Power System Protection

A Novel Approach for Tuning of Power
System Stabilizer Using Genetic

Algorithm
A Thesis
Submitted for the Degree of
in the Faculty of Engineering
By
Ravindra Singh
Department of Electrical Engineering

INDIAN INSTITUTE OF SCIENCE
Bangalore 560012, (INDIA)
July 2004
Abstract
The problem of dynamic stability of power system has challenged power system engineers
since over three decades now. In a generator, the electromechanical coupling between the
rotor and the rest of the system causes it to behave in a manner similar to a spring mass
damper system, which exhibits an oscillatory behaviour around the equilibrium state, following any disturbance, such as sudden change in loads, change in transmission line parameters,
fluctuations in the output of turbine and faults etc. The use of fast acting high gain AVRs
and evolution of large interconnected power systems with transfer of bulk power across weak
transmission links have further aggravated the problem of low frequency oscillations. The
oscillations, which are typically in the frequency range of 0.2 to 3.0 Hz, might be excited by
the disturbances in the system or, in some cases, might even build up spontaneously. These
oscillations limit the power transmission capability of a network and, sometimes, even cause
a loss of synchronism and an eventual breakdown of the entire system.
The application of Power System Stabilizer (PSS) can help in damping out these oscillations and improve the system stability. The traditional and till date the most popular solution to this problem is application of conventional power system stabilizer (CPSS). However,
continual changes in the operating condition and network parameters result in corresponding
change in system dynamics. This constantly changing nature of power system makes the
design of CPSS a difficult task.
Adaptive control methods have been applied to overcome this problem with some degree of
success. However, the complications involved in implementing such controllers have restricted
their practical usage.
In recent years there has been a growing interest in robust stabilization and disturbance
attenuation problem. H control theory provides a powerful tool to deal with robust stabilization and disturbance attenuation problem. However the standard H control theory
does not guarantee robust performance under the presence of all the uncertainties in the
power plants.
This thesis provides a method for designing fixed parameter controller for system to ensure
robustness under model uncertainties. Minimum performance required of PSS is decided a
priori and achieved over the entire range of operating conditions.
A new method has been proposed for tuning the parameters of a fixed gain power system stabilizer. The stabilizer places the troublesome system modes in an acceptable region
in the complex plane and guarantees a robust performance over a wide range of operating
conditions. Robust D-stability is taken as primary specification for design. Conventional
lead/lag PSS structure is retained but its parameters are re-tuned using genetic algorithm
(GA) to obtain enhanced performance. The advantage of GA technique for tuning the PSS
parameters is that it is independent of the complexity of the performance index considered.
It suffices to specify an appropriate objective function and to place finite bounds on the optimized parameters. The efficacy of the proposed method has been tested on single machine
as well as multimachine systems. The proposed method of tuning the PSS is an attractive
alternative to conventional fixed gain stabilizer design as it retains the simplicity of the conventional PSS and still guarantees a robust acceptable performance over a wide range of
operating and system condition.
The method suggested in this thesis can be used for designing robust power system stabilizers for guaranteeing the required closed loop performance over a prespecified range of
operating and system conditions. The simplicity in design and implementation of the proposed stabilizers makes them better suited for practical applications in real plants.
Acknowledgements
The completion and compilation of this thesis is the outcome of inspiring guidance of Dr.
Indraneel Sen. His keen interest in the progress of this work and patience to read through
my script are greatly acknowledged. I am thankful for his suggestions and discussions.
A special word of thank is due to Prof. K. R. Padiyar for his excellent teaching and who
influenced me to create a deep interest in the area of Power System Dynamics.
The help and cooperation of the chairman and staff of the Department of Electrical Engineering is gratefully acknowledged.
Perhaps words cannot express the gratitude I owe all my seniors like Mr. Anup Kumar
Singh, Mr. Maneesh Tewari, Mr. Nagesh Prabhu, Ms. Bijuna and Ms. Divya and friends
like Raghvendra Gupta, Raghvendra Pandey, Ashish, Ritwik, Amit and Vishal who helped
me in every way.
It, probably, goes without saying that I owe the biggest thank to my parents and family
members who have been a constant source of help and encouragement.
Finally, I thank everyone who have directly or indirectly helped me during the course of
this work.
Ravindra Singh
Contents
List of Tables
iv
List of Figures
1 Introduction
1.1
Low Frequency Oscillations in Power System . . . . . . . . . . . . . . . . . .
1.2
Fixed Parameter Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.1
Conventional Stabilizers . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.2
Other Fixed Parameter Controllers . . . . . . . . . . . . . . . . . . .
1.2.3
The Drawbacks of Conventional Fixed Parameter Controllers . . . . .
1.3
Adaptive Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4
Fuzzy Logic Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.5
Robust Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.6
Application of Genetic Algorithms to PSS Design . . . . . . . . . . . . . . .
1.7
Robust PSS design using Genetic Algorithms: the present approach . . . . .
10
1.8
Performance Requirements of Power System Damping Controllers . . . . . .
11
1.8.1
How Much Damping Do We Need? . . . . . . . . . . . . . . . . . . .
12
1.8.2
Performance Evaluation of a PSS . . . . . . . . . . . . . . . . . . . .
13
Scope of Present Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
1.10 Organization of Chapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
1.9
2 Mathematical Modelling of Power System
16
2.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16
2.2
SMIB Model in Non-Linear Form . . . . . . . . . . . . . . . . . . . . . . . .
17
2.2.1
17
Rotor Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
i
Contents
ii
2.2.2
Stator Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
2.2.3
Network Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.3
Excitation System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
20
2.4
PSS Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
2.5
SMIB Test System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
2.6
Modelling of Multimachine System . . . . . . . . . . . . . . . . . . . . . . .
23
2.6.1
Rotor Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
24
2.6.2
Stator Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
26
2.6.3
Inclusion of Generator Stator in the Network . . . . . . . . . . . . . .
26
2.6.4
Load Representation . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
2.6.5
Network Equations for Multimachine . . . . . . . . . . . . . . . . . .
28
2.7
Multimachine Test System . . . . . . . . . . . . . . . . . . . . . . . . . . . .
29
2.8
Linearized 1.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
3 Genetic Algorithm: An Overview
32
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
32
3.2
What is Genetic Algorithm? . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.3
Working Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.3.1
Coding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.3.2
Fitness Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.3.3
GA Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
3.3.4
Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
3.4
Implementation of genetic algorithm . . . . . . . . . . . . . . . . . . . . . .
36
3.5
Mathematical Model of SGAs . . . . . . . . . . . . . . . . . . . . . . . . . .
38
3.6
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
41
4 Proposed Stabilization Technique: Single Machine System
42
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
42
4.2
Objective Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
4.3
Proposed Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
4.4
Application to SMIB System . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
4.4.1
Control Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
4.4.2
GA Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
46
Contents
4.5
iii
4.4.3
Operating Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
4.4.4
GA Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
4.4.5
Performance Analysis of Proposed GA Based PSS . . . . . . . . . . .
48
4.4.6
Robustness Test and Eigen Value Plots . . . . . . . . . . . . . . . . .
48
4.4.7
Simulation Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.4.8
Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .
52
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
5 Proposed Stabilization Technique: Multimachine System
64
5.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
64
5.2
Performance Evaluation of the Stabilizer in Multimachine System . . . . . .
64
5.2.1
Control Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . .
64
5.2.2
GA Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
5.2.3
Loading Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
5.2.4
GA Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
66
5.2.5
Robustness Test and Eigen Value Plots . . . . . . . . . . . . . . . . .
67
5.2.6
Operating Points For Simulation Studies . . . . . . . . . . . . . . . .
69
5.2.7
Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . .
70
5.2.8
Computational Requirements . . . . . . . . . . . . . . . . . . . . . .
73
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
73
5.3
6 Conclusions
85
A Calculation of Initial Conditions
87
B Heffron-Philips Model of the SMIB System
88
C Data for SMIB and Multimachine System
90
D Tuning Guidelines for the CPSS
92
E Mapping From a Binary String to a Real Number
98
F Derivation of Equation 4.1
99
References
101
List of Tables
4.1
Control Parameters Bounds . . . . . . . . . . . . . . . . . . . . . . . . .
46
4.2
GA Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
47
4.3
Initial and Final Values of PSS Parameters . . . . . . . . . . . . . . .
48
5.1
Control Parameters Bounds . . . . . . . . . . . . . . . . . . . . . . . . .
65
5.2
GA Parameters For Multimachine Case . . . . . . . . . . . . . . . . .
65
5.3
Loading Range of 3 Machine, 9 Bus System
. . . . . . . . . . . . . .
66
5.4
Optimal stabilizer parameters of PGAPSS . . . . . . . . . . . . . . . .
67
5.5
Operating points of generators on a 100 MVA base . . . . . . . . . .
69
C.1 Generator Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
C.2 AVR Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
90
C.3 Generator Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
C.4 AVR Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
iv
List of Figures
1.1
D-contour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
12
2.1
External two port network . . . . . . . . . . . . . . . . . . . . . . . . . . . .
17
2.2
Excitation system block diagram. . . . . . . . . . . . . . . . . . . . . . . . .
21
2.3
Block diagram of PSS
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
2.4
Single machine infinite bus system . . . . . . . . . . . . . . . . . . . . . . . .
22
2.5
Schematic of a multimachine system
. . . . . . . . . . . . . . . . . . . . . .
23
2.6
Generator equivalent circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . .
27
2.7
3 machine, 9 bus power system model, single line diagram. . . . . . . . . . .
29
3.1
Single point crossover operation . . . . . . . . . . . . . . . . . . . . . . . . .
35
3.2
A single mutation operation . . . . . . . . . . . . . . . . . . . . . . . . . . .
35
3.3
The general structure of genetic algorithms . . . . . . . . . . . . . . . . . . .
37
4.1
Flow Chart representation of the proposed method of tuning stabilizer
. . .
45
4.2
Open loop poles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
4.3
Closed loop poles with CPSS . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
4.4
Closed loop poles with PGAPSS . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.5
A step change of Tm = 0.1 pu, St = 0.5 + j0.1, Xe = 0.3 . . . . . . . . . . .
55
4.6
55
4.7
55
4.8
56
4.9
56
4.10 A step change of Tm = 0.1 pu, St = 0.5 + j0.1, Xe = 0.6
. . . . . . . . . .
56
4.11 A step change of Tm = 0.1 pu, St = 0.8 + j0.2, Xe = 0.6 . . . . . . . . . . .
57
List of Figures
vi
57
57
58
58
58
59
59
59
4.20 A step change of Tm = 0.1 pu, St = 0.5 j0.2, Xe = 0.3 . . . . . . . . . . .
60
60
60
61
4.24 A 3 to ground fault for 100 ms at generator terminal, St = 1.0+j0.2, Xe = 0.3 61

4.25 A 3 to ground fault for 100 ms at generator terminal, St = 0.8j0.2, Xe = 0.3 61
4.26 A 3 to ground fault for 100 ms at generator terminal, St = 1.0+j0.5, Xe = 0.6 62
4.27 A step change of Tm = 0.1 pu, St = 1.0 + j0.2, Xe = 0.3, H 0 = H/4 . . . .
0
62
4.28 A step change of Tm = 0.1 pu, St = 0.8 j0.2, Xe = 0.3, H = H/4 . . . .
62
4.29 A step change of Tm = 0.1 pu, St = 1.0 + j0.5, Xe = 0.6, H 0 = H/4 . . . .
63
4.30 A step change of Tm = 0.1 pu, St = 0.6 j0.15, Xe = 0.65 . . . . . . . . .
63
5.1
Open loop poles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
5.2
Closed loop poles with CPSS . . . . . . . . . . . . . . . . . . . . . . . . . . .
68
5.3
Closed loop poles with PGAPSS . . . . . . . . . . . . . . . . . . . . . . . . .
69
5.4
A step change of Tm1 = 0.1 pu at unit 1, under SOP. . . . . . . . . . . . .
75
5.5
A step change of Tm1 = 0.1 pu at generator 1, under SOP. . . . . . . . . .
75
5.6
75
5.7
76
5.8
76
5.9
76
5.10 A 3- to ground fault for 100 ms at P, under SOP. . . . . . . . . . . . . . .
77
5.11 A 3- to ground fault for 100 ms at P, under SOP. . . . . . . . . . . . . . .
77
5.12 A step change of Tm1 = 0.1 pu at generator 1, under HOP. . . . . . . . . .
77
78
List of Figures
vii
78
78
79
79
5.18 A 3- to ground fault for 100 ms at P, under HOP. . . . . . . . . . . . . .
79
5.19 A 3- to ground fault for 100 ms at P, under HOP. . . . . . . . . . . . . .
80
5.20 A step change of Tm1 = 0.1 pu at generator 1, under LOP. . . . . . . . . .
80
80
81
81
81
82
5.26 A 3- to ground fault for 100 ms at P, under LOP. . . . . . . . . . . . . . .
82
5.27 A 3- to ground fault for 100 ms at P, under LOP. . . . . . . . . . . . . . .
82
5.28 A step change of Tm2 = 0.1 pu at generator 2, under OOP. . . . . . . . . .
83
83
83
84
84
84
B.1 Heffron-Philips model of the SMIB system . . . . . . . . . . . . . . . . . . .
88
D.1 Phase angle plots of GEP(s), PSS(s) and GEP(s).PSS(s) . . . . . . . . . . .
93
D.2 Phasor diagram representation of synchronizing and damping torques . . . .
93
D.3 A typical root locus plot for SMIB system with the lead Compensator . . . .
94
D.4 Relationship governing VT (s), P SS(s), and EXC(s)
. . . . . . . . . . . . .
96
D.5 Phase shift of Tpss for a speed input PSS . . . . . . . . . . . . . . . . . .
97
F.1 D-contour in x y plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
99
Chapter 1
Introduction
1.1
Low Frequency Oscillations in Power System
Small oscillations in power systems were observed as far back as the early twenties of
this century. The oscillations were described as hunting of synchronous machines. In a
generator, the electro-mechanical coupling between the rotor and the rest of the system
causes it to behave in a manner similar to a spring-mass-damper system which exhibits
oscillatory behaviour following any disturbance from the equilibrium state.
Small oscillations were a matter of concern, but for several decades power system engineers
remained preoccupied with transient stability. That is the stability of the system following
large disturbances. Causes for such disturbances were easily identified and remedial measures
were devised. In early sixties, most of the generators were getting interconnected and the
automatic voltage regulators(AVRs) were more efficient. With bulk power transfer on long
and weak transmission lines and application of high gain, fast acting AVRs, small oscillations
of even lower frequencies were observed. These were described as Inter-Tie oscillations. Some
times oscillations of the generators within the plant were also observed. These oscillations
at slightly higher frequencies were termed as Intra-Plant oscillations.
The combined oscillatory behaviour of the system encompassing the three modes of oscillations are popularly called the dynamic stability of the system. In more precise terms it is
known as the small signal oscillatory stability of the system.
A power system is said to be small signal stable for a particular steady-state operating
condition if, following any small disturbance, it reaches a steady state operating condition
which is identical or close to the pre-disturbance operating condition.
1
Chapter 1. Introduction
The oscillations, which are typically in the frequency range of 0.2 to 3.0 Hz., might be
excited by disturbances in the system or, in some cases, might even build up spontaneously.
These oscillations limit the power transmission capability of a network and, sometimes, may
even cause loss of synchronism and an eventual breakdown of the entire system. In practice,
in addition to stability, the system is required to be well damped i.e. the oscillations, when
excited, should die down within a reasonable amount of time.
Reduction in power transfer levels and AVR gains does curb the oscillations and is often
resorted to during system emergencies. These are however not feasible solutions to the
problem.
The stability of the system, in principle, can be enhanced substantially by application of
some form of close-loop feedback control. Over the years a considerable amount of effort
has been extended in laboratory research and on-site studies for designing such controllers.
There are basically three following ways by which the stability of the system can be improved,
(1) Using supplementary control signals in the generator excitation system.
(2) Making use of fast valving technique in steam turbine.
(3) Impedance Control-resistance breaking and application of the FACTS devices, etc.
The problem, when first encountered, was solved by fitting the generators with a feedback
controller which sensed the rotor slip or change in terminal power of the generator and fed
it back at the AVR reference input with proper phase lead and magnitude so as to generate
an additional damping torque on the rotor [1]. This device came to be known as a Power
System Stabilizer (PSS).
Damping power oscillations using supplementary controls through turbine, governor loop
had limited success. With the advent fast valving technique, there is some renewed interest
in this type of control [2].
There can also be other kinds of controls applied to the system for counteracting the oscillatory behaviour - for instance FACTS devices can be fitted with supplementary controllers
which improve the system stability.
Power system stabilizers are now routinely used in the industry. However, the complex,
constantly changing nature of power systems has severely restricted the efficacy of these
devices.
1.2
Fixed Parameter Controllers
Over the years, a number of techniques have been developed for designing PSSs and other
damping controllers [3]. Some of these stabilizing methods have been briefly described in
this section. The main motivation for including this rather brief exposition of the existing
techniques is to introduce the need for the application of robust control techniques in power
systems. Some of references cited here include a more comprehensive coverage of the topic.
1.2.1
Conventional Stabilizers
The earlier stabilizer designs were based on concepts derived from classical control theory
[4-8]. Many such designs have been physically realized and widely used in actual systems.
These controllers feedback suitably phase compensated signals derived from the power, speed
and frequency of the operating generator either alone or in various combination as input
signals so as to generate an additional rotor torque to damp out the low frequency oscillations.
The gain and the required phase lead/lag of the stabilizers are tuned by using appropriate
mathematical models, supplemented by a good understanding of the system operation.
The principles of operation of this controller are based on the concepts of damping and
synchronizing torques within the generator. A comprehensive analysis of these torques has
been dealt with by deMello and Concordia in their landmark paper in 1969 [1]. These
controllers have been known to work quite well in the field and are extremely simple to
implement. However, the tuning of these compensators continues to be a formidable task
especially in large multimachine systems with multiple oscillatory modes. Larsen and Swann,
in their three part paper [6], describe in detail the general tuning procedure for this type of
stabilizers.
PSS design using this method involves some amount of trial and error and experience on
part of the designer. Further these controllers are tuned for a particular operating conditions and with change in operating conditions they require re-tuning. Robustness issues are
also not adequately addressed in this classical setting. The problem associated with these
controllers is more fully described later in this chapter.
1.2.2
Other Fixed Parameter Controllers
There have also been numerous attempts at applying various other control strategies in particular -modal control [9-11] and LQ optimal control [12, 13] techniques for designing
damping controllers. These attempts exemplify the growing preference for algorithmic controller design methods as opposed to the classical intuitive ones. They call for a lesser amount
of engineering judgement and experience on part of the designer. The ill-suitedness of the
quadratic performance index used in LQR/LQG to the problem has motivated researchers to
define alternative performance indices which aptly capture the magnitude of system damping
[14, 15]. Such indices can be optimized using standard numerical optimization techniques
[16].
These techniques have the advantage of being straight forward and algorithmic with little ambiguity in the recommended procedure. A few extensions of these methods tried to
incorporate some robustness by optimizing some additional index such as eigen value sensitivities. Sensitivity minimization in this form, though, quite helpful as a means of providing
robustness in the absence of better methods is essentially a qualitative approach and hence
does not guarantee performance preservation in the face of modal inaccuracies [17].
1.2.3
The Drawbacks of Conventional Fixed Parameter Controllers
The main drawback of the above controllers is their inherent lack of robustness. Power
systems continually undergo changes in the load and generation patterns and in the transmission network. This results in an accompanying change in small signal dynamics of the
system. The fixed parameter controllers, tuned for a particular operating condition, usually
give good performance at that operating condition. Their performance, at other operating
conditions, may at best be satisfactory, and may even become inadequate when extreme
situations arise. However such stabilizers have been very useful in system that could be
represented by single machine infinite bus models. In interconnected multimachine systems
the dynamic instability can manifest itself in the form of poorly damped oscillation of one
particular unit with the rest of the system or a group, or a group of machines oscillating
against another group of machines. Thus, a generating unit in a multimachine environment
often participates in both local and inter-area modes of oscillations simultaneously. The
spectral and temporal distributions of these modes are largely determined by the rest of the
system. As the operating conditions and system configuration are constantly changing in
actual power system the performance of the fixed parameter stabilizers can not be always
guaranteed.
1.3
Adaptive Controllers
The problem of changing system dynamics due to changes in the operating conditions can
be handled by the application of adaptive control [18, 19]. The power system can be continuously monitored and the controller parameters can be updated in real time to maintain
specified performance inspite of changes in the system dynamics. All three standard methods
of adaptive control listed below have been tried for designing power system stabilizers.
(a) Model reference adaptive control (MRAC) [20, 21]
(b) Self tuning control (STC) [22-24]
(c) Gain scheduling adaptive control (GSAC) [25]
In MRAC, the desired behavior of the closed loop system is incorporated in a reference
model. With the plant and the reference model excited by the same input, the error between
the plant output and the reference model output is used to modify the controller parameters,
such that the plant is driven to match the behavior of the reference model.
In STC, at every sampling instant, the parameters of an assumed model for the plant are
identified using some suitable algorithms, such as Recursive least squares (RLS) or Maximum
likelihood estimator etc. The identified parameters are then used in control laws which could
be based on various popular techniques such as pole-shifting, pole placement etc.
In GSAC, the gains of the controller are adjusted according to a variety of innovative
control strategies depending upon the plant operating conditions and important system
parameters. The gains could be computed either off-line or on-line.
A few non standard adaptive control schemes have also been reported [26, 27] which do
not fit into any of the above categories. These schemes have been shown to work quite well
through simulations and laboratory experiments.
Adaptive controllers totally avoid the problem of tuning since that is taken care of by the
adaptation algorithm. The trade off is the larger on-line computational requirement. The
stabilizers are difficult to design and are also susceptible to problems like non-convergence
of parameters and numerical instability. Due to these reasons practical implementation of
adaptive stabilizers in actual plants has not been popular.
There have been numerous non conventional approaches including feedback linearization,
variable structure or sliding mode control and, in more recent times, schemes involving neural
networks, fuzzy systems and rule based systems [3] for designing stabilizers. Many of these
non-conventional approaches have been shown to work quite well in simulated power system
models.
Some of the above approaches have also been applied for designing supplementary stabilizing controls for FACTS devices. Most of the modern control theoretical techniques use a
black box model for the plant. Hence, identical procedures can be adopted for the design of
power system stabilizers and other damping controllers.
1.4
Fuzzy Logic Controllers
In recent years, Rule based [28, 29], Artificial Neural Network (ANN) based [30, 31] and
Fuzzy Logic based [32-37] controllers have been suggested for PSS design. These are modelfree controllers i.e. precise mathematical model of the controlled system is not required. Here
control strategy depends upon a set of rules which describe the behavior of the controller.
Here lies, both the strength and weakness of this design philosophy. FLC controllers are
well-suited for PSS design as system and its interrelations are not precisely known as they
keep constantly changing with changes in both system and operating conditions. However,
as the design is rule and experience based, there can not be a unique design procedure.
1.5
Robust Control
The last 15 years have seen major developments in the field of robust control. This topic
deals with the analysis and design of feedback systems subject to incomplete knowledge of
the plant dynamics and accompanying uncertainties in the model of the plant. Such an
uncertainty in the plant model could arise due to various reasons, for instance - deliberate
approximations in the modelling procedure, measurement inaccuracies, parameter drifts and

