Вы находитесь на странице: 1из 159

Robust Nondynamic Multirate

Output Feedback Technique based


Power System Stabilizers
submitted in partial fulllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
by
Rajeev Gupta
(Roll No. : 00423301)
Supervisor(s)
Prof. B.Bandyopadhyay
Prof. A.M.Kulkarni
Interdisciplinary Programme in
Systems and Control Engineering
Indian Institute of Technology, Bombay
Mumbai-400076
2003
APPROVAL SHEET
Thesis entitled Robust Nondynamic Multirate Output Feedback Technique
based Power System Stabilizers by RAJEEV GUPTA is approved for the degree of
Doctor of Philosophy.
Examiners

Supervisor(s)

Chairman

Date:
Place
INDIAN INSTITUTE OF TECHNOLOGY, BOMBAY, INDIA
CERTIFICATE OF COURSE WORK
This is to certify that Mr. RAJEEV GUPTA was admitted to the candidacy of the
Ph.D. Degree on 20
th
July 2000, after successfully completing all the courses required for
the Ph.D. programme. The details of the course work done are given below.
Sr. No. Course No. Course Name Credits
1. SC 601 Modeling of Dynamic Systems 6.00
2. SC 621 Quantitative Feedback Theory I 6.00
3 SCS 801 Credit Seminar 4.00
I.I.T Bombay Dy. Registrar (Academic)
Dated:
To my parents, my wife Bharti
and
daughters Rupal and Disha
Abstract
Power System Stabilizers (PSS) are added to excitation systems, to enhance the damping
of electric power system during low frequency oscillations. For large scale power systems
comprising of many interconnected machines, the power system stabilizer (PSS) parameter
tuning is a complex exercise due to the presence of several poorly damped modes of
oscillation. The problem is further complicated by continuous variation in power system
operating conditions.
In the simultaneous tuning approach, exhaustive computational tools are required
to obtain optimal parameter settings for the PSS, while in case of sequential tuning, al-
though the computational burden is lesser, evaluating the tuning sequence is an additional
requirement. There is a further problem of eigenvalue drift.
Based on a recent convex programming algorithm for simultaneous stabilization by
linear state feedback, periodic output feedback and fast output sampling feedback are the
two types of multirate techniques for stabilizing a family of systems, when either a si-
multaneous stabilizing state feedback gain or a simultaneous stabilizing output injection
matrix exist, and complete state information is not available. In the periodic output
feedback technique the rate of applying control input signal is higher than the output
sampling rate, whereas in fast output sampling feedback technique the output signal is
sampled at a faster rate than that of the control input applied. These controllers provide
superior alternative to observer based state feedback control. Nondynamic feedback of
output samples allows a particularly exible design.
This thesis proposes the applications of periodic output feedback and fast output
sampling feedback techniques for designing robust power system stabilizers for a single
machine innite bus system. The powerful properties of these methods are used to op-
timize the PSS parameter settings. The nonlinear model of a machine is linearized at
dierent operating points and linear models are obtained. In periodic output feedback
method, for each of these linear models, an output injection gain is obtained using DLQR
technique. A robust periodic output feedback gain which realizes these output injection
gains is obtained using LMI approach. This robust periodic output control is applied to
the nonlinear model of the single machine at dierent operating (equilibrium) points.
In fast output sampling feedback method, for all of these linear models, a single
stabilizing state feedback gain is obtained. A robust fast output sampling feedback gain
which realizes this state feedback gain is obtained using LMI approach. This method does
not require the state of the system for feedback and is easily implementable. This robust
fast output sampling control is applied to nonlinear model of a single machine at dierent
operating (equilibrium) points.
These methods are further extended for designing centralized and decentralized con-
troller for multimachine power system. Design of robust decentralized power system stabi-
lizers for multimachine power system via reduced order model by using both the multirate
output feedback methods have been also proposed. These methods give very good results
for the design of Power System Stabilizers for single machine and multimachine power
system. Investigations reveal that the classical approach does provide satisfactory per-
formances for operating conditions up to the nominal but gives deteriorating responses
when the load increases. Also the classically tuned PSS fails to stabilize the system at
some operating conditions.
Keywords: Periodic output feedback, fast output sampling feedback , power systems,
small-signal stability, power system stabilizers design, single machine power system, mul-
timachine power system, decentralized control, nonlinear simulation, robust control.
Contents
1 INTRODUCTION 1
1.1 Contributions of this thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2 Review of literature 11
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Power system stabilizers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.1 Basic concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Performance objectives . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Classical stabilizer implementation procedure . . . . . . . . . . . . 13
2.3 Modeling of Single machine system . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1 Small signal analysis with block diagram representation . . . . . . . 14
2.3.2 Rotor-mechanical equations and torque-angle loop . . . . . . . . . . 15
2.3.3 Representation of ux decay . . . . . . . . . . . . . . . . . . . . . . 15
2.3.4 Representation of Excitation System . . . . . . . . . . . . . . . . . 16
2.3.5 System representation . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3.6 Computation of Heron-Philips constants for lossless
network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.7 State space model of single machine system . . . . . . . . . . . . . 18
2.4 Multimachine power system analysis . . . . . . . . . . . . . . . . . . . . . 18
2.4.1 State space model of multimachine system
(Machine model 1.0) . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.4.2 State space model of multimachine system
(machine model 1.1) . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Review on periodic output feedback . . . . . . . . . . . . . . . . . . . . . . 25
2.6 Review on fast output sampling feedback . . . . . . . . . . . . . . . . . . . 27
i
3 Design of PSS for Single Machine Innite Bus System 31
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Periodic output feedback technique for multimodel system . . . . . . . . . 32
3.3 Case study I: Robust PSS design by periodic
output feedback technique . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.4 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5 Fast output sampling feedback technique for
multimodel system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.6 Case Study II: Robust PSS design by fast output sampling technique . . . 42
3.7 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 43
3.8 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4 Design of PSS for Multimachine Power System 51
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Periodic output feedback technique . . . . . . . . . . . . . . . . . . . . . . 52
4.3 Case study I: Centralized PSS design for
multimachine power system (4 machine 10 bus
system) by periodic output feedback technique . . . . . . . . . . . . . . . . 53
4.4 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 56
4.5 Decentralized periodic output feedback technique for multimodel system . 57
4.6 Robust decentralized PSS design for
multimachine power system by periodic output
feedback technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.6.1 Case study II: 4 machine 10 bus system . . . . . . . . . . . . . . . . 59
4.6.2 Case study III: 10 machine 39 bus system . . . . . . . . . . . . . . 62
4.7 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 67
4.7.1 4 Machine 10 bus system . . . . . . . . . . . . . . . . . . . . . . . . 67
4.7.2 10 Machine 39 bus system . . . . . . . . . . . . . . . . . . . . . . . 77
4.8 Fast output sampling feedback technique . . . . . . . . . . . . . . . . . . . 82
4.9 Case study IV: Centralized PSS design for
multimachine power system (4 machine 10 bus)
by fast output sampling feedback technique . . . . . . . . . . . . . . . . . 82
4.10 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 84
4.11 Decentralized fast output sampling feedback
technique for multimodel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
ii
4.12 Robust decentralized PSS design for
multimachine power system by fast output
sampling feedback technique . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.12.1 Case study V: 4 machine 10 Bus system . . . . . . . . . . . . . . . 87
4.12.2 Case study VI: 10 machine 39 bus system . . . . . . . . . . . . . . 88
4.13 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 89
4.13.1 4 Machine 10 bus system . . . . . . . . . . . . . . . . . . . . . . . 89
4.13.2 10 Machine 39 bus system . . . . . . . . . . . . . . . . . . . . . . . 94
4.14 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5 Design of PSS for Multimachine Power System via Reduced order Model101
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.2 Model order reduction based on dominant modes retention . . . . . . . . . 102
5.2.1 Davison technique . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.3 Decentralized periodic output feedback control for multimodel system via
reduced order model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.4 Case study I: Robust decentralized PSS design for multimachine power
system by periodic output feedback technique via reduced order model . . 106
5.5 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 108
5.6 Decentralized fast output sampling feedback
technique via reduced order model for multimodel
system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
5.7 Case study II: Robust decentralized PSS design for multimachine power
system by fast output sampling feedback technique via reduced order model 114
5.8 Simulation with nonlinear model . . . . . . . . . . . . . . . . . . . . . . . . 116
5.9 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
6 Conclusions and Future work 123
6.1 Salient features of the present work . . . . . . . . . . . . . . . . . . . . . . 124
6.2 Scope of future work in this area . . . . . . . . . . . . . . . . . . . . . . . 125
iii
List of Figures
2.1 Block diagram of PSS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Single machine connected to an innite bus system . . . . . . . . . . . . . 14
2.3 Excitation system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Block diagram of single machine innite bus system . . . . . . . . . . . . . 17
2.5 Block diagram of 4 machine and 10 bus System . . . . . . . . . . . . . . . 20
2.6 Block diagram of 10 machine and 39 bus System . . . . . . . . . . . . . . . 21
2.7 Block diagram including excitation system and PSS . . . . . . . . . . . . . 24
2.8 Fast output sampling feedback method . . . . . . . . . . . . . . . . . . . . 28
3.1 Open and closed loop impulse responses for linearized system for various
operating conditions using periodic output feedback controller . . . . . . . 36
3.2 Open and closed loop impulse responses for linearized system for various
operating conditions using periodic output feedback controller . . . . . . . 36
3.3 Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller . . . . . . . . . . . . 37
3.4 Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller . . . . . . . . . . . . 37
3.5 Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller . . . . . . . . . . . . . 38
3.6 Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller . . . . . . . . . . . . . 38
3.7 Open and closed loop impulse responses for nonlinear system for various
operating conditions using fast output sampling feedback controller . . . . 44
3.8 Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller . . . . . . . . . . 45
3.9 Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller . . . . . . . . . . 45
iv
3.10 Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller . . . . . . . . . 46
3.11 Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller . . . . . . . . . 46
3.12 Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller . . . . . . . . . . 47
3.13 Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller . . . . . . . . . . 47
3.14 Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller . . . . . . . . . . 48
4.1 Open and closed loop responses using centralized periodic output feedback
controller(4 machine and 10 bus system) . . . . . . . . . . . . . . . . . . . 56
4.2 Open and closed loop responses with fault at bus 3 using centralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 57
4.3 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 69
4.4 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 69
4.5 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 70
4.6 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 70
4.7 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 71
4.8 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 71
4.9 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 72
4.10 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(4 machine and 10 bus system) . . . . . . . 72
4.11 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
v
4.12 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.13 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.14 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.15 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.16 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.17 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.18 Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback controller(4 machine and 10
bus system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.19 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(10 machine and 39 bus system) . . . . . . 78
4.20 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(10 machine and 39 bus system) . . . . . . 78
4.21 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(10 machine and 39 bus system) . . . . . . 79
4.22 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(10 machine and 39 bus system) . . . . . . 79
4.23 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(10 machine and 39 bus system) . . . . . . 80
4.24 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(10 machine and 39 bus system) . . . . . . 80
vi
4.25 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller(10 machine and 39 bus system) . . . . . . 81
4.26 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller (10 machine and 39 bus system) . . . . . . 81
4.27 Open and closed loop responses using centralized fast output sampling
feedback controller(4 machine and 10 bus system) . . . . . . . . . . . . . . 84
4.28 Open and closed loop responses with fault at bus 3 using centralized fast
output sampling feedback controller(4 machine and 10 bus system) . . . . 85
4.29 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 90
4.30 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 90
4.31 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 91
4.32 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 91
4.33 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 92
4.34 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 92
4.35 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 93
4.36 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system) . . . . . . . . 93
4.37 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 94
4.38 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 95
4.39 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 95
4.40 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 96
4.41 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 96
vii
4.42 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 97
4.43 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 97
4.44 Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system) . . . . . . . . 98
5.1 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 109
5.2 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 109
5.3 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 110
5.4 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 110
5.5 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 111
5.6 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 111
5.7 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 112
5.8 Open and closed loop responses with fault using robust decentralized peri-
odic output feedback controller via reduced order model . . . . . . . . . . 112
5.9 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 116
5.10 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 117
5.11 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 117
5.12 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 118
5.13 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 118
5.14 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 119
viii
5.15 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 119
5.16 Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model . . . . . . . . . . . 120
ix
List of Tables
3.1 Model parameters used in periodic output feedback method . . . . . . . . 35
3.2 Model parameters used in fast output sampling feedback Method . . . . . 42
4.1 Machine data: 4 machine and 10 bus system . . . . . . . . . . . . . . . . . 53
4.2 Line data: 4 machine and 10 bus system . . . . . . . . . . . . . . . . . . . 54
4.3 Load ow data: 4 machine and 10 bus system . . . . . . . . . . . . . . . . 54
4.4 Eigen values of the closed loop system (
N
+KC) . . . . . . . . . . . . 55
4.5 Model parameter variations: 4 machine and 10 bus system . . . . . . . . . 60
4.6 Machine data: 10 machine and 39 bus system . . . . . . . . . . . . . . . . 62
4.7 Load ow data I: 10 machine and 39 bus system . . . . . . . . . . . . . . 63
4.8 Load ow data II: 10 machine and 39 bus system . . . . . . . . . . . . . . 64
4.9 Transformer data: 10 machine and 39 bus system . . . . . . . . . . . . . . 64
4.10 Line data I: 10 machine and 39 bus system . . . . . . . . . . . . . . . . . 65
4.11 Line data II: 10 machine and 39 bus system . . . . . . . . . . . . . . . . . 66
4.12 Location of fault: 4 machine and 10 bus system . . . . . . . . . . . . . . . 68
4.13 Location of fault: 10 machine and 39 bus system . . . . . . . . . . . . . . 77
4.14 Eigen values of the closed loop system (
N
+ LC) . . . . . . . . . . . . . 83
x
List of Symbols
A state (model) matrix of the system
B control input matrix
D damping coecient
P active power
Q reactive power
Q state weighting matrix
R control weighting matrix
H inertia constant
I
d
, I
q
direct and quadrature components of armature current
M inertia coecient
rotor angle
angular speed
x
d
, x
q
synchronous reactances in d and q axes, respectively
x

