial deri\ ati\res of V(.v).
Regi017~R, RI, R,
A11 the regioris ;ire assuincd to contain the origin ;is an interior point. R: i < , ;I
\Libregion ot' K , \+,hicti is ;I siibi-egion of R : K: I K , I K.
Theorem
Let V ( x )be a positiise definite function with continuous partial deri\rriti\ves i n a region R, then
The origin of the system described by Ey. (5.3) is stable if i ' ( .v) I0 i n
a subregion R, I R.
The system is asymptotically stable in the region if i t is stable and i'(.v) =0 (identically zero) takes place only at the region i n a subregion R.
I RI. The origin is globally asymptotically stable if the sqstem is asymptoti-
cally stable. R2 is the whole space and V(.v) + 00 as 11.v11 -+0.
I *Let be the srnaller one between a given E and N specified i n the
lemma. That is. I - = Mill [ E . N I .
Continuity of V(.\-)assures that there is a m i n i m u m of 1T.v) on the
sphere ll.vll = t*. Let j* be among all the 11's that yield the minimum.
V ( I - , )= I U . then V(I*,,)2 111 must hold for any 1 1 . Monotonicity tells that
V(PI,) = 111 for a p i n the interval 0 I P I I - , and hence \'(.v = PI,)= i i i is
enclosed by the sphere ~ ~ . r ~ ~= I",and also i t is a closed contour since I I
is any uni t vector.
Let 6 be the minimum norm of the points on the closed contour \/(.I-)
= 1 1 1 , then since ~ I x , ~ ~ ~< 6 is enclosed by the contour, V(.v,,) < 1 1 1 ~ O I I O M ~ S by monotonocity. Thus, any trajectory initiated from .v,) cannot possi-
bly cross the sphere II.vll = t - 5 E due to the fact that \'(.v) is non-increas-
ing and V(.v) 2 1 1 1 on the sphere II.YII = I - .
This completes the proof of part I of the theorem.
Part 2 can be shown by observing V(.\-)can be identically zero only at the
origin. Hence, V(.v) keeps on decreasing except at a countable number of points
at which i t stops decreasing momentarily. This implies that .v +0 L \'L' I - + -
since V ( 0 )= 0 only when .v = 0 in R2. I t seems ob\.ious t o have part 3 verified by the same reasoning used i n part
2. This is true except for the case when V ( x )approaches a finite value as ll.v]l + 00 when .v,) is allowed to be any point in the whole space. The assumption that
V(.v) 4 a s \l.vll +- excludes this possibility which completes the proof of the00
the ore ni . Except for relying on experience, there is no systematic method to find the
Lyapunov function as required by the theorem. It has been shown that any stable
and constant linear system has ii V ( x )but no one has yet shonm its existence i n
ge ne ra 1 for non I i near sy stems.
I14 Chciptrr 5
It can be hhown that instability and asymptotic stability of the system dt:-
scribed by Eq. (5.3) are the same as its linearized system at the origin, which is
where A =J;(O) is a constant and n-square matrix. That is, the system describd
by Eq. (5.3) is unstable if at least one eigenvalue of A has ii positive real px t
and is asymptotically stable if all eigenvalues have negative real parts. Asymptotic stability as judged from the linearized system is simple but 0 1
less practical use. I t is valid only in a sufficiently small region which is not
easily known. The Lyapunov function contains more information on stability i n
the regions R , and R2.For instance. global asymptotic stability tells us that the
trajectory initiated from anywhere in the whole space, converges to the origiii.
5.1.2 Application of Lyapunov's Method to the Simple Pendulum
We consider the dynamics of a pendulum as a prototype for exploring stability
of an electric power system.
The motion of a pendulum with friction is described by
where -n: < n: < rl is the angle, C I is the damping constant and dh is the undamped
angular velocity. We regard the problem as a mathematical one without rest-ic-
tion on 8. To convert the system to the standard form of Eq. (5 .3) . we detine
that .vI = 8 and .i,= .Y? to get:
First, we want to check if i t is possible to find a L Y ~ ~ U I W V function for the
problem. The matrix A for the linearized system is:
Both eigenvalues of A have negative and real parts when N > 0. This sugge;ts a
possibility of finding a Lyapunov function.
The Lyapunov function is sometimes referred to as a generalized energy
function. The name comes from the fact that i t has an initial positive kralue and does not increase as time goes on. This is the case of physical systems that iiiok'e
without interference from the outside such as the pendulum system. To see this,
let M. J , p and I be the mass. inertia, damping constant and the length of the
pendulum, then one dynamic equation becomes
J 6 + pi 6 + M ~ Isine = o
Multiply by iv and then integrate with respect to t :
The total energy of the pendulum (sum of kinetic and potential energjr) is
The time derivative is
Therefore, the proposed Lyapunov function is given by
with
V ( x )= -ari
Thus, the Lyapunov function is actually the total energy per unit inertia of
the pendulum system.
Now. we have
1 7
V(.r) = h( 1 - cosx,) + - .rj2
in the region
R = { -271: <x < 271:;free .r2)
The derivative of 11 is
1 . The origin is stable since the condition of part 1 is satisfied by choosiiig
R I = R. This is also true for the choice ( I = 0. which is the case o f ii
frictionless pendulum.
2. V ( . v ) = -(i.vi =0 implies that .v2 0 and hence .i:= 0. This makes .i-,=
-u.v2 - h sinv, = 0 which yields sinvl = 0 or .vl = /in. Therefore. \f(.v) = 0 only at the origin by choosing R, = { - n < .vl < n: ,ftw. v 2 ) . Thus, t ie
origin is asymptotically stable since the condition o f part 2 of the thco-
rein is satisfied in R2.
3. Part 3 is not applicable for the follcnving t M ' o reasons:
( i ) \/(.I-) = 0 at .vl = 2 1 m i n the whole space,
( i i ) .vl + 00 Mfith.v2 = 0 makes V(.v) = h( I - cos.^,) 5 2 b which does i r o t
approach infinity iis required.
For demonstration purposes, M'L' employ the proposed \'(.v) to find ;I 6 5 E
iis required by the definition of stabilit}!. To this end. let I ( = [ l r , , I ( : ] ' be any u n i t
i w t o r started from the origin, then
3 7 7
on the circle .I + .v> = 1.- o r \l.vll = I - clV/t/.v,= h siiul - .I I < 0 for h I I and .vl > 0. Since \ / ( .I ) is a decreasing function, i t ha\ a niinuinum at .t l = + r - and .I = 0.
Therefore, ui = h( I - CO\/-). The expression for the minimuin becomes coirpli-
cated for h > 1 .
on the contour \'(.v) = 1 1 1 . c/ /A/ .v I = 2(.vl - h sin.\l)> 0 for- .vl > 0 and h I I .
Since the minimuin occ~trslit .vl = 0 and .v2 = &U, we haie the rninitwrn
norm
The inequality holds for the reason that
117
For the case b > I , we may choose an independent \variable a = wf \ii t h CO' = 1 7 .
Thus. the original dynamic equation becomes
where 0 = &/da and CI' = cdw. Hence, the results of (b) and ( c ) are \Valid because
h = I . Probably. there exists a better V(.t-) to yield the same results of ( b ) and
(c) without considering h 5 1 and h > 1 separately. Although different Lyapunov
functions may serve the same purpose on stability. they may be different from
other points of view such as estimating 6 and other control applications.
3. The condition of part 2 guarantees that the closed contour
keeps on decreasing in R,, as time goes on unt i l C = 0 as a l i m i t . Being
nonnegatiLte. each term of V(.r) must be zero siniultaneouslj \idien C =
0. This suggests that xI 40 and x2+0 as C = 0 and hence asjiniptotic
stability for the origin is established.
5.2. SYSTEM MODEL DESCRIPTION
5.2.1 Real Power Supplied by a Generator
For a power network consisting of 12-generators connected together by mutual
admittances as shown in Fig. 5.3, we may write in matrix form:
Figure 5.3 A general rt-Generator system.
I 18
where [ I ] is the injected current vector, [ E ]is generator internal voltage vector,
and [ V ] is the system admittance matrix in which the generator impedances art:
included.
for all i, j = 1.2. . , . n. The matrix element and complex voltage are specified by
-
Y,,= G,,+ jR,, and E,= E, 6, (5.f 1
The real power supplied by generator i to the network is
( 5 . 7 ,
where
PI,= C,,sin(6,- 6,)+ D,,cos(6,- 6 , ) with C,,= E, E, H , , and Ill,= E, E, G,,
The power P,, is the real power delivered by generator i to j ; i t may 3e
positive or negative. P,,= E ; G,, is the power delivered to the local load at gencr-
ator I .
From Eq. (5.8) it is clear that f,,depends on difference\ between phiw
angles rather than individual phase angles. This result suggests that one may
choose an arbitrary reference for the angles without affecting the resulting q , . Indeed, we will first choose a reference rotating at synchronous speed to cle-
\tribe rotor dynamic\ and then a center of inertia to minimize the kralue of rotor
hinetic energy.
5.3 STABILITY OF A SINGLE-MACHINE SYSTEM
Consider ii generator connected to an infinite bus with voltage V L 0" throirgh
a pure reactance X as shown in Fig. 5.4. If the internal reactaxe and EMF of
Assessitig Aiigle Stcihility \in Transiertt Eiiergy Fiirictiori I I9
21
v L 0
Infinite
Bus
Figure 5.4 Single generator and infinite bus system.
the generator are given by X , and E L 6 respectively, we have from Eq. (4.7)
that
P,. = A ( s ) sin6
with
E VA ( x ) = -x, + x
Multiply on both sides (4-7) by w6t = d6 gives
M O = (P,,,- PJd6 (5.9)
Let S,,= (6,,,0), S, = (&, a,) and S,,,= (S,,,, 0) be the start. fault clearing and
maximum states of the fault respectively as shown in Fig. 5.5. Let S = (6, w) be
any state on the P-6 curve generated by X , then integration of Eq. (5.9)from S
to S, yields
-1
M d - -1 MO$= P,,,(6- 6,)-A(X)(COSG,- ~0.46) (5 .10)
2 2
(a) During the hult: X = X ,
For S = S,,,we obtain from the above equation
-1
Mw,! = PJ 6,)- 6,) -A(X,)(COSG,- COS^,,) A I (5 .1 1 )2
(b) After the clearance: X = X ,
For S = S,,,,we have from Eq. (5.1 I ) that
I30
Power
P (P.U.:
A -Prefault
B- Post-fault
C -During
fault
I III
Ip , i i --Lx! I I
I II II + TorqueTorque
n Angle, S (radims)
Thus, ~ . ' econclude froin ( a ) and ( b ) that
Therefore. 6,,,can be found by \olving the nonlinex algebraic Eq. ( 5 .I2 1
o r graphically from Fig. 5.5 to judge stability. Thi\ i \ hno\\n ;is the eyual-.iic.a
criterion for \tability \ t i d y of power system\.