time varying nature of certain systems. In the case of power systems, the nonlinear system
equations when linearized about an operating point result in a linear model with parameters
which vary with the operating condition.
Some of the major approaches to robust control are:
1. Loop transfer recovery for LQG designs [38]
2. H optimal control [39-48]
3. analysis and synthesis [49-50]
4. 11 optimal control [51]
5. Quantitive feedback theory (QFT) [52]
6. Parameter space methods [53]
In all the above approaches, except the first, the uncertainty in the plant is explicitly
modeled and is incorporated into the design process so as to guarantee good performance
in the presence of model uncertainties. Methods 2 and 4 above deal with norm bounded
descriptions of the uncertainty whereas 5 and 6 deal with bounds on parameter variations
in the system. analysis encompasses both kinds of descriptions.
Francis B.A. provides a theoretical basis to deal with uncertainties in a system control
design [39]. The parameter uncertainty was first addressed by Kartinov [54]. Doyle in his
frame work of H brought out a procedural approach to handle perturbations which are
norm bounded and time invariant [40].
H is a optimization technique which can be used to optimize the PSSs parameters. H
control theory provides a powerful tool to deal with robust stabilization and disturbance
attenuation problem. However the standard H control theory does not guarantee robust
performance under the presence of all uncertainties in the plant. This is specially true
in power systems where the plant parameters may change considerably with variations in
operating conditions.
There has been some effort in uncertainty modelling and treatment of plant with structured uncertainties. The problem is to find a controller such that infinity norm for the closed
loop system is satisfied for all the uncertainties in a given bounded set [55]. In the context
of power system the system uncertainties have to be identified, modelled and bounded, before guaranteed robust stabilizer can be designed. H optimal control design minimizes
the worst case energy gain(H norm) of certain suitably weighted closed loop transfer matrix. With properly selected weighting functions, the controllers have good performance in
case of uncertainties in plant modelling and/or disturbances; moreover, trade-offs between
performance and robustness can be studied in this framework. Chen and Malik [43] have
developed a PSS based on H optimization method with an uncertainty description which
represents the possible perturbation of a synchronous generator around its normal operating
point.
Ashgharian [44] applies H theory to guarantee non degradation of torsional phenomenon
by considering the high frequency unmodelled dynamics in the system. Ohtsuka et.al. [41]
apply H optimization theory to improve the disturbance attenuation performance of LQ
optimal controllers. The changes in the operating conditions are not considered and there is
no explicit uncertainty modelling.
Chen and Malik [50] have applied synthesis for PSS design. The uncertainty in the
system is modelled in terms of variations in the values of the parameters K1 to K6 in the
Heffron-Philips model of a single machine infinite bus system. Bounds on these variations
are found and a controller is synthesized using the D-K iteration technique [56].
Almost all the above references are concerned only with robust stability of the closed loop
system which criterion is not sufficient for power system applications [45, 47]. Some of them
include disturbance attenuation specifications. Such specifications are not very relevant in
this application and are introduced to fit the problem to existing theory which has been
developed primarily for applications other than power system control.
Gibbard [57] suggests a PSS tuning method which is shown to be robust through an
example. The argument in this paper depends strongly on the invariance of the P-Vr charac-
teristics of the generators in spite of the variations in the operating conditions. Fatehi et.al.
[58] have applied loop transfer recovery to obtain a robust controller for power systems.
Khammash et.al. [59, 60] have used a non-negative matrices test for checking robust
stability in the presence of variations in the elements of the system matrices. Pai et.al. [61]
apply a Hurwitzness test for interval matrices to check the robust stability of power systems
in the presence of parameter variations. Werner et.al. [62] use LMI techniques for robust
PSS design. These papers deal primarily with robustness analysis of power systems.
Rao and Sen [63-65], have proposed a method based on quantitative feedback theory
(QFT) for designing a robust controller for a power system in a single input single output
(SISO) framework. These authors extended their work to multi-variable case also [66].
The increasing presence of FACTS devices in power systems now provides an alternative
control loop for further improving the stability of the system. It is known that well designed
controller of any FACTS device can enhance the system damping [67, 68]. The simultaneous application of PSSs and FACTS devices can be used to further enhance the small
signal dynamic performance of a power system. However, the distributed nature of power
systems requires the application of a decentralized control strategy wherein only locally measurable signals are used for feedback at the various control inputs to the system. A robust
decentralized damping controller has been proposed by Rao and Sen [66].
Robust controllers are less sensitive to changes in operating conditions than conventional
controllers. They provide adequate damping over a wide operating range of power system.
There have also been a few other miscellaneous publications dealing with robustness issues
in power systems which are relevant to the present work.
1.6
Application of Genetic Algorithms to PSS Design
Genetic algorithm has recently attracted the attention of Power System Stabilizer designers [69-75]. The advantage of GA technique is that it is independent of the complexity of
the performance index considered. It suffices to specify the objective function and to place
finite bounds on the optimized parameters. GA provides greater flexibility regarding controller structure and objective function considered. Further more, GA based optimization
10
problem can readily accomplish control performance constraints, such as required closed
loop minimum performance. Introduction of GA helps to obtain an optimal tuning for all
PSS parameters simultaneously, which thereby takes care of interaction between different
PSSs, hence eigen value drift problem associated with sequential tuning methods can be
eliminated.
Several techniques of tuning of PSS using genetic algorithms have been reported in recent literature. Magid and Abido [70] have applied GA to tune the hybridizing rule based
PSS. Advantage of rule based PSS is its robustness, less computational burden and ease of
realization.
Taranto et.al. [72] have presented a method for simultaneous tuning of damping controllers
using modified GA operators. Tuning of fixed structure conventional PSS is reported in this
paper.
Zhang and Coonick [73] have proposed a new method based on the method of inequalities applied to GA for the coordinated synthesis of PSS parameters in multimachine power
systems.
Andreoiu and Kankar Bhattacharya [74] have proposed Lyapunovs method based genetic
algorithm for robust PSS design.
Robust stability of closed loop system can be achieved using genetic algorithms. AbdelMagid et.al. [75] apply genetic algorithms to tune the parameters of a PSS such that robust
stability is achieved over a range of operating conditions. Taranto et.al. [76] have applied
parameter optimization using genetic algorithms for synthesizing a robust controller for
power systems. These papers focus upon the robust closed loop rotor mode location as is
the case in this thesis.
1.7
Robust PSS design using Genetic Algorithms: the

present approach
In this thesis a new method has been proposed for tuning of PSS using genetic algorithm.
Proposed method guarantees a robust performance over a set of operating conditions. A
more elegant approach to robust stabilizer design is used, in which fixed gain robust PSSs
11
have been designed to guarantee a minimum performance inspite of variations in the plant
operations, due to changes in load, line switching, transformer tap-changing and other occasional disturbances. Based on system experience minimum performance requirements of
PSS have been decided and an attempt has been made to achieve it over a wide range of
operating conditions. The performance requirements of the PSSs are more fully described
in the next section.
In the present approach the power system operating at various loading is treated as a
finite set of plants. The problem of selecting the parameters of PSS which simultaneously
stabilize this set of plants is converted to a simple optimization problem which is solved by
genetic algorithm and an eigen value based objective function.
1.8
Performance Requirements of Power System Damping Controllers
There exists considerable ambiguity in current literature about the performance requirements of stabilizers and other damping controllers. This section attempts to establish, in
clear and precise term, the closed loop specifications required of any power system damping
controller.
Practical considerations merely require that the troublesome low frequency oscillations,
when excited, die down within a reasonable amount of time. No advantage is gained by
having excessive damping for these system modes. In fact, it has been noted [6] that aggressive damping of oscillations can have detrimental effects on the system. Hence, rather than
maximizing the damping at some particular operating condition, it seems more appropriate
to decide upon the minimum amount of damping or minimum performance required of the
closed loop and attempt to achieve this over the entire range of operating conditions which
the system experiences. This set of operating conditions, which any given power system
might experience, is always known a priori in terms of maximum and minimum values of
power generations, transmissions and loads and all possible values of the network impedances.
It is therefore possible to model this bounded variation in the system as an uncertainty and
attempt to synthesize a PSS delivering the required performance over this entire range of
variation.
1.8.1
12
How Much Damping Do We Need?
A damping factor of around 10% to 20% for the troublesome low frequency electromechanical mode is considered adequate. For a second order system =10% results in system
oscillations decaying to within 15% of the initial amplitude in 3 cycles of the oscillations. (for
=20%, the decay is to within 2.1% of initial amplitude in 3 cycles.) A damping factor of
10% would be acceptable to most utilities and can be adopted as the minimum requirement.
Further, having the real part of rotor mode eigen value restricted to be less than a value,
say , guarantees a minimum decay rate . A value = - 0.5 is considered adequate for an
acceptable settling time. The closed loop rotor mode location should simultaneously satisfy
these two constraints for an acceptable small disturbance response of the controlled system.
The frequency of oscillation is related to synchronizing torque and hence the imaginary
part of the rotor mode eigen value should not change appreciably due to feed back.
If any new modes arise as a result of closing the controller loop (e.g. exciter mode), these
should also be well damped i.e. they should satisfy the same constraints on the real part
and damping factor as the rotor mode. Real poles close to the origin can result in a sluggish
response and persistent deviations of the system variables from their steady state values and
hence should be avoided.
20
15
=10%
10
imag
= 0.5
0
5
10
15
20
5
2
real
Figure 1.1: D-contour
13
If all the closed loop poles are located to the left of the contour shown in Figure 1.1,
then the constraints on the damping factor and the real part of rotor mode eigen values
are satisfied and a well damped small disturbance response is guaranteed. This contour is
referred as the D-contour [63]. The system is said to be D-stable if it is stable with respect
to this D-contour, i.e. all its pole lie on the left of this contour. This property is referred to
as generalized stability in control literature. This generates a neat specification- the closed
loop should be robustly D-stable i.e. D-stable for the entire range of operating and system
conditions. Hence, in this thesis a system is said to be robust, if, inspite of changes in
system and operating conditions, the closed loop poles remain on the left of the D-contour
for specified range of system and operating conditions.
1.8.2
Performance Evaluation of a PSS
Many of the design methods suggested in literature have been accompanied by comparisons
between different types of stabilizers. Such comparisons usually consider the amount of
damping enhancement provided by each PSS. It is clear from the discussion in the previous
section that a more aggressive damping is not particularly beneficial. In fact, in view of
the other considerations, it would be more fruitful to have the rotor mode damping closer
to the minimum requirements. Thus a comparison of two different stabilizers on grounds
of the amount of damping they contribute at some particular operating condition is not
very appropriate. A better PSS would be one which guarantees the minimum acceptable
performance over a wider range without adversely affecting the large disturbance response
of the system.
1.9
Scope of Present Work
The objective of the present work is to show that even a properly tuned fixed parameter
controller can guarantee a robust minimum performance over a wide range of operating
conditions, if it is properly tuned. Since fixed parameter PSS is simple in structure and
widely used by most utilities, an attempt is made to tune the fixed parameter PSS to ensure
its robustness.
A new method has been proposed for robust PSS design, which includes several operating conditions and system configurations simultaneously in the design process and works
well with equal effectiveness in single and multimachine environments. PSS parameters are
obtained using genetic algorithm.
14
A simple objective function based on eigen values is formulated for robust PSS design in
which robust D-stability of the closed loop is taken as primary specification.
The efficacy of the proposed PSS in damping out low frequency oscillations have been
established by extensive simulation studies on single and multimachine systems. The details
of the proposed method are given in Chapter 4.
1.10
Organization of Chapters
The thesis chapters are organized as under,