d
direct axis transient reactance
x

q
quadrature axis reactance
E
fd
equivalent excitation voltage (eld circuit voltage)
E

q
internal voltage behind transient reactance x

d
E

d
internal voltage behind transient reactance x

q
f frequency
K
E
, T
E
AVR gain and time constant, respectively
T
w
washout time constant
T
m
, T
e
mechanical and electrical torques, respectively
T

d0
open circuit transient time constant
T

q0
open circuit subtransient time constant
V
d
, V
q
direct and quadrature components of terminal voltage
Y admittance matrix
s Laplace operator
xi
t time
u stabilizing signal
T transpose
, sampling rates for discrete system

state (model) matrix of the discrete system sampled at

control input matrix of the discrete system sampled at


state (model) matrix of the discrete system sampled at
control input matrix of the discrete system sampled at
C output matrix
C ctitious measurement matrix
C
0
measurement matrix of lifted system
D
0
transmission matrix of lifted system
F state feedback gain
G output injection gain
K periodic output feedback gain
L fast output sampling feedback gain
K
1
K
6
Heron-Philips constants
K
s
classical stabilizer gain
T
1
T
4
classical stabilizer time constants
damping factor
C
1
C
8
linearizing constants
V
t
generator terminal voltage
S
m
slip
V
s
correction voltage
V
ref
reference voltage
x state vectors
y output vectors
xii
spectral radius
N no of gain sequences
controllability or observability index
T transformation matrix
x
e
external reactance
R
e
external resistance
X
T
transformer reactance
R
T
transformer resistance
R
s
series resistance
X
s
series reactance
B
c
shunt reactance
B
l
shunt susceptance
angle of terminal voltage
Z transformed state (plant) matrix of the system
P, V transformation matrix
S stabilizing matrix for reduced model
xiii
List of Acronyms
ac Alternating current
AVR Automatic Voltage Regulator
dc Direct current
Diag Diagonal element
DLQR Discrete Linear Quadratic Regulator
FILT(s) Filter(s)
FOS Fast Output Sampling Feedback
HVDC High Voltage Direct Current
LHP Left Hand Poles
LMI Linear Matrix Inequalities
LQR Linear Quadratic Regulator
LTI Linear Time Invariant
MIMO Multi Input Multi Output
POF Periodic Output Feedback
PSSs Power System Stabilizer(s)
SISO Single Input Single Output
SMIB Single Machine Innite Bus
TGR Transient Gain Reduction
xiv
Chapter 1
INTRODUCTION
In the late 1950s most of the new generating units added to the electric utility systems
were equipped with continuously acting voltage regulators. As these units became a larger
percentage of generating capacity, it became apparent that the voltage regulator action
had a detrimental impact upon the dynamical stability (small signal stability) of the
power system. Oscillations of small magnitude and low frequency often persist for long
periods of time and in some cases it can cause limitations on the power transfer capability.
Power system stabilizers (PSSs) were developed to aid in damping these oscillations via
modulation of the generator excitation. The art and science of applying power system
stabilizers has been developed over the past thirty ve years, since the rst widespread
application to the Western systems of the United States. The development has evolved
the use of various tuning techniques and input signals and learning to deal with turbine
generator shaft torsional mode of vibrations [1].
Power System Stabilizers are added to excitation systems to enhance the damping
of electric power system during low frequency oscillations. Several methods are used for
the design of PSSs. Tuning of supplementary excitation controls for stabilizing system
modes of oscillation has been the subject of much research during the past 35 years.
Two basic tuning techniques have been successfully utilized with power system stabilizer
applications: phase compensation method and the root locus method. A commonly used
approach is based around the conventional PSS structure which is composed of a wash
out circuit and a cascade of two-phase lead networks. A number of PSS input signals,
such as terminal voltage, rotor speed, accelerating power, electric power etc., and linear
combinations of these have been extensively investigated and recommendations regarding
their use have been reported in the literature. Phase compensation consists of adjusting
the stabilizer to compensate for the phase lags through the generator, excitation system
1
and power system, such that, the stabilizer path provides torque changes which are in
phase with the speed changes [2] - [9]. This is the most straightforward approach, easily
understood and implemented in the eld and is most widely used. The design of such
PSSs requires the determination(or tuning) of few parameters for each machine viz. the
overall dc gain, the wash out circuit time constant, and the various constants for the two-
lead networks. A number of sequential and simultaneous approaches for tuning of these
parameters have been reported in literature[10] - [15]. Although the above approaches
have been used and have produced satisfactory results regarding the damping of local
modes of oscillation, their outcome may not be considered the best possible. This is
because of the restrictive assumptions made and the intuitive nature of the design process
[16].
Synthesis by root locus involves the shifting of eigenvalues associated with power sys-
tem modes of oscillation, by adjusting the stabilizer pole and zero locations in the s-plane
[17]. This approach gives an additional insight to the performance, by working directly
with the closed-loop characteristics of the systems, as opposed to the open loop nature of
the phase compensation technique. But it is more complicated to apply, particularly in
the eld. More over, the performance of these stabilizers considerably degrades with the
changes in the operating condition during normal operation. It is also known that for a
multimachine system, eigenvalue assignment is often too involved and complex for simul-
taneous stabilization of multivariable systems and may not provide satisfactory results for
sequential multivariable systems applied as SISO systems. Not much attempts have been
made for designing the power system stabilizers for multimachine power system using
multivariable control theory. The complexity stems from the fact that insucient degree
of freedom is available to the designer in assigning eigenstructure by xed gain output
feedback method. Moreover, even if a sucient degree is available or a dynamic output
feedback stabilizer is sought, numerical problems often arise regarding the solution of sets
of high dimensional nonlinear algebraic equations, for which a solution may or may not
exist. It is also well known that, in application of multivariable Nyquist array methods
to multimachine power system, many diculties arrive for the attainment of necessary
diagonal dominance condition [18].
Since eigenvalue assignment and Nyquist array approaches have proved to be cumber-
some, modern control methods have been used by several researchers to take advantage
of the optimal control techniques. These methods utilize a state space representation of
power system model and calculate a gain matrix which when applied as a state feedback
2
control will minimize a prescribed objective function. Successful application of the opti-
mal control to power system stabilizers requires that the constraints imposed by power
system nonlinearities be used eectively and that a limited number of feedback signals be
included [19]. As a result, reports have appeared in the literature concerning the applica-
tion of LQR theory for the design of power system stabilizers(PSSs). First LQR theory
was applied to single machine innite bus(SMIB) system. Later, this was extended to
multimachine case. It is quite easy to realize that neither of these approaches can be suc-
cessful while dealing with real-life power systems which, in general, may have thousands
of state variables [20]- [24]. The reason is that all the states may not be available for mea-
surement or may be dicult to measure. In this case, the optimal control law requires
to design the state observer. This increases the implementation cost and reduces the
reliability of control system. Another disadvantage of the observer based control system
is that, even a slight variations of the model parameters from their nominal values may
result into signicant degradation of the closed loop performance. Hence, it is desirable
to go for an output feedback design method.
In recent years there have been several attempts at designing power system stabilizer
using H

based robust control techniques [25]- [28]. In this approach, the uncertainty in
the chosen system is modeled in terms of bounds on frequency response. A H

optimal
controller is then synthesized which guarantees robust stability of the closed loop sys-
tem. Other performance specications such as disturbance attenuation criteria are also
imposed on the system. However, it should be noted that the main objective of using a
PSS is to provide a good transient behaviour. Guaranteed robust stability of the closed
loop, though necessary, is not adequate as a specication in this application. In addi-
tion to this, the problem of the poorly damped pole-zero cancellations and the choice of
weighting functions used in design, limit the usefulness of this technique for PSS design.
H