I t i \ important to note that the excessive energy A , created during fault to
con\eer-teci to rolor kinetic energy a l clearance. Thi\ re\ult in\pire\ the ii\e o ' the
transient t .~it.rg~ffunction (TEF)after clearing. That i \ . to obtain f r o m Eq. (f . 12)
U ith X = X ,
I'I
where
E = -1 MO2 + P J 6 - 6,) (5.13)
2
and
V = A ( X , COS^, - COS^)
5.3.1 Transient Swing
Let II = do/clt denote the acceleration of the system, then Eq. (5.13) together
with Fig. 5.5 show that C I < 0 above the P,,, line and that ( I > 0 belour i t after friult
clearance. There are two angles: 6,= sin-' P,, , /A(X,) and 6,,= TI: - 6, that corre-
spond t o II = 0 . Figure (5 .5 ) reveals that 6 increases from 6,,with O) = 0 due t o
II > 0 until reaching 6,,,< 6,,at which o = 0. Since ( I < 0 at 6,,,, 6 begins to de-
crease until o = 0 at an angle less than 6, and then comes back because ( I > 0.
As such. the poclrer angle swings back and forth around 6,. This is the case for
E - 1' < 0 at 6,, because 6,, is unreachable (0 is imaginary). Ho\be\,er. delayed
fault clearance may result in large A , which makes 6 cross 6,,uith 03 2 0. Then.
6 increases further without return due to C I > 0. This is thc case for ( E - V ) 2 0. We will consider the transient to be stable if the p o u w angle s\\Iings around
6, and is otherwise unstable. This definition makes i t possible to ;isscss the
stability by means of the (TEF) as follows:
( a ) The transient is stable if ( E- V) < 0 at 6,,; large magnitude yields better stability.
( b ) The transient is unstable if ( E - V) 2 0 at 6,,.
5.4 STABILITY ASSESSMENT FOR IFGENERATOR SYSTEM BY THE TEF M E T H O D
Consider a power system consisting of ri-generators. The dynamics of each is
described by Eq. (5.11 ) . We have
de, = (It
for i = I , 2, . . . , 1 1 . The dynamics are second order and nonlinear differential
equations coupled together through the phase angles contained i n the expression
I22 Chcipter .i
of e,.Although the numerical solution can be determined, a closed form of solution is impossible to obtain. We need not discuss the solution of the system of equations as stability that does not require the exact solution.
Another reference with velocity CO,,is to be chosen to minimize the integral- square error
for any t , and t:. The necessary and sufficient condition for this purpose is
M , = M,o,, with M , = cM , 1- I 1- I
which indicates that the reference is the center of inertia (COI).The CO1 has a phase angle 6,,satisfying
With respect to 6,,,all generators have phase angles
0, = 6,- 6,)
In terms of 8,,Eq. (5.14) is now expressed by
M , ~-+ M ,(10 (!!!!U = U,,,, -c ,= p, - P:, (5.15)
t l t t l t
with
and
where j f i signifies summation o f j from 1 to t i except i. All the angles 6,- 6, in P, , may be replaced by 8,,= 8,- 8, because
Multiply Eq. (5.11 ) by 8, and then sum i from 1 to U :
I23
(5.17)
The second term of Eq. (5.17) is zero because
The states at clearance angle 6, and unstable equilibrium 6,,are to be speci-
fied by 11 pairs of (angle, velocity), that is.
Integration of Eq. (5.15) with clt from S, to S,, gives
(5.18)
where
and
I 1 1'
v = c j P:, e,dt (5.30) ,=I ,
with c' and 14 denoting S, and &. We know from Eq. (5.19)that E is the total energy input plus kinetic energy
of rotors. The electrical energy stored and dissipated in the system is V given
by Eq. (5.20). The energy here is referred to power integration with respect to phase angle (not time).
The stored energy in V is path independent but the dissipated energy i n V
is path dependent. To show this, we multiply Eq. (5.17) by 6, and then sum
from i = 1 to 11 to obtain:
(5.2 I )
The preceding equation can be verified by carrying out the summation. Substitu-
tions of 6, from Eq. (5.8) gives
c,e,+ ~ , , 6 ,C , sine,,e,,+ D,, cos(8,+ 8,) (5 .22)=
Therefore, the integration of Eq. (5.2 ) results in
- 8 , )- co\ (8 - 8/11+ I ( 5 .33 )
We can see from Eq. (5.23)that the first part is the stored energy and I \
independent of the path of integration. But, the second part denoted by I dcpenci\ OII the path of 8,. Sonie kind of approximation has to be used to evaluatc sincc O , ( t )cannot be found analytically.
Lct 0, be qqxoximated for all i = I . 2,. . . . 1 2 , by
behere C,and K, are constants but may change with i andf7t) is the only one f ) r all I . For con\mience, we use a parameter I / to make
where t , and I , , are the times at the clearance and unstable equilibrium. Then
0, = C + K f ( t )= C' + K, f l t + ( I , - t ) ' I = c' + K g( " )
for t, Ir It,,and 0 5 I I I I . The two constants are required to meet the bound:q cond i t ion s
Solving the twro constants gives
Based on this approximation, we haw
and hence i t follows that
with
This result makes i t possible to integrate I as
I t is interesting to note thatfit) or g(rr) need not be known and that anj J t )
yields the same approximation for I as indicated by Eq. (5.28).Thus. one may
perceive.f(r) to be the Col:
which naturally minimizes the integral-square error /M,(O,- f'fdt This is proba-
I ,
bly the best choice of theJs. For example, if g ( u ) is chosen as the combination:
then g ( 0 ) = g, g( 1 ) = a + p and
In concept, one may regard a system as stable if the kinetic energy accuniu-
lated at the instant of clearance can be absorbed by the electrical components
of the system. Thus. i t is the kinetic energy that determines the stabilitjr. I t is
usually during the fault that some generators are affected and tend to separate
from the rest that are coherent with the system Col. So far as the kinetic energqr
is concerned. the gross motion of the separating generators (say the first k ) may
be considered as a single generator with inertia and velocity the same as that o f
their COI. That is,
i i
M , = C M , and MR = M , 6, I I I - I
To be coherent with the system's COI, the rest of the generators must ha1 e
zero velocity; 8, = 0 for all i = k + I , k + 2, . . . 1 1 . Coherence suggests that the
rest of the generators rnay be regarded as an infinite bus. As such, the whole
system beha\res like a single generator and an infinite bus and hence the restilt
obtained before may be applied t o the multiple generator system.
We modify Eqs. (5.18) and (5.19)according to the conclusion and updi te
of Eq. (5.19) below.
1E - V = - M,n:,
-
where
and
Note that R,, a,,,(€I,,), and (€I,,),, denote the velocities and angles at the clearance
and unstable state.
Comparing Eq. (5.26) with Eq. (5.1 I ) enables u s to draw the same con( lu-
+ion as being made for single generator systems. That is. the system is stab12 if
E < V and unstable if E > V. It is inconclusive to talk about E = V \ince n e
have made approxiniations in evaluating E and V.
5.5 APPLICATION TO A PRACTICAL POWER SYSTEM
The application of the direct method to actual power systems is quite difficult.
A number of simplifying assumptions are necessary. To date, the analysis has
been mostly limited to power system representation with generators represented
by classical models and loads modeled as constant impedances. Recently. there
have been several attempts to extend the method to include more detailed load
models.
In a multi-machine power system, the energy function V describing the total
system transient energy for the postdisturbance system is given by:
(5 .30)
where
(3: = angle of bus i at the postdisturbance SEP
J, = 2 H , o , ,= per unit moment of inertia of the it’’ generator The transient energy function consists of the following four terms:
1 . 1/2 Z 1,~::change in rotor kinetic energy of all generators in the CO1
reference frame
2. C P,:,(e,- (3:): change in rotor potential energy of all generators relatiLre
to CO1
3. CC C,/(cos@,- cos0,;): change in stored magnetic energy of all branches
4. ZZ D,,cos0,,d(O,+ (3,): change in dissipated energy of all branches
The first term is called the kinetic energy (En,,)and is a function of only
generator speeds. The sum of terms 2, 3, and 4 is called the potential energy
(E,,,,)and is a function of only generation angles.
The transient stability assessment procedure involves the following steps:
Step 1 Calculation of the critical energy V,,.
Step 2 Calculation of the total system energy at the instant of fitult-clear-
ing Vd Step 3 Calculation of stability index: V,,- V,,.The system is stable if the
stability index is positive.
Time-domain simulation is run up to the instant of fault clearing to obtain
the angles and speeds of all the generators. These are used to calculate the total
system energy (V,,)at fault clearing. The flowchart of TEF for transient stability
analysis is shown in Fig. (5.6).
5.6 B O U N D A R Y OF THE R E G I O N OF STABILITY
The calculation of the boundary of the region of stability, V ( ? ,is the most diffi-
cult step in applying the TEF method. Three different approaches are briefly
described here,
Input system data b~~~~~~~ ~I Power now calculation
Con ting ency apecifica tion
Form during fault Y matrix and reduced Y matrix
Form post-fault Y matrix and reduced Y matrix
C alcglate post fault SEP
C alcolate critical energy V .I
Calculate .yatem total energy V, at clearing time
Hes 2
1 . The Closest Unstc?ble Equilibrium Point (UEPI Approxh
Early papers on the application of the TEF method for transient stability analysis
used the following approach to determine the smallest V , ,
Step 1 Determine all the UEPs. This is ach ie id by solving the postdis-
t u rbance system steady -s tat e equations wi t h di ffe re n t i n i t i a1 va 1lies
of bus angles.
Step 2 Calculate system potential energy at each of the UEPs obtained i n
step 1 . The critical energy V ( ris gi\fen by the system at the UEP.
which results in the minimum potential energy.
This approach computes the critical energy by implicitly assuming the \I or\t
fault location. hence, the results are very conservative.
2. The Controlling UEP Approach
The degree of conservatism introduced by the closest UEP approach is such that
the results are usually of little practical value. The controlling UEP approach
removes much of this conservatism by computing the critical energy depending on the friult location. This approach is based on the obser\ation that the system
trajectories for all critically stable cases get close to those UEPs that are closelj,
related to the boundary of system separation. The UEPs are called the control-
ling or relevant UEPs.