Chapter 1
This chapter introduces the problem of low frequency oscillations and defines the closed loop
performance requirements for power system damping controller.
Chapter 2
In this chapter mathematical models of power system have been developed. Non-linear differential equations required for more accurate simulation of single machine infinite bus system
and multimachine system are given.
Chapter 3
Chapter 3 reviews the basic ideas of genetic algorithms , genetic operators and mathematical
model of simple genetic algorithm which are needed to support controller design tuning of
power system stabilizer.
Chapter 4
This chapter deals with the formulation of objective function based on D-contour and minimum performance requirement criterion. A new method is proposed and robustness of the
PSS is tested on the single machine infinite bus system.
Chapter 5
This chapter illustrates the application of proposed method to multimachine power system.
The performance of the stabilizer has been promising over a range of system and test conditions. Due to its simple structure and ease of design, the proposed stabilizer appears to be
well suited for application in real plants.
15
Chapter 6
This concluding chapter gives a brief summary of the work done and also includes a section
on the scope of future work relating to design of power system stabilizers.
Chapter 2
Mathematical Modelling of Power
System
2.1
Introduction
For stability assessment of power system adequate mathematical models describing the
system are needed. The models must be computationally efficient and be able to represent
the essential dynamics of the power system.
A realistic power system is seldom at steady-state, as it is continuously acted upon by
disturbances which are stochastic in nature. The disturbances could be a large disturbance
such as tripping of generator unit, sudden major load change and fault switching of transmission line etc. The system behavior following such a disturbance is critically dependent
upon the magnitude, nature and the location of fault and to a certain extent on the system
operating conditions. The stability analysis of the system under such conditions, normally
termed as Transient-stability analysis is generally attempted using mathematical models
involving a set of non-linear differential equations.
In contrast to this disturbance-specific transient instability, there exists another class of
instability called the Dynamic Instability or more precisely Small Oscillation Instability,
described in Chapter 1. As the small oscillation stability concerns itself with small excursions
of the system about a quiescent operating point, the system can be sometimes approximated
by a linearized model about the particular operating point. Once valid linearized model is
available, powerful and well established techniques of the linear control theory can be applied
for stability analysis and performance evaluation of various power system stabilizers.
16
Chapter 2. Mathematical Modelling of Power System
17
Nonlinear models on the other hand have more realistic representation of the power systems. Designing controllers for such nonlinear systems are understandably more difficult.
In this chapter, non-linear models of single and multimachine power systems have been
developed. Linear models have been obtained from these nonlinear models for designing
conventional power system stabilizers that are used for comparative performance analysis.
2.2
SMIB Model in Non-Linear Form
Consider the system shown in Figure 2.1. This shows the external network with two ports.
One port is connected to the generator terminals while the second port is connected to a
voltage source Eb 6 0. Assuming both the magnitude Eb and phase angle of the voltage source
to be constant, and neglecting the network transient, the system can be modelled using rotor
mechanical equations, rotor electrical equations and excitation system model.
.
^
Ia
External
^
Vt
Two Port
+
Eb
Network
Figure 2.1: External two port network
2.2.1
Rotor Equations
Rotor Mechanical Equations

The mechanical equations in per unit can be expressed as
M
where, M =
2H
,
B
d2
d
+ D 0 = Tm Te
2
dt
dt
(2.1)
and H, D0 , , Tm and Te are inertia constant, rotor damping, rotor angle,
mechanical and electrical torques respectively. The above equation can be expressed as two
18
first order differential equations as:

d
= B (Sm Smo ) = o
dt
dSm
= D(Sm Smo ) + Tm Te
dt
where, per unit damping D and generator slip Sm are given by:
2H
D = B D 0
B
Sm =
B
(2.2)
(2.3)
(2.4)
(2.5)
o and B are the synchronous and the base speed of the system.
Rotor Electrical Equations
Since the stator equations 2.12 and 2.13, stated later, are algebraic (neglecting stator transients) and rotor windings either remain closed (damper windings) or closed through finite
voltage source (field winding), the flux linkages of these windings cannot change suddenly.
Hence, it is not possible to choose stator currents id and iq as state variables (state variables
have to be continuous functions of time). The obvious choice for state variables are rotor
flux linkages or transformed variables which are linearly dependent on the rotor flux linkages
(Chapter 6 of [84]).
In a report published in 1986 by an IEEE Task Force [85], many machine models are
suggested based on varying degrees of complexity. Higher order models of machine in general
provide better results but it is adequate to use model (1.1) if the data is correctly determined.
In case studies cited in this report, only 1.1 model has been considered where two electrical
circuits are considered on the rotor i.e. a field winding on the d-axis and one damper winding
on q-axis. Differential equations for rotor and the electrical torque and are:
i
dEq0
1 h
0
0
)i
+
E
+
(x
x
=
E
fd
d
d d
q
0
dt
Tdo
(2.6)
i
dEd0
1 h
0
0
E
(x
x
)i
=
q
d
q q
0
dt
Tqo
(2.7)
Te = d iq q id
= Ed0 id + Eq0 iq + (x0d x0q )id iq
(2.8)
(2.9)
19
where, vd , vq =d-q components of generator terminal voltage

id , iq =d-q components of armature current
Ef d =voltage proportional to field voltage
Ed0 =voltage proportional to damper winding flux
Eq0 =voltage proportional to field flux
0
Tdo
=d-axis transient time constant
0
=q-axis transient time constant.
Tqo
2.2.2
Stator Equations
The stator equations in Parks reference frame are expressed in per unit, these are
1 d
q Ra id = vd
B dt
B
(2.10)
1 q
d Ra iq = vq
B dt
B
(2.11)
It is assumed that the zero sequence in the stator are absent. If stator transients are to
be ignored, it is equivalent to ignoring the the pd and pq terms in above equations. In
addition it is also advantageous to ignore the variations in the rotor speed . If the armature
flux linkage components (pD and pQ ), with respect to a synchronously rotating frame, are
(rotating at speed o ) constants, then transformer e.m.f. terms (pd and pq ) and terms
induced by the variations in the rotor speed cancel each other (chapter 6 of [84]). Then the
above equations 2.10 and 2.11 are reduced to
(1 + Smo )q Ra id = vd
(2.12)
(1 + Smo )d Ra iq = vq
(2.13)
where, Smo is the initial operating slip, which, in most of the cases is assumed to be zero.
For the 1.1 model of the generator (field circuit with one equivalent damper winding on the
q-axis) the flux linkages are given by:
d = x0d id + Eq0
(2.14)
q = x0q iq Ed0
(2.15)
20
Neglecting stator transients and letting Smo = 0, and substituting equation 2.15 in 2.12 and
equation 2.14 in 2.13, we get:
2.2.3
vd = Ed0 x0q iq Ra id
(2.16)
vq = Eq0 + x0d id Ra iq
(2.17)
Network Equations
It is assumed that the external network connecting the generator terminals to the infinite
bus is linear two port. The loads are assumed to be of constant impedance type. The voltage
there can be expressed as:
Vt = h11 Ia + h12 Eb = VQ + jVD
h11 = zR + jzI ,
(2.18)
h12 = h1 + jh2
(2.19)
where, h11 is the short circuit self impedance of the network, measured from the generator
terminals, and h12 is a hybrid parameter (open circuit voltage gain). Equation 2.18 is
multiplied with ej which can be expressed as:
where, Eb0 =
(vq + jvd ) = (zR + jzI )(iq + jid ) + Eb0 ej(h )
(2.20)
(h21 + h22 )Eb , and tan h = h2 /h1 .
Equating real and imaginary parts of equation 2.20 separately, we can get:
zR
zI
zI zR
id
iq
vd
vq
+ E0
b
sin( h )
cos( h )
(2.21)
From Equation 2.21 we can get d-q component of stator currents. By using all the equations
in Section 2.2 model (1.1) can be simulated.
2.3
Excitation System Model
The excitation system is represented by a first order model. Let Ka and Ta be the AVR gain
and its time constants respectively. The block diagram of AVR is shown in figure 2.2 and
the equation describing it can be written as:
1
dEf d
= [Ka (Vref + Vs Vt ) Ef d ]
dt
Ta
Ef dmin Ef d Ef dmax
(2.22)
(2.23)
21
VS
max
E fd
Ka
Vref
E
fd
1 + sTa
min
E
fd
Vt
Figure 2.2: Excitation system block diagram.
2.4
PSS Model
For the simplicity a conventional PSS is modelled by two stage (identical), lead/lag network
which is represented by a gain KS and two time constants T1 and T2 . This network is
connected with a washout circuit of a time constant Tw , as shown Figure 2.3.
VSmax
Sm
sTw
1 + sTw
Ks
1 + sT1 2
VS
1 + sT2
VSmin
Figure 2.3: Block diagram of PSS
2.5
SMIB Test System
For the SMIB test system, the synchronous machine is assumed to be connected to an
infinite bus of voltage Eb through a transmission line of impedance Ze = jXe , as shown in
Figure 2.4. Since Re = 0 for this system hence ZR = 0.0, Zi = Xe , h1 = 1.0, h2 = 0.0,
h = 0.0.
22
P,Q
Xe
Eb
Efd
Vt
AVR
Infinite bus
Control Input
Figure 2.4: Single machine infinite bus system

Considering Ra = 0, the dynamic equations of the SMIB system considered can be summarized as :
d
= B Sm
dt
(2.24)
dSm
1
=
[DSm + Tm Te ]
dt
2H
(2.25)
i
dEd0
1 h
0
0
=
E
(x
x
)i
q
q
d
q
0
dt
Tqo
(2.26)
i
dEq0
1 h
0
0
=
E
+
(x
x
)i
+
E
d
d
f
d
q
d
0
dt
Tdo
(2.27)
dEf d
1
=
[Ka (Vref + Vs Vt ) Ef d ]
dt
Ta
vd
vq
id
iq
Ed0
Eq0
0 Xe
Xe
0 x0q
x0d
vd
vq
id
(2.28)
(2.29)
iq
Eb0
sin
cos
(2.30)
23
Te = Ed0 id + Eq0 iq + (x0d x0q )id iq
2.6
(2.31)
Modelling of Multimachine System
Figure 2.5 shows the schematic of a multimachine system. This section describes the
dynamic equations represented by each block shown in the ith machine and external network.
It is assumed that power system consists of n number of generators and generators feed local
loads which are constant.
Loads
^
Ii
^
Ij
^
Vi
^ ^ ^
I=Y V
N
To Other Machines
^
Vj
V
ref, i
AVR
E fdi
V ,V
Di Qi
Vdi , Vqi
Interface
Di
, I
Qi
Machine
(Electrical)
I di , I qi
Machine
i
Figure 2.5: Schematic of a multimachine system
In multimachine system without infinite bus, it is necessary to take a reference angle to

compare all other rotor angles of generators. Conventionally the rotor angle of machine
24
having highest inertia is taken as a reference. Another reference which is also considered
very often is the center of inertia (COI) angle and speed deviation 0 and 0 and these are
defined as:
where, MT =
COI =
n
1 X
Mi i
MT i=1
(2.32)
0 =
n
1 X
Mi i
MT i=1
(2.33)
Mi is total inertia of n number of generators. In the case study the rotor
angle and slip of the machine having highest inertia are taken as reference.
2.6.1
Rotor Equations
Rotor Mechanical Equations

The mechanical equations for multimachine in per unit can be expressed as:
2H
di
= B (Smi Smio ) = i io
dt
(2.34)
dSmi
= Di (Smi Smio ) + Tmi Tei
dt
(2.35)
where, H, Tmi and Tei are machine inertia, mechanical and electrical torque respectively of
ith machine. Per unit damping (Di ), generator slip (Smi ), and electrical torque (Tei ) are
given by:
Di = B Di0
i B
Smi =
B
0
0
Tei = Edi idi + Eqi
iqi + (x0di x0qi )idi iqi
(2.36)
(2.37)
(2.38)
In matrix form we can rewrite the equations 2.34 and 2.35 as:
d[]
= B [Sm ] = [] [o ]
dt
2[H]
d[Sm ]
= {[D][Sm ] + [Tm ] [Te ]}
dt
(2.39)
(2.40)
25
where,
[H] = diag
[D] = diag
h
h
H1 H2 ... Hk ... Hn
Sm1 Sm2 ... Smk ... Smn

h
[Tm ] =
Tm1 Tm2 ... Tmk ... Tmn

h
[Te ] =
[] =
1 2 ... k ... n
(2.43)
it
(2.44)
(2.45)
it
1 2 ... k ... n
h
it
it
Te1 Te2 ... Tek ... Ten

h
[] =
(2.42)
D1 D2 ... Dk ... Dn
[Sm ] =
(2.41)
(2.46)
it
(2.47)
Rotor Electrical Equations

For 1.1 model, differential equations for the rotor flux linkages and voltages for rotor windings
for multimachine can be written as:
0
i
dEqi
1 h
0
0
=
E
+
(x
x
)i
+
E
di
f di
qi
di di
0
dt
Tdoi
0
i
1 h
dEdi
0
0
=
E
(x
x
)i
qi
di
qi qi
0
dt
Tqoi
(2.48)
(2.49)
Above equations in matrix form are,

0
[Tdo
]
n
o
d[Eq0 ]
= [Eq0 ] + ([xd ] [x0d ])[id ] + [Ef d ]
dt
(2.50)
0
[Tqo
]
n
o
d[Ed0 ]
= [Ed0 ] ([xq ] [x0q ])[iq ]
dt
(2.51)
where,
0
[Tdo
] = diag
0
[Tqo
] = diag
[Ef d ] =
[Ed0 ] =
[Eq0 ] =
[id ] =
[iq ] =
h
h
Tdo1 Tdo2 ... Tdok ... Tdon

Tqo1 Tqo2 ... Tqok ... Tqon
Ef d1 Ef d2 ... Ef dk ... Ef dn
h
h
h
0
0
0
0
... Edn
... Edk
Ed2
Ed1
0
0
0
0
... Eqn
... Eqk
Eq2
Eq1
id1 id2 ... idk ... idn

h
iq1 iq2 ... iqk ... iqn
it
it
it
it
it
(2.52)
(2.53)
(2.54)
(2.55)
(2.56)
(2.57)
(2.58)
26
[xd ], [xq ], [x0d ], [x0q ] and [Ra ] are diagonal matrices of same size, and one of them is shown as
below
[Ra ] = diag
2.6.2
Ra1 Ra2 ... Rak ... Ran
(2.59)
Stator Equations
Stator equations are expressed in per unit with assumption of neglecting zero sequence and
stator transients, as in section 2.2.2, we have the equations:
(1 + Smio )qi Rai idi = vdi
(2.60)
(1 + Smio )di Rai iqi = vqi
(2.61)
where, subscript i stands for ith machine; Smo is the initial operating slip, which, in most of
the cases is assumed to be zero and is defined as:
io B
Smio =
(2.62)
B
Neglecting stator transients and letting Smo = 0, equations 2.16 and 2.17 are rewritten for
multimachine as:
0
vdi = Edi
x0qi iqi Rai idi
(2.63)
0
vqi = Eqi
+ x0di idi Rai iqi
(2.64)
The above two equations can be represented in matrix form as:
[vd ]
[vq ]
[Ed0 ]
[Eq0 ]
[Ra ]
[x0q ]
[x0d ] [Ra ]
[id ]
[iq ]
(2.65)
where,
[vd ] =
[vq ] =
2.6.3
vd1 vd2 ... vdk ... vdn

h
vq1 vq2 ... vqk ... vqn
it
it
(2.66)
(2.67)
Inclusion of Generator Stator in the Network
The generator equivalent circuit can be drawn as in the Figure 2.6. It can be represented
in terms of a current source Ig and its internal admittance Yg such that armature current,
Ia = Ig Yg Vt . The equivalent circuit shown in the figure can easily be merged with the AC
network external to the generator.
27
Ia
Ig
Yg
Vt
Figure 2.6: Generator equivalent circuit.