design, being essentially a frequency domain techinique does not provide much con-
trol over transient behavior and closed loop pole location. It would be more desirable to
have a robust stabilizer which, in addition guarantees an acceptable level of small signal
transient performance. Moreover, this will lead to dynamic output feedback, which may
be feasible but leads to a higher order feedback system [29].
The static output feedback problem is one of the most investigated problems in con-
trol theory. The complete pole assignment and guaranteed closed loop stability is still
not obtained by using static output feedback [30]- [32]. Another approach to pole place-
ment problem is to consider the potential of time-varying periodic output feedback. It
3
was shown by Chammas and Leondes [33] that a controllable and observable plant was
discrete time pole assignable by periodically time-varying piecewise constant output feed-
back. Since the feedback gains are piecewise constant, their method could be easily
implemented and indicate a new possibility. Such a control law can stabilize a much
larger class of systems than the static output feedback [34]-[40]. In this thesis periodic
output feedback method has been used to design robust power system stabilizer for single
machine innite bus system. This method has also been extended in this thesis to design
the power system stabilizers for multimachine power systems. In periodic output feed-
back technique, the gain matrix of centralized power system stabilizers for multimachine
power system is generally full [36]. This results in the control input of each machine
being a function of outputs of all machines. Also centralized periodic output feedback
PSSs require transmission of signals among the generating units. This requirement in
itself no longer constitutes a problem from practical and technical view points. This is
due to the rapid advancement in optical ber communication and their adoption by the
power utilities. However, if a completely decentralized PSS can be found, so that no sig-
nicant deterioration in the system performance is experienced as compared to the state
and centralized periodic output feedback based schemes, then such a scheme would be
more advantageous, in terms of practicability and reliability. In such schemes, not only is
the cost of implementation drastically reduced, but also, the risk of loss of stability due to
signal transmission failure is minimized [41]- [44]. Due to the geographically distributed
nature of power systems, a decentralized control scheme may be more feasible than a cen-
tralized control scheme. In the decentralized power system stabilizer, the control input
for each machine should be a function of the output of that machine alone. This can be
achieved by designing a decentralized PSS using periodic output feedback technique in
which the gain matrix should have all o-diagonal terms zero or are very small compare
to the diagonal terms. In a decentralized PSS, to activate the proposed controller at same
instant, a proper synchronization signal is required to be sent to all machines. All PSSs
can be applied simultaneously to the respective machines. So the decentralized stabilizer
design problem can be translated into a problem of diagonal gain matrix design for mul-
timachine power system. This thesis proposes the design of a decentralized robust power
system stabilizer for multimachine system using periodic output feedback.
For large power systems, the order of the state matrix may be quite large. It would
be dicult to work with these complex systems in their original form [45]. In particular
the computation of the output injection gain which is needed to obtain the decentralized
4
periodic output feedback based power system stabilizer becomes very tedious for large
power systems. One of the ways to overcome this diculty is to develop a reduced order
model for large power systems. Then an output injection gain can be computed from
the reduced model of the power system, and using the aggregation techniques, an output
injection gain can be obtained for the higher order (actual) model. The decentralized
periodic output feedback gain which realizes this output injection gain can be obtained
for the actual model. This thesis proposes the design of a decentralized robust power
system stabilizer for multimachine system using periodic output feedback via reduced
order model.
Another approach for solving pole placement problem is to consider the potential
of time-varying fast output sampling feedback. With fast output sampling approach
proposed by Werner and Furuta [38], it is generically possible to simultaneously realize
a given state feedback gain for a family of linear, observable models. This approach
requires to increase the low rank of the measurement matrix of an associated discretized
system, which can be achieved by sampling the output several times during one input
sampling interval, and constructing the control signal from these output samples. Since
the feedback gains are piecewise constant, their method could be easily implemented and
indicated a new possibility. Such a control law can stabilize a much larger class of systems
than the static output feedback [34]-[40]. As in periodic output feedback technique, the
fast output sampling feedback method has been used here to design robust power system
stabilizer for single machine innite bus system. This method has also been extended to
design of centralized and robust decentralized power system stabilizers for multimachine
power system.
The decentralized power system stabilizers based on fast output sampling method has
been also designed via reduced order model in this thesis. In this case a state feedback gain
from the reduced model of the power system is obtained and then, using the aggregation
techniques, a state feedback gain is obtained for the higher order (actual) model. The
decentralized fast output sampling feedback gain which realizes this state feedback gain
can be obtained for the actual model.
1.1 Contributions of this thesis
The main contributions of this thesis are summarized as follows:
A new application is presented to design a robust power system stabilizer for a
5
single machine connected to an innite bus system using periodic output feedback.
In this method slip signal is taken as output and the rate of applying control input
signal is higher than the output sampling rate. This method is more general than
static output feedback and the control input required for these models are of smaller
magnitudes. It is found that the designed robust controller provides good damping
enhancement for various operating points of a single machine connected to an innite
bus system. The system dynamic performances with PSS tuned using the proposed
techniques are satisfactory for dierent load conditions and system congurations.
This method succeeds in achieving a robust tuned PSS parameter set.
Another new application is presented to design of a robust power system stabilizer
for a single machine connected to an innite bus system using fast output sampling
feedback. In this method also slip signal is taken as output and the rate of applying
control input signal is slower than the output sampling rate. This method is also
more general than static output feedback. It is found that the designed robust con-
troller provides good damping enhancement for various operating points of a single
machine connected to an innite bus system. The system dynamic performances
with PSS tuned using the proposed techniques are satisfactory for dierent load
conditions and system congurations. This method also succeeds in achieving a
robust tuned PSS parameter set.
The design of centralized power system stabilizers for a multimachine power system
using periodic output feedback is presented. In this method, the periodic output
feedback gain matrix is full, which makes control input as a function of output of all
machines. Power system stabilizers for all machines can be applied simultaneously
to all machines. It is found that the designed controller provides a good damping
enhancement for a multimachine power system.
A new algorithm is presented to design robust decentralized power system stabilizers
for a multimachine power system using periodic output feedback. In this method,
control input to each machine becomes a function of the output of that machine only.
This can be achieved by designing robust decentralized PSS using periodic output
feedback technique in which the gain matrix should have all o-diagonal terms zero
or very small as compared to the diagonal terms. It is found that the designed robust
controller provides good damping enhancement for various operating points of a
multimachine power system. In the case of delay in synchronization, the proposed
6
controller can still damp out the oscillations, but it deteriorates the performance.
This method is more advantageous, in terms of practicability and reliability.
Another new application is presented for designing of centralized power system
stabilizers for a multimachine power system using fast output sampling feedback.
As in the case of periodic output feedback technique here also control input is a
function of output of all machines. It is found that the designed controller provides
a good damping enhancement for a multimachine power system.
A new algorithm is presented to design robust decentralized power system stabilizers
for a multimachine power system using fast output sampling feedback. Like in the
periodic output feedback method, here also control input to each machine becomes
a function of the output of that machine only. This can be achieved by designing
robust decentralized PSS using fast output sampling feedback technique in which
the gain matrix should have all o-diagonal terms zero or very small as compared
to the diagonal terms.
A new algorithm is presented to design robust decentralized power system stabilizers
for a multimachine power system using periodic output feedback via reduced order
model. In particular, the computation of the output injection gain, which is needed
to obtain the decentralized periodic output feedback based power system stabilizer,
becomes very tedious for large power system. One of the ways to overcome this
diculty is to develop a reduced order model for large power system. The output
injection gain is then computed from the reduced model of the power system, and
using the aggregation techniques, an output injection gain can be obtained for the
higher order (actual) model. The decentralized periodic output feedback gain which
realizes this output injection gain can be obtained for the actual model. It is found
that the designed robust controller via reduced order model provides good damping
enhancement for various operating points of a multimachine power system.
In a similar way, a new algorithm is presented for the design of robust decentralized
power system stabilizers for a multimachine power system using fast output sampling
feedback via reduced order model. The computation of the state feedback gain,
which is needed to obtain the decentralized fast output sampling feedback based
power system stabilizer, becomes very tedious for large power system. Here a state
feedback gain is computed from the reduced model of the power system and using
the aggregation techniques, a state feedback gain can be obtained for the higher
7
order (actual) model. The robust decentralized fast output sampling feedback gain
which realizes this state feedback gain, can be obtained for the actual model.
1.2 Outline of the thesis
This thesis examines the application of periodic output feedback method and fast output
sampling feedback method in order to design a globally decentralized PSSs that will
ensure a stable and robust operation of a single machine connected to an innite bus and
multimachine power system, for each operating point within a wide range.
Chapter 1 introduces the problem of small-signal stability in power systems, with em-
phasis on the low frequency oscillation phenomena occurring due to small disturbances
and its mitigation by means of PSS. It also presents a review of literature which dis-
cusses the relevant work in this area of tuning of PSS and lays down the motivations and
objectives of the work.
Chapter 2 presents classical power system stabilizer in brief. It also presents the
small-signal stability models of a single machine connected to an innite bus (SMIB)
and multimachine power system. The mathematical formulations have been detailed for
the interested readers. The periodic output feedback method and fast output sampling
feedback method are introduced as the method chosen to perform the simultaneous tuning
of PSS parameters. The machine models have been formulated in state-space form and
their open-loop (uncontrolled mode) characteristics have been examined.
Chapter 3 contains the design of robust power system stabilizer using periodic output
feedback control method and fast output sampling feedback control method for single
machine connected to an innite bus system. The system is represented in a discrete
state-space form and the inuence of the sampling time on the PSS parameter tuning is
investigated. These methods ensure that for any operating condition within a pre-dened
domain, the system remains stable when subjected to small perturbations. The system
dynamic performances with PSS tuned using the proposed techniques are satisfactory for
dierent load conditions and system congurations. These methods succeed in achieving
a robust simultaneously tuned PSS parameter set.
Chapter 4 presents the design of centralized and robust decentralized power system
stabilizers using periodic output feedback control method and fast output sampling feed-
back control method for multimachine power system. Here also, the system is represented
in a discrete state-space form and the inuence of the sampling time on the PSS parameter
8
tuning is investigated. These methods also ensure that for any operating condition within
a pre-dened domain, the system remains stable when subjected to small perturbations.
The system dynamic performances with PSS tuned using the proposed techniques are also
satisfactory for dierent load conditions and system congurations. These methods also
succeed in achieving a robust simultaneously tuned PSS parameter set.
Chapter 5 contains the design of power system stabilizer of multimachine power system
via reduced model using periodic output feedback and fast output sampling feedback
methods. These have been carried out by computing the output injection in case of
periodic output feedback and state feedback gain in case of fast output sampling feedback
from the reduced model of multimachine power system.
Chapter 6 highlights the signicant contributions of the present work and draws the
scope for future work in this area.
9
.
10
Chapter 2
Review of literature
2.1 Introduction
Over the last four decades, a considerable amount of research have been done in the
area of PSS. Research has been directed towards obtaining such a PSS that can provide
an optimal performance for a wide range of machines and system parameters. Various
control strategies and optimization techniques have found their applications in this area
as also various degrees of system modeling have been attempted. The review of required
literature for the present work is discussed in this chapter. This chapter is organized as
follows.
Section 2.2 presents classical power system stabilizer in brief. Section 2.3 and 2.4
present the small-signal stability models of single machine connected to an innite bus
(SMIB) and multimachine power system. The mathematical formulations have been de-
tailed for the interested readers. The models have been formulated in state-space form and
their open-loop (uncontrolled mode) characteristics have been examined. The periodic
output feedback method and fast output sampling feedback method are introduced in the
Section 2.5 and 2.6, as these methods are chosen to perform the simultaneous tuning of
PSS parameters.
2.2 Power system stabilizers
Small-signal stability is the ability of the power system to maintain synchronous operation
when subjected to small disturbances. Since the disturbance is considered to be small, the
equations that describe the resulting dynamics of the system may be linearized. Instability
that may result can be of two types:
11
a) steady increase in generator rotor angle due to lack of synchronizing torque;
b) rotor oscillations of increasing amplitude due to lack of sucient damping torque.
In todays practical power systems, the small-signal stability problem is usually one of
insucient damping of system oscillations.
For the analysis of small-signal stability, linearized models are generally considered to be
adequate for representation of the power system and its various components.
2.2.1 Basic concept
The basic function of a power system stabilizer is to extend the stability limits by modu-
lating generator excitation, to provide damping to the oscillation of synchronous machine
rotors relative to one another. The oscillations of concern typically occur in the frequency
range of approximately 0.2 to 3.0 Hz, and insucient damping of these oscillations may
limit the ability to transmit the power. To provide damping, the stabilizer must produce
a component of motor slip which is in phase with reference voltage variations. For input
signal, the transfer function of the stabilizer must compensate for the gain and phase of
excitation system, the generator and the power system, which collectively determine the
transfer function from the stabilizer output to the component of mechanical speed. This
can be modulated via excitation system [1].
2.2.2 Performance objectives
Power system stabilizers can extend power transfer stability limits which are characterized
by lightly damped or spontaneously growing oscillations in the 0.2 to 3.0 Hz frequency
range. This is accomplished via excitation control to contribute damping to the system
modes of oscillations. Consequently, it is the stabilizers ability to enhance damping under
the least stable conditions, i.e., the performance conditions, which is important. Addi-
tional damping is primarily required under the conditions of weak transmission and heavy
load as occurs, for example, when attempting to transmit the power over long transmission
lines from the remote generating plants or relatively weak tie between systems. Contin-
gencies, such as line outage, often precipitate such conditions. Hence, system normally
having adequate damping can often benet from stabilizers during such conditions.
12
T(s) FILT(s)
LIMITER COMPENSATOR
DYNAMIC
FILTER
TORSIONAL
WASHOUT
u
u
sT
w
1+s T
w
V
s
Figure 2.1: Block diagram of PSS
2.2.3 Classical stabilizer implementation procedure
It is important to realize that the stabilizer is intended to provide damping for small
excursions about a steady-state operating point, and not to enhance transient stability,
i.e., the ability to recover from a severe disturbance. In fact, the stabilizer will often
have deleterious eect on transient stability by attempting to pull the generator eld out
of ceiling too early in response to a fault. The stabilizer output is generally limited to
prevent serious impact on transient stability, but stabilizer tuning also has a signicant
impact upon the system performance following a large disturbance [2].
The block diagram used in industry is shown in Fig. 2.1. It consists of a washout
circuit, dynamic compensator, torsional lter and limiter. The washout circuit is provided
to eliminate steady-state bias in the output of PSS which will modify the generator
terminal voltage. The PSS is expected to respond only to transient variations in the
input signal (rotor slip) and not to the dc oset in the signal. The washout circuit acts
essentially as a high pass lter and it must pass all frequencies that are of interest.
Implementation of a power system stabilizer implies adjustment of its frequency char-
acteristic and gain to produce the desired damping of the system oscillations in the fre-
quency range of 0.2 to 3.0 Hz. The transfer function of a generic power system stabilizer
having washout circuit and a dynamic compensator may be expressed as
H(s) = K
s
T
w
s (1 +sT
1
) (1 +sT
3
)
(1 +T
w
s) (1 +sT
2
) (1 +sT
4
)
, (2.1)
where K
s
represents stabilizer gain.
The stabilizer frequency characteristic is adjusted by varying the time constant T
w
, T
1
,
T
2
, T
3
and T
4
. The output of PSS must be limited to prevent the PSS acting to counter
the action of AVR.
A number of sequential and simultaneous approaches for the tuning of the these pa-
13
G
R
X
E
e
e
b
Figure 2.2: Single machine connected to an innite bus system
rameters have been reported in literature. Although these approaches have been used
and produce satisfactory results regarding the damping of local modes of oscillation, their
outcome may not be the optimal. This is due to the restrictive assumption made and the
intuitive nature of the design process.
A power system stabilizer can be made more eective if it is designed and applied with
the knowledge of associated power system characteristics. Power system stabilizer must
provide adequate damping for a range of frequencies of the power system oscillation modes.
To begin with, simple analytical models, such as that of a single machine connected to an
innite bus system, can be useful in determining the frequencies of local mode oscillations.
Power system stabilizer should also be designed to provide stable operation for the weak
power system conditions and associated loading. A designed stabilizer must ensure for the
robust performance and satisfactory operation with an external system reactance ranging
from 20% to 80% on the unit rating [46].
2.3 Modeling of Single machine system
2.3.1 Small signal analysis with block diagram representation
Consider a single machine system as shown in Fig. 2.2. For simplicity, we assume a
synchronous machine represented by model 1.0 neglecting damper windings both in the d
and q axes. Also, the armature resistance of the machine is neglected and the excitation
system is represented by a single time-constant system as shown in Fig. 2.3 [47].
The algebraic equations of stator are
E