The essence of the controlling UEP method is to use the constant energy surfrice through the controlling UEP to approximate the rele\mt part of the
stability boundary (stable manifold of the controlling UEP) to which the f i i u l t -
on trajectory is heading.
For any fault-on trajectory q ( t ) starting from a point p ~ A ( . v , )with V ( p )< V(.i-), if the exit point of the fault-on trajectory lies in the stable manifold of .i-,
the fault-on trajectory must pass through the connected constant energy surfiice
AV((.?) before it passes through the stable manifold of .t(W’(.i-))(thus exits the
stability boundary AA(-t,J). Therefore, the connected constant energy surfuce A\’,
(.P) can be used to approximate the part of the stability bondary AA(,?,)for the
fault-on trajectory .v,( I). The computation process in this approach consists of
the following steps:
Step 1 Determine the controlling UEP, A-,,, for the fault-on trajectory .v,(I).
Step 2 The critical energy V, is the value of the energy function V ( * )a t
the controlling UEP, that is, V, = V(x, , , ) .
Step 3 Calculate the value of the energy function \I( .) at the time of fiiult
clearance (say, I,,) using the fault-on trajectory I{, = V ( . v , ( I ( , ) ) .
Step 4 If V,< V , , then the postfault system is stable. Otheritrise, i t is u n -
stable.
The key element of the controlling UEP method is how to find the controllinp
UEP for a fault-on trajectory. Much of the recent work in the controlling UEP
method is based on heuristics and simulations. A theory-based algorithm to
find the controlling UEP for the classical power system model with transfer
conductance G, is presented now.
The energy function is of the form:
JI
where M , = M,, .Y' = (6 ' ,0) is the stable equilibrium point (SEP) under consij-
eration. ' = I
Algorithm to Find the Controlling UEP The reduced system is
The algorithm for finding the controlling UEP consists of the following steps:
Step 1 From the fault-on trajectory (6 ( f ) ,w( r ) ) ,detect the point 6* at which
the projected trajectory 6(r) reaches the first local maximum : ) t E,,(*).Also, compute the point 6- that is one step ahead of 6" along 6(r).and the point 6' that is one step after 6*.
Step 2 Use the point 6" as initial condition and integrate the postfault I1
reduced system Eq. (5.31) to find the first local minimum of I : II.f;(s>),say at 6:.
Step 3 Use 6- and 6' as initial conditions and repeat Step 2 to find the
corresponding points, say 6, and 6; respectively.
Step 4 Compare the values of If(6)i at 6,, 6:, and 6;. The one with the
smallest lralue is used a s the initial guess to solve Eq. (5.3I ),.f;(6) = 0, say the solution is 6,,,.
Step 5 The controlling UEP with respect to the fault-on trajectory is (6,(#. 0).
The proposed algorithm finds the controlling UEP Lria the controlling UEP
of the reduced system Eq. (5.31) with respect to the projected frlult-on trajectory
6(r).Steps 1-4 find the controlling UEP of the reduced system and step 5 relates
the controlling UEP of the reduced system to the controlling UEP of the original
system. Theoretical justification of the proposed algorithm can be found in work
done by Chiang.
3. The Boundary of Stability-Region-Based Controlling
UEP (BCU) Method
Earlier UEP methods faced serious convergence problems when solving for the
controlling UEP, especially when the system is highly stressed or highly un-
stressed, or when the mode of system instability is complex. These problem\
usually arise if the starting point for the UEP solution is not sufficiently close
to the exact UEP. Some of the convergence problems can be otm-come by the
BCU method which has the capability of producing a much better \tarting point
for the UEP solution.
CONCLUSION
One of the major innovations in stability assessment is based on the energy
function concept. which is an offshoot of Lyapunov stability criteria. This chap-
ter introduced the fundamental Lyapunov stability thought and the procedure of
constructing Lyapunov stability function.
The main thrust of this chapter is to utilize concepts of system modeling to
evaluate system stability using the energy function method. The reader is re- ferred to the list of references and the annotated glossary of terms for further
information on the subject matter.
Voltage Stability Assessment
INTRODUCTION
Voltage stability studies evaluate the ability of a power system to maintain ac-
ceptable voltages at all nodes under normal conditions and after being subject :cl to contingency conditions. A power system is said to have entered ii state of
voltage instability when a disturbance causes a progressive and uncontrollable
decline in voltage values. Inadequate reactive power support from generators.
reactive sources, and transmission lines ciiri lead to kd tage instability o r e ~ w i
\.oltage collapse, which have resulted i n several major system fiiilures (b1ac.k-
outs) such as:
I . September I9 10, New York Power pool.
2. Northern Belgium System and Florida System disturbances of I982
3. Swedish system disturbance in December 1983.
3. French and Japanese system disturbance5 in 1987.
5 . Recently, i n the late nineties, i n the U.S. and other parts of the wor d.
The literature and background studies reLriewed indicate that \roltage i n <ta-
bility or collapse are characterized by a progressike fall of idtage which can
take several forms. The main factor is the inability of the network to meet the
demand of reactive power. The process of instability may be triggered by seine
forni o f di\turbance, resulting i n changes in the reactikre poucr requirement. '-he
disturbance may either be small or large changes in essential load. The conse-
quence of voltage instability may, however. have uidespread impact on the
system.
Power system voltage stability involves generation. transmission, and distri-
bution. Voltage stability is closely associated with other aspects of poM-er system
steady-state and dynamic performance. Voltage control, reacti1.e power compen-
sation and management, rotor-angle stability, protective relaying. and control
center operations all influence voltage stability.
Voltage stability studies involve a wide range of phenomena. Because of
this, voltage stability means different things to different people. I t is a fast phe-
nomenon for engineers involved with the operation of induction motors. air
conditioning loads, or HVDC links. I t is a slow phenomenon (iri\.ol\.ing, for
example. mechanical tap changing) for others. (Appropriate analysis methods
have been discussed with the debate centering on \%!hether the phenomena of'
voltage stability are static or dynamic). Voltage instabilitjr or collapse is a dqr-
namic process. The term "stability" implies that a dynamic system is being
discussed. A power system is a dynamic system. In contrast to rotor angle ( s j w
chronous) stability, the dynamics involve mainly the loads and the means for
voltage control. Voltage stability has alternatively been called load stability.
The loss of lines or generators can sometimes cause \.ohage quality degra-
dation. This phenomenon has equally been attributed to the lack of sufficient
reactikre reserve when the power system experiences a he;\\,} load or a se\.ere
~contingency. Thus, voltage instability is characterized i n such ;i I J that \dtage
magnitude of the power system decreases gradually and then rapidly near the collapsing point. Voltage stability is classified as either static voltage stability
or dynamic voltage stability. The latter is further classified into small signal
stability and large disturbance stability problems. A unified frame\%.ork related
to voltage stability problems will be shown in a proceeding section.
In dynamic voltage stability analysis, exact models of transformers. Static
Voltage Compensating devices (SVCs), induction motors and other types ot'
loads are usually included in problem formulations i n addition to models 01'
generators, exciters. and other controllers. Small signal voltage stabiliti, prob-
lems are formulated as a combination of differential and algebraic equations
that are linearized about an equilibrium point. Eigen analysis methods are used
to analyze system dynamic behavior. Small signal analysis can provide uset'ul
information on modes of voltage instability and is instructik'e in locating VAR
compensations and in the design of controllers. Large disturbance \wltage stabil-
ity is approached mainly by using numerical simulation techniques. Since s ~ r s -
tem dynamics are described by nonlinear differential and algebraic equations
that cannot be linearized in nature. Voltage collapse is analyzed based on ii
center manifold voltage collapse model.
I24
6.1 WORKING DEFINITION OF VOLTAGE COLLAPSE STUDY TERMS
Voltage stability has been viewed as a steady-state “viability” problem suitable
for static analysis techniques. The ability to transfer reactive power from pro-
duction sources to consumption sinks during steady operating conditions is a
major aspect of voltage stability. The following definitions are often used i n
voltage stability studies.
Voltage collapse incidents in the U.S., Europe, and Japan ha\x led to differ-
ent explanations. interpretations. and concerns. To achieve a unifying franre-
work, a working definition for detection and prevention of Ldtage collapse I-iis
been constructed (EPRI RP 2473-36).
6.1.1 Classification of Voltage Collapse Detection
Detection of VC is based on determination of imminence of power system volt-
age violating its limits which value may lead to system instability and conse-
quently voltage collapse. The phenomenon in many instances is due to a def cit
in reactikte power generation, loss of critical lines or degradation of control on
key buses.
6.1.2 Classification of Voltage Collapse Prevention
Voltage collapse prevention is any action taken to reduce the likelihood of
power system degradation due to the violation of operating limits. Prevention
scheme includes the use of optimal power tlow strategy and other measure:* to
minimize voltage deviations.
Other definitions of voltage collapse or stability are documented in the l i ter-
ature. For example, according to the IEEE working group in voltage stability.
voltage collapse/stability is defined as follows:
Voltage stability is the cihility of a system to maintain kroltage so that when
load admittances are increased, load power will increase, so that t o t h
power and voltage are controllable. Voltage collapse is the pr.oc.t..s.s by which voltage instability leads to 10s.; of
voltage in a significant part of the system. Voltage degradation may
also lead to “angle stability” as well, especially if the preventive mea-
sures are not enforced quickly enough. Sometimes only careful post
incident analysis can discover the primary cause of voltage collapse.
Voltcige Stcrhility A.s.ses.sment 1-35
6.2 TYPICAL SCENARIO OF VOLTAGE COLLAPSE
Assume that a power system undergoes a sudden increase of reactive power
demand following a system contingency, the additional demand is met by the
reactive power reserves carried by the generators and compensators. It is possi-
ble, because of a combination of events and system conditions. that the addi-
tional reactive power demand may lead to voltage collapse, causing a major
breakdown of part or all of the system. A typical sequence of events leading to
a voltage collapse can be as follows:
1. The power system is experiencing abnormal operating conditions with
large generating units near the load centers being out of service. Some EHV
lines are heavily loaded and reactive power resources are low.
2. A heavily loaded line is lost which causes additional loading on the
remaining adjacent lines. This increases the reactive power losses in the lines
causing a heavy reactive power demand on the system. (Reactive power ab-
sorbed by a line increases rapidly for loads above surge impedance loading).
3. Immediately following the loss of the line, a considerable reduction of
voltage takes place at adjacent load centers due to extra reactive power demand. This causes a load reduction, and the resulting reduction in power flow through
the lines would have a stabilizing effect. The generator AVRs would, however,
quickly resolve terminal voltages by increasing excitation. The resulting addi-
tional reactive power flow through the inductances associated with generator
transformers and lines would cause increased voltage drop across each of these elements. At this stage, generators are likely to be within the limits of P-(2
output capabilities, i.e., within the armature and field current heating limits. The
speed governors would regulate frequency by reducing MW output.