Treatment of Transient Saliency
When transient saliency is neglected then the stator can be represented by a voltage source
(Eq0 + jEd0 ) behind an equivalent reactance (Ra + jx0 ). But if transient saliency is considered
then a stator cannot be represented by a single phase equivalent circuit. A generator can
be represented by a dependent current source, which is a function of the field and damper
winding flux and , to treat saliency.
The generator stator voltage can be re-expressed as a single equation in phasor quantities:
0
Vt = (vq + jvd )ej = [Eq0 + j(Ed0 + Edc
)]ej (Ra + jx0d )Ia
where,
0
Edc
= (x0d x0q )iq
(2.68)
(2.69)
Equation 2.68 can be rearranged to represent the equivalent circuit of Figure 2.6 as:
where,
Ig = Yg Vt + Ia
(2.70)
0
Ig = Yg [Eq0 + j(Ed0 + Edc
)]ej
1
Yg =
Ra + jx0d
(2.71)
(2.72)
This requires an iterative solution for the dependent current source and this problem of
iterative solution can be eliminated by considering a rotor dummy coil on q-axis which links
0
as a state variable. The differential
only with q-axis coil in the armature and considering Edc
0
can be expressed as:
equation for Edc
0
i
dEdc
1 h 0
0
=
(xd x0q )iq Edc
dt
Tc
(2.73)
28
where, Tc is the open circuit time constant of the dummy coil, which can be arbitrarily
selected. Tc should be small and it can be 0.01 sec for acceptable accuracy. This is of a
similar order as the time constant of high resistance damper winding.
2.6.4
Load Representation
Loads are represented as static voltage dependent models given by

PL = fP (VL ) = a0 + a1 VL + a2 VL2
(2.74)
QL = fQ (VL ) = b0 + b1 VL + b2 VL2
(2.75)
If load is represented by constant impedances then a0 = a1 = b0 = b1 = 0, and Yl is given by

Yl =
PLo jQLo
2
VLo
(2.76)
where subscript o indicates operating values.
2.6.5
Network Equations for Multimachine
The AC network consists of transmission lines, transformers, shunt reactors, capacitors in

series and shunt. It is assumed that the network is symmetric. Hence single phase representation (positive sequence network) is adequate. The network equations can be expressed
using bus admittance matrix YN as
IN = [YN ]V
(2.77)
where, V is a vector of complex bus voltages and IN is vector of current injections. The
generator and load equivalent circuits at all the buses can be integrated into the AC network
and the overall system algebraic equations can be obtained as follows:
I = [Y ]V
(2.78)
where [Y] is the complex admittance matrix which is obtained from augmenting [YN ] by inclusion of the shunt admittance Yg (from generator equivalent circuit) and Yl at the generator
and load buses. Element Ygj or Ylj corresponding to j th bus is added to YN jj element of the
admittance matrix YN to obtain [Y]. I is the vector of complex current sources. Equations
29
2.78 can be rewritten as:

V = [Y ]1 I = [Z]I
(2.79)
[VQ + jVD ] = [ZR + jZI ][IQ + jID ]
VD
VQ
ZR
ZI
ID
Load C
0.0085+j0.072
0.0119+j0.1008
B/2=j0.0745
B/2=0.1045
Load A
(2.80)
j0.0586 13.8 kV
G3
B/2=j0.179
B/2=j0.153
0.039+j0.170
230/13.8 3
9
j0.0576
16.5/230
230 kV
B/2=j0.079
B/2=j0.088
0.010+j0.085 0.032+j0.161
2 18/230
0.017+j0.092
G2
IQ
ZI ZR
18 kV j0.0625
Load B
4
16.5 kV
G1
Figure 2.7: 3 machine, 9 bus power system model, single line diagram.
2.7
Multimachine Test System
The multimachine configuration considered for the purpose of study consists of 3 generators
[86, 87] interlinked as shown in Figure 2.7.
30
Substituting n = 3 in the equations developed in Section 2.6, the dynamic equations

representing this system can be summarized as :
3
1 X
Mi i
MT i=1
(2.81)
3
1 X
Mi i
=
MT i=1
(2.82)
COI =
COI
d[]
= B [Sm ] = [] [o ]
dt
(2.83)
d[Sm ]
= {[D][Sm ] + [Tm ] [Te ]}
dt
(2.84)
0
[Tdo
]
n
o
d[Eq0 ]
= [Eq0 ] + ([xd ] [x0d ])[id ] + [Ef d ]
dt
(2.85)
0
[Tqo
]
n
o
d[Ed0 ]
= [Ed0 ] ([xq ] [x0q ])[iq ]
dt
(2.86)
0
n
o
dEdc
0
= ([x0d ] [x0q ])[iq ] [Edc
]
dt
(2.87)
2[H]
[Tc0 ]
[Ta ]
d[Ef d ]
= [[Ka ]([Vref ] + [Vs ] [Vt ]) [Ef d ]]
dt
[id ]
[iq ]
VD
VQ
[Ra ]
[x0q ]
[x0d ] [Ra ]
ZR
ZI
ZI ZR
ID
[Ed0 ] [vd ]
[Eq0 ] [vq ]
(2.88)
(2.89)
IQ
0
0
iqi + (x0di x0qi )idi iqi
idi + Eqi
Tei = Edi
0
Ig = Yg [Eq0 + j(Ed0 + Edc
)]ej
1
Yg =
Ra + jx0d
(2.90)
(2.91)
(2.92)
(2.93)
where, [ZR + jZI ] = [Z] = [Y ]1 and [Y] is the complex admittance matrix which is obtained
31
from augmenting bus admittance matrix YN by shunt admittance Yg of generator and load
admittances at the generator and load buses Yl .
2.8
Linearized 1.1 Model
Linearized 1.1 model for both, single machine and multimachine system was obtained using LINMOD facility available in MATLAB. Details of this model are given in ref. [84].
These models have been used in later chapter for simulating power systems equipped with
conventional and proposed PSSs to analyze the performance of the controllers at various
system and operating conditions.
Chapter 3
Genetic Algorithm: An Overview
3.1
Introduction
In the open access environment, the power utilities are often forced to work their system
far away from predesigned conditions. In this situation, the systems may be operating near
their stability limits. It is therefore necessary to re-approach the problem related to power
system stability with this perspective. Several recent major system blackouts in different
countries and voltage collapses have clearly indicated the need for better stabilization efforts
in the interconnected power systems. Conventional power system stabilizers are designed for
particular system and operating conditions and are therefore not effective throughout the
expected range of operation of such systems.
In contrast, application of Genetic Algorithms (GA) in power system stabilizer design
is an attractive proposition as it provides greater flexibility regarding controller structure
and objective function. In addition to the constraints on the parameter bounds, the GA
based optimization problem can readily accomplish control performance constraints, such
as required closed-loop minimum performance. Further more, GA helps to obtain an optimal tuning for all PSS parameters simultaneously, which takes care of interactions between
different PSSs.
This chapter gives a brief and quick introduction to Genetic Algorithm. This is needed for
a better understanding of the GA based stabilizer design process dealt in the later chapters.
32
Chapter 3. Genetic Algorithm: An Overview
3.2
33
What is Genetic Algorithm?
Genetic Algorithms are adaptive methods which may be used to solve search and optimization problems. Over many generations, natural populations evolve according to the
principles of natural selection and survival of the fittest. By mimicking the process, genetic
algorithms are able to evolve solutions to real world problems, if they have been suitably
encoded.
3.3
Working Principles
The basic principles of GAs were first laid down rigourously by Holland [77], in mid sixties.
Thereafter, many researchers have contributed to developing this field. To date, most of the
GA studies are available through a few texts [78-81]. There are many variations of the
genetic algorithm but the basic form is the simple genetic algorithm. The working principle
[82] of SGA can be described as:
3.3.1
Coding
Before a GA can run, a suitable coding for the problem must be devised. It is assumed
that a potential solution to a problem may be represented as a set of parameters. These
parameters (known as genes) are joined together to form a string of values (often referred
as chromosome or Individual). Binary coded strings having 1s and 0s are mostly used.
For example, if 10 bits are used to code each variable in a two-variable function optimization
problem, chromosome would contain two genes, and consists of 20 binary digits. Decoding
technique of binary coded strings in to function variables is given in Appendix E.
3.3.2
Fitness Function
As pointed out earlier, GAs mimic the survival of the fittest principle of nature to make
a search process. Therefore, GAs are naturally suitable for solving maximization problems.
Minimization problems are usually transformed in to maximization problems by suitable
transformation. In, general, a f itness function is first derived from the objective function
and used in successive genetic operations. Certain genetic operators require that the fitness
function be nonnegative, although certain operators do not have this requirement. For
34
maximization problems, the fitness function can be considered to be the same as the objective
function. For minimization problems, the fitness function is an equivalent maximization
problem chosen such that the optimum point remains unchanged.
3.3.3
GA Operators
The GA works with a set of individuals comprising the population. The initial population consists of N randomly generated individuals where, N is the size of population. At
every iteration of the algorithm, the fitness of each individual in the current population is
computed. The population is then transformed in stages to yield a new current population
for the next iteration. The transformation is usually done in three stages by sequentially
applying the following genetic operators:
(1) Selection : In the first stage, the selection operator is applied as many times as there
are individuals in the population. In this stage every individual is replicated with a
probability proportional to its relative fitness in the population. The population of N
replicated individuals replaces the original population.
(2) Crossover: In the next stage, the crossover operator is applied with a probability pc ,
independent of the individuals to which it is applied. Two individuals (parents) are
chosen and combined to produce two new individuals (offsprings). The combination is
done by choosing at random a cutting point at which each of the parents is divided into
two parts; these are exchanged to form the two offsprings which replace their parents
in the population. This is known as single point crossover. Figure 3.1 illustrates the
single point crossover operation.
(3) Mutation : In the final stage, the mutation operator changes the values in a randomly
chosen location on an individual with a probability pm . Figure 3.2 shows the mutation
operation.
3.3.4
Convergence
If the GA has been correctly implemented, the population will evolve over successive
generations so that the fitness of the best and the average individual in each generation
increases towards the global optimum. The algorithm converges after a fixed number of
iterations and the best individual generated during the run is taken as the solution.
35
Crossover Point
Parent 1
1 0 1 0 0 1 0 1 1 0
Parent 2
1 0 1 1 1 0 1 0 1 0
Crossover
Offspring 1
1 0 1 0 0 1 1 0 1 0
Offspring 2
1 0 1 1 1 0 0 1 1 0
Figure 3.1: Single point crossover operation
Offspring
1 0 1 0 0 1 1 0 1 0 .
Mutation
Mutated
Offspring
1 0 1 0 0 1 0 0 1 0 .
Figure 3.2: A single mutation operation
3.4
36
Implementation of genetic algorithm
The implementation of the simple genetic algorithm is as follows:

Input:
l: length of each solution string
N : population size, number of strings in a population
pc : probability of crossover
pm : probability of mutation
MAXGEN: maximum number of generations
output:
x : best string from the current population
Algorithm:
1. Generate N strings, each of length l, randomly to form the initial population.

2. Evalute each string in the current population and assign a fitness value to each
string.
3. Select a highly fit string using selection operator and repeat this process N times
to generate a new population of N strings for next generation.
4. Randomly choose pairs of these selected strings and perform crossover with a
probability pc to generate children strings. Crossover exchanges bit values between
the two strings at one or more locations.
5. Randomly choose some bit positions with a probability pm and mutate the bit
values. That is change 1 to a 0 and 0 to a 1.
6. Steps 2-5 constitute a generation. Repeat steps 2-5 till the number of generations
is MAXGEN and stop. Output the best string from the current population.
Figure 3.3 shows the general structure of the genetic algorithms.
37
Xover
1100101010
1011101110
Chromosomes
encoding 1100101010
1011101110
Solutions
0011011001
1100110001
1100101110
1011101010
Mutation
0011011001
0011001001
New
population
Selection
Evaluation
Offspring
1100101110
1011101010
Roulette
wheel
0011001001
Decoding
Solutions
Fitness
computation
Figure 3.3: The general structure of genetic algorithms
3.5
38
Mathematical Model of SGAs
This section describes the exact mathematical model [81] of simple genetic algorithms.
This model is based on above mentioned algorithm, with only difference that only one
offspring from each crossover survives.
If we define vectors
p (t) and
s (t), each of length 2l , where vector
p (t) exactly specifies
the composition of the population at generation t, and s (t) reflects the selection probabilities
under the fitness function, then these are connected via fitness.
Let F be a two-dimensional matrix such that Fi,j = 0 for i 6= j and Fi,i = f (i). Diagonal
elements (i, i) of F which are nonzero give the fitness of the corresponding string i. Under
proportional selection,
F
p (t)
s (t) = P2l 1
j=0 Fjj pj (t)
(3.1)
Where, pj (t) is the j t h component of vector

p (t). Thus, given
p (t) and F ,
s (t) can be
easily found, and vice versa.
Now , define a single operator G such that applying G to

s (t) will exactly mimic the
expected effects of running the GA on the population at generation t to t + 1:
s (t + 1) = G
s (t)
(3.2)
Then iterating G on
p (0) will give an exact description of the expected behaviour of the
GA.
Let GA be operating with selection alone (no crossover or mutation). Let E(x) denote the
expectation of x. Then, since si (t) (ith component of vector

s (t)), is the probability that i
will be selected at each selection step,
E(
p (t + 1)) =
s (t)
(3.3)
Let
x
y denotes the scalar difference between
x and
y , i.e.
x = k
y , where, k is a
scalar. Then, from Equation 3.1, we have
39
s (t + 1) F
p (t + 1)
which implies
E(
s (t + 1)) F
s (t)
This is the type of relation of the form in Equation 3.2, with G = F for this case of selection
alone.
Crossover and mutation can be included in the model by defining G as the composition of
the fitness matrix F and a recombination operator M that mimics the effects of crossover
and mutation. One way to define M is to find ri,j (k), the probability that string k will be
produced by a recombination event between string i and string j, given that i and j are
selected to mate. If ri,j (k) were known, we could compute
E(pk (t + 1)) =
si (t)sj (t)ri,j (k)
(3.4)
i,j
Once ri,j (0) is defined, it can be used to define the ri,j (k)
Term ri,j (0) can be expressed as a sum of two terms: the probability that crossover does
not occur between strings i and j and the selected offspring (i or j) is mutated to all zeros
(first term) and probability that crossover does occur and the selected offspring is mutated
to all zeros (second term).
The probability that string i will be mutated to all zeros can be given by:
l|i|
p|i|
m (1 pm )
(3.5)
where, |i| is the number of ones in a string i of length l

Incorporating the above expression, the first term in the expression for ri,j (0) can be
written as
1
l|j|
l|i|
]
+ p|j|
ri,j (0)1 = (1 pc )[p|i|
m (1 pm )
m (1 pm )
2
(3.6)
where,
40
pc = The probability that crossover occurs between strings i and j

1 pc = The probability that crossover does not occur between strings i and j
pm = The probability that mutation occurs at each bit in the selected offspring
1 pm =The probability that mutation does not at each bit in the selected offspring
The factor
1
2
indicates that each of the two offsprings has equal probability of being
selected.
Let h and k denote the two offspring produced from a crossover at point c (counted from
the right-hand side of the string). Since there are l 1 crossover points, so the probability
of choosing point c is 1/(l c).
Second term can be expressed as:
ri,j (0)2 =
l1
1 pc X
l|k|
[p|h| (1 pm )l|h| + p|k|
]
m (1 pm )
2 l 1 c=1 m
(3.7)
Let i1 be the substring of i consisting of l c bits to the left of point c, let i2 be the
substring consisting of the c bits to the right of point c, and let j1 and j2 be defined likewise
for string j. Then |h| and |k| can be given by:
|h| = |i| |i2 | + |j2 |
(3.8)
|k| = |j| |j2 | + |i2 |
(3.9)
Expression for i2 and j2 can be written as:

|i2 | = |(2c 1) i|
(3.10)
|j2 | = |(2c 1) j|
(3.11)
where denotes bitwise and. Since 2c 1 represents the string with l c zeros followed
by c ones, |(2c 1) i(orj) returns the number of ones in the rightmost c bits of i(orj).
If we define:
i,j,c = |i2 | |j2 | = |(2c 1) i| |(2c 1) j|
(3.12)
41
Then
|h| = |i| i,j,c
(3.13)
|k| = |j| + i,j,c
(3.14)
Now, a complete expression for ri,j (0) can be written as:

ri,j (0) =
l1
l1
(1 pm )l |i|
pc X
pc X
[ (1 pc +
i,j,c ) + |j| (1 pc +
i,j,c )] (3.15)
2
l 1 c=1
l 1 c=1
These results give expectation values only; in any finite population. In the limit of an
infinite population, the expectation results are exact. Let G(

x ) = F M(
x ) for vectors
x , where is the composition operator. Then, in the limit of an infinite population,
G(
s (t))
s (t + 1)
Define Gp as
Gp (
x ) = M(F
x /|F
x |)
(3.16)
where |F
x | denotes the sum of the components of vector F
x . Then in the limit of an
infinite population,
G p (
p (t)) =
p (t + 1)
(3.17)
G and Gp act on different representations of the population, but one can be transformed
into other by simple transformation.
3.6
Conclusions
The simple genetic algorithm described in this chapter is applied for tuning the PSS
parameters for both single machine and multimachine power systems, discussed in Chapter
2. Strings are represented by binary digits and single-point crossover and single mutation is
used.
Chapter 4
Proposed Stabilization Technique:
Single Machine System
4.1
Introduction
Objective of Power System Stabilizers is to produce additional damping in the system at

all operating and system conditions. Though several stabilizers have been proposed, conventional PSSs are most commonly used by various utilities to provide additional damping. This
is due to the stabilizers simple structure and the ease of integration with existing system.
Conventional PSSs are however, tuned for a particular operating and system condition and
as the operating conditions change they can not maintain the quality of performance and
may require re-tuning of stabilizer parameters. In this context, an attractive alternative to
PSS design would be to retain the structure of the conventional PSS but tuned in a way
such that it guarantees a minimum performance over a wide range of system and operating
condition.
In this chapter a new method has been proposed for tuning of fixed parameter PSS,
which can explicitly handle the robust D-stability specification mentioned in Chapter 1 and
guarantee a robust performance over a wide range of operating condition. Hence, robust
D-stability is taken as primary specification for design. Conventional lead/lag PSS structure
is retained but its parameters are re-tuned using genetic algorithm to obtain the enhanced
performance. Effectiveness of the proposed method has been tested on a single machine
infinite bus system and has been extended to multimachine system in the next chapter. The
proposed method of tuning the PSS is an attractive alternative to the conventional fixed gain
stabilizer design. However, there are a number of factors, both objective and subjective that
42
Chapter 4. Proposed Stabilization Technique: Single Machine System
43
have to be taken taken into consideration to achieve the desired performance over the entire
expected range of operation of power system. The choice of the control variables, objective
function and the tuning of the controller parameters significantly affect the performance of
a stabilizer.
4.2
Objective Function
According to Appendix F the equation of D-contour shown in Figure 1.1 is given by

f (z) = Re(z) min[|Im(z)|, ] = 0
(4.1)
where z C, is a point on D-contour, C represents the complex plane.