q
+x

d
i
d
= v
q
, (2.2)
x
q
i
q
= v
d
. (2.3)
14
V
V
E
K
1 + s T
fd
t
s
E
E
+
+

Figure 2.3: Excitation system


Linearizing Eqns.(2.2-2.3) as discussed in [47], the following equations are obtained
v
q
= x

d
C
1
+
_
1 +x

d
C
2
_
E

q
, (2.4)
v
d
= x
q
C
3
+x
q
C
4
E

q
. (2.5)
2.3.2 Rotor-mechanical equations and torque-angle loop
The rotor mechanical equations are
d
dt
=
B
(S
m
S
mo
) , (2.6)
2H
dS
m
dt
= DS
m
+T
m
T
e
, (2.7)
T
e
=
_
E

qo
i
q

_
x
q
x

d
_
i
d
i

q
_
. (2.8)
Linearizing Eqns. (2.6-2.8) and by taking Laplace transform, we obtain
=

B
s
S
m
=

B
s
, (2.9)
S
m
=
1
2Hs
[T
m
T
e
DS
m
] . (2.10)
2.3.3 Representation of ux decay
The equation for the eld winding can be expressed as
T

do
dE

q
dt
= E
fd
E

q
+
_
x
d
x

d
__
C
1
+C
2
E

q
_
, (2.11)
Taking Laplace transform of Eqn.(2.11), we get
_
1 +sT

do
K
3
_
E

q
= K
3
E
fd
K
3
K
4
. (2.12)
15
2.3.4 Representation of Excitation System
The block diagram of the excitation system is as shown in Fig. 2.3. The linearized
equations of the excitation system can be represented by omitting the limiter. For present
analysis we can ignore the auxiliary signal V
s
. The perturbation in the terminal voltage
V
t
can be expressed as
V
t
=
v
do
v
t0
v
d
+
v
qo
v
t0
v
q
, (2.13)
V
t
= K
5
+K
6
E

q
. (2.14)
The coecient K
1
to K
6
are termed as Heron-Philips constant. They are dependent
on machine parameters and operating conditions.
2.3.5 System representation
The state-space representation is concerned not only with input and output properties,
but also with its complete internal behavior. In contrast, the transfer function represen-
tation species only the input/output behavior. If state-space representation of a system
is known, the transfer function is uniquely dened. In this sense, the state-space repre-
sentation is a more complete description of the system. The block diagram of the system,
consisting of the representation of the rotor swing equation, ux decay and excitation
system is shown in Fig. 2.4. Here the damping term in the swing equation is neglected
for convenience.
The electrical torque component T
e2
is related to by the following relation.
From the Fig. 2.4, we have
T
e2
(s) =
K
2
K
4
T

do
F(s),
where
F(s) =
_
_
s +
1
T
E
_
1 +
K
5
K
E
K
4
_
s
2
+s
_
1
T
E
+
1
K
3
T

do
_
+ (1 +K
3
K
6
K
E
) /K
3
T

do
T
E
_
_
(s). (2.15)
For static exciter, T
E
is very small. If T
E
0,
T
e2

can be approximated as
T
e2


K
2
K
4
T

do
_
1 +
K
5
K
E
K
4
_
_
s + (1 +K
3
K
6
K
E
) /K
3
T

do
. (2.16)
For large value of K
E
Eqn.(2.16) can be further approximated as
T
e2


K
2
K
5
K
E
_
T

do
s +K
6
K
E
_ =
K
2
K
5
K
6
T

do
s/ (K
6
K
E
) + 1
. (2.17)
16
+

+
_
+
_
1
2Hs
B

K
K
K
K
1 + s T
K
1 + s T K
K
K

T
T

6
3 d0
5
E
q
2
4
1
e1
e2

v
S
m
5
ref
fd
E
Figure 2.4: Block diagram of single machine innite bus system
The slip S
m
of machine is taken as output. The transfer function of the system is
approximated as
S
m
=
1
2Hs
_
T
m

_
K
2
K
5
K
6
T

do
s/ (K
6
K
E
) + 1
+K
1
_

_
, (2.18)
taking T
m
small,
S
m
=
1
2Hs
_
K
2
K
5
K
6
T

do
s/ (K
6
K
E
) + 1
+K
1
_
.
2.3.6 Computation of Heron-Philips constants for lossless
network
For R
e
= 0, the expressions for the constants K
1
to K
6
are simplied. As the armature
resistance is already neglected, this refers to a lossless network on the stator side. The
expressions are given below
K
1
=
E
b
E
qo
cos
0
(x
q
+x
e
)
+
_
x
q
x

d
_
_
x

d
+x
e
_E
b
i
qo
sin
0
,
K
2
=
(x
q
+x
e
)
_
x

d
+x
e
_i
qo
=
E
b
sin
0
_
x

d
+x
e
_,
17
K
3
=
_
x

d
+x
e
_
(x
d
+x
e
)
,
K
4
=
_
x
d
x

d
_
_
x

d
+x
e
_E
b
sin
0
,
K
5
=
x
q
v
do
E
b
cos
0
(x
q
+x
e
) v
to

d
v
qo
E
b
sin
0
_
x

d
+x
e
_
v
to
,
K
6
=
x
e
_
x

d
+x
e
_
_
v
qo
v
to
_
.
2.3.7 State space model of single machine system
From the block diagram as shown in Fig. 2.4, the following state space representation of
the entire system can be derived using Heron-Phillips model [47, 48, 49] as
.
x= [A] x + [B] (V
ref
+V
s
) , (2.19)
where
x
T
=
_
S
m
E

q
E
fd
_
, (2.20)
[A] =
_

_
0
B
0 0

K
1
2H

D
2H

K
2
2H
0

K
4
T

do
0
1
T

do
K
3

1
T

do

K
E
K
5
T
E
0
K
E
K
6
T
E

1
T
E
_

_
, (2.21)
[B]

=
_
0 0 0
K
E
T
E
_
, (2.22)
[C] =
_
0 1 0 0
_
. (2.23)
The damping term D is included in the swing equation. The eigenvalues of the ma-
trix should lie in LHP of the s-plane for the system to be stable. The eect of various
parameters (for example K
E
& T
E
) can be examined from eigenvalue analysis. It is to be
noted that the elements of [A] are dependent on the operating condition.
2.4 Multimachine power system analysis
Analysis of practical power system involves the simultaneous solution of equations con-
sisting of synchronous machines and the associated excitation system and prime movers,
interconnecting transmission network, static and dynamic load (motor loads), and other
18
devices such as HVDC converters, static var compensators. The dynamics of the ma-
chine rotor circuits, excitation systems, prime mover and other devices are represented
by dierential equations. The result is that the complete system model consists of large
number of ordinary dierential and algebraic equations [50].
2.4.1 State space model of multimachine system
(Machine model 1.0)
Generator equations
The machine equations ( for k
th
machine ) are
pE

qk
=
1
T

d0k
[E

qk
+ (x
dk
x

dk
)i
dk
+E
fdk
], (2.24)
p
k
=
B
(S
mk
S
mk0
), (2.25)
pS
mk
=
1
2H
[D
k
(S
mk
S
mk0
) +P
mk
P
ek
. (2.26)
4 Machine and 10 bus power system
Model 1.0 is assumed for synchronous machines by neglecting the damper windings. In
addition, the following assumptions are made for simplicity [47].
1. The loads are represented by constant impedances.
2. Transients saliency is ignored by considering x
q
= x

d
.
3. Mechanical power is assumed to be constant.
4. E
fd
is single time constant AVR.
The state space model of a 4-machine 10 bus system as shown in Fig.2.5 can be
obtained using machine data, line data and load ow data as given in [47] as
.
x=
_
A

x +
_
B
_
V
ref
+V
s
_
, (2.27)
y = [C]x, (2.28)
where
x = [x
1,
x
2,
x
3
, x
4
]
T
, and y = [y
1,
y
2,
y
3
, y
4
]
T
.
x
k
(k=1,4) denotes the states of k
th
machine, and y
k
(k=1,4) denotes the output of the
k
th
machine.
19
5
6
9 10
7
8 3
2 1
To Load
AREA 1 AREA 2
4
Figure 2.5: Block diagram of 4 machine and 10 bus System
The elements (sub matrices of 4 4) of matrix A are depend on the machine and
network parameters. Thus we have
A =
_

_
a
11
a
12
a
13
a
14
a
21
a
22
a
23
a
24
a
31
a
32
a
33
a
34
a
41
a
42
a
43
a
44
_

_
, B =
_

_
B
1
0 0 0
0 B
2
0 0
0 0 B
3
0
0 0 0 B
4
_

_
, (2.29)
C =
_

_
C
1
0 0 0
0 C
2
0 0
0 0 C
3
0
0 0 0 C
4
_

_
,
_
V
ref
+V
s
_
=
_

_
(V
ref1
+V
s1
)
(V
ref2
+V
s2
)
(V
ref3
+V
s3
)
(V
ref4
+V
s4
)
_

_
.
10 Machine and 39 bus power system
The state space model of a 10-machine 39 bus system as shown in Fig. 2.6 can be obtained
using machine data, line data and load ow data as given in [47] as
.
x=
_
A

x +
_
B
_
V
ref
+V
s
_
, (2.30)
y = [C]x, (2.31)
20
26
27
35
8
8
11
12
10
10
2
2
13
14
25
38
17
18
7
7
15
19
4 5
37
16 32
1
1
3
31
34
3
4
20
30
33
5
39
6
9
21
24
28
29
36
22
23
6
9
W
W
W
W
W W
W
W
W
W
W
W
43
42
Figure 2.6: Block diagram of 10 machine and 39 bus System
21
where
x = [x
1,
x
2,
, . . . , x
10
]
T
, and y = [y
1,
y
2,
, . . . , y
10
]
T
.
x
k
(k=1,10) denotes the states of k
th
machine, and y
k
(k=1,10) denotes the output of the
k
th
machine.
The elements (sub matrices of 4 4) of A matrix depend on the machine and network
parameters. Thus we have
A =
_