4. The EHV level voltage reduction at load centers would be reflected into
the distribution system. The ULTCs of substation transformers would restore
distribution voltages and loads to prefault levels in about 2 to 4 minutes. With
each tap change operation, the resulting increment in load on EHV lines would
increase the line XI’ and RI’ losses, which in turn would cause a greater drop
in EHV lines. If the EHV line is loaded considerably above the SIL, each MVA
increase in line flow would cause several MVAs of line losses.
5. As a result, with each tap-changing operation, the reactive output of
generators throughout the system would increase. Gradually. the generators
would exceed their reactive power capability limits (imposed by maximum al-
lowable continuous field current) one by one. When the first generator reached
its field current limit, its terminal voltage would drop. At the reduced terminal
voltage for a fixed MW output, the armature current would increase. This may
further limit reactive output to keep the armature current within allowable limits.
Its share of reactive loading would be transferred to other generators, leading to
overloading of more and more generators. With fewer generators on automatic
excitation control, the system would be much more prone to voltage instability.
This \.+wild likely be conipounded by the r:duced effectiLmess of shunt coni-
pensators at low voltages. The process will eventually lead to voltage collap ;e
o r avalanche, possibly leading to loss of synchronism of generating.
Sometimes. the term 'voltage security' is used. This means the ability of ;I
system not only to operate i n a stable inode but also to remain stable follohiiig
credible contingencies o r load increases. I t often means the existence o f a con-
siderable margin from an operating point to the voltage instability point fc)l lo\\ ,-
ing credible contingencies.
6.3 TIME-FRAME VOLTAGE STABILITY
Voltage instability incidents in \.arious places around the world has been c.c-
scribed iis ii function o f time. The time ranges from seconds to tenths of minutcs.
Three time frames and sconarios described by Carson Taylor are sumniari;red
;is follo\+fs:
I . Transient cultage stability: This occurs bet\+.een0 to 10 seconds.
2. Classical Voltage Stability: This occurs betureen I to 5 minutes. This
i s t he c1iiss ic sccnario i n 1'0 I 1,i ng aiit0111 at ic on-I oad tap change r , d i st r i bu t ion \,(1I t -
age regi i 1at ion . x ic l gener;i t or c Urre 11t I i mi t i n g. Th i s scen ario i n \'oI \re s high 1o;i Js. high p o ~ v e rimports from ii remote generator, and a large distribution. Tap
changcr action is also significant, i t i t can be beneficial o r detritnental depend,ng on load characteristics and location of the tap changer. Collapse in one area c-an
at'tect ;I much larger area, thus, leading to a mtijor black-out. This occurrenct' i \
tj.pical on the East Coast. the pacific West, and in Japan.
3. Long-term Voltage Stability: This in\rol\res se\feriil minutes. Se\.t ral
VC incidents (the Tokyo I977 blackout. the Shieden blackout) are cxperien:cs
o f VC. The factors may include overload time, limit of transmission, loss load
di\,ersitJ,due to low Lroltage (theoretically controlled loads), timeliness of appI>'-
ing reuc*ti\.epower. and other operating interventions such iis load shedding.
Reported idtage i nstabi I i t y i tic ident s \v i t h and L + i i t hoii t vo I tage are su 111I na-
r ixd i n Figure 6. I . For classical \,()Itage instability. the phenomenon \+r i l l occur
at the onset o f the \,oltagc collapse. For long-term stabilitjr, the shorter-t. me
frame phenoniena w i l l occur once \dtage begins to sag leading to \ultage c:ol-
lapse.
6.4 MODELING FOR VOLTAGE STABILITY STUDIES
Voltage stability studies involve the solution to algebraic and differential equa-
tions that map the system behavior under steady-state and transient state\. Belo\v
are tjpical \ ector\ encountered and the notation used.
--- -11
Classical Long-Term (Large Disturbance) (Large Disturbance) (Load Buildup)
GeneratorExc. Dynamics LTCsi-Rime Mover Control Load DiversityJrhermostat
0 0 0
Max. Exciter Limiter 0
Mechanically Switched Capacitors Linnransformer Overload -;-y- Gen. ChangdAGC 0 4
Inertia Dynamics Boiler Dynamics 0 0 1
Operation Intervention I
DC Converter LTCs Gas Turbines F- - 1
I min 10 min I h r 1 I I I 1 1 1 1 1 I 1 I I I I 1 I1 I I I 1
Time in seconds
Figure 6.1 Time-iramc of voltage stability (courtesy of C a r ~ nTa)rlor).
1 . Dynamic state \rector. ~ ( t )
\+!here
8( t ) = Rotor angle
E’(t ) = Voltage components of synchronous machine
S ( r ) = Dynamics in load bus
p ( t ) = Other dynamic states (exciter, governor)
2. Algebraic state vector, y ( t )
where
~ ( t ) .6 ( t)= Bus voltage niagnitude and angle
Chtrprrr 6
Q ( t ) = Nonscheduled reactive power
q ( r ) = Other algebraic variables
3. Parameter vector, p ( t )
where
P,(r 1 = Turbine shaft power
P,( t 1, Q r ( t )= Scheduled load power
6 d ( t ) = Controlled voltage or set points
CT(t 1 = Other similar parameters
6.5 VOLTAGE COLLAPSE PREDICTION METHODS
The framework for voltage stability studies can be simplified to fit the time
span of the analysis. The categories of interest are as follows:
6.5.1 Static Stability
Assume all time derivatives equal zero at some operating point H ( . v , j - . p )= 0
6.5.2 Dynamic Stability
At some operating point, small perturbations (local)
6.5.3 Extended Stability
Simulation through time (up to hours)
6.6 CLASSIFICATION O F VOLTAGE STABILITY PROBLEMS
Voltage problems are distinguished in three categories:
1 . Primary phenomena related to system structure. These retlect the auton-
oinous response of the system to reactive supply/demand imbalanccs.
139 Voltuge Stcrhility Assessrtierit
2. Secondary phenomena related to control actions. These reflect the
counterproductive nature of some manual or automatic control actions.
3. Tertiary phenomena resulting from interaction of the above.
This classification of voltage quality problems implies that the problems
involve both static and dynamic aspects of system components. Voltage collapse
dynamics span a range in time from a fraction of a second to tens of minutes.
Time frame charts are used to describe dynamic phenomena which show time
responses from equipment that may affect voltage stability. The time frame chart
is shown in Table 6. I , where .q is a state vector representing transient dynamics.
I, is a state vector representing long-term dynamics.
Y is a state vector representing very fast transient dynamics related to net-
work components and P is a system parameter vector. Then the time frames to
be considered become very fast transient, transient, and long term. The main
characteristics of the three time frames are as follows:
1. A very fast transient voltage collapse involLres network RLC compo-
nents having very fast response. The time range is from microseconds to milli-
seconds.
2. A transient voltage collapse involves a large disturbance and loads ha\.-
ing a rapid response. Motor dynamics following a fault are often the main con-
cern. The time frame is one to several seconds.
3. A long-term voltage collapse usually involves a load increase or a
power transfer increase. Within this time frame, a voltage collapse shows load restoration by tap-changer and generator current limiting. Manual actions by
system operators may be important. The time frame is usually 0.5 to 30 minutes.
Since voltage stability is affected by various system components in a wide time
range, in order to tackle this problem, one must consider proper modeling and
analysis methods. Currently, voltage stability approaches mainly include static
and dynamic, i.e., transient voltage collapse and long-term voltage collapse.
Table 6.1 Time Frame and Relevant Models in Voltage
Stability Assessment
Voltage stability models and time scale
Micro to milli seconds A few seconds Minutes
I40
6.7 VOLTAGE STABILITY ASSESSMENT TECHNIQUES
The loss o f lines o r generators can sometimes cause degradation in voltage. This
phenomenon hiis equally been attributed to the lack of sufficient reacti\,e reset ve
when the pouer system experiences a heavy load o r severe contingency. Thus.
\-oltage stability is characterized in such a way that voltage magnitude o f the
p o ~ wsystem decreases gradiially and then rapidly in the neighborhood o f I he
collapsing point. Voltage stability is classified as static voltage stability and
dq'naiiiic \.oltage stability. The latter is further di\ridt.d into srnall signal stabilitJr
ancl large disturbance stability problems.
I n dy n a i n ic vo1t age st 11bi I i t y ana1y si s, exac t t node I ing of t ran sformc rs ,
SVCs. induction motors, and other types of loads are usually included in prob-
lem formulations i n addition to models of generators, exciters. and other control-
lers. Small signal voltage stability problems are fimiiulated as a combination o t
differential rind algebraic equations that are linearized about an equilibri .in1
point. Eigen analysis methods are used to analyze system dynamic beha\rior.
Small signal rinalysis can provide useful information on modes of \voltage in:,ta-
bility and is instructikre i n locating VAR compensations and in the design o f
controllers. On the other hand, large disturbance voltage stability is mainly dorilt
\+,ith bj. numericul simulation techniques. since system dynamics are descrilxx!
by nonlinear differential and algebraic equations that caniiot be linearized i n
nature. The mechanism o f \,()Itage collapse has been explained as saddle node
bifurcation i n some literature. Voltage collapse is anrilyzed based on ii ceriter
manifold \ d t a g e collapse model.
Static Lroltage stability analysis is based on po\+w system load tlow e q u -
tions. Indices characterizing the proximity of ;in operating state to the collapse
point are cie\~eloped.The degeneracy of the load tlo\ss Jacobian matrix has bcen
used ;is ;in iritlex of p o \ + w system steady-state stabilitjr. I:nder certain coiidi-
tions, ;I change i n the sign o f the determinant o f the Jacobian niatris diiiing
c*ontinuoiis \rariations o f pnrameters means that ;I real eigeiivalue o f the liii1:ar-
i d sb+ring equations crosses the imaginary axis in to the right hal f o f the coin-
ples plane and stability is lost. Various researchers ha\.e considered th;it ;I
change i n the sign o f the Jacobian matrix may probubljf not indicate the losj of
steady-state stabilitj, urhen e\wi niirnber eigen\Aues \+!hose real part cross the
imaginury axis. Voltage stability is also related to iiiiiltiple load flow soluti~ms.