Define J as:
J = max[Re(i ) min{|Im(i )|, }]

i
(4.2)
i = 1, 2, ..., n
where n is the number of eigen values. i is the ith eigen value of the system at an operating
point. A negative value of J implies that all the eigen values lie on the left of the D-contour.
Similarly some or all eigen values will lie on the right of the D contour if J is positive.
On the basis of these facts, objective function F is defined as:
if J 0
if J > 0
F =
(4.3)
where is large positive constant.

The optimization Problem can be stated as:
Minimize F
Subject to:
max
min
Ksj Ksj
Ksj
max
min
T1j T1j
T1j
max
min
T2j T2j
T2j
j = 1, 2, ..., m
where m is the no. of machines. Ksj , T1j and T2j are PSS parameters of j th machine.
(4.4)
4.3
44
Proposed Method
Proposed method consists of two parts, in first part, the PSS parameters are obtained by
solving the constraint optimization problem 4.4 using genetic algorithm. Once parameters
are obtained they are tested for the robustness and D-stability in the second part. The
proposed method of tuning essentially involves the following steps.
Step 1. Start with an initial operating condition, preferably suggested by Larsen and Swann[6],
i.e. for speed input PSS, strong system with heavy loading.
Step 2. Solve the constrained optimization problem given by Equation 4.4 in Section 4.2
using genetic algorithm and obtain the PSS parameters.
Step 3. Once PSS parameters are obtained check for robustness with these parameters. For
this generate a set of loading/system condition.
Step 4. Run load flow for each loading/system condition, eliminate those conditions for
which load flow does not converge. Obtain the operating conditions from load flow.
Step 5. For each operating condition evaluate J given by Equation 4.2.
Step 6. If J 0 for all operating conditions then all the eigen values lie on the left of the
D-contour, hence PSS parameters obtained in Step 2 guarantee desired minimum performance of PSS with robustness.
Step 7. If J > 0 for some operating conditions then choose the operating condition for which
J is maximum positive as an initial condition. Go to Step 2 and repeat the procedure
till the criteria given in Step 6 is satisfied for all operating conditions.
45
Flow chart for the above mentioned process is shown in Figure 4.1. It is clear from the flow
chart that each set of PSS parameter undergoes the robustness screening. After screening
of all the sets only that set is selected which satisfies the minimum performance criteria for
the entire set of operating conditions. This set of PSS parameters is said an optimal set.
Start
Set Initial System
and Loading Condition
LC = 1
Set Jmax = 0
Set Initial
Operating Condition
Generate Initial
Population for
K s , T1 , T2
Run Load Flow
LC = LC+1
Linearize System and
Evalute Eigen Values
Load
Flow
Converged
?
Evalute
Fitness
No
LC = LC+1
Yes
Linearize System and Evalute
Eigen Values
Next
Generation
No
Yes
Converged?
Evalute J
No
Is
No
Perform GA Operations
* Selection
* Xover
* Mutation
All
LC
Converged
?
Yes
0
Yes
No
Is
J > Jmax
Is
Jmax = 0
?
No
Yes
?
Solution
Guarantees
Robustness
Yes
Set Jmax = J
Store Operating
Condition
Take Operating
Condition Corresponding to
Jmax
Stop
Figure 4.1: Flow Chart representation of the proposed method of tuning stabilizer
4.4
46
Application to SMIB System
The proposed method was applied for designing a PSS in a single machine infinite bus
system. Figure 2.4 shows the schematic of SMIB system. Modelling details of this system
are given in Chapter 2, and system data are given in Appendix C.1. Linearized 1.1 model
was used for obtaining the system eigen values.
4.4.1
Control Parameters
As shown in Figure 2.3, a two stage lead/lag compensator structure was chosen for the
PSS. Hence all three parameters Ks , T1 and T2 were taken as control parameters. Bounds
on these parameters are shown in Table 4.1. These parameter-bounds were defined on the
basis of conventional control design for nominal operating condition.
Table 4.1: Control Parameters Bounds
4.4.2
Bounds
Ks
T1 (sec) T2 (sec)
Min
0.01
0.01
0.01
Max
50
0.7
0.7
GA Parameters
GA parameters utilized are given in in the Table 4.2. These parameters were selected on
the basis of system experience and main criteria for selection were taken as:
1. Accuracy in the solution,
2. Convergence
3. Small changes in these parameters should not significantly affect the GA performance.
47
Table 4.2: GA Parameters

Encoding
Population Size
200
Genes Length
30
Chromosome Length
90
Crossover Probability (Pc )
0.95
Mutation Probability (Pm )
0.033
Maximum Generation
4.4.3
Binary
400
Operating Conditions
The operating condition for the above system is completely defined by the values of real
power, P , the reactive power, Q, at the generator terminals and the equivalent transmission
line impedance, Xe . P , Q and Xe were assumed to vary independently over following ranges:
P : 0.4 to 1.0
Q : -0.2 to 0.5
Xe : 0.2 to 0.7
This encompasses almost all practically occurring operating conditions and a very weak to
very strong transmission network.
4.4.4
GA Results
PSS parameters were obtained by solving the constrained optimization problem 4.4, along
with the robustness test under above mentioned operating conditions. Parameters of conventional PSS were taken as the initial values for the GA routine. Table 4.3 shows the initial
values of the PSS parameters and the final values arrived at by the proposed method.
48
Table 4.3: Initial and Final Values of PSS Parameters
4.4.5
Ks
T1 (sec)
T2 (sec)
Initial
12.50
0.0738
0.0280
Final
24.81
0.1024
0.0213
Performance Analysis of Proposed GA Based PSS
A number of studies involving variety of tests at different system and operating conditions
have been conducted to evaluate the efficacy of the proposed stabilizer. All results are
compared with the performance of a conventional PSS having the following transfer function:
5s
k(s) = 12.5
1 + 5s
1 + .0738s
1 + .028s
(4.5)
An illustrative set of results are presented in the following section. For simplicity from
now onwards, the conventional PSS has been referred to as CPSS and the proposed GA
based PSS as PGAPSS.
CPSS Parameter Selection
The operating point corresponding to P = 0.8, Q = 0.4 Xe = 0.2 was chosen as nominal
operating point. The CPSS was designed for this nominal operating point using the tuning
guidelines given in Appendix D.1.
The parameters of both the PGAPSS and the CPSS were fixed at their designed values,
i.e. they were kept the same for all tests described in the following sections. Output limits
(Vsmax and Vsmin ) of both the PSSs were taken as 0.15
4.4.6
Robustness Test and Eigen Value Plots
The fixed-parameter CPSS normally needs to be redesigned for each power system application and has to be re-tuned if operating condition or the system configuration changes.
However, the PGAPSS, irrespective of system and operating conditions with the specified
bounds is expected to meet the D-stability criterion as defined earlier. The PGAPSS must
49
therefore meet the guaranteed minimum performance. Thus changes in the system configuration or parameters should have a minor effect on the performance of the PGAPSS.
To test the robustness of the PGAPSS P , Q and Xe were independently varied over the
range of operating conditions given in subsection 4.4.3 in a step size of 0.05 and system
eigen values were plotted for the entire range of operating conditions. Some of the operating
conditions within this range having leading power factor and high values of Xe did not yield
a steady state load flow solution and were eliminated.
Figure 4.2 shows the open loop poles of the system for the specified set of operating
conditions. The desired D-contour plot is also shown in the figure. As seen, the system is
poorly damped outside the D-contour or unstable for most of the operating conditions.
20
15
10
imag
5
0
5
10
15
20
14
12
10
6
real
Figure 4.2: Open loop poles
Figure 4.3 shows the closed loop poles with CPSS. It is clear from the figure that D-
50
stability can not be achieved by application of CPSS as the system is unstable for few of the
operating conditions. Local modes are poorly damped for most of the operating conditions.
Hence CPSS does not guarantee robust performance for entire set of operating conditions.
20
15
10
imag
5
0
5
10
15
20
14
12
10
6
real
Figure 4.3: Closed loop poles with CPSS
Figure 4.4 shows the closed loop poles for the entire set of operating conditions with the
PGAPSS. As seen, robust D-stability of the closed loop has been achieved. Hence PGAPSS
guarantees desired performance for entire set of operating conditions.
51
25
20
15
10
imag
5
0
5
10
15
20
25
20
15
10
real
Figure 4.4: Closed loop poles with PGAPSS
4.4.7
Simulation Studies
The time domain performance of the PSSs were studied using nonlinear 1.1 simulation
model of the system. The following test cases were considered for simulations.
(a) Lagging Power Factor Operation Test:
With the generator operating at wide range of output power at lagging power factors, stepchanges in input torque were applied. The response of the system without stabilizer shows
large oscillations. Also three phase-to-ground fault at generator terminal were simulated.
Per unit speed deviation (slip) response of the system with PGAPSS and CPSS were plotted
to compare the relative performance of the two PSS.
(b) Unity Power Factor Operation Test:
With the generator operating at different output power with unity p.f., step-changes in
52
input torque were applied. Again, the response of the system with PGAPSS and CPSS were
plotted for comparison.
(c) Leading Power Factor Operation Test:
With the generator operating at leading power factor, step-changes in input torque as well
as three phase-to-ground fault at generator terminal were simulated and the response of the
system with PGAPSS and CPSS were plotted.
4.4.8
Results and Discussion
The power system model of Figure 2.4 is simulated for 8 sec. The disturbances are given
at t = 0.5 sec. System response in the form of slip (Sm ) are plotted here. The following
types of disturbances have been considered.
(i) A step change of 0.1 pu in input mechanical torque.
(iii) A three phase-to-ground fault for 100 ms at generator terminal.
(iv) A Step change of 0.1 pu in input mechanical torque with inertia H 0 = H/4.
Followings are the results under different system operating conditions followed by above
disturbances
(1) Figures 4.5 - 4.14: system responses in lagging p.f. operation test at different output
power.
(a) Figures 4.5 - 4.9: responses with strong transmission system.
(b) Figures 4.10 - 4.14: responses with weak transmission system.
(2) Figures 4.15 - 4.19: system responses in unity p.f. operation test at different output
power.
(b) Figures 4.18, 4.19: responses with weak transmission system.
(3) Figures 4.20 - 4.23: system responses in leading p.f. operation test at different output power.
(b) Figure 4.23: response with weak transmission system.
53
(4) Figures 4.24- 4.26: System responses with A three phase-to-ground fault for 100 ms
at generator terminal.
(5) Figures 4.27- 4.29: responses with machine inertia constant H 0 = H/4.
(6) Figure 4.30: case when system with CPSS is unstable.
The responses with two stabilizers, CPSS and PGAPSS and without controller are plotted
simultaneously. The system without stabilizer is highly oscillatory. Both the controllers are
able to damp the oscillations reasonably well or to stabilize the system at most of the
Figures 4.10 to 4.14 are the responses corresponding to the lagging power factor operating
condition with weak transmission system. As seen, in Figures 4.11 to 4.14, system without
stabilizer is unstable and highly oscillatory. In Figures 4.10 and 4.14 with disturbance in the
form of Tm =0.1 pu, overshoots with PGAPSS is considerably less compared to CPSS.
Figures 4.5 to 4.9 shows the responses of same operating condition but with strong transmission system. System is more stable in this case, following any disturbance. Both the
controllers improves its dynamic stability considerably and PGAPSS shows its superiority
over CPSS. Similarly for other operating points we can see that although the CPSS is effective in damping the oscillations, but settling time is slightly less and overshoot is much
less with PGAPSS. As system shifts from lagging to leading p.f. operation, its oscillatory
behavior increases and dynamic stability decreases.
Figures 4.24 to 4.26 refer to a three-phase to ground fault at generator terminal , where we
can see that PGAPSS works much better than CPSS in terms of settling time but overshoot
of PGAPSS is same as that of CPSS.
Figures 4.27 to 4.29 shows a disturbance of Tm =0.1 pu with machine inertia H 0 = H/4.
In this case the system oscillation frequency is almost doubled and we find that PGAPSS
still damps out the oscillation more efficiently than CPSS.
54
Figure 4.30 shows the case when CPSS is unstable. In this case PGAPSS shows the
satisfactory performance.
4.5
Conclusions
This chapter has presented a new technique for the design of power system stabilizers.
The proposed design technique is seen to provide the desired closed loop performance over
the prespecified range of operating conditions. The performance evaluation of the proposed
stabilizer on single machine system shows that this increased robustness could be achieved
by application of genetic algorithms to stabilizer design. The design procedure is simple and
bears much potential for practical implementation.
It may be mentioned that the conventional PSS design is quite complex and requires
considerable expertise. Considerable effort was also extended towards designing the very
well tuned CPSS used for the comparative analysis.
Conventional stabilizer design for multimachine systems are even more complex. The GA
based stabilizer design technique proposed here could be quite helpful under such circumstances. This is being investigated in the next chapter.
55
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1
0
4
time(sec)
Figure 4.5: A step change of Tm = 0.1 pu, St = 0.5 + j0.1, Xe = 0.3

3
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1.5
0
4
time(sec)

3
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1.5
0
4
time(sec)
56
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1.5
2
0
4
time(sec)

3
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1.5
0
4
time(sec)

3
2.5
x 10
without PSS
with CPSS
with PGAPSS
2
1.5
1
Sm
0.5
0
0.5
1
1.5
2
0
4
time(sec)
57
2.5
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1.5
0
4
time(sec)

3
2.5
x 10
without PSS
with CPSS
with PGAPSS
2
1.5
1
Sm
0.5
0
0.5
1
1.5
2
2.5
0
4
time(sec)

3
x 10
without PSS
with CPSS
with PGAPSS
Sm
4
0
4
time(sec)
58
x 10
without PSS
with CPSS
with PGAPSS
4
3
2
Sm
1
0
1
2
3
4
5
0
4
time(sec)

3
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1.5
0
4
time(sec)

3
2.5
x 10
without PSS
with CPSS
with PGAPSS
2
1.5
1
Sm
0.5
0
0.5
1
1.5
2
2.5
0
4
time(sec)
59
x 10
without PSS
with CPSS
with PGAPSS
Sm
4
0
4
time(sec)

3
x 10
without PSS
with CPSS
with PGAPSS
2
Sm
3
0
4
time(sec)

3
x 10
without PSS
with CPSS
with PGAPSS
Sm
4
0
4
time(sec)
60
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm
0.5
0.5
1.5
2
0
4
time(sec)
Figure 4.20: A step change of Tm = 0.1 pu, St = 0.5 j0.2, Xe = 0.3

3
x 10
without PSS
with CPSS
with PGAPSS
Sm
4
0
4
time(sec)

3
x 10
without PSS
with CPSS
with PGAPSS
4
3
2
Sm
1
0
1
2
3
4
5
0
4
time(sec)
61
x 10
without PSS
with CPSS
with PGAPSS
4
Sm
6
0
4
time(sec)

0.02
without PSS
with CPSS
with PGAPSS
0.015
0.01
0.005
Sm
0
0.005
0.01
0.015
0.02
0.025
0
0.5
1.5
2.5
time(sec)
3.5
4.5
Figure 4.24: A 3 to ground fault for 100 ms at generator terminal, St = 1.0+j0.2, Xe = 0.3
0.02
without PSS
with CPSS
with PGAPSS
0.015
0.01
0.005
Sm
0
0.005
0.01
0.015
0.02
0.025
0
0.5
1.5
2.5
time(sec)
3.5
4.5
Figure 4.25: A 3 to ground fault for 100 ms at generator terminal, St = 0.8j0.2, Xe = 0.3
62
0.03
without PSS
with CPSS
with PGAPSS
0.02
Sm
0.01
0.01
0.02
0.03
0
0.5
1.5
2.5
time(sec)
3.5
4.5
Figure 4.26: A 3 to ground fault for 100 ms at generator terminal, St = 1.0+j0.5, Xe = 0.6
3
x 10
without PSS
with CPSS
with PGAPSS
Sm
4
0
4
time(sec)
Figure 4.27: A step change of Tm = 0.1 pu, St = 1.0 + j0.2, Xe = 0.3, H 0 = H/4
3
x 10
without PSS
with CPSS
with PGAPSS
4
3
2
Sm
1
0
1
2
3
4
5
0
4
time(sec)
Figure 4.28: A step change of Tm = 0.1 pu, St = 0.8 j0.2, Xe = 0.3, H 0 = H/4
63
0.01
without PSS
with CPSS
with PGAPSS
0.008
0.006
0.004
Sm
0.002
0
0.002
0.004
0.006
0.008
0.01
0
4
time(sec)
Figure 4.29: A step change of Tm = 0.1 pu, St = 1.0 + j0.5, Xe = 0.6, H 0 = H/4
3
x 10
without PSS
with CPSS
with PGAPSS
Sm
6
0
4
time(sec)
Chapter 5
Proposed Stabilization Technique:
Multimachine System
5.1
Introduction
Single machine infinite bus power system models are frequently used in stability studies to
validate the performance of power system stabilizers. However, it is rarely possible to model
an actual power system by a SMIB model. A more realistic power system model can be
obtained by considering the case of multimachine power system. This chapter presents an
extension of the proposed stabilization technique to multimachine case, so that the efficacy of
the proposed method in damping the multimodal oscillations in more realistic power system
could be examined.
5.2
Performance Evaluation of the Stabilizer in Multimachine System
The proposed method was applied for PSSs design in a 3 machine 9 bus power system
model. Single line diagram of the system is shown in Figure 2.7. AVR and machine data
are given in Appendix C.2. Each machine was considered to be equipped with an AVR and
PSS as shown in Figures 2.2 and 2.3.
5.2.1
Control Parameters
All three machines were considered to be equipped with PSSs. The structure of each
PSS remains the same as discussed in Chapter 4. Hence, a total of nine parameters were
64
Chapter 5. Proposed Stabilization Technique: Multimachine System
65
considered as control parameters. Bounds on these parameters are shown in Table 5.1.
Table 5.1: Control Parameters Bounds
5.2.2
Bounds
Ks (1)
T1 (1) T2 (1)
Ks (2) T1 (2)
T2 (2)
Ks (3) T1 (3)
T2 (3)
Min
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.01
Max
50
0.8
0.8
50
0.8
0.8
50
0.8
0.8
GA Parameters
Table 5.2 shows the GA parameters considered for the optimization routine.
Table 5.2: GA Parameters For Multimachine Case