_
a
00
a
01
a
02
a
03
a
04
a
05
a
06
a
07
a
08
a
09
a
10
a
11
a
12
a
13
a
14
a
15
a
16
a
17
a
18
a
19
a
20
a
21
a
22
a
23
a
24
a
25
a
26
a
27
a
28
a
29
a
30
a
31
a
32
a
33
a
34
a
35
a
36
a
37
a
38
a
39
a
40
a
41
a
42
a
43
a
44
a
45
a
46
a
47
a
48
a
49
a
50
a
51
a
52
a
53
a
54
a
55
a
56
a
57
a
58
a
59
a
60
a
61
a
62
a
63
a
64
a
65
a
66
a
67
a
68
a
69
a
70
a
71
a
72
a
73
a
74
a
75
a
76
a
77
a
78
a
79
a
80
a
81
a
82
a
83
a
84
a
85
a
86
a
87
a
88
a
89
a
90
a
91
a
92
a
93
a
94
a
95
a
96
a
97
a
98
a
99
_

_
, (2.32)
B =
_

_
B
1
0 0 0 0 0 0 0 0 0
0 B
2
0 0 0 0 0 0 0 0
0 0 B
3
0 0 0 0 0 0 0
0 0 0 B
4
0 0 0 0 0 0
0 0 0 0 B
5
0 0 0 0 0
0 0 0 0 0 B
6
0 0 0 0
0 0 0 0 0 0 B
7
0 0 0
0 0 0 0 0 0 0 B
8
0 0
0 0 0 0 0 0 0 0 B
9
0
0 0 0 0 0 0 0 0 0 B
10
_

_
, (2.33)
22
C =
_

_
C
1
0 0 0 0 0 0 0 0 0
0 C
2
0 0 0 0 0 0 0 0
0 0 C
3
0 0 0 0 0 0 0
0 0 0 C
4
0 0 0 0 0 0
0 0 0 0 C
5
0 0 0 0 0
0 0 0 0 0 C
6
0 0 0 0
0 0 0 0 0 0 C
7
0 0 0
0 0 0 0 0 0 0 C
8
0 0
0 0 0 0 0 0 0 0 C
9
0
0 0 0 0 0 0 0 0 0 C
10
_

_
. (2.34)
2.4.2 State space model of multimachine system
(machine model 1.1)
Generator equations
The machine equations ( for k
th
machine ) are
pE

qk
=
1
T

d0k
[E

qk
+ (x
dk
x

dk
)i
dk
+E
fdk
], (2.35)
pE

dk
=
1
T

q0k
[E

dk
(x
qk
x

qk
)i
qk
], (2.36)
p
k
=
B
(S
mk
S
mk0
), (2.37)
pS
mk
=
1
2H
[D
k
(S
mk
S
mk0
) +P
mk
P
ek
. (2.38)
Model 1.1 is assumed for synchronous machines by neglecting some damper windings.
In addition, the following assumptions are made for simplicity [47].
1. The loads are represented by constant impedances.
2. Transients saliency is ignored by considering x

q
= x

d
.
3. Mechanical power is assumed to be constant.
4. E
fd
is single time constant AVR.
4 Machine and 10 bus Power System
The state space model of a 4-machine 10 bus system as shown in Fig. 2.5 can be obtained
using machine data, line data and load ow data as given in [47] as
.
x=
_
A

x +
_
B
_
V
ref
+V
s
_
, (2.39)
23
PSS
AVR
Mech. System Network
Field
Winding
T
e
T
m
S
m
V
ref
V
s
E
fd
i
d
i
q
E
q

Figure 2.7: Block diagram including excitation system and PSS


y = [C]x, (2.40)
where
x = [x
1,
x
2,
x
3
, x
4
]
T
, and y = [y
1,
y
2,
y
3
, y
4
]
T
.
x
k
(k=1,4) denotes the states of k
th
machine, and y
k
(k=1,4) denotes the output of the
k
th
machine.
The elements (sub matrices of 5 5) of matrix A are depend on the machine and
network parameters. Thus we have
A =
_

_
a
11
a
12
a
13
a
14
a
21
a
22
a
23
a
24
a
31
a
32
a
33
a
34
a
41
a
42
a
43
a
44
_

_
, B =
_

_
B
1
0 0 0
0 B
2
0 0
0 0 B
3
0
0 0 0 B
4
_

_
,
C =
_

_
C
1
0 0 0
0 C
2
0 0
0 0 C
3
0
0 0 0 C
4
_

_
,
_
V
ref
+V
s
_
=
_

_
(V
ref1
+V
s1
)
(V
ref2
+V
s2
)
(V
ref3
+V
s3
)
(V
ref4
+V
s4
)
_

_
.
The block diagram including excitation system and power system stabilizers is given
in Fig. 2.7.
24
2.5 Review on periodic output feedback
The problem of pole assignment by piecewise constant output feedback was studied by
Chammas and Leondes [33] for linear time-invariant systems with infrequent observation.
They showed that, by use of periodically time-varying piecewise constant output feedback
gain, the poles of the discrete-time control system could be assigned arbitrarily (within the
natural restriction that they be located symmetrically with respect to real axis) [38],[40].
Consider a discrete time invariant system with sampling interval sec.
x(k + 1) =

x(k) +

u(k) , (2.41)
y (k) = Cx(k) , (2.42)
where x R
n
, u R
m
, y R
p
and

and C are constant matrices of appropriate


dimensions. The following control law is applied to this system. The output is measured
at the time instant t = k, k = 0,1, . We will consider constant hold function because
they are more suitable for implementation. An output sampling interval is divided into
N subintervals of length = /N, and the hold function is assumed constant on these
subintervals. Thus the control law becomes
u(t) = K
l
y (k) , (2.43)
k +l t k + (l + 1) , K
l+N
= K
l
, (2.44)
for l = 0, 1, , N 1.
Note that a sequence of N gain matrices {K
0,
K
1
, , K
N1
} when substituted in
Eqn.(2.44) generates a time-varying, piecewise constant output feedback gain K(t) for
0 t . Consider the following system obtained by sampling the system in Eqn.(2.42)
at sampling interval = /N and which is denoted by (, , C) :
x(k + 1) = x(k) + u(k) , (2.45)
y (k) = Cx(k) . (2.46)
A useful property of the control law in Eqn.(2.43) is given by the following lemma.
Assume (

, C) is observable and (, ) is controllable with controllability index


such that N , then it is possible to choose a gain sequence K

such that the closed


loop system, sampled over , takes desired self-conjugate set of eigenvalues.
Dene
K =
_
K
0
K
1
K
N1
_
T
, (2.47)
25
u(k) = Ky (k) =
_

_
u(k)
u(k +)
.
.
.
u(k + )
_

_
, (2.48)
then a state space representation for the system sampled over is
x(k +) =
N
x(k) +u, (2.49)
y (k) = Cx(k) , (2.50)
where
= [
N1
, , ].
Applying periodic output feedback in Eqn.(2.43), i.e., Ky (k) is substituted for
u(k) , the closed loop system becomes
x(k +) =
_

N
+KC
_
x(k) . (2.51)
The problem has now taken the form of static output feedback problem. Eqn.(2.51)
suggests that an output injection matrix G be found such that

N
+GC
_
< 1, (2.52)
where () denotes the spectral radius. By observability one can choose an output injection
gain G to achieve any desired self-conjugate set of eigenvalues for the closed loop matrix
_

N
+GC
_
, and from N it follows that one can nd a periodic output feedback gain
which realizes the output injection gain G by solving
K =G, (2.53)
for K.
The controller obtained from the above equation will give desired behavior, but might
require excessive control action. To reduce this eect we relax the condition that K
exactly satisfy the above linear equation and include a constraint on the gain K. Thus
we arrive at the following inequations
K <
1
, KG <
2
. (2.54)
_
_

2
1
I K
K
T
I
_
_
< 0,
_
_

2
2
I (KG)
(KG)
T
I
_
_
< 0. (2.55)
In this form, the LMI Toolbox of MATLAB can be used for synthesis [51].
26
2.6 Review on fast output sampling feedback
The problem of pole assignment by fast output sampling was studied by Werner and
Furuta [38]. It was shown that fast output sampling feedback technique has the features of
static output feedback and makes it possible to arbitrarily assign the system poles. Unlike
static output feedback, fast output sampling feedback always guarantees the stability of
the closed loop system. In fast output sampling, each sampling period is subdivided
into N subintervals = /N. N must be chosen greater than or equal to the observability
index. The last N output samples are measured at the time instant t = l, l = 0, 1,
and a constant control signal is applied over a period . The control signal is constructed
as a linear combination of the last output samples. Since the feedback gains are piecewise
constant, their method could easily be implemented and indicated a new possibility. The
method is explained briey in the following.
Consider a plant described by a linear model
.
x= Ax +Bu, (2.56)
y = Cx, (2.57)
with (A,B) controllable and (C,A) observable. Assume the plant is to be controlled by
a digital controller, with sampling time and zero order hold, and that a sampled data
state feedback design has been carried out to nd a state feedback gain F such that the
closed loop system
x(k +) = (

F)x(k) , (2.58)
has desired properties. Hence

= e
A
and

=
_

0
e
As
ds B. Instead of using a state
observer, the following sampled data control can be used to realize the eect of the state
feedback gain F by output feedback. Let = /N, and consider
u(t) = [L
0
, L
1
, , L
N1
]
_

_
y (k )
y (k +)
.
.
.
y (k )
_

_
= Ly
k
, (2.59)
for k t < (k + 1), where the matrix blocks L
j
represent output feedback gains, and
the notation L, y
k
has been introduced for convenience. Note that 1/ is the rate at which
the loop is closed, whereas output samples are taken at the N-times faster rate 1/. This
control law is illustrated in Fig. 2.8.
27
0 2
+
y(t)
u(t)
t
t
L L L
2 1 0
=

Figure 2.8: Fast output sampling feedback method


To show how a fast output sampling controller Eqn. (2.59) can be designed to realize
the given sampled-data state feedback gain, we construct a ctitious, lifted system for
which Eqn. (2.59) can be interpreted as static output feedback. Let ( , , C ) denote the
system Eqn. (2.58) at the rate 1/. Consider the discrete-time system having at time t
= k the input u
k
= u(k), state x
k
= x(k) and output y
k
as
x
k+1
=

x
k
+

u
k
, (2.60)
y
k+1
= C
0
x
k
+D
0
u
k
, (2.61)
where
C
0
=
_

_
C
C
.
.
.
C
N1
_

_
, D
0
=
_

_
0
C
.
.
.
C

N2
j=0

j

_
.
Assume that the state feedback gain F has been designed that (

F) has no
eigenvalues at the origin. Then, assuming that in intervals k t < k +
u(t) = Fx(k), (2.62)
28
one can dene the ctitious measurement matrix
C(F, N) = (C
0
+D
0
F)(

F)
1
, (2.63)
which satises the ctitious measurement equation y
k
= Cx
k
. For L to realize the eect
of F, it must satisfy
LC = F. (2.64)
Let denote the observability index of (, ). It can be shown that for N ,
generically C has full column rank, so that any state feedback gain can be realized by a
fast output sampling gain L.
The controller obtained from the above equation will give desired behavior, but might
require excessive control action. To reduce this eect we relax the condition that L exactly
satisfy the above linear equation and include a constraint on the gain L. Thus we arrive
at the following inequations
L <
1
,
LD
0
F

<
2
,
LCF <
3
. (2.65)
This can be formulated in the framework of Linear Matrix Inequalities as follows
_
_

2
1
I L
L
T
I
_
_
< 0,
_
_

2
2
I (LD
0
F

)
(LD
0
F

)
T
I
_
_
< 0,
_
_

2
3
I (LCF)
(LCF)
T
I
_
_
< 0. (2.66)
In this form, the LMI Tool Box MATLAB can be used for synthesis [51].
If the initial state is unknown, there will be an error u
k
= u
k
- Fx
k
in constructing
the control signal under state feedback. One can verify that the closed loop dynamics are
governed by
_
_
x
k+1
u
k+1
_
_
=
_
_

0 LD
0
F

_
_
_
_
x
k
u
k
_
_
. (2.67)
29
To see this, apply the coordinate transformation
T =
_
_
I 0
F I
_
_
, (2.68)
to the equation
_
_
x
k+1
u
k+1
_
_
=
_
_