A proximity indicator t o r Lroltage collapse (VCPI) was defined for ;I bus, ;in
area. o r the complete system ;is ii Lector of ratios o f the incremental gener.ited
reactive powcr at a generator to a given reactive load d e n i d increase. A dil'ter-
ent indicator ( L index) is calculated from normal load tlow results with reaxm-
;IbIe coinput iit ions. The 111 i ni in u 111 s i n g u 1ar va1ue of the J iicob i an was prop )sed
;is ii \coltage security index. since the magnitude o f the rninimiim singular \ d u e
coincides Lsrith the degree of Jacobian ill-conditioning and the proximity to col-
lapse point. Based on a similar concept, the condition number of the Jacobian
is also applied as an alternative voltage instability indicator by pioneers i n the
field.
Bifurcation theory is used to analyze static stability and voltage collapse.
Static bifurcation of power flow equations were associated Lvith either di\wgent-
type instability or loss of casualty. Researchers ha\.e described necessary and
sufficient conditions for steady-state stability based on the concept of feasibility
regions of power flow maps and feasibility margins but with high computational
efforts. A security measure is derived to indicate system \~ulnerability to Lwltage
collapse using an energy function for system models that include idtage \,aria-
tion and reactiire loads. I t is concluded that the key to applications of the energy
method is finding the appropriate T\pe-l low voltage solutions.
In addition to the above methods for direct coinputation of stability index.
some indirect approaches, based on either the continuation method or optimiza-
tion methods haire been developed to compute the exact point of collapse. I n
a p p I y i n g the co n t i n u at ion methods , assumptions about 1oad ch an g i 11g pat t e1-17s
are needed.
In summary, the methods for static voltage instability analysis are based on
multiple load flow solutions (voltage instability proximity indicator [ VIPI], en-
ergy method), load flow results ( L index, VCPI), or eigen\Aues of the Jacobian
matrix (iiiinimum singular value and condition number). While studies on djr-
namic voltage collapse shed light on control strategy design (off-line applica-
tions), static \.oltage stability analysis can provide operators with guideline in-formation on the proximity of the current operating state to the collapse point
(on-line applications ). In this case, an index, which can @\re ad\wice utarning about the proximity to the collapse point, is useful.
The next discussion will be that of the formulation of selected \x)lta, c’e sta-
bility indices. Of the wide range of techniques available. we shall discuss the
VIPI method. a method based on singular value decomposition. condition nuin-
ber of the Jacobian, and the method based on the Energy Margin.
6.7.1 Voltage Instability Proximity Indicator Method
The \.oltage instability proximity indicator (VIPI) was developed by Y. Tamura
et al. based on the concept of multiple load flow solutions. A pair of load f l o ~
solutions .vl and .v2 are represented by two vectors ci and 17 as followfs:
which are equivalent to:
where .vl is the normal (high) power tlow solution and .v2 its corresponding IOYA.
voltage power tlow solution; LI is a singular vector in the space of node voltages
and h is a margin vector in the same space.
We now define two other vectors Y , and Y(c i ) .called singular vectors in t k
space of node specifications. The relationship between these \rectors is shou n
in Figure 6.2.
VIPI is defined by the following equation:
where vector Y,, consists of bus injections computed with respect to .vl but the
injection kAues corresponding to reactive powers of PV buses are replaced by
the squared values of voltage magnitudes. Y(cc)consists of bus injections with
respect to vector ( I : I( .v ( 1 is the /,-norm of vector, .v. The computation of VIPI is
easy once the relevant low voltage power tlow solutions are obtained. Generallj,
speaking. finding all the relevant low voltage solutions are time-consuming tor
practical size systems.
6.7.2 Minimum Singular Value (U,,,,,,) Method
When an operating state approaches the collapse point, the Jacobian matrix of
the power tlon equations ( J ) , approaches singularity. The minimum singu ;ir
Figure 6.2 Concept o f VIP1 in the node specification space.
value of the Jacobian matrix expresses the closeness of Jacobian singularity.
The singular value decomposition method is used to solve the minimum singular
value for static voltage stability analysis.
According to the theory of singular value decomposition. power flow Jacob-
ian can be decomposed as:
J = LEV' (6.11 )
where: J E R'J1*'J1is the power flow Jacobian matrix; C = diag(o,, (J?, - , oJJ)
with (J,,,,, = (J, 2 o22 2 (J,~= (J,,,,,,2 0. If matrix J has rank ) - ( I " 5211).its singu- lar values are the square roots of the I" positive eigenvalues of A7A (or A A ' ) .
U and V are orthonormal matrices of order 211, and their columns contain the
eigenvectors of AA' and ATA respectively. From Eq. (6.9). i t can be obtained
that
A V,= CJ,~(,
(6.13)
A ' U , = (3,V, (6.13)
We define
E, = u,V: (6.14)
Then Eq. (6.I 1 ) can be written as:
(6.16)
then, as far as the /?-norm of the J matrix is concerned, J' is a matrix of rank
11 - 1 nearest to the J matrix of rank n. This means that the smallest singular
value of a matrix is a measure of the distance between matrices J and J'. As for
the power flow equations, its minimum singular value expresses the proximity
of the Jacobian to singularity. It can be used as an index for static voltage
stability.
6.7.3 Condition Number of the Jacobian Method
The condition number is used in numerical analysis to analyze the propagation
of errors in matrix A or vector 6 in solving variable vector x for the linear
equation Ax = h. If matrix A is ill-conditioned, even very small \xiations i n
vector 17 (or A ) may result in significant changes in solution \rector .v.
For the Iinearized load flow equations, the condition number of the Jacobi.in
matrix can be iised to measure its conditioning and whether any small variations
in i'ector 11 ( o r A ) inay result in significant changes in solution vector .r.
For the Iinearized load tlow equations, the condition number of the Jacobi,in
matrix can be used to measure its conditioning and whether any small variations
i n loads niay lead to large changes i n bus voltages. If the condition number is
greater than a specified threshold, this will Iiieiin that the current operating state
is close to the collapsing point. A precise measure of the sensitivity of a linear sq'stem solution with resptxt
to matrix A or vector h ciiii be defined as:
For pobrer t locb Jacobian iiiatrix J . the \ alue of Cond, ( J ) can gile an
indication of the condition of J "with respect to inversion." A small ~ ~ a l u eof'
Cond2( J ) ( 1 - 10) refer\ to a bell-conditioned Jacobian matrix (relatitrely large
voltage stability margin): a large value of Cond2(J ) (>100) mean\ that the opi*ra-
ting 5tute is \ c ry close to the point of Jacobian \ingiilarity and has a I O N colt ige
\tiibilitj iiiargin. The extreme condition is that J is singular and Cond,(Jr is
infinite. Hence. the condition nuniber Cond?(J ) ciin be used to iiieii\ure the
proxiiiiitj of the operating \tates to voltage collap\e.
6.7.4 Energy Margin-Based Method
The energy method uses an energy function, dericed from a clo\ed form \o:tor
integration of the real iiiid reactive mismatch equations betu een the oper;ihle
pou er flou \olution and ;i I w oltage power tlou \olution, to prokide ii yuiinti-
tatice tiieasiire ol' I ~ O M clo\e the \y\tem i \ to toltage instability. The point of
t oltage in\tability correspond\ to the \addle node bifurcation point definec hq
;i \iiigiiliir potver tlou Jacobian N i t h ~ e r ocnergj margin.
The encrgq function t'or boltage \tabilit> analysi\ i \ defined as:
1-45
(6.19)
with real and reactive mismatches defined as:
where: .I-' = (a' ,V')is the normal operable power tlow solution (o r the stable
equilibrium point, SEP): xf'= (a",V")is the relevant low Yoltage po\\.er ilow
solution with respect to x ' (or unstable equilibrium point, UEP).
A large energy value indicates a high degree o f \rollage stabilitjr ndiilc ;i
small \yalue indicates a low degree of voltage stability. I n applj'ing the c n e r g ~ ~
method. the key is finding the relevant UEPs. Since the number o f rele\rant l o ~ i p
\.o~tage power tlow solutions is very large (Y'- I for a practical syteni. the
exhausti\ve approach is not feasible. There is an impro \4 technique to compute
all the Type- I UEPs based on the results that sho\i' for tjpical po\i'er sj'stems. the system always loses steady-state stability by a saddle node bifurcation bc-
tureen the operable solution and a Type- 1 low-voltage solution. That condition
restricts the computation of relevant UEPs only corresponding to sj'sterii PQ
buses, or practically PQ load buses. After finding all thc rele\mt UEPs. the
buses corresponding to which the energy function has the lowest \ dues arc
buses \sulnerable to voltage instability. Similar to the VIP1 method. the energy
methods depend on the low-voltage power flow solutions. urhere the Ne\s,ton-
Raphson method with the optimal multiplier can be used.
6.8 ANALYSIS TECHNIQUES FOR STEADY-STATE VOLTAGE STABILITY STUDIES
6.8.1 Introduction to the Continuation Method
I n its early stages, \dtage collapse studies were mainly concerned \\!it11 stead>.-
state voltage behavior. The voltage collapse is often described ;is a problcm that
results when a transfer limit is exceeded. The transfer l i m i t of an elcctrical
pom'er network is the tnaximal real or reactive power that the system c;in deli\w
from the generation sources t o the load area. Specit'icallqr. the transfer l i m i t i h
136 Clttipter 6
the maximal amount of power that corresponds to at least one power-tlow sol J -
tion. From the well-known P-V or Q-V curves, one can observe that the volta;;e
gradually decreases as the power transfer amount is increased. Beyond the maxi-
mum power transfer limit, the power-tlow solution does not exist, which implies
that the system has lost its steady-state equilibrium point. From an analytical
point of view, the criteria for detecting the point of voltage collapse is the point
where Jacobian of power-flow equations become singular.
The steady-state operation of the power system network is represented ‘ ~ y
power-tlow equations given in equation (6.20).
where 8 represents the vector of bus voltage angles and V represents the vec or of bus voltage magnitudes. h is a parameter of interest we wish t o b a y . In
general the dimension of F will be 211,,, + t i / )u, where upsand / I / ’ \ are the number
of PQ and PV buses, respectively.
From equation (6.20) one obtains the fundamental equation of sensitii i t )
an a I y sis
Let .v’ = [e,V]’.From Eq. (6.2 1 ), one can obtain an ODE system
(62 2 )
For a specific variation of the parameter h, the corresponding variatioIi t o
the solution x is calculated by evaluating the Jacobian (dF/d.4-].I t should be
emphasized that the singularity of the power flow Jacobian dF/d.v is necesaary
but not a sufficient condition to indicate voltage instability. The method pro-
posed to observe the voltage instability phenomenon is closely related to tn111ti-
ple power flow solutions. which are caused by the nonlinearity of power flow
solutions. The drawback of the method is that i t relies on the Newton-Raphson
method of power flow analysis, which is unreliable in the vicinity of the vollage
stability limit. As such, researchers have developed a technique knourn as the
cont i n u ati on met hod.