Encoding
Binary
Population Size
500
Genes Length
30
Chromosome Length
270
Crossover Probability (Pc )
0.95
Mutation Probability (Pm )
0.033
Maximum Generation
400
All nine parameters were tuned using the genetic algorithm for the specified range of
system conditions.
5.2.3
Loading Conditions
A range of operating conditions for this system was obtained by varying the system loads
and generations from a small fraction to a large percentage of the base case. Table 5.3 shows
nominal (base case) and the minimum and maximum loading conditions of the system.
66
Table 5.3: Loading Range of 3 Machine, 9 Bus System

Bus No.
5.2.4
Power injection(pu)
Minimum
Nominal
Maximum
2+
0.570
1.63
3.994
3+
0.296
0.85
2.083
0.000
0.000
0.00
0.00
0.000
0.000
-0.438
-0.175
-1.25 -0.50
-3.060
-1.225
-0.315
-0.105
-0.90 -0.30
-2.205
-0.735
0.000
0.000
0.00
0.00
0.000
0.000
-0.350
-0.123
-1.00 -0.35
-2.450
-0.858
0.000
0.000
0.00
0.000
0.000
slack bus,
0.00
PV bus.
GA Results
Parameters of proposed PSSs were obtained by solving the constrained optimization problem 4.4 followed by robustness test given in Chapter 4. Optimal parameters of proposed
PSS along with initial guess for these parameters are shown in Tables 5.4(a) and 5.4(b). The
optimal parameters are again the parameters of the proposed PSSs that guarantee robust
D-stability performance.
CPSS Parameter Selection
Performance of the proposed PSS was compared with CPSS employed on each machine
with the following transfer function.
Vpss
10s
= 2.0
1 + 10s
1 + 0.2218s
1 + 0.05s
Sm
(5.1)
CPSS parameters were obtained by using the tuning guideline given in Appendix D.2. Out-
67
Table 5.4: Optimal stabilizer parameters of PGAPSS

(a) Initial Guess
(b) Final Values
Generator
KS
T1 (sec)
T2 (sec)
Generator
KS
T1 (sec)
T2 (sec)
G1
38.0
0.15
0.05
G1
23.641
0.1245
0.0244
G2
6.0
0.20
0.05
G2
2.910
0.3215
0.0895
G3
1.5
0.29
0.05
G3
3.732
0.1704
0.0364
put limits of both CPSS and PGAPSS were taken as 0.15.
5.2.5
Robustness Test and Eigen Value Plots
From the loading conditions given in Table 5.3, approximately 60,000 operating conditions
were generated to test the performance of the proposed PSS for a wide range of operating
conditions. Eigen value plotes for these operating conditions are shown in Figures 5.1 to
5.3.
Figure 5.1 shows the open loop poles of the system for the range of load variation.The
low frequency oscillatory modes are seen to be poorly damped or undamped for most of the
Figure 5.2 shows that by using the CPSS these modes can be shifted in the left of the
D-contour, up to some extent. But still system is unstable for some operating conditions.
Figure 5.3 shows the system poles with proposed PSS. As seen, The low frequency modes
have been shifted into the acceptable region. This shows the robust performance of the
PGAPSS over the entire range of operating conditions.
Figure 5.1: Open loop poles
Figure 5.2: Closed loop poles with CPSS
68
69
Figure 5.3: Closed loop poles with PGAPSS
5.2.6
Operating Points For Simulation Studies
Four operating conditions for the time response studies are as shown in Table 5.5. All
quantities are on 100 MVA base. The loads YLA , YLB and YLC are represented by the constant
admittances.
Table 5.5: Operating points of generators on a 100 MVA base

(a) Standard Operating Point (SOP)
Gen.
Pto (pu)
Qto (pu)
Vto (pu)
Load Admittances
(Real Power)
(React. Power)
(Term. Voltage)
at A, B and C
G1
0.7135
0.2746
1.040
YLA
1.2610 - j0.5044
G2
1.6297
0.0679
1.025
YLB
0.8777 - j0.2926
G3
0.8522
-0.1077
1.025
YLC
0.9690 - j0.3391
70
(b) Heavily Loaded Operating Point (HOP)

Gen.
Pto (pu)
Qto (pu)
Vto (pu)
Load Admittances
(Real Power)
(React. Power)
(Term. Voltage)
at A, B and C
G1
2.2074
1.0924
1.040
YLA
2.3135 - j0.9254
G2
1.9200
0.5653
1.025
YLB
2.0318 - j0.6773
G3
1.2800
0.3599
1.025
YLC
1.5837 - j0.6335
(c) Lightly Loaded Operating Point (LOP)

Gen.
Pto (pu)
Qto (pu)
Vto (pu)
Load Admittances
(Real Power)
(React. Power)
(Term. Voltage)
at A, B and C
G1
0.3623
0.1663
1.040
YLA
0.6403 - j0.5418
G2
0.8000
-0.1069
1.025
YLB
0.4306 - j0.3349
G3
0.4500
-0.2033
1.025
YLC
0.4719 - j0.2359
(d) Other Operating Point (OOP)

Gen.
Pto (pu)
Qto (pu)
Vto (pu)
Load Admittances
(Real Power)
(React. Power)
(Term. Voltage)
at A, B and C
G1
1.8489
0.7318
1.040
YLA
3.3833-j0.1935
G2
3.8514
0.9308
1.025
YLB
2.0043-j0.1088
G3
2.0057
0.2279
1.025
YLC
2.5027-j0.1256
5.2.7
Simulation Results and Discussion
The system is simulated for 10 sec, for various operating conditions as shown in Table
5.5. The responses of the system with proposed stabilizer have been compared with system
equipped with CPSS and system without PSS. Machine 1 has the highest inertia constant
and this is taken as reference. The responses of Sm21 (= Sm2 Sm1 ) and Sm31 (= Sm3 Sm1 )
have been plotted for each disturbance which is initiated at t = 1.0 sec. The following cases
71
have been considered.

(a) A step change of 0.1 pu in the input mechanical torque.
(b) A three-phase to ground fault for 100 ms between buses 6 and 9, and then removal
of line between these buses.
Followings are the results under four different system operating conditions followed by above
disturbances
(1) Figures 5.4 - 5.11: responses at standard operating condition (SOP).
(a) Figures 5.4 & 5.5: responses following a step change in input torque at generator 1.
(b) Figures 5.6 & 5.7: responses following a step change in input torque at generator 2.
(c) Figures 5.8 & 5.9: responses following a step change in input torque at generator 3.
(d) Figures 5.10 & 5.11: responses to a 3-phase to ground fault for 100 ms at point P,
middle of the line between bus number 6 and 9.
(2) Figures 5.12 - 5.19: responses at heavily loaded operating condition (HOP).
(3) Figures 5.20 - 5.27: responses at lightly loaded operating condition (LOP).
72
(4) Figures 5.28 - 5.33: responses at other operating condition (OOP).
Figures 5.4 to 5.11 illustrate the simulation results under step changes in input mechanical
torque and 3-phase fault in the system under standard operating point (SOP), shown in Table
5.5(a). Generator 1 has greater inertia than the rest two generators 2 and 3 and hence the
oscillations in generator 1 is smaller when disturbances are given at generator 1 compared to
when disturbances are given at generator 2 or 3. So change in slip Sm21 and Sm31 are smaller
for generator 1 (Figures 5.4 & 5.5) when disturbance is given in generator 1, in comparison
to generator 2 or 3 when disturbances are given at generator 2 or 3 (Figures 5.6 - 5.9).
With both the stabilizers CPSS and PFPSS (conventional and proposed) damping is improved considerably but with PGAPSS settling time is much less than that could be achieved
with CPSS. Peak overshoots in case of PGAPSS are close to the CPSS case. Figures 5.10 &
5.11 illustrate the situation when there is a three-phase to ground fault for 100 ms at point
P and the line between bus no. 6 and 9 is removed after fault. The system is unstable for
this disturbance without stabilizers but it is stable with both stabilizers. Settling time and
overshoots are much improved with PGAPSS.
torque and 3-phase fault in heavily loaded operating point (HOP) of the system, shown in
Table 5.5(b). Figures 5.12 to 5.17 show the simulations results under step change in input
torque at each generator at a time and system response with both the stabilizers. The
overshoots and settling time are again less with PFPSS. Figures 5.18 & 5.19 illustrate the
situation when there is a three-phase to ground fault for 100 ms at point P and the line
between bus no. 6 and 9 is removed. The system would be unstable for this disturbance
without stabilizers.
73
torque and 3-phase fault in lightly loaded operating point (LOP) of the system, shown in Table
5.5(c). Figures 5.20 to 5.25 show the simulations results under step change in input torque
of 0.1 pu at each generator at a time. Comparing with the case of conventional PSS, the
settling time with PGAPSS is more than that with CPSS. The response is slightly sluggish
after the oscillations have died down. When disturbance is given at Generator 1, the peak
overshoot with PGAPSS is also more as compared to CPSS. Hence time domain performance
of proposed PSS is slightly inferior under very lightly loaded operating conditions. The Dstability criterion however has been met. Figures 5.26 & 5.27 illustrate the situation under
3-phase fault for 100 ms as described above. The system is stable for this disturbance
without stabilizers. However, considerable improvement in performance can be noticed with
PGAPSS.
Figures 5.28 to 5.33 illustrate the simulation results under step changes in input mechanical torque for a loading condition, shown in Table 5.5(d). This operating condition has
been chosen to show the robustness of PGAPSS over CPSS. Figures 5.28 to 5.33 show the
simulations results under step change in input torque at each generator at a time and system
response with both the stabilizers. In this case the CPSS fails to stabilize the system.
5.2.8
Computational Requirements
The PSS tuning performed on 3 machine 9 bus system was done on a Pentium-III PC,
using a program coded in MATLAB 6.1. The CPU time spent in one generation of GA is
approximately 45 seconds on average. Most of the time is spent in obtaining 1.1 linearized
model of this system and eigen value evaluation.
5.3
Conclusions
The results presented in this chapter showed that in a multimachine system, fixed-structure
damping controllers can be tuned to provide satisfactory damping performance over a prespecified set of operating conditions. The GA based tuning process has shown robustness in
achieving controllers satisfying the design criteria in a large-scale realistic power system.
74
The computational time required by algorithm can be considered adequate for a design
study. It can be reduced considerably if a through code optimization process is performed.
For larger power systems, the use of high performance computing i.e. parallel processing
may be of great help in keeping the computational time within tolerable limits.
75
x 10
without PSS
with CPSS
with PGAPSS
2.5
2
1.5
1
Sm21
0.5
0
0.5
1
1.5
2
2.5
0
10
time(sec)
Figure 5.4: A step change of Tm1 = 0.1 pu at unit 1, under SOP.
x 10
without PSS
with CPSS
with PGAPSS
1.5
1
0.5
Sm31
0
0.5
1
1.5
2
2.5
0
10
time(sec)
Figure 5.5: A step change of Tm1 = 0.1 pu at generator 1, under SOP.
10
x 10
without PSS
with CPSS
with PGAPSS
8
6
4
Sm21
2
0
2
4
6
8
0
time(sec)
10
76
x 10
without PSS
with CPSS
with PGAPSS
Sm31
8
0
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
4
Sm21
6
0
10
time(sec)
1.5
x 10
without PSS
with CPSS
with PGAPSS
1
Sm31
0.5
0.5
1
0
time(sec)
10
77
x 10
8
without PSS
with CPSS
with PGAPSS
Sm21
10
time(sec)
Figure 5.10: A 3- to ground fault for 100 ms at P, under SOP.
x 10
without PSS
with CPSS
with PGAPSS
8
6
4
Sm31
2
0
2
4
6
8
0
10
time(sec)
Figure 5.11: A 3- to ground fault for 100 ms at P, under SOP.
x 10
2.5
2
1.5
1
Sm21
0.5
0
0.5
1
1.5
2
2.5
0
5
time(sec)
10
Figure 5.12: A step change of Tm1 = 0.1 pu at generator 1, under HOP.
78
x 10
without PSS
with CPSS
with PGAPSS
1.5
1
0.5
Sm31
0
0.5
1
1.5
2
2.5
0
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
0.8
0.6
0.4
Sm21
0.2
0
0.2
0.4
0.6
0.8
1
0
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
Sm31
8
0
10
time(sec)
79
x 10
without PSS
with CPSS
with PGAPSS
4
Sm21
6
0
10
time(sec)
1.5
x 10
without PSS
with CPSS
with PGAPSS
1
Sm31
0.5
0.5
1
0
10
time(sec)
without PSS
with CPSS
with PGAPSS
0.01
Sm21
0.005
0.005
0.01
0
time(sec)
Figure 5.18: A 3- to ground fault for 100 ms at P, under HOP.
10
80
0.015
without PSS
with CPSS
with PGAPSS
0.01
Sm31
0.005
0.005
0.01
0.015
0
10
time(sec)
Figure 5.19: A 3- to ground fault for 100 ms at P, under HOP.
x 10
without PSS
with CPSS
with PGAPSS
1.5
1
0.5
Sm21
0
0.5
1
1.5
2
2.5
3
0
10
time(sec)
Figure 5.20: A step change of Tm1 = 0.1 pu at generator 1, under LOP.
1.5
x 10
without PSS
with CPSS
with PGAPSS
0.5
Sm31
0.5
1.5
2.5
0
10
time(sec)
81
10
x 10
without PSS
with CPSS
with PGAPSS
Sm21
5
0
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
2
Sm31
6
0
10
time(sec)
10
x 10
without PSS
with CPSS
with PGAPSS
Sm21
5
0
10
time(sec)
82
x 10
without PSS
with CPSS
with PGAPSS
2
Sm31
6
0
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
1.5
Sm21
0.5
0.5
1.5
0
10
time(sec)
Figure 5.26: A 3- to ground fault for 100 ms at P, under LOP.
x 10
without PSS
with CPSS
with PGAPSS
Sm31
3
0
time(sec)
Figure 5.27: A 3- to ground fault for 100 ms at P, under LOP.
10
83
x 10
without PSS
with CPSS
with PGAPSS
2
Sm21
3
0
10
time(sec)
Figure 5.28: A step change of Tm2 = 0.1 pu at generator 2, under OOP.
x 10
without PSS
with CPSS
with PGAPSS
2
Sm31
3
0
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
2
Sm21
10
time(sec)
84
x 10
without PSS
with CPSS
with PGAPSS
2
Sm31
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
2
Sm21
3
0
10
time(sec)
x 10
without PSS
with CPSS
with PGAPSS
2
Sm21
10
time(sec)
Chapter 6
Conclusions
An attempt has been made in this work to develop a simple but robust PSS with particular
emphasis on achieving a minimum closed loop performance over a wide range of operating
and system conditions. The minimum performance requirements of the controller has been
decided and this performance has been obtained using genetic algorithm. The PSS is thus
robust to changes in the plant parameters, brought about due to changes in system and
operating conditions, guaranteeing a minimum performance.
Extensive simulation results have been presented for a large number of operating and
system condition in single machine and multimachine systems to establish the efficacy of the
proposed stabilization technique.
Performance evaluation of proposed PSSs on single and multimachine systems shows that
robust fixed parameter stabilizers are indeed a viable solution to the problem of low frequency
oscillations. Eigen value analysis and extensive simulation studies show that the proposed
method for PSS design provides the desired closed loop performance over the prespecified
range of operating conditions. Furthermore, robust performance of the proposed PSS over
a widely varying operating conditions shows its superiority over existing stabilizers.
The attractive features of the proposed stabilizer are its simple structure and design procedure. Conventional lead/lag PSS structure is retained but its parameters are re-tuned using
genetic algorithm. Due to simple design procedure and robust performance the proposed
stabilizer bears much potential for practical implementation.
85
Chapter 6. Conclusions
86
Scope for Future Work