LC
0
LD
0
_
_
_
_
x
k
u
k
_
_
, (2.69)
and use Eqn. (2.63). Thus, one can say that the eigenvalues of the closed loop system
under a fast output sampling control law Eqn. (2.59) are those of (

F) together
with those of (LD
0
F

).
One feature of fast output sampling control that makes it attractive for robust con-
troller design, is the fact that a result similar to the above can be shown to hold when the
same state feedback is applied simultaneously to a family of models representing dierent
operating conditions of the plant.
In the next chapter, the periodic output feedback technique and fast output sampling
feedback technique will be used to design a robust power system stabilizer for single
machine connected to an innite bus system.
30
Chapter 3
Design of PSS for Single Machine
Innite Bus System
3.1 Introduction
The Nonlinear dierential equations governing the behavior of a power system can be
linearized about a particular operating point, to obtain a linear model which represents
the small signal oscillatory response of a power system. Variations in the operating
condition of the system result in the variations in the parameters of the small signal
model. A given range of variations in the operating conditions of a particular system
thus generates a set of linear models, each corresponding to one particular operating
condition. Since, at any given instant, the actual plant could correspond to any model
in this set, a robust controller would have to import adequate damping to each one
of this entire set of linear models. This chapter is organized as follows. In Section 3.2,
periodic output feedback technique for multimodel system is presented whereas in Section
3.3, design of robust PSS for single machine connected to an innite bus using periodic
output feedback technique is presented. In Section 3.4, simulations with nonlinear model
using robust periodic output feedback controller are presented. Section 3.5 discusses fast
output sampling feedback technique for multimodel system whereas Section 3.6 contains
the design of robust PSS for single machine connected to an innite bus using fast output
sampling feedback technique. In Section 3.7 simulations with nonlinear model using robust
fast output sampling feedback controller are presented followed by concluding remarks.
31
3.2 Periodic output feedback technique for multimodel
system
For multimodel representation of a plant, it is necessary to design controller which will
robustly stabilize the multimodel system. Multimodel representation of plants can arise in
several ways. When a nonlinear system has to be stabilized at dierent operating points,
linear models are sought to be obtained at those operating points. Even for parametric
uncertain linear systems, dierent linear models can be obtained for extreme points of
the parameters. The models are used for stabilization of the uncertain system.
Let us consider a family of plant S = {A
i,
B
i,
C
i
}, dened by
.
x= A
i
x +B
i
u, (3.1)
y = C
i
x, i = 1, , M. (3.2)
By sampling at the rate of 1/ we get a family of discrete systems {
i
,
i
, C
i
} .
Assume each
_

N
i
, C
i
_
is observable. Then we can nd output injection gains G
i
such that
_

N
i
+G
i
C
i
_
has required set of poles. Now consider the augmented system
dened as

=
_

1
0 0
0
2
0
.
.
.
.
.
.
.
.
.
.
.
.
0 0
M
_

_
,

=
_

2
.
.
.

M
_

_
,

G
=
_

_
G
1
G
2
.
.
.
G
M
_

_
. (3.3)
The linear equation
_

N1

_
_

_
K
0
.
.
.
K

N1
_

_
=

G, (3.4)
has a solution if
_

_
is controllable with controllability index , and

N / (number
of inputs or outputs). This periodic output feedback gain realizes the designed G for all
plants of the family. It has been shown in the [38], the controllability of individual plants
generically implies the controllability of the augmented system.
The controller is obtained may be alright, but from the practical point of view it may
require excessive eort. To reduced this eect, we can approximate the condition given
in Eqn. (3.4) with a constraint on the gain.
32
Thus we consider the following inequations
K <
1
,

i
KG
i
<
2i
, i = 1....M. (3.5)
This can be formulated in the framework of Linear Matrix Inequalities as follows
_
_

2
1
I K
K
T
I
_
_
< 0,
_
_

2
2i
I (
i
KG
i
)
(
i
KG
i
)
T
I
_
_
< 0. (3.6)
Here, the LMI Toolbox of MATLAB [51] can be used for the design of K.
The robust periodic output feedback controller obtained by the above method requires
only constant gains and hence is easier to implement.
3.3 Case study I: Robust PSS design by periodic
output feedback technique
The sixteen models corresponding to linearized model at dierent operating conditions of
a single machine connected to an innite bus system are considered for designing robust
periodic output feedback controller using LMI approach. Then controller gains are applied
to simulate the nonlinear plant of a single machine connected to an innite bus system at
dierent operating points.
The discrete models of dierent linear models with dierent power (P
g0
) and dierent
external line inductance(x
e
) are obtained for the sampling time = 0.5 sec. and are given
in Table 3.1.
The following parameters are used for simulation of sixteen models:
H = 5, T

do
= 6 sec., D = 0.0, K
E
= 100, T
E
= 0.02 sec., x
e
= 0.2 p.u.
The output matrix for slip is given as
C =
_
0 1 0 0
_
. (3.7)
Using the method discussed in Section 3.2, stabilizing output injection gain matrices
are obtained for all sixteen models. Using LMI approach, Eqns.(3.5-3.6) are solved for
33
dierent values of to nd the gain matrix K. The robust periodic output feedback gain
K is obtained as
K =
_
75.5647 105.3353 68.8612 16.6909 0.9479
_
T
.
The eigen values of (
i
+
i
KC
i
) are found to be within the unit circle and no eigen-
values at the origin. The open loop and closed loop responses with impulse disturbance
for some of the linear models are as shown in Fig. 3.1 and Fig. 3.2.
3.4 Simulation with nonlinear model
A SIMULINK based block diagram including all the nonlinear blocks is generated. The
slip of the machine is taken as output. The PSS output signal which has input signal
derived from slip signal with gain K and a limiter is used to provide additional damping
torque without aecting the synchronizing torque at critical frequencies. The PSS output
is added to V
ref
signal. This is used to damp out the small signal disturbances via
modulating the generator excitation. The output must be limited to prevent the PSS
acting to counter action of AVR. Dierent operating points are taken as the dierent
models. The disturbance considered is a self clearing fault at generator terminal cleared
after 0.1 second. So the applied control law takes more time to stabilize the output in
comparison to linear model because a limiter 0.20 is used with input signal in nonlinear
model simulations.
Simulation results at dierent operating points (P
go
) and dierent external line induc-
tance (x
e
) are shown in Fig. 3.3 to Fig. 3.6 without controller and with controller along
with the required control input as modulating voltage.
As shown in plots, the proposed controller is able to damp out the oscillations in 4 to 8
seconds after clearing the fault up to active power, P
go
= 1.0 with external line inductance
from x
e
= 0.2 to x
e
= 0.4 and 5 to 20 seconds up to active power, P
go
= 2.0 with external
line inductance, x
e
= 0.2.
34
Table 3.1: Model parameters used in periodic output feedback method
P
g0
K
1
K
2
K
3
K
4
K
5
K
6
x
e
Model 1 0.4 1.0170 1.1239 0.2889 1.4385 0.0643 0.3291 x
e
Model 2 0.5 1.1624 1.2983 0.2889 1.6618 0.0577 0.3083 x
e
Model 3 0.6 1.4336 1.5855 0.2889 2.0294 0.0194 0.2628 x
e
Model 4 0.75 1.1624 1.2983 0.2889 1.6618 0.0577 0.3308 x
e
Model 5 1.0 1.5772 1.7402 0.2889 2.2274 -0.0309 0.2300 x
e
Model 6 1.2 1.6342 1.8588 0.2889 2.3792 -0.1130 0.1988 x
e
Model 7 1.4 1.6263 1.8752 0.2889 2.4003 -0.1328 0.1939 x
e
Model 8 1.5 1.6117 1.8883 0.2889 2.4170 -0.1524 0.1898 x
e
Model 9 1.6 1.6263 1.8752 0.2889 2.4003 -0.1328 0.1939 x
e
Model 10 1.7 1.6117 1.8883 0.2889 2.4170 -0.1524 0.1898 x
e
Model 11 1.8 1.5911 1.8987 0.2889 2.4303 -0.1717 0.1866 x
e
Model 12 2.0 1.4977; 1.9174 0.2889 2.4543 -0.1878 0.1805 x
e
Model 13 1.0 1.4414 1.6126 0.3081 2.0641 -0.0469 0.2681 1.25x
e
Model 14 1.0 1.3197 1.5042 0.3263 1.9254 -0.0641 0.3020 1.5x
e
Model 15 1.0 1.2087 1.4107 0.3436 1.8057 -0.0820 0.3326 1.75x
e
Model 16 1.0 1.3288 1.7402 0.3600 1.7009 -0.1002 0.3608 2x
e
35
0 5 10 15 20
0.1
0.05
0
0.05
0.1
0.15
Time(secs)
S
l
i
p
Model 1
0 5 10 15 20
0.5
0
0.5
Time(secs)
S
l
i
p
Model 5
0 5 10 15 20
0.5
0
0.5
Time(secs)
S
l
i
p
Model 12
0 5 10 15 20
0.5
0
0.5
Time(secs)
S
l
i
p
Model 16
P
g0
=0.4,x
e
=0.2
P
g0
=1.0,x
e
=0.2
P
g0
=2.0,x
e
=0.2 P
g0
=1.0,x
e
=0.4
with controller
with controller
with controller
with controller
without controller
without controller
without controller
without controller
Figure 3.1: Open and closed loop impulse responses for linearized system for various
operating conditions using periodic output feedback controller
0 5 10 15 20
0.1
0.05
0
0.05
0.1
0.15
Time(secs)
S
l
i
p
Model 4
0 5 10 15 20
0.5
0
0.5
Time(secs)
S
l
i
p
Model 8
0 5 10 15 20
0.5
0
0.5
Time(secs)
S
l
i
p
Model 9
0 5 10 15 20
0.4
0.2
0
0.2
0.4
Time(secs)
S
l
i
p
Model 12
P
g0
=0.75, x
e
=0.2
P
g0
=1.5, x
e
=0.2
P
g0
=1.6, x
e
=0.2 P
g0
=1.00, x
e
=0.25
with controller
with controller
with controller
with controller
without controller
without controller
without controller
without controller
Figure 3.2: Open and closed loop impulse responses for linearized system for various
operating conditions using periodic output feedback controller
36
0 5 10
4
2
0
2
4
x 10
3
S
l
i
p
Time in Seconds
Model 1
0 5 10 15 20
0.1
0.05
0
0.05
0.1
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 1:Control input
0 5 10 15
0.02
0.01
0
0.01
0.02
S
l
i
p
Time in Seconds
Model 5
0 10 20 30
0.1
0.05
0
0.05
0.1
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 5:Control input
P
g0
=0.4,x
e
=0.2
P
g0
=0.4,x
e
=0.2
P
g0
=1.0,x
e
=0.2
with controller
with controller
without controller
without controller
Figure 3.3: Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller
0 5 10
0.01
0.005
0
0.005
0.01
S
l
i
p
Time in Seconds
Model 4
0 5 10 15 20
0.1
0.05
0
0.05
0.1
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 4:Control input
0 10 20 30
0.1
0.05
0
0.05
0.1
S
l
i
p
Time in Seconds
Model 8
0 10 20 30
0.2
0.1
0
0.1
0.2
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 8:Control input
P
g0
=0.75, x
e
=0.2
P
g0
=1.5, x
e
=0.2
P
g0
=0.75, x
e
=0.2
with controller
without controller
without controller
with controller
Figure 3.4: Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller
37
0 10 20 30
0.2
0.1
0
0.1
0.2
S
l
i
p
Time in Seconds
Model 12
0 10 20 30 40
0.2
0.1
0
0.1
0.2
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 12:Control input
0 5 10
0.05
0
0.05
S
l
i
p
Time in Seconds
Model 16
0 10 20 30
0.1
0.05
0
0.05
0.1
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 16:Control input
P
g0
=2.0,x
e
=0.2
P
g0
=1.0,x
e
=0.4
with controller
without controller
with controller
without controller
Figure 3.5: Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller
0 10 20 30
0.2
0.1
0
0.1
0.2
S
l
i
p
Time in Seconds
Model 9
0 10 20 30 40
0.2
0.1
0
0.1
0.2
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 9:Control input
0 5 10
0.05
0
0.05
S
l
i
p
Time in Seconds
Model 13
0 10 20 30
0.1
0.05
0
0.05
0.1
M
o
d
u
l
a
t
i
n
g

v
o
l
t
a
g
e
Time in Seconds
Model 13:Control input
0 5 10
0.05
0
0.05
S
l
i
p
Time in Seconds
P
g0
=1.6, x
e
=0.2
P
g0
=1.0, x
e
=0.25
with controller
with controller
without controller
without controller
Figure 3.6: Open and closed loop responses for nonlinear system for various operating
conditions using periodic output feedback controller
38
3.5 Fast output sampling feedback technique for
multimodel system
For multimodel representation of a plant, it is necessary to design controller which will
robustly stabilize the multimodel system. Multimodel representation of plants can arise in
several ways. When a nonlinear system has to be stabilized at dierent operating points,
linear models are sought to be obtained at those operating points. Even for parametric
uncertain linear systems, dierent linear models can be obtained for extreme points of
the parameters. The models are used for stabilization of the uncertain systems.
Now consider a family of plant S = {A
i,
B
i,
C
i
}, dened by
.
x= A
i
x +B
i
u, (3.8)
y = C
i
x, i = 1, , M. (3.9)
By sampling at the rate of 1/, we get a family of discrete-time systems {
i
,
i
, C
i
}.
Now consider the augmented system dened below.