137 Voltcige Stcthility Assessriierlt
6.8.2 Continuation Method and Its Application to
Voltages Stability Assessment
Consider the power flow equation defined in Eq. (6.20). The vector function F
consists of \I scaler equations defining a curve in the I I + 1 diniensional (.v,h) space. Continuation means tracing this curve. For a convenient graphical repre-
sentation of the solution (.v.h)of Eq. (6.20) we need a one-dimensional measure
of x. The frequently used measures are:
( i ) 1 .v 1 = E','=,xf (square of the Euclidean norm),
( i i ) j s I = max 1 .I-, (maximum norm), 1 I=I I f
( i i i ) 1 s 1 =xi for some index k , 1 I k I n .
In power systems generally we use the measure of ( i i i ) . As can be seen
from Fig. 6.3 we have a type of critical solution for h = k*, where for h > h* there are no solutions. For h < h* we have two solutions (one is the high voltage
state variable
parameter
Figure 6.3 The fold type curve including predictor-corrector step.
solution and the other is the low voltage solution). When h approaches h '~0 .< h+).both solutions merge. At this point the Jacobian of the power f l o ~ ,solution
is singular. In the mathematical literature these points are called turning points.
fold points. or bifiircation points. An algebraic featiire of the turning point is
given by F , below
F , ( . V ' ~ ' . ~ ' ~ )is singular for rank < 11.
F ( p:,A:!:)/F.\(, .v * .A:]: h iis ;I ful l rank I I and satisfies some nondegenericitj,1 - cl: )
conditions.
Several techniques ha\ e been proposed to calculate these point\. Thew
methods based their iinaIy\is on t\vo approache\ referred to a\ direct and indii cct
inet hods .
6.8.3 Detection of Voltage Collapse Points Using
the Continuation Method
Direct Methods
This approach tries to find the maximum allowdAe wriation of h: that is. an
operating point ( .P,A'; :) of the equation:
such that the Jiicobian at this point is singulur. I t solves the tolluct~ing \>stem o f
eq11at i ons
This procedure basically augments the original set of pobrer tlow eqiiat ions
=F(.v,h)= 0 by F,(.v,)L)/I0 where h is ;in u-\fector n,ith / I , = 1 , The disadewmges
of this approach are:
The dimension of the nonlinear set of equations to be solcved is tnice that
for the concmtional power flow.
The approach requires good estimate of the vector h.
The advantage is that. convergence of the direct method is \,ery fast i ' the
initial operating point is close to the turning point. The enlarged system is sol\wl
in such ii way that i t requires the solution of four I I x / I ( 1 1 is the dimensicln of
the Jacobian F,(.v,h))linear systems, each w i t h the same matrix. requiring only
one LU decomposition.
indirect Method (Continuation Methods)
Assuming that the first solution (.q,,&,) of F(.u.h)= 0, is available. the continua-
tion problem is to calculate further solutions, ( .~ , .h , ) , unti l one reaches a(.v2.h>). target point. say at h = A*. The ith continuation step starts from an approsima-
tion of (x,,h,)and attempts to calculate the next solution. However, there is an
intermediate step in between. With predictor, corrector type continuation, the
step i 3 i + I is split into two parts. The first part tries to predict a solution.
and the second part tries to make this predicted part to coni'erge to the required
solution:
Continuation method\ differ among other thing\, i n the follo\J ing: ( 1 ) choice of predictor, (2) type of the parameterization \trategy. ( 3 ) type of correc-
tor method. ( 3 )step length control. All four aspects uill be explained through
the formulation of the power flow equations.
I n order to apply the continuation method to the poner tlou problem, the
power flow equation\ must be reformulated to include a load parameter (A). This can be done by expressing the load and the generation at a bus as ;I function
of the load parameter (h) .The general form of the ne\+ equation\ a\sociated
with each bus i is:
where the subscripts L,, G,, and T, denote bus load, generation, and power out
of a bus respectively. The voltage at bus i is V, 8, and Y,/ a,/ is the (i,j)th element of the system admittance matrix [ Yut ,5) .P, l (h)and Q,,(h ) terms depend
on the type of load model. For example for the constant power load:
(6.26)
For the nonlinear model
I so
In addition, for any type of load model. the active power generation term
can be modified to obtain
where the following definitions are made
PI , Ql,,,= Original load at bus i, active and reactive respectively r <
= Multiplier to designate the rate of load change at bus i as h changes
= Power factor angle of load change at bus i
= Apparent power which is chosen to provide appropriate scaling of h
= Active generation at bus i in the base case
= Constant to specify the rate of change in generation as h varies
= Initial voltage at the bus
= Frequency dependent fraction of active power load
= Voltage exponent for frequency-dependent active power load
= Voltage exponent for nonfrequency-dependent active power load
= Ratio of uncompensated reactive power load to active power load
= Voltage exponent for uncompensated reactive power load
= Voltage exponent for reactive power compensation
Now if F is used to denote the whole set of equations, then the problem
can be expressed as il set of nonlinear algebraic equations given by Eq. (6.20).
The predictor, corrector continuation process can then be applied to those equa-
tions.
The fir\( task in the predictor \tep is to calculate the tangent vector. Tlii\
tector can be obtained from factorizing Eq. (6.21).i.e.,
(6 .18)
On the left side of the equation is a matrix of partial derivatives multipl ed
by vector of differentials. The former s the conventional power flow Jacobian
Voltage Stcihility Assrsstnerit 151
augmented by one column ( F j , ) ,while the latter t = [de,dV,dh]' is the tangent
vector being sought. A normalization has to be imposed in order to give t a
nonzero length. One can use for example
e:t = tL= I (6.29)
where ek is an appropriately dimensioned row vector with all elements equal to
zero except the kfh one, which equals one. If the index k is chosen properly.
letting tk = + I .O imposes a nonzero norm on the tangent vector which guarantees
that the augmented Jacobian will be nonsingular at the point of maximum possi-
ble system load. Thus the tangent vector is determined as the solution of the
1i near sy s tem
Once the tangent vector has been found by solving Eq. (6.30), the prediction
can be made as follows:
(6.31 )
=[E] [ ;]+ U [i] where "*" denotes the predicted solution and (3 is a scalar that designates the
step size.
6.9 PARAMETERIZATION
The branch consisting of solutions of Eq. (6.20) forming a curve i n the (.LA) space has to be parameterized. A parameterization is a mathematical way of
identifying each solution on a branch. A parameterization is a kind of measure
along the branch. There are many different kinds of parameterization. For in-
stance, by looking at a PV curve, one sees that the voltage is continually de-
creasing as the load nears maximum. Thus, the voltage magnitude at some par-
ticular bus could be changed by small amounts and the solution is found for
each given value of the voltage. Here the load parameter would be free to take
on any value i t needed to satisfy the equations. This is called local parameteriza-
tion. I n local parameterization the original set of equations is augmented by one
equation that specifies the value of one of the state variables. In equation form
this can be expressed as follows:
1-52
\%=[q =o , (6.31 I
where q is an appropriate \~aluefor the kth element of J*. N o w once ii suitablc
index k and the value o f 11 itre chosen. a slightly modified Ne~\'ton-Rapliioii
(N-R) power flow method (altered only in that one additional equation and oiic
iiddirional state sariable are involved) can be iised to sol\.e the set of equatior-s.
This pro\.ides the corrector needed to modify the predicted solution found i n tlic
prc \,iou s section.
The algorithm tor static assessnient is shourn i n Figure 6.4. We ciin use ;I
simple example to explain the static ~inulysis proccdure.
6.9.1 Static Assessment: A Worked Example
Consider ;i \j'steiii is represented by
U here h is ;I irariation parameter from h,,= 0 to h, = A,,, To begin, s,ol\.e the s j stem equations at h = 0 , ~e ha\re
Input System Data 1 f
Select Contingencies
* Select Continuation Parameter 1
&
Solve base load flow .
I I Choose stepsize o I
* Calculate Stability index A p OT * Q
Figure 6.4 The algorithm for static assessment.
2-r -1 0 d.r
2 I - 1 1 [& = 0[ 0 1 0 clh
Since s,, = 1. substituting into the above equation we have:
2 - I 0 c1.r
2 1 4][" = O[ 0 1 0 dh
and
t1.r 0.5[ ."];[ 1.0
dh 2.0
Therefore,
Choosing 0 = 0. I , one gets
.?, = .c*= 1 .o + 0.05 = 1 .OS
f, = j'" = -2 + 0.1 = - I .9
h, = h* = 0 + 0.2 = 0.2
where .t,y, and are the approximated solutions.
In order to find the solution of F(.u,~.,h)= 0. we need to solve the equation
where r\ is an appropriate value of y.
Choose r\ = y* = -1.9. we have the solution of
'1
Voltcige Stcihility Assessriieiit
XI = 1 .OS
=-1.9
h, = 0.2
Based on the solution of (x ,y ,L) , we can get the solution of ( . I - ~ . ~ ! ~ , X ~ ) ,we
have
Choose CJ = 0. I , one gets
, ;, = [ I:i: + 11 0.2 + 9
21
Choose q = -1.8, we have the solution of
X I = fl = -1.8
h,= 2 0 - 1.8
Using the same procedure until the target system is reached. The modal analysis procedure is given in the following. System linearization equation is
given by
where p represents the variation parameter.
At .Y,~= I , y,, = -2, pi ,= 0, the above equation can be reduced to
I56
6.10 THE TECHNIQUE OF MODAL ANALYSIS
The inodal o r eigen\ralue analy\is method i \ a kind ot' sensitivity analy\is but
the modal \eparation provides additional insight. The \ystem partitional niatri-,
equations of the Newton-Raphson method can be reu ritten as
(6 .33 )
M here the partitioned Jacobian retlects ;I solb~edpower tlow condition and i n -
clude\ enhanced de\.ice modeling. By letting AP = 0. we can write
Uhere J K , is a reduced Jacobian matrix of the system. J K directly relates the h i \
voltage magnitude and bus reactive power injection.
Let h, be the ith eigenwlue of J , with cl and q, being the corresponding
col urn n right eiget n ~ ~tor and row left e i gen\.ect or, re spect i \re1y .
The ith modal reactise power Lwiation is
where A': cci,= I with c,l the j t h element o f 5,. The corresponding ith mocal
~ v ltage \wiit t ion is
The magnitude of each eigenvalue h, determines the weakness o f the corie-
sponding modal voltage. The smaller the magnitude of h, the uteaker the c o r ~e-
sponding modal voltage. If h,= 0. the ith modal voltage will collapse because
any change in that modal power will cause infinite modal voltage irariation.