It has been well established, by extensive computer simulation studies, that fixed gain
robust controllers could be an ideal choice for both single and multimachine power systems.
The efficacy of these stabilizers in damping out single and multimodal oscillations has to be
now established by actual field trials on real systems.
The proposed method utilizes simple genetic algorithm (SGA), which is time consuming.
Application of highly efficient genetic algorithms can reduce the computational time considerably. For larger power systems, the use of high performance computing i.e. parallel
processing may be of great help in keeping the computational time within tolerable limits.
The present formulation of the objective function guarantees a minimum performance under varying system conditions. It may be possible to specify more complex performance
requirements by appropriately formulating the objective functions. The CPSS structure has
been adopted for the present stabilizer design keeping practical implementations in mind.
The PSS structure could also be appropriately chosen to obtain better GA based performance.
Appendix A
Calculation of Initial Conditions
1. Compute Iao from
Pto jQto
Iao = Iao 6 o =
Vto 6 o
(A.1)
Eqo 6 o = Vto 6 o + (Ra + jxq )Iao 6 o
(A.2)
2. Compute Eqo and o from
3. Compute
ido = Iao sin(o o )
iqo = Iao cos(o o )
vdo = Vto sin(o o )
vqo = Vto cos(o o )
4. Compute
Ef do = Eqo (xd xq )ido
0
0
Eqo
= Ef do + (xd x0d )ido = Eqo
(xq x0d )ido
0
Edo
= (xq x0q )iqo
0
= (x0d x0q )iqo
Edc
(f or the multimachine)
0
0
iqo + (x0d x0q )ido iqo
ido + Eqo
Teo = Edo
87
Appendix B
Heffron-Philips Model of the SMIB
System
K1
Te1
Tm
Ms
Te2
PSS(s)
D
K4
K2
K5
E q
E fd
K3
Ka
1 + sTa
1+ sK 3T do
Vt
VS
V ref
K6
GEP(s)
Figure B.1: Heffron-Philips model of the SMIB system
Constants K1 to K6 are known as Heffron-Philips constants.For lossless network zR = 0,

zI = xe , the expression for the constants K1 to K6 are simplified. For the lossless network
88
Appendix B. Heffron-Philips Model of the SMIB System
89
on the stator side Ra = 0. The expressions are given below [1, 84].
"
K1
Eb
(xq x0d ) 2
sin o
=
Eqo cos o + Eb
(xe + xq )
(xe + x0d )
(B.1)
K2 =
Eb sin o
(xe + x0d )
(B.2)
K3 =
xe + x0d
xe + xd
(B.3)
K4 =
(xd x0d )
Eb sin o
(xe + x0d )
(B.4)
"
K5
Eb xq vdo cos o x0d vqo sin o

=
+
vto (xe + xq )
(xe + x0d )
K6 =
vqo
xe
vto (xe + x0d )
(B.5)
(B.6)
Appendix C
Data for SMIB and Multimachine
System
C.1 SMIB System Data
Table C.1: Generator Constants

f(Hz)
50
H(sec) D(pu) Ra (pu) xd (pu)

3.25
0.0
0.0
2.0
0
xq (pu) x0d (pu) x0q (pu) Tdo
(sec)
1.91
0.244
Table C.2: AVR Parameters

Ka
Ta (sec)
E f dmin
E f dmax
50.0
0.05
-7.0
7.0
90
0.244
4.18
0
Tqo
(sec)
0.75
Appendix C. Data for SMIB and Multimachine System
91
C.2 Multimachine System Data
Table C.3: Generator Constants

Generator
G1
G2
G3
Type
Hydro
Steam
Steam
Rated MVA
247.5
192.0
128.0
kV
16.5
18.0
13.8
f(Hz)
60
60
60
H(sec)
23.64
6.4
3.01
D(pu)
0.01
0.01
0.01
Ra (pu)
0.0
0.0
0.0
xd (pu)
0.146
0.8958
1.3125
xq (pu)
0.0969
0.8645
1.2578
x0d (pu)
0.0608
0.1198
0.1813
x0q (pu)
0
Tdo
(sec)
0
Tqo
(sec)
0.0969
0.1969
0.25
8.96
6.0
5.89
0.31
0.535
0.6
All reactances are in pu on a 100MVA system base.
Table C.4: AVR Parameters

Ta (sec) E f dmin
E f dmax
Gen.
Ka
G1
100.0
0.05
-7.0
7.0
G2
100.0
0.05
-7.0
7.0
G3
100.0
0.05
-7.0
7.0
Appendix D
Tuning Guidelines for the CPSS
D.1 SMIB Case
The conventional, lead compensation type of PSS continues to be most popular with the
industry due to its simplicity and well understood operational principles. This Appendix
explains the tuning procedure for this type of PSS [6] used in this thesis for comparison.
Consider the Heffron-Philips model of the SMIB system shown in Figure B.1. It is required
to inject a signal derived from the rotor speed at the AVR voltage reference input such that
a damping component of the electrical torque is generated. As seen, the stabilizing signal
passes through a part of the system before arriving at the torque summing junction. This
subsystem termed GEP(s) is shown by dotted lines in Figure B.1. The signal suffers a phase
lag due to this block. Figure D.1 shows a typical phase angle plot of the transfer function of
GEP(s). For the signal at output of GEP(s) to be in phase with rotor speed it is necessary
to to provide a phase lead to the injected signal equal to the phase lag of block GEP(s) at
the oscillating frequency. Figure D.1 also shows the phase plot of such a lead compensator
PSS(s) and the resulting compensated phase angle of the cascade GEP(s).PSS(s).
The transfer function of the lead compensator is taken as
P SS(s) = Ks
1 + sT1
1 + sT2
A value for the compensated phase angle of GEP(s).PSS(s) between 0 and -90 degrees results
in additional damping. Overcompensation results in a decrement of the net synchronizing
torque and hence should be avoided (Figures D.2(a) and D.2(b)). For a practical implementation, it is recommended [6] to keep the phase angle of GEP(s).PSS(s) between 0 and -40
92
Appendix D. Tuning Guidelines for the CPSS
93
90
PSS(s)
45
GEP(s).PSS(s)
45
GEP(s)
90
135
180
1
10
10
10
10
Figure D.1: Phase angle plots of GEP(s), PSS(s) and GEP(s).PSS(s)

degrees at the oscillating frequency and between 0 and -90 degrees over as wide a frequency
range as possible. The compensator parameters T1 and T2 are chosen so as to maintain the
phase angle of the GEP(s).PSS(s) within these limits.
T
D
Te
Te2
Te1
a
Te2
TS
Te
TD
TS
Te1
Figure D.2: Phasor diagram representation of synchronizing and damping torques
94
The root locus of the system with the chosen lead compensator is then plotted. The gain
of the PSS is chosen such that the damping of the rotor mode eigen values is maximized.
If any other modes (like the exciter mode) tend to get destabilized with an increase in the
gain, the damping of these other modes is also considered. Figure D.3 shows a typical root
locus of a linearized SMIB system with the lead compensator and the closed loop poles with
the chosen gain.
35
30
25
imag
20
PSS Gain
15
10
14
12
10
real
Figure D.3: A typical root locus plot for SMIB system with the lead Compensator
The process can be summarized as:
1. Form the linearized model of the plant at the chosen nominal operating condition.
2. Compute the transfer function of GEP(s).
3. Plot the phase angle of GEP(s) and choose the compensator parameters T1 and T2
so as to provide a phase lead equal to the lag of GEP(s) at oscillating frequency and
95
to maintain the compensated angle between 0 and -40 degrees at the frequency of
oscillation.
4. Plot the root locus of the open loop system and choose value for the gain which simultaneously maximizes the damping of the rotor and exciter mode.
D.2 Multimachine Case[10]

The time response expression [8] corresponding to a single complex eigen value may be
simplified to:
y = et cos t
(D.1)
dy/dt = sy = et cos t et sin t = ( + j)et cos t = y
(D.2)
s = + j
(D.3)
Differentiating,
Hence,
The machine acceleration torque can be written as:

Ta = Tm Te
(D.4)
Assuming damping constant D = 0 and using Equation D.3:

Ta = Tm Te = 2Hs = 2H
(D.5)
where is change in slip (Sm ).

It is the design objective of a PSS to create an extra electrical damping torque component
using the exciter such that it is in phase with . The relationship governing , EXC(s)
and PSS(s)(with as the input signal), together with the terminal voltage Vt and exciter
output Ef d , is shown in Figure D.4, which can be described by:
Ef d = VT (s)EXC(s)Vt + P SS(s)EXC(s)
(D.6)
96
where VT (s), EXC(s) and P SS(s) are transfer functions of voltage transducer, Exciter and
PSS respectively. Considering 1.0 machine model the component of Te can be obtained
from the equations given in Chapter 2:
0
0
Te = Eq iq + iq [(xq x0d ) + (xd x0d )/(1 + sTdo
)]id + [iq /(1 + sTdo
)]Ef d
(D.7)
Substituting Equation D.6 into Equation D.7 to eliminate Ef d ,

Te = Tq + Td + Tv + Tpss
(D.8)
where
Tq = Eq iq
Vt
(D.9)
0
Td = iq [(xq x0d ) + (xd x0d )/(1 + sTdo
)]id
(D.10)
0
Tv = VT (s)EXC(s)[iq /(1 + sTdo
)]Vt
(D.11)
0
Tpss = P SS(s)EXC(s)[iq /(1 + sTdo
)]
(D.12)
V ref
VT (s)
PSS(s)
EXC(s)
E fd
E pss
Figure D.4: Relationship governing VT (s), P SS(s), and EXC(s)
Therefore Te can be considered to be consisting of four where the component Tpss is

directly contributed by the PSS. PSS(s) does not appear explicitly in the other components.
In accordance with Equation D.12, for a speed input PSS, the phase shift of Tpss /
can be accounted for as shown in Figure D.5. For perfect compensation , according to the
97
more conventional interpretation, Tpss should be in phase with and the overall phase
shift condition should be
pss + exc 90 = 0 ( input)
(D.13)
where pss and exc are the phase shifts of PSS() and EXC() respectively. For a powerinput PSS, to make Tpss in phase with will require an additional 90 phase shift, and
the phase shift condition with respect to Tpss /Te should then be
pss + exc 90 = 90
(D.14)
pss + exc 180 = 0 (Pe input)
(D.15)
that is
If the acute angle sum in Equation D.14 or D.15 happens to be other than 0 , say , Tpss
will lead by .
pss
exc
PSS(s)
EXC(s)
90o
iq
T pss
1 + s Tdo
Figure D.5: Phase shift of Tpss for a speed input PSS
In the above , the equations describing the electrical torque composition, Equations D.8
to D.12, and the conditions for perfect PSS compensation, Equations D.13 and D.15, are
both are established on the basis of more realistic parameters as compared to the case of
SMIB system.
On the basis of above the lead lag components of PSS can be obtained. Gain of the PSS
can be selected by root locus technique given in the case of SMIB system.
Appendix E
Mapping From a Binary String to a
Real Number
Consider an optimization problem having n decision variables and domain of variable xj
(j=1,2...n) is [xmin
, xmax
]. Let, the required precision is k places after the decimal point.
j
j
The precision requirement implies that the range of domain of each variable should be
divided into at least (xmax
xmin
) 10k size ranges. The required bits (denoted with mj )
j
j
for a variable is calculated as follows:
2mj < (xmax
xmin
) 10k 2mj 1
j
j
(E.1)
The mapping from a binary string to a real number for variable xj can be computed as:
xj =
xmin
j
+ decimal(substringj )
!
max
xj xmin
j
2mj 1
(E.2)
where decimal(substringj ) represents the decimal value of substringj for decision variable
xj .
98
Appendix F
Derivation of Equation 4.1
Consider the Figure F.1. As shown, the equation of lines ABO and CDO in x y plane
20
15
yaxis
10
Slope
|dy/dx|=m
10
15
20
5
xaxis
Figure F.1: D-contour in x y plane