=
_

1
0 0
0
2
0
.
.
.
.
.
.
.
.
.
.
.
.
0 0
M
_

_
,

=
_

1
0 0
0
2
0
.
.
.
.
.
.
.
.
.
.
.
.
0 0
M
_

_
,

C
=
_
C
1
C
2
C
M
_
.
(3.10)
Consider the family of discrete-time systems(3.8, 3.9) having at time t = k the input
u
k
= u(k), state x
k
= x(k) and output y
k
as
x
k+1
=
i
x
k
+
i
u
k
, (3.11)
y
k+1
= C
0i
x
k
+D
0i
u
k
, (3.12)
where
C
0i
=
_

_
C
i
C
i

i
.
.
.
C
i

N1
i
_

_
, D
0
=
_

_
0
C
i

i
.
.
.
C
i

N2
j=0

j
i

i
_

_
.
Assume each (
i
,
i
) is controllable. Then we can nd a robust state feedback gains
F such that (
i
+
i
F) has no eigenvalues at the origin.
Then, assuming that in intervals k t < k +
u(t) = Fx(k), (3.13)
39
one can dene the ctitious measurement matrix
C
i
(F, N) = (C
0i
+D
0i
F)(
i
+
i
F)
1
, (3.14)
which satises the ctitious measurement equation y
k
= C
i
x
k
. For robust fast output
sampling gain L to realize the eect of F, it may satisfy
LC
i
= F, i = 1, , M. (3.15)
The Eqn.(3.15) can be written as
L

C
=

F
, (3.16)
where

C
=
_
C
1
C
2
C
M
_
,

F
=
_
F F F
_
.
Let

denotes the observability index of (

C
). It can be shown that for N

/
(number of inputs or outputs), generically

C
has full column rank, so that robust state
feedback gain can be realized by a fast output sampling gain L.
When this idea is realized in practice i.e. fast output sampling gain L have been
obtained by realizing the state feedback gain F, two problems are required to be addressed.
The rst one is apparent from(2.67). With this type of controller, the unknown states
are estimated implicitly, using the measured output samples and assuming that initial
control is generated by state feedback. If initial state causes an estimation error, then
decay of this error will be determined by the eigenvalues of the matrix (LD
0i
F
i
) which
depends on L and whose dimension equals the number of control input. For stability these
eigenvalues have to be inside the unit disc, and for fast decay they should be as close to
the origin as possible. This problem must be taken into account while designing L.
The second problem is that the gain matrix L may have elements with large mag-
nitude. Because these values are only weights in linear combination of output samples,
large magnitudes do not necessarily imply large control signal, and in theory noise free
simulation pose no problem. But in practice they amplify measurement noise, and it is
desirable to keep these values low. This objective can be expressed by an upper bound
on the norm of the gain matrix L.
When trying to deal with these problems, it is better not to insist on an exact solution
to the design(3.15): one can allow a small deviation and use an approximation LC
i
F,
40
which hardly aects the desired closed-loop dynamics, but may have considerable eect
on the two problems described above. Instead of looking for an exact solution to the
equalities, the following inequalities are solved
L <
1
,
LD
0i
F
i
<
2i
, i = 1, , M,
LC
i
F <
3i.
(3.17)
Here three objectives have been expressed by upper bounds on matrix norms, and
each should be as small as possible. The value of
1
is small means low noise sensitivity,
the value of
2
is small means fast decay of estimation error, most important -
3
-small
means that fast output sampling controller with gain L is a good approximation of the
originally designed state feedback controller. If
3
= 0 then Lis an exact solution.
Using the Schur complement, it is straight forward to bring these conditions in the
form of LMI (Linear Matrix Inequalities).
_
_

2
1
I L
L
T
I
_
_
< 0,
_
_

2
2i
I (LD
0i
F
i
)
(LD
0i
F
i
)
T
I
_
_
< 0,
_
_

2
3i
I (LC
i
F)
(LC
i
F)
T
I
_
_
< 0. (3.18)
In this form, the function mincx() of the LMI control Toolbox for MATLAB can
be used to minimize a linear combination of
1
,
2
,
3
. The following approach turned
out to be useful. If the actual measurement noise is known, the magnitude of L is xed
accordingly. Likewise eigenvalues of (LD
0i
F
i
) less than 0.05 cause no problem. So
we can x
1
,
2
and only
3
is minimized subject to these constraints. This is illustrated
in the following section.
Here, the LMI Tool Box of MATLAB can be used for the design of L [51].
The fast output sampling feedback controller obtained by the above method requires
only constant gains and hence is easier to implement.
41
3.6 Case Study II: Robust PSS design by fast output
sampling technique
Here sixteen discrete models of a single machine connected to an innite bus system are
obtained for dierent power (P
g0
) for = 0.5 sec. In this case the external line inductance
(x
e
) varies from 0.2 to 0.8 p.u. These model parameters are given in Table 3.2.
The following parameters are used for simulation of sixteen models:
H = 5, T