If all eigenkralues are positive, the system is considered voltage stable. This
is a dift'erent dynamic system where eigenvalues with negative real parts i re
stable. The relationship between system bdtage stability and eigen\ralues of the
J , matrix is best understood by relating the eigen\dues with Q-\' sensiti\.ity o f
each bus. J , can be taken as ;t syminetric matrix and therefore the eigenvuli es
of J , are close to being purely real. I f all the eigenc~alues are positiLe J K is
positi\re definite and the V-Q sensitivities are also positiLre. indicating that the
system is voltage stable.
The system is considered voltage unstable if at least one of the eigenvalues
is positive. A zero eigenvalue of J , means that the system is on the \yerge of
voltage instability. Furthermore, small eigenvalues of J , determine the proximity
of the system to be voltage unstable.
The participation factor of bus k to mode i is defined as
For all the small eigenvalues, bus participation factors determine the areas
close to cdtage instability. In addition to the bus participations, modal analysis
also calculates branch and generator participations. Branch participations indi-
cate which branches are important in the stability of a g i \ m mode. This proLides
insight into possible remedial actions as well as contingencies. which inay result
in loss of voltage stability. Generator participations depict which machines niust
retain reactit,e reserves to ensure stability of a given mode. Figure 6.5 depicts
the technique static voltage stability assessment using modal analysis.
For a practical system with several thousand buses it is impractical and
unnecessary to calculate all the eigenvalues. Calculating only the minimum ei-
genvalue of J K is not sufficient because there are usually more than one \ \ ~ a k
modes associated with different parts of the system. and the mode associated
with the minimum eigenvalue may not be the most troublesome mode as the
system is stressed. The I I I smallest eigenvalues of J , are the I I I least stable modes
of the system. If the biggest of the vz eigen\dues, say mode I I I . is a strong
enough mode, the modes that are not computed can be neglected because they
are known to be stronger than mode ni. An implicit inixrse lopsided simultane-
OLIS iteration technique is used to compute the I I I smallest eigenvalues of J , and
the associated right and left eigenvectors.
Similar to sensitivity analysis, modal analysis (see the worked example at
the end of this chapter) is only valid for the linearized model. Modal analj~sis
can, for example. be applied at points along P-V cur\.es or at points i n lime of
a dynamic s i mU I at i on.
6.11 ANALYSIS TECHNIQUES FOR DYNAMIC
VOLTAGE STABILITY STUDIES
I t is only recently that the effects of system and load dynamics are being in\.esti-
gated in the context of voltage collapse. The dynamics that are being considered
are:
1.58
r
Obtain system architecture and network data
t I . Solve Base Case power flow +-2. Do Contingency Analysis 3. Select a desired set
+ Detailed
I . Fonn Full Jacobian Matrix. J Analysis?
Determine weaker voltage areas
based on eigenvalues I Perform participation factor
analysisI I
Compute
A V = g/\-'qAQ
Plot P-V and P-QCurves
L--l_--l Figure 6.5 Static voltage stability assessment using modal analysis.
1 . Machine and excitation system dynamics including power system stabi-
lizer (PSS).
2. Load dynamics.
3 . Dynamics of SVC controls and FACTS devices.
4. Tap-changer dynamics. 5. Dynamics due to load frequency control. AGC, etc.
While 1, 2, and 3 involve fast dynamics, 4 and 5 represent slow dynarrics.
A classification process of dynamic voltage stability vis-a-vis static stability is shown in Figure 6.6. Here "load" implies demand and "U" represents set pcints
of LFC, AGC, and voltage/VAr controls at substations. .Y, represents the slow
1
Volfuge SfcJhi1if.Y Assessriierif
, Subsystems: Subsystem F x&(x *x , u * I - d ) k f + x $xFut Load)
k $ ( x *XF"U*Load)
N (Voltage Collapse) Type.11
Instability
Both Subsystems S and F are Stable I I
Figure 6.6 Classification of voltage instabilities.
variables such as the state variables belonging to tap-changing transformers,
AGC loop and center of angle variables in the case of a multi-area representa-
tion. .rf. represents the fast variables belonging to the generating unit including
PSS and governor, induction motor load dynamics, SVC dynamics, and so forth.
The overall mathematical model is of the form:
(6.38)
(6.39)
(6.40)
(6.411
Ignoring the more slower AGC dynamics and the faster network transients
(60 Hz) we can categorize the variables appearing in Eq. (6.38)-(6.41).
x\ = [ n , ] i = 1, . . . ,I?
where
1 1 , = transfortner tap ratio
rectangular \ ariables o f ith bii4 \ oltage o r [ O f ]\:j
/, = niachinc terminal currents in machine rcference franie i = I , . . . . 111
de\ired real power of ith generator cle\ired \wltage at ith gcneratoi bu\
de~tredI oltagc at the bu\ controlled b j tap-changer i
i = 1, 2. . . . . 111
i = 1, 2. . . . . I I I
pr = c'ector of load parameters to be defined. The state \w-iables of the static VAr system ( S V C ) control and induct ,011
motor \ + r i l l appear in .v, i f included i n the o\rerall model. As an example \%re giife belour the equations for ;I I I I machine 11 bus s q w m halting p tap-changing t ran s formers. On Iy the sy tich ronoii s mach i ne 11nd t ap-c hanger d y nani i cs arc i n -c1Uded .
6.1 1.1 Equations of Slow and Fast Subsystems
For an ur-rnachine, 11-bus system ha\ing I - tap-changing transformerh. the follou~-ing equations are applicable
Slo\. v Subsystem
f a s t Subsystem
~ , , , c ' % = -E;, - ( x(/,- x;,) + E,,,, i = I . . . . . tit
-T\,'& = -vRl+ K \J,, E,,,,+ K ,,( v,,.,,- v,) i = I . . . . . 111
rlt Tl I
KI I T / , S= - RI,+ ~ E , , , , i = I , . . . . 111 (6.43)
tlt TI I
The algebraic equations for the stator and network can be used to andjrze
the system
6.11.2 load Flow and Equilibrium Point
The equilibrium point is calculated for a given set of reference points, \',',, ,.Tl,,,,l',,
and a given demand P I ,and QL,and then solving the follonfing equations for the
\w-iables 8?, .8,,. V,,,+], ,V,,.
I62 Cticipter- ri
We may alternatively combine Eq. (6.44) in a compact way as
, I
P:"' = ~ v , v , Y , , c o s ( ~ ,- 8, - a,k)= o i = I , . . . , I I
i-I
Q:"'
(1
= CV,V,Y, ,~~- e, - ~(O,a,h)= o i = I , . . . , 11 (6.3: )
i-I
and
The parameter vector p l can be defined in terms of PI,,,,Q,l,l,tt,,,, IZ,,,, etc.
The equilibrium point is calculated for a given set of reference points V, , ,
,.T,,,. V,,,, and a given demand P I ,and Ql,.The load flow equations are extracted
from Eqs. (6.45) and (6.46)as follows
1 . Specify bus voltage magnitudes numbered 1 to m. 2. Specify bus voltage angle number 1 (slack bus).
3. Specify net injected real power P'",'= P I ,and Q'"/'= Ql , at all buses num-
bered 111 + 1 to 1 1 .
Solve the following equations for the variables
e?,. . . ,elf. v,,,,,.. . . . v,,.
The standard load-flow Jacobian matrix involves the linearization of E q.
(6.47) with respect to 6:. . . . . 611rVlll+lr. . . , V,,. After the solution using Newtcm's
method. compute
(6.48)
In the above load flow problem one can include inequalities on Q genera-
tion at P-V buses, switching Var sources, etc. From the load flow solution, the
initial conditions of state variables in Eq. (6.48)can be computed systematically.
The initial value of V, is V,,,.
L inearization
Define 9' = [9:0;] corresponding to generator and load buses. Also define
.v' = [s: 1 .4= [.v;,.Y;, . . . J,!] where
s:= [HI, . . . , l l , , ]
and
.I-; = [G,,o,,E;,.E,,,EI,,,, VR,*R,,]i= I . . . . 1 111
and the algebraic variables as I,, Vq, V,. Also let
s,:= (Pi,,( V,l.Qi,( V , ) )
The linearized equations corresponding to Eqs. (6.32)-(6.43) can be ex-
pressed as
A , ,
0
0
0
0
In Eq. (6.48) the variations corresponding AVl in the nonlinear load charac- teristic is contained in ASLeand ASL,.
6.1 1.3 Static Stability (Type I Instability)
In Eq. (6.46), suppose that both Ais= AtI.= 0. Then we have a static situation
with all equations being algebraic. Let all the voltage deviations in AO? and At!,
be denoted by AP. then the rest of the algebraic variables can be eliminated
(assuming con\tant power load) to express A P = J I HAp,. I f det (1,)-+0 ;I\
load is increased i t is referred to a s Type I static in\tability. i.e., the \y\tem i \
not able to handle the increased load.
6.1 1.4 Dynamic Stability (Type II Instability)
Eliminating the algebraic ~ariables i n Eq. (6.39)atid assuming Ail E 0. i t can bt.
expressed a s
6.1 1.5 Slow Instability
Theoretically i t shoiild be possible to eliminate AY,in Eq. (6.50)using the singii-
lar perturbation theory and obtain the linearized slow system as At, = A,A.\.,.
The time scale o f the phenomena is so large that linearized results may ii8.)t
retlect the true picture. For such a time intensi\re phenomena. nonlinear siii1ul.i-
t ion is recolnmended.
6.11.6 Fast Instability
Fint u e rearrange the Lrar.iable\ [Al,.Av,,A$',] a r [AZ,.&.AV,, . . . .AV,,l 1 A02.AOj.
. . . .AO,,,AV,,,+,,. . . = [A:.Avl. Next b e a\\ume I , a\ con\tant and load pa-
r:inieter\ a\ con\tant which implie\ Ap, = 0 . We get
For the constant power case, both AS, and AS2 are = 0. Otherwise. AS, =
AS,,(V,) and AS2,= AS,,(V,). For a given voltagedependent load, AS,, and
can be computed. Only the appropriate diagonal elements of B:, C,, and C; M i l l
be modified and we obtain the system
Now cqis the load flow Jacobian J L I and B2 8' = J , / . The system matrix. A c. c,I
is obtained as
Atl = A,,,Avl + E h
Using drastic assumptions about voltage control and load characteristic\ that the
steady-state stability associated with the system matrix, A,,, can be determined
by examining the load flow Jacobian, J , / .