can be written as:
|y| = m.x
x = m1 |y|
x+
1
|y|
m
99
=0
(F.1)
Appendix F. Derivation of Equation 4.1
100
dy
where m is the +ve slope of the line (i.e. m = | dx
|). Now equation of line BC can be
written as:
x=
x=0
(F.2)
Combining Equations F.1 and F.2, the equation of D-contour ABCD can be written as:
x min{
Defining
1
m
1
|y|, } = 0
m
(F.3)
= , and considering the case of complex plane, Equation F.3 can be written
as:
f (z) = Re(z) min[|Im(z)|, ] = 0
where z = x + iy (i.e. x = Re(z) & y = Im(z)).
(F.4)
References
[1] F. P. Demello and C. Concordia , Concepts of Synchronous Machine Stability as Effected by Excitation Control, IEEE Transactions on Power Apparatus and Systems,
Vol. PAS-88, No. 4, pp. 316-329, April 1969.
[2] A. Klofenstein, Experience with System Stabilizing Excitation Controls on the Generation of the Southern California Edition Company, IEEE Transactions on PAS, Vol.
90, No. 2, pp. 698-706, March/April 1971.
[3] IEEE Working Group Annotated Bibliography on Power System Stability Controls
1986-1994 , IEEE Transactions on Power Systems, Vol. 88, No. 4, pp. 794-804, May
1996.
[4] K. Bollinger, A. Laha, R. Hamilton and T. Harras, Power System Stabilizer Design
Using Root-Locus Methods, IEEE Transaction on PAS, Vol. 94, No. 5, pp. 1484-1488,
Sep/Oct. 1975.
[5] F.P. Demello, P.J. Nolan, T.F. Laskowski, and J.M. Undrill, Coordinated Application
of Stabilizers in Multi-machine power systems, IEEE Transactions on PAS, Vol. 99,
pp. 892-901, 1980.
[6] R.V. Larsen and D.A. Swann, Applying Power System Stabilizers, I, II and III, IEEE
Transactions on PAS, Vol. 100, No.6, pp. 3017-3046, June 1981.
[7] J.S. Czuba, L.N. Hannett and J.R. Willis, Implementation of Power System Stabilizer
at the Ludington Pumped Storage Plant, IEEE Transactions on Power Systems, Vol.
PWRS-1, pp. 121-128, Feb. 1986.
101
References
102
[8] C.T. Tse, S.K. Tso, Approach to the Study of Small-perturbation Stability of Multimachine Systems , IEE Proceedings, Generation, Transmission and Distribution, Vol.
135, No. 5, pp. 396-405, 1988.
[9] C. Chen, Y. Hsu, Coordinated Synthesis of Multimachine Power System Stabilizers
Using an Efficient Decentralized Modal Control Algorithm, IEEE Transactions on
Power Systems, Vol. 2, No. 3, pp. 543-551, Aug. 1987.
[10] C.T. Tse, S.K. Tso, Refinement of Conventional PSS Design in Multimachine System
by Modal Analysis, IEEE Transactions on Power Systems, Vol. 8, No. 2, pp. 598-605,
May 1993.
[11] Yao-Nan Yu and Qing-Hua Li, Pole Placement Power System Stabilizers Design of an
Unstable Nine-Machine System, IEEE Transactions on PAS, Vol. 5, No. 2, May 1990.
[12] Y.N. Yu, K.Vongsuriya, and I.N. Wedman, Application of an Optimal Control Theory
to a Power System, IEEE Transactions on PAS, Vol. 89, No. 1, pp 55-61, 1970.
[13] Y. Lee and C. Wu , Damping of Power System Oscillations With Output Feedback
and Strip Eigenvalue assignment , IEEE Transactions on Power Systems, Vol. 10, No.
3, pp 1620-1626, Aug 1995.
[14] M.R. Khaldi, A.K. Sarkar, K.Y. Lee, and Y.M. Park , The Modal Performance Measure for Parameter Optimization of Power System Stabilizers, IEEE Transactions on
Energy Conversion, Vol. 8, No. 4, pp 660-666, Dec 1993.
[15] J.B. Simo, I. Kamwa, G. Trudel, and S. Tahan, Validation of a New Modal Performance Measure for Flexible Controllers Design, IEEE Transactions on Power Systems,
Vol. 11, No. 2, pp 819-826, May 1996.
[16] A.J. Urdaneta, N.J. Bacalao, B. Feijoo, L. Flores, and R. Diaz, Tuning of Power System Stabilizers Using Optimization Techniques, IEEE Transactions on Power Systems,
Vol. 6, No. 1, pp 127-134, Feb. 1991.
[17] Jan Lunze,Robust Multivariable Feedback Control, Prentice Hall International, U.K.,
1989.
References
103
[18] D.A. Pierre, A Perspective on Adaptive Control of Power Systems, IEEE Transactions
on PWRS, Vol. 2, No. 2, pp. 387-396, May 1987.
[19] A. Ghosh, G. Ledwich, O.P. Malik and G.S. Hope, Power System Stabilizer Based on
Adaptive Control Techniques, IEEE Transactions on PAS, Vol. 103, pp. 1983-1989,
March/April 1984.
[20] O.P. Malik, G.S. Hope and Ramanujan, Real Time Model Reference Adaptive Control
of Synchronous Machine Excitation, IEEE PES Winter Meeting Paper, 178, 297-304,
1976.
[21] A. Ghandakly and P. Idowu, Design of a Model Reference Adaptive Stabilizer for the
Exciter and Governor Loops of Power Generators, IEEE Transactions on PAS, Vol. 5,
pp. 887-893, March/April 1990.
[22] Shi-Jie Cheng, O.P. Malik and G.S. Hope, Self Tuning Stabilizer for a Multimachine
Power System, IEE Proceedings, Part C, No. 4, pp. 176-185, 1986.
[23] A. Chandra, O.P. Malik and G.S. Hope A self-Tuning Controller for the Control of
Multimahine Power Systems, IEEE Transactions on PAS, Vol. 3, No. 3, pp. 1065-1071,
Aug. 1988.
[24] N.C. Pahalawaththa, G.S. Hope and O.P. Malik, Multivariable Self-Tuning Power System Stabilizer Simulation and Implementation Studies, IEEE Transactions on Energy
Conversion, Vol. EC-6, No. 2, pp. 310-316, June 1991.
[25] F. Ghosh and Indraneel Sen, Design and Performance of a Local Gain Scheduling Power
System Stabilizer for Interconnected System, TENCON-91, International Conference
on EC3-Energy, Computer, Communication and Control Systems, Vol. 1, pp. 355-359,
Aug. 1991.
[26] D.P. Sen Gupta, N.G. Narhari, I. Boyd and B.W. Hogg, An Adaptive Power System
Stabilizer which Cancels the Negative Damping Torque of a Synchronous Generator,
Proceedings of IEE, Part. C, pp. 109-117, 1985.
[27] G. Bandyopadhyay and S.S. Prabhu, A New Approach to Adaptive Power System
Stabilizers , Electrical Machines and Power System, 14, pp. 111-125, 1988.
References
104
[28] T. Hiyama, Application of Rule-Based Stabilizing Controller to Electrical Power System, IEE Proceedings, Part C, Vol. 136, No. 3, pp. 175-181, 1989.
[29] T. Hiyama, Rule-Based Stabilizer for Multi-Machine Power System, IEEE Transactions on PWRS, Vol. 5, No. 2, pp. 403-411, May 1990.
[30] Y. Zhang, G.P. Chen, O.P. Malik, and G.S. Hope, An Artificial Neural Network Based
Adaptive Power System Stabilizer, IEEE Transactions on Energy Conversion, Vol. 8,
No. 1, pp. 71-77, 1993.
[31] Y. Zhang, O.P. Malik, G.S. Hope and G.P. Chen, Application of an Inverse InputOutput Mapped ANN as a PSS, IEEE Transactions on Energy Conversion, Vol. 9,
No. 3, pp. 433-441, 1994.
[32] T. Hiyama and C.M. Lim, Application of Fuzzy Logic Control Scheme for Stability
Enhancement of a Power System, IFAC International Symposium on Power Systems
and Power Plant Control, Korea, pp. 313-316, Aug. 1989.
[33] Y. Hsu, C. Cheng, Design of Fuzzy Power System Stabilizer for Multi-Machine Power
Systems, IEE Proceedings, Part C, Vol. 137, No. 3, pp. 233-238, May 1990.
[34] M.A.M. Hasan, O.P. Malik and G.S. Hope, A fuzzy Logic Based Stabilizer for a Synchronous Machine, IEEE Transactions on Energy Conversion, Vol. 6, No. 3, pp. 407413, September 1991.
[35] M. Braae and D.A. Rutherford, Selection of Parameters for a Fuzzy Logic Controller,
Fuzzy Sets and Systems, Vol. 2, No. 3, pp. 185-199.
[36] K.A. El-Metwally and O.P. Mallik, Parameter Tuning for a Fuzzy Logic Controller,
Proceedings of the IFAC Twelfth World Congress on Automatic Control, Sydney, Australia, Vol. 2, pp. 581-584, July 18-23, 1993.
[37] A.K. Singh,A Novel Fuzzy Logic Based Power System Stabilizer, M.Sc.(Engg.) Thesis,
Indian Institute of Science, Bangalore , Feb. 2003.
[38] J. Doyle, and G. Stein, Multivariable Feedback Design : Concepts for a Classical/Modern Synthesis, IEEE Transactions on Automatic Control, Vol. 26, No. 1, pp.
4-16, Feb 1981.
References
105
[39] Bruce A. Francis, A Course in H Control Theory , Lecture Notes in Control and
Information Sciences, No. 88, Spring-Verglas New York, 1987.
[40] J.C. Doyle, K.Glover, P.P. Khargonerkar and B.A. Francis, State Space Solutions to
Standard H 2 and H Control Problems, IEEE Transactions on Automatic Control,
Vol. 34, No. 8, pp. 831-847, Aug 1989.
[41] K. Ohtsuka, T. Taniguchi, T. Sato, S. Yokokawa, and Y. Ueki, An H Optimal TheoryBased Generator Control System, IEEE Transactions on Energy Conversion, Vol.
7, pp. 108115, 1992.
[42] Huibert Kwakernaak, Robust Control and H Optimization - Tutorial Paper, Systems and Control Letters, Vol. 29, No. 2, pp. 255-273, Jan. 1993.
[43] S. Chen, and O.P. Malik, H Optimization-based Power System Stabilizer Design,
IEE Proceedings, Generation, Transmission and Distribution, Vol. 142, No. 2, pp.
179184, March 1995.
[44] R. Ashgharian, and S.A. Tavakoli, A Systematic Approach to Performance Weights
Selection in Design of Robust H PSS Using Genetic Algorithms, IEEE Transactions
on Energy Conversion, Vol. 11, No. 1, pp. 111-117, March 1996.
[45] S.S. Ahmed, L. Chen, and A. Petroianu, Design of Suboptimal H Excitation Controllers, IEEE Transactions on PWRS, Vol. 11, pp. 312318 1996.
[46] T.C. Yang, Applying H Optimization Method to Power System Stabilizer Design,
Part 1 : Single Machine Infinite Bus Systems, International Journal of Electrical Power
and Energy Systems, Vol. 19, No. 1, pp. 29-35, Jan. 1997.
[47] F.A. Komla, N. Yorino, and H. Sasaki, Design of H PSS Using Numerator Denominator Uncertainty Representation, IEEE Transactions on Energy Conversion, Vol. 12,
No. 1, pp. 45-50, March 1997.
[48] T.C. Yang, Applying H Optimization Method to Power System Stabilizer Design,
Part 2 : Multimachine Power Systems, International Journal of Electrical Power and
Energy Systems, Vol. 19, No. 1, pp. 37-43, Jan. 1997.
References
106
[49] J. Doyle, Analysis of Feedback Systems with Structured Uncertainties , IEEE Proceedings, Part D, Vol. 129, No. 6, pp. 242-250, Nov. 1982.
[50] S. Chen, and O.P. Malik, Power System Stabilizer Design Using Synthesis, IEEE
Transactions on Energy Conversion, Vol. 10, pp. 175181, March 1995.
[51] M.A. Dahleh and M.H. Khammash, Controller Design for Plants with Structured Uncertainty, Automatica, Vol. 129, No. 1, pp. 37-56, 1993.
[52] I. Horowitz, Quantitative Feedback Theory, IEE Proceedings, Part D, Vol. 129, No.
6, pp. 215-226, Nov. 1982.
[53] D.D. Siljak, Parameter Space Methods for Robust Control Design : A Guided Tour,
IEEE Transactions on Automatic Control, Vol. 34, No. 7, pp. 674-688, July 1989.
[54] V.L. Kartinov, Asymptotic Stability of an Equilibrium Position of a Family of Systems
of Linear Differential Equation , Diffenti nye Uraveniya 14:99 1483-1485, 1978.
[55] J.H. Chow, M.A. Kale, H.A. Othman, J.J. Sanchez and G.E. Terwillinger, Robust
Control Design of a Power System Stabilizer Using Multivariable Frequency Domain
Techniques, Proceedings of the 29th Conference on Decision and Control Honolulu,
Hawaii , Vol. 4, pp. 2067-2073, Dec. 1990.
[56] J. Doyle and A. Packard, The Complex Structured Singular Value, Automatica, Vol.
129, No. 1, pp. 71-109, 1993.
[57] M.J. Gibbard, Robust Design of Fixed Parameter Power System Stabilizers Over a
Wide Range of Operating Conditions, IEEE Transactions on Power Systems, Vol. 6,
No. 2 pp. 794800, May 1991.
[58] F. Fatehi, J.R. Smith, and D.A. Pierrie, Robust Power System Controller Design
Based on Measured Models, IEEE Transactions on Power Systems, Vol. 11, No. 2 pp.
774780, May 1996.
[59] M.H. Khammash, V. Vittal, and C.D. Pawloski, Analysis of Control Performance for
Stability Robustness of Power Systems, IEEE Transactions on Power Systems, Vol.
11, No. 1 pp. 18611867, Feb. 1994.
References
107
[60] S. Venkataraman, M.H. Khammash, and V. Vittal, Analysis and Synthesis of HVDC
Controls for Robust Stability of Power Systems, IEEE Transactions on Power Systems,
Vol. 10, No. 4 pp. 19331938, May 1996.
[61] M.A. Pai, C.D. Vournas, A.N. Michel, and H. Ye, Applications of Interval Matrices in
Power System Stabilizer Design, International Journal of Electrical Power and Energy
Systems, Vol. 19, No. 3, pp. 179-184, March 1997.
[62] M. Werner, P. Korba, and T.C. Yang, Robust Tuning of Power System Stabilizer Using
LMI-Techniques, IEEE Transactions on Control Systems Technology, Vol. 11, No. 1,
pp. 147-152, Jan 2003.
[63] P.S. Rao,On The Design of Robust Power System Damping Controllers , Ph.D Thesis,
Indian Institute of Science, Bangalore , May 1998.
[64] P. Srikant Rao and Indraneel Sen, A QFT Based Robust SVC Controller for Improving
the Dynamic Stability of Power System, Electric Power Systems Research, Vol. 46, No.
3, pp. 213-220, Sept. 1998.
[65] P. Srikant Rao and Indraneel Sen, Robust Tuning of Power System Stabilizers Using
QFT, IEEE Transactions on Control Systems Technology, Vol. 7, No. 4, pp. 478-486,
July 1999.
[66] P. Srikant Rao and Indraneel Sen, Robust pole placement stabilizer design using linear
matrix inequalities , IEEE Transactions on Power Systems, Vol. 15, No. 1, pp. 313-319,
Feb. 2000.
[67] A.E. Hammad, Analysis of Power System Stability Enhancement by Static VAR Compensators, IEEE Transactions on PWRS, Vol. 1, No. 4, pp. 222-227, 1986.
[68] E.V. Larsen, J.J. Sanchez Gasca, and J.H. Chow, Concepts of Design of FACTS Controllers to Damp Power Swings, IEEE Transactions on PWRS, Vol. 10, No. 2, pp.
948-956, May 1995.
[69] A. Magid, Bettayah and M.M. Dawound, Simultaneous Stabilization of Power System
Using Genetic Algorithms, IEE Proceedings, Generation, Transmission and Distribution, Vol. 144, No. 1, pp. 39-44, Jan. 1997.
References
108
[70] M.A. Abido, and Y.L. Abdel-Magid, Hybridizing Rule-Based Power System Stabilizers
With Genetic Algorithms, IEEE Transactions on Power Systems, Vol. 14, No. 2, pp.
600-607, May 1999.
[71] Y.L. Abdel-Magid, M.A. Abido, S.A. Baiyat, A.H. Mantawy, Simultaneous Stabilizeation of Multimachine Power Stystems via Genetic Algoriths, IEEE Transactions
on Power Systems, Vol. 14, No. 4, pp. 1428-1439, Nov 1999.
[72] A.L.B. do Bomfim, G.N. Taranto, and D.M. Falcao,Simultaneous Tuning of Power
System Damping Controllers Using Genetic Algorithms, IEEE Transactions on Power
Systems, Vol. 15, No. 1, pp. 163-169, Feb 2000.
[73] P. Zhang, and A.H. Coonick,Coordinated Synthesis of PSS Parameters in MultiMachine Power Systems Using the Method of Inequalities Applied to Genetic Algorithms,IEEE Transactions on Power Systems, Vol. 15, No. 2, pp. 811-816, May 2000.
[74] A. Andreoiu, and K. Bhattacharya,Robust Tuning of Power System Stabilizers Using a Lyapunov Method Based Genetic Algorithm, IEE Proceedings on Generation
Transmission and Distribution, Vol. 149, No. 5, pp. 585-592, Sep. 2002.
[75] Y.L. Abdel-Magid, and M.A. Abido, Optimal Multiobjective Design of Robust Power
System Stabilizers Using Genetic Algorithms, IEEE Transactions on Power Systems,
Vol. 18, No. 3, pp. 1125-1132, Aug. 2003.
[76] G.N. Taranto, and D.M. Falcao,Robust Decentralised Control Design Using Genetic Algorithms in Power Systems Damping Control, IEE Proceedings on Generation, Transmission and Distribution, Vol. 145, No. 1, pp. 1-6, Jan. 1998.
[77] J.H. Holland,Adaptation in Natural and Artificial Systems, MIT Press, 1975.
[78] L. Davis, Genetic Algorithms and Simulated Annealing, Pitman, 1987.
[79] Goldberg, Genetic Algorithm In search, Optimization and Machine Learning, Addison
Wesley Reading, 1989.
[80] L. Davis, Handbook of Genetic Algorithms, Van Nostrand Reinhold, 1991.
[81] M. Mitchell, An Introduction to Genetic Algorithms, Prentice Hall, India , 2002.
References
109
[82] D. Beasley, D.R. Bull, and R.R. Martin, An Overview of Genetic Algorithms: Part 1,
Fundamentals, University Computing, Vol. 15, No. 2, pp. 58-69, 1993.
[83] M. Gen, and R. Cheng, Genetic Algorithms & Engineering Design, John Wiley &
Sons, Inc, 1997.
[84] K.R. Padiyar, Power System Dynamics, Stability and Control, Second Edition, Hyderabad, B.S. Publication, 2002.
[85] IEEE Task Force, Current Usage and Suggested Practices in Power System Stability
Simulations for Synchronous Machines, IEEE Transactions on Energy Conversion, Vol.
1, No. 1, pp. 77-93, 1986.
[86] P.M. Anderson and A.A. Faud, Power System Control and Stability, Iowa State University Press, 1977.
[87] P.W. Sauer and M.A. Pai, Power System Dynamics and Stability, Pearson Education
Asia, 2002.

Power System Protection

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Power System Protection

Uploaded by

Copyright:

Available Formats

A Novel Approach for Tuning of Power

System Stabilizer Using Genetic

in the Faculty of Engineering

Department of Electrical Engineering

Low Frequency Oscillations in Power System . . . . . . . . . . . . . . . . . .

Fixed Parameter Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . .

Other Fixed Parameter Controllers . . . . . . . . . . . . . . . . . . .

The Drawbacks of Conventional Fixed Parameter Controllers . . . . .

Fuzzy Logic Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Application of Genetic Algorithms to PSS Design . . . . . . . . . . . . . . .

Robust PSS design using Genetic Algorithms: the present approach . . . . .

Performance Requirements of Power System Damping Controllers . . . . . .

How Much Damping Do We Need? . . . . . . . . . . . . . . . . . . .

Performance Evaluation of a PSS . . . . . . . . . . . . . . . . . . . .

Scope of Present Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1.10 Organization of Chapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 Mathematical Modelling of Power System

SMIB Model in Non-Linear Form . . . . . . . . . . . . . . . . . . . . . . . .

Excitation System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

SMIB Test System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Modelling of Multimachine System . . . . . . . . . . . . . . . . . . . . . . .

Inclusion of Generator Stator in the Network . . . . . . . . . . . . . .

Network Equations for Multimachine . . . . . . . . . . . . . . . . . .

Multimachine Test System . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Linearized 1.1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 Genetic Algorithm: An Overview

What is Genetic Algorithm? . . . . . . . . . . . . . . . . . . . . . . . . . . .

Implementation of genetic algorithm . . . . . . . . . . . . . . . . . . . . . .

Mathematical Model of SGAs . . . . . . . . . . . . . . . . . . . . . . . . . .

4 Proposed Stabilization Technique: Single Machine System

Application to SMIB System . . . . . . . . . . . . . . . . . . . . . . . . . . .

Performance Analysis of Proposed GA Based PSS . . . . . . . . . . .

Robustness Test and Eigen Value Plots . . . . . . . . . . . . . . . . .

Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .

5 Proposed Stabilization Technique: Multimachine System

Performance Evaluation of the Stabilizer in Multimachine System . . . . . .

Robustness Test and Eigen Value Plots . . . . . . . . . . . . . . . . .

Operating Points For Simulation Studies . . . . . . . . . . . . . . . .

Simulation Results and Discussion . . . . . . . . . . . . . . . . . . . .

A Calculation of Initial Conditions

B Heffron-Philips Model of the SMIB System

C Data for SMIB and Multimachine System

D Tuning Guidelines for the CPSS

E Mapping From a Binary String to a Real Number

F Derivation of Equation 4.1

Control Parameters Bounds . . . . . . . . . . . . . . . . . . . . . . . . .

Initial and Final Values of PSS Parameters . . . . . . . . . . . . . . .

Control Parameters Bounds . . . . . . . . . . . . . . . . . . . . . . . . .

GA Parameters For Multimachine Case . . . . . . . . . . . . . . . . .

Loading Range of 3 Machine, 9 Bus System

Optimal stabilizer parameters of PGAPSS . . . . . . . . . . . . . . . .

Operating points of generators on a 100 MVA base . . . . . . . . . .

C.1 Generator Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

C.2 AVR Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

C.3 Generator Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

C.4 AVR Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

External two port network . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Excitation system block diagram. . . . . . . . . . . . . . . . . . . . . . . . .

Block diagram of PSS

Single machine infinite bus system . . . . . . . . . . . . . . . . . . . . . . . .

Schematic of a multimachine system

Generator equivalent circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . .

3 machine, 9 bus power system model, single line diagram. . . . . . . . . . .

Single point crossover operation . . . . . . . . . . . . . . . . . . . . . . . . .