6.1 1.7 Voltage Stability Assessment
The algorithm for Lwltage collapseholtage stability assessment includes static
and dynamic assessment. The algorithm for dynamic stabilitl- assessnient is
shown in Fig. 6.7.
6.1 1.8 VSTAB-Voltage Stability Assessment (EPRI)
A more promising method with the trade name VSTAB. uses po~7el- tlow and
modal analysis techniques. I t provides assessment of the proximity to Lvltage
instability and determines the mechanism of voltage instability. I n this method.
the proximity to \voltage instability is evaluated by conducting a series of p o ~ ' e r
tlow solutions with load increase until load tlow diLwgsiice is encountered.
When load flow divergence is encountered, the step size for load increase i x
reduced and the power flows are continued. The voltage stability limit is consid-
ered to have been reached when the step size reaches the cutoff due specified
by the user. The load level at this point is the maximum loadabilitjf. This proce-
dure is carried out simultaneously for the intact system as well as for contingen-
cies. Load increase can be carried out with or without generation scaling. The
slack bus generation is not scaled. Loading can be by area or by zonc.
The mechanism of voltage instability is studied in VSTAB by using modal
analysis. Modal analysis employing V-Q sensitivities can identifqr areas that
have potential problems and provide information regarding the mechanism of
Lroltage collapse. The method is briefly discussed as follows.
The usual power tlow equations can be expressed in the linearized form.
( 6 . 5 3 )A \IAel
where
AP = incremental change in bus real power
I66
* Select Contingencies
+ Select Continuation Parameter P(Q)
* Sdve base load flow
rk Increase P(Q)
+ Run the power flow
I
< Converge?
Yes
I I Compute initial conditions of state variable
-I Calculate eigenvalue A 1
Figure 6.7 Summary of the dynamic voltage stability assessment technique.
AQ = incremental change in bus reactive power injection
A 0 = incremental change in bus voltage angle
A V = incremental change in bus voltage magnitude The system dynamic behavior can be expressed by the first order diffei-en-
tial equation,
167 Voltage Stability As.se.ssineiit
where
X = state vector of the system
V = bus voltage vector
For the steady-state condition X = 0, using the enhanced device models used
in Eq. (6.54). Equation (6.53) can be rewritten as:
(6 .55)
where
Af,, = incremental change in device real power output
AQ,,= incremental change in device reactive power output
AV,, = incremental change in device voltage magnitude
A0(,= incremental change in device voltage angle
The terms A l l , AI? , and AZ2represent a modified form of J,w. J\>,, JVH.
J,-, in the terms associated with each device. We can study the Q-V sensititrity
while keeping f constant. For this analysis we can substitute Af = O in Eq.
(6.53) to give us upon simplification,
AV = Jk
A0 (6.56)
where
By analyzing the eigenvalues and eigenvectors of the reduced Jacobian J R .
we arrive at
1) = A-’ x q (6.58)
or
1 v, = -9,A,
(6.59)
where
h is the iIhmodal voltage.
I hS
U\ing modal analy\is. these relative bu\ participation and branch particip.1-
(ion factor\ c m be computed for the i"' mode. The complete procedure for \tatic
\ oltage \tabilit> a\se\snient L ia rnodal analy\i\ is outlined in Fig. 6.8.
6.1 1.9 Preventive Control of Voltage Stability
There are t ~ olet els of lroltage stability enhancement, the fir\t le\rel uith del ic:e
ba\eci control. the \econd level i \ i n the form of operation-based control. The
Obtain base case
Set for pre-contingency
solution
Solve the load flow
0: contingenc
Generate QV curves 1 b+,
- New load level or change
Figure 6.8 Thc VSTAB algorithm (Cl EPRI) .
voltage stability is improved by optimal system operation conditions. The static
analy\is method is used for the determination of prcbmitive control scheme.
System operation conditions are determined by F ( 8 . 1’. h )= 0. The design of a
broltage stability preventive control scheme includes the \teps outlined i n Fig.
6.9.
CONCLUSION
Pourer sy s t e111 i.01tag e st abi I i t y i n vol ves generat i on, t ran sni i ss ion and d i s t ri h i -
tion. So to maintain the voltage stabilitr is crucial tor thc normal operatioti
Input System Data
4 Select Critical Contingencies
1 Use Optimal Power Flow
to do Contingency
4 Identify and Rank contingencies with
low stability limit using VSTAB
+I Select the first contingency 1
Any other contingencies ?
Select the next contingency from the list
4 Output results
Incorporate the selected contingencies in the Contingency Constrained OPF for
expanding the lowest stability limit
c Adjust control parameters to reflect
optimized values
a I
Figure 6.9 VoI t age 5 tabi 1it y pre\.ent i v e cont ro I \c he iiie.
I 70 Chccprer 6
of a power system. In adequate reactive power support from generators arid
transmission lines lead to voltage instability or voltage collapse which halie
resulted in several major system failures (blackouts) such as the massive Tokyo
blackout in July 1987.
In order to prevent the stability limit being reached or exceeded during a
given contingency, remedial actions need to be recommended. I t is well knov,Tn
that in all cases, voltage instability is caused by inadequate transmission capac-
ity at a given operating condition due to a contingency, which the system cannot
withstand. Based on contingencies that occur, the distribution of plant gene1 a- tors, transmission tlows and load to meet given stability criteria is usually done
by using effective/economical control actions.
Future work in the determination of adequate remedial measures for stabil-
i ty enhancement have been proposed in past publications, where the correctiLre
control action is handled as an optimization problem. The two-stage formulati .In
to achieve the desired stability enhancement utilizes the concepts of Chapta S and this chapter. The first stage handles voltage stability enhancement while [he
second-stage optimization scheme deals with angle stability enhancement. The
process will lead to a unified index. Hence, when carrying out stability enhance-
ment based on a selected list of contingencies, only enhancement of the app-o-
priate problem (either voltage or angle) needs to be carried out, thus saving
labor and computational time. Future work in unifying the indices while inc,.>r-
porating the irarious available controls is still a challenge. The reader is intrir ed
to research further literature in selected references located at the end of .he
book. Also. the annotated glossary of terms supports the chapter.
M O D A L ANALYSIS: WORKED EXAMPLE
Consider the SO0 kV. 322 km (200 miles) lines transmission system shown in
Fig. I O(a) below supplying power to a radial load from a 'strong' power system
represented by an infinite bus. The line parameters, as shown in Fig. 10(b). are
expressed in their respective per unit values on a common system base of 00
MVA and 500 kV.
1 . 1 Compute the full admittance matrix of the two-bus system and write the
power flow equations from the sending end to the receiving end in the form:
I .2 Hence or otherwise, write down the expressions for the four (4) sub-matri-
ces of the Jacobian in the linearized load tlow equations as defined by:
Voltage Stcihility Assswiierit I7i
(a) Infinite Bus Load Bus Bus I Bus 2
pZ-JQz,
Transmissicm Line
(3c1Load Shunt Load
Qsh -7 j
(b) Infinite Bus Load Bus Bus I Bus 2
v, = 1 .OLOO "2 =I v, IY, = 2.142 -J24.973
Figure 6.10 The SS0 kV, 370 km (230 miles) line tranwii\\ion \ystem \uppl>4ng a
radial load: (a) schematic diagram of the transmission system and (b) the equi\alt.nt
WYE circuit repmentation of the transmission line.
1.3 When P2= 1500 MW, calculate the eigenvalues of the reduced Q-V Jacob-
ian matrix and the V-Q sensitivities with the following different reactive power
injections for each of the corresponding two voltages on the Q-V curve.
a. Q, = 500 MVAR.
b. Q,=400 MVAR.
c. Values of Q, close to the bottom of the V-Q curve
1.4 Determine the voltage stability of the system by computing the eigenval-
ues of the reduced V-Q Jacobian matrix for the following cases:
a. P = 1500 MW, Q,= 450 MVAR.
b. P = 1’300 MW. Ql= 950 MVAR.
(Aswnie that the reactiLne pouw Qi i \ wpplied by ii \hunt capacitor).
Solution
From the figure. the admittance matrix of the 2-bus system is
2.142 - j22.897 -2.142 + j24.973
-2.142 + i24.973 2.142 - j23.897
The expression tor P and (2 at any bus k are gibfenby:
where
Hence. are interested in only P1 and Q2b v i t h V, = 0 p i i .
Hence the expressions tor the Jacobian terms are give by:
a P+J,ll = = -2.132cose + 24.973sine + 4.284\’> a v,
(a) The linearized power flow equations are
with
The expression for J,,. J,,, Jpo. and Jl,, were given before. For this simple
system, J K is a I x 1 matrix. The eigenvalue lambda of the matrix is the same
as the matrix itself. The Q-V sensitivity is equal to the inlverse of the eigenkdue.
For each of the Q,s, there are two solutions for the recei\,ing end \*oltage.
Table 6.2 summarizes the V, 8, h, and dV/dQ with P = 1500 MW and Q = 500.
400, 306, and 305.9 MVAr. For each case the eigenvalue and W Q sensiti\,ity
are both negatikre at the low voltage solutions, and are both p0sitii.e at high
jroltage solutions. With Q = 305.9 MVAr close to the bottom of the Q-\’ c~11-k~
tlVIdQ is large and h is very small.
Table 6.2 Results for Modal Analysis Worked Example
High Voltage Solution Lou Voltage Solution
500.0 1.024 -37.3 17.03 0.059 0.67 1 -66.7 -39.87 -0 .02S
400.0 0.956 -40.1 12.4 I 0.081 0.706 -60.3 -20.96 -0.048 306.0 0.820 -48.2 0.52 1.923 0.812 -48.8 -0.9SO - I . O S 3
305.5 0.184 -48.7 0.02 50.10 0.815 -48.6 -0.700 -1.134
( b ) With the shunt capacitor connected at the receiving end of the line, the
self admittance is:
Y,, = 2.142 -j(22.897 - H , )
with P = IS00 MW and A 450 MVAr reactive shunt capacitor. V, =
0.98I , 8 = -39.1 degrees. Since B, = 4.5 pu., then Y2:= 2.132 -
j(22.897 - 4.5) = 2.132 -j18.397. With this new value of Y2:, the r t s -
duced Q-V Jacobian matrix is J , = 5.348, and J , is positive indicating
that the system is voltage stable.
with P = I900 MW and a 950 MVAr reactive shunt capacitor, V, =
0.995, 8 = -52.97 degrees. Since B, = 9.5 pi'.. then YJ2= 2.142 -
j(22.897 - 9.5)= 2.142 -jl.397. With t h i h new value of Y?:. the re- duced Q-V Jacobian matrix is J , = -13.683, and J , is negative indicat-
ing that the system is voltage iinstable.