d0
= 6 sec., D = 0.0, K
E
= 100, T
E
= 0.02 sec., x
e
= 0.2 p.u.
Table 3.2: Model parameters used in fast output sampling feedback Method
P
g0
Q
g0
K
1
K
2
K
3
K
4
K
5
K
6
x
e
Model 1 0.4 0.0160 1.0170 1.1239 0.2889 1.4385 0.0643 0.3291 x
e
Model 2 0.5 0.0251 1.1624 1.2983 0.2889 1.6618 0.0577 0.3083 x
e
Model 3 0.6 0.0361 1.4336 1.5855 0.2889 2.0294 0.0194 0.2628 x
e
Model 4 0.75 0.0566 1.1624 1.2983 0.2889 1.6618 0.0577 0.3308 x
e
Model 5 1.0 0.1010 1.5772 1.7402 0.2889 2.2274 -0.0309 0.2300 x
e
Model 6 1.2 0.1461 1.6342 1.8588 0.2889 2.3792 -0.1130 0.1988 x
e
Model 7 1.4 0.2000 1.6263 1.8752 0.2889 2.4003 -0.1328 0.1939 x
e
Model 8 1.5 0.2303 1.6117 1.8883 0.2889 2.4170 -0.1524 0.1898 x
e
Model 9 1.6 0.2629 1.6263 1.8752 0.2889 2.4003 -0.1328 0.1939 x
e
Model 10 1.7 0.2979 1.6117 1.8883 0.2889 2.4170 -0.1524 0.1898 x
e
Model 11 1.8 0.3352 1.5911 1.8987 0.2889 2.4303 -0.1717 0.1866 x
e
Model 12 2.0 0.4174 1.4977 1.9174 0.2889 2.4543 -0.1878 0.1805 x
e
Model 13 1.0 0.1535 1.3197 1.5042 0.3263 1.9254 -0.0641 0.3020 1.5x
e
Model 14 1.0 0.2087 1.3288 1.7402 0.3600 1.7009 -0.1002 0.3608 2x
e
Model 15 1.0 0.3333 0.7430 1.0776 0.4180 1.3794 -0.1739 0.4561 3x
e
Model 16 1.0 0.5000 0.3903 0.8904 0.4667 1.1397 -0.2468 0.5380 4x
e
The output matrix is given as
C =
_
0 1 0 0
_
. (3.19)
Using the method discussed in Section 3.5 common stabilizing state feedback gain
matrix F is obtained for all sixteen models as given below.
F
T
=
_
0.4101 19.3152 0.4879 0.0016
_
. (3.20)
42
Using LMI approach, Eqns.(3.17-3.18) are solved for dierent values of
1
,
2
and
3
to nd the gain matrix L. The robust fast output sampling feedback gain L is obtained
as
L =
_
L
1
L
2
_
, (3.21)
where
L
1
=
_
18.7233 14.0543 7.6172 31.0312 46.6674
_
,
L
2
=
_
48.7778 34.0972 0.9996 51.0546 121.8813
_
.
The eigen values of (
i
+
i
LC
i
) are found to be within the unit circle.
3.7 Simulation with nonlinear model
A SIMULINK based block diagram including all the nonlinear blocks is generated [47].
The slip of the machine is taken as output and the initial value of slip is taken as zero.
The PSS output signal which is has input signal derived from slip signal with gain L and
a limiter is used to provide additional damping torque without aecting the synchronizing
torque at critical frequencies. The PSS output is added to V
ref
signal. This is used to
damp out the small signal disturbances via modulating the generator excitation. The
output must be limited to prevent the PSS acting to counter action of AVR. Dierent
operating points are taken as the dierent models. The disturbances considered is a self
clearing fault at generator terminal cleared after 0.1 second. The limits of PSS output
are taken as 0.25 and exciter output limits (E
fd
) are taken 6 pu.
The conventional PSS is designed as compensator with washout circuit [47]. The
transfer function of PSS compensator is selected as
H(s) =
16(1 + 0.08s)
(1 + 0.027s)
(3.22)
and washout circuit constant is selected as 2.0 seconds.
Simulation results at dierent operating points (P
go
) and dierent external line induc-
tance (x
e
) are shown in Fig. 3.7 to Fig. 3.14 without controller and with the proposed
controller. The control input as modulating voltage is also shown in these gures. The
responses are compared with the conventional power system stabilizer. The simulation
results show that conventional power system stabilizer is able to damp out the oscillations
up to active power, P
go
= 1.5 with external line inductance, x
e
= 0.2 and up to active
power, P
go
= 1.0 with external line inductance, x
e
= 0.4 whereas the proposed fast output
43
0 5 10 15
5
0
5
10
x 10
3
Model 2
S
l
i
p
Time in Seconds
with proposed controller
with conventional PSS
without controller
0 5 10 15
0.4
0.2
0
0.2
0.4
Model 2: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0= 0.5, xe=0.2
Pg0= 0.5, xe=0.2
Figure 3.7: Open and closed loop impulse responses for nonlinear system for various
operating conditions using fast output sampling feedback controller
sampling feedback controller is more robust and a single controller is able to damp out
the oscillations for all cases.
As shown in plots, the proposed controller is able to damp out the oscillations in 2 to 3
seconds after clearing the fault up to active power, P
go
= 1.0 with external line inductance
from x
e
= 0.2 to x
e
= 0.8 and 5 to 15 seconds up to active power, P
go
= 2.0 with external
line inductance, x
e
= 0.2.
3.8 Concluding remarks
In this chapter, a design scheme of the robust power system stabilizer for single machine
connected to an innite bus system using periodic output feedback method has been
developed. The slip signal is taken as output and the periodic output feedback control is
applied at a faster than output sampling rate.
The chapter also presents, a design scheme of robust power system stabilizer for single
machine connected to an innite bus using fast output sampling feedback technique. In
this also the slip signal is taken as output and the fast output sampling feedback control
is applied at a slower rate than output sampling rate. This method is also compared
with conventional power system stabilizer. The proposed method results better response
44
0 5 10 15
0.01
0.005
0
0.005
0.01
Model 4
S
l
i
p
Time in Seconds
with proposed controller
with conventional PSS
without controller
0 5 10 15
0.4
0.2
0
0.2
0.4
Model 4: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0= 0.75, xe=0.2
Pg0= 0.75, xe=0.2
Figure 3.8: Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller
0 5 10 15
0.04
0.02
0
0.02
0.04
Model 5
S
l
i
p
Time in Seconds
with proposed controller
without controller
with conventional PSS
0 5 10 15
0.4
0.2
0
0.2
0.4
Model 5: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0= 1.00, xe=0.2
Pg0= 1.00, xe=0.2
Figure 3.9: Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller
45
0 5 10 15
0.1
0.05
0
0.05
0.1
Model 8
S
l
i
p
Time in Seconds
with proposed controller
with conventional PSS
without controller
0 5 10 15
0.2
0.1
0
0.1
0.2
Model 8: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0= 1.50, xe=0.2
Pg0= 1.50, xe=0.2
Figure 3.10: Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller
0 5 10 15
0.1
0.05
0
0.05
0.1
Model 10
S
l
i
p
Time in Seconds
with proposed controller
with conventional PSS
without controller
0 5 10 15
0.2
0.1
0
0.1
0.2
0.3
Model 10: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0=1.7, xe=0.2
Pg0=1.7, xe=0.2
Figure 3.11: Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller
46
0 5 10 15 20 25
0.1
0.05
0
0.05
0.1
Model 12
S
l
i
p
Time in Seconds
with proposed controller
with conventional PSS
without controller
0 5 10 15 20 25 30 35
0.2
0.1
0
0.1
0.2
0.3
Model 12: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0=2.0, xe=0.2
Pg0=2.0, xe=0.2
Figure 3.12: Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller
0 5 10 15
0.05
0
0.05
Model14
S
l
i
p
with proposed controller
without controller
with conventinal PSS
0 5 10 15
0.4
0.2
0
0.2
0.4
Model 14: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0=1.0, xe=0.4
Pg0=1.0, xe=0.4
Time in Seconds
Figure 3.13: Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller
47
0 5 10 15
0.05
0
0.05
S
l
i
p
Time in Seconds
Model 16
with proposed controller
with conventional PSS
without controller
0 5 10 15
0.2
0.1
0
0.1
0.2
0.3
Model 16: Control Input
V
o
l
t
a
g
e
Time in Seconds
with proposed controller
with conventional PSS
Pg0=1.0, xe=0.8
Pg0=1.0, xe=0.8
Figure 3.14: Open and closed loop responses for nonlinear system for various operating
conditions using fast output sampling feedback controller
behaviour to damp out the oscillations. The conventional power system stabilizer is
dynamic in nature and is required to be tuned according to the power characteristics.
Whereas, the proposed controller gains are static in nature and one controller structure
is able to damp out the oscillations for all models.
Both the methods are more general than static output feedback. It is found that
the designed robust controllers provide good damping enhancement for various operating
points of single machine connected to an innite bus system.
It is observed from the simulations that The fast output sampling feedback method
provides better damping enhancement for various operating points of single machine con-
nected to an innite bus system compared to periodic output feedback method. It is also
found from the simulations the control input for these models are required of small mag-
nitudes in fast output sampling feedback method compared to periodic output feedback
method. As shown in plots, the proposed periodic output feedback controller is able to
damp out the oscillations in 4 to 8 seconds after clearing the fault up to active power, P
go
= 1.0 with external line inductance from x
e
= 0.2 to x
e
= 0.4 and 5 to 20 seconds up to
active power, P
go
= 2.0 with external line inductance, x
e
= 0.2, whereas, the proposed fast
output sampling feedback controller is able to damp out the oscillations in 2 to 3 seconds
after clearing the fault up to active power, P
go
= 1.0 with external line inductance from
48
x
e
= 0.2 to x
e
= 0.8 and 5 to 15 seconds up to active power, P
go
= 2.0 with external
line inductance, x
e
= 0.2. In the next chapter, these methods have been extended to the
multimachine system.
49
.
50
Chapter 4
Design of PSS for Multimachine
Power System
4.1 Introduction
Analysis of practical power system involves the simultaneous solution of equations con-
sisting of synchronous machines and the associated excitation system and prime movers,
interconnecting transmission network, static and dynamic load (motor loads), and other
devices such as HVDC converters, static var compensators. The dynamics of the ma-
chine rotor circuits, excitation systems, prime mover and other devices are represented
by dierential equations. The result is that the complete system model consists of large
number of ordinary dierential and algebraic equations [50].
The Nonlinear dierential equations governing the behavior of a power system can be
linearized about a particular generating power, bus structure and load distributions to
obtain a linear model which represents the small signal oscillatory response of a power
system. Variations in the generating power, bus structure and load distribution of the
system result in the variations in the parameters of the small signal model. A given range
of variations in the generating power, bus structure and load distribution of a particular
system thus generates a set of linear models each corresponding to one particular gener-
ating power, bus structure and load distribution. Since, at any given instant, the actual
plant could correspond to any model in this set, a robust controller would have to import
adequate damping to each one of this entire set of linear models. This chapter is organized
as follows.
In Section 4.2 periodic output feedback technique is presented and in Section 4.3 de-
sign of centralized PSS for multimachine power system using periodic output feedback
51
technique is presented. In Section 4.4 simulations with nonlinear model using proposed
periodic output feedback controller are presented. Section 4.5 contains decentralized peri-
odic output feedback technique for multimodel system. Section 4.6 presents the design of
robust decentralized PSS for multimachine power system viz 4 machine 10 bus system and
10 machine 39 bus system using periodic output feedback technique. Section 4.7 presents
the simulations with nonlinear model using the proposed decentralized periodic output
feedback controller. Section 4.8 discusses the fast output sampling feedback technique
whereas in Section 4.9 design of centralized PSS for multimachine power system using
fast output sampling feedback technique is presented. Section 4.10 presents the simula-
tions with nonlinear model using proposed fast output sampling feedback controller. In
Section 4.11 decentralized fast output sampling feedback technique for multimodel system
is discussed. Section 4.12 contains the design of robust decentralized PSS for multima-
chine power system viz 4 machine 10 bus system and 10 machine 39 bus system using fast
output sampling feedback technique. In Section 4.13 simulations with nonlinear model
using the proposed decentralized fast output sampling feedback controller are presented
followed by concluding remarks.
4.2 Periodic output feedback technique
Consider a linear continuous time-invariant 4 machine 10 bus, MIMO power system,
represented by
x = Ax +Bu,
y = Cx.
(4.1)
A discrete time invariant system with sampling interval sec. can be represented as
x(k + 1) =

x(k) +

u(k), (4.2)
y(k) = Cx(k).
The adjoint or the dual for the above system would be
x(k + 1) =
T

x(k) +C
T
u(k), (4.3)
y(k) =
T

x(k).
Let u(k) = S x be a stabilizing control for the model in Eqn.(4.3). Thus closed loop
model
_

+C
T
S
_
becomes stable. But the eigenvalues of (
T

+C
T
S) and
_

+C
T
S
_
T
are the same. So
_

+S
T
C
_
will also be stable. Thus S
T
G is the output injection
gain for the system in Eqn.(4.2).
52
Using this G, LMI constraints given in Eqns.(2.54-2.55) are solved using dierent
values of to nd the centralized gain matrix.
4.3 Case study I: Centralized PSS design for
multimachine power system (4 machine 10 bus
system) by periodic output feedback technique
The Nonlinear dierential equations governing the behavior of 4 machine system has been
linearized about a particular operating point to obtain a linear model which represents
the small signal oscillatory response of a power system.
The 4 machines, 10 bus multimachine power system data are considered for designing
periodic output feedback controller using LMI approach of MATLAB. The single line
diagram of the system is shown in Fig. 2.5. The machine data, line data, AVR data and
load ow data are given in Table 4.1 to Table 4.3 [47].
Table 4.1: Machine data: 4 machine and 10 bus system
Gen R
a
x
d
x

d
x
q
x

q
H T

do
T

qo
X
l
D K
E
T
E
(pu) (pu) (pu) (pu) (pu) (pu) (pu) (pu) (pu) (pu) (pu) (pu)
1 .00028 0.2 .033 0.19 .061 54.0 8.0 0.4 .022 0.0 200 0.05
2 .00028 0.2 .033 0.19 .061 54.0 8.0 0.4 .022 0.0 200 0.05
3 .00028 0.2 .033 0.19 .061 63.0 8.0 0.4 .022 0.0 200 0.05
4 .00028 0.2 .033 0.19 .061 63.0 8.0 0.4 .022 0.0 200 0.05
The above multimachine system is modeled using SIMULINK Toolbox of MATLAB
using machine model 1.0. The linear state space model is obtained. Then discrete model
is obtained for the sampling time = 0.1 sec.
Using the method discussed in Section 4.2, stabilizing output injection gain matrix G
(16x4) is obtained using DLQR theory [41, 52, 53].
Using LMI approach, Eqns.(2.54-2.55) are solved for dierent values of to nd the
gain matrix, K. Thus the centralized periodic output feedback gain matrix K ( 16x4) is
obtained as
53
Table 4.2: Line data: 4 machine and 10 bus system
From Bus To Bus Series Series Shunt
Number Number Resistance Reactance Reactance
(R
s
) pu (X
s
) pu (B
c
) pu
1 6 0.01 0.012 0.0
2 5 0.01 0.012 0.0
9 10 0.022 0.22 0.33
9 10 0.022 0.22 0.33
9 10 0.022 0.22 0.33
9 5 0.002 0.02 0.03
9 5 0.002 0.02 0.03
3 8 0.001 0.012 0.0
4 7 0.001 0.012 0.0
10 7 0.002 0.02 0.03
10 7 0.002 0.02 0.03
6 5 0.005 0.05 0.075
6 5 0.005 0.05 0.075
8 7 0.005 0.05 0.075
8 7 0.005 0.05 0.075
Table 4.3: Load ow data: 4 machine and 10 bus system
Bus. V P
G
Q
G
P
L
Q
L
B
l
No. (pu) (degrees) (pu) (pu) (pu) (pu) (pu)
1 1.03 8.2154 7.0 1.3386 0.0 0.0 0.0
2 1.01 -1.5040 7.0 1.5920 0.0 0.0 0.0
3 1.03 0.0 7.2122 1.4466 0.0 0.0 0.0
4 1.01 -10.2051 7.0 1.8083 0.0 0.0 0.0
5 1.0108 3.6615 0.0 0.0 0.0 0.0 0.0
6 0.9875 -6.2433 0.0 0.0 0.0 0.0 0.0
7 1.0095 -4.6997 0.0 0.0 0.0 0.0 0.0
8 0.9850 -14.9443 0.0 0.0 0.0 0.0 0.0
9 0.9761 -14.4194 0.0 0.0 11.59 2.12 3.0
10 0.9716 -23.2977 0.0 0.0 15.75 2.88 4.0
54
K=
_

_
65.2479 69.1969 59.8482 58.1485
68.9798 61.6213 59.6293 58.8642
68.9332 68.7500 79.2084 75.3349
67.6619 68.1050 76.7574 71.9817
77.7411 53.1078 63.9723 60.1125
60.9788 67.5728 61.6212 57.1467
70.5518 67.6737 85.4084 61.4865
72.6057 68.9402 73.5482 78.4378
0.5915 2.5590 1.8597 1.6127
2.3962 0.4752 2.0133 1.7171
1.3771 1.1191 0.0479 1.9259
0.8511 0.5546 1.1651 0.6582
0.4832 0.3438 0.4022 0.3769
0.3952 0.4207 0.3904 0.3631
0.4525 0.4335 0.5445 0.4029
0.4601 0.4373 0.4730 0.4930
_

_
.
The closed loop responses with this centralized gain K for the linearized model are
satisfactory and able to stabilize the outputs. The eigen values of (
N
+KC) are found
to be within the unit circle and given in Table 4.4.
Table 4.4: Eigen values of the closed loop system (
N
+KC)
0.4501 + j 0.4864 0.4501 -j 0.4864
0.5117 + j 0.4386 0.5117 -j 0.4386
0.8201 + j 0.4824 0.8201 -j 0.4824
0.8092 + j 0.5006 0.8092 -j 0.5006
0.6245 + j 0.1562 0.6245 -j 0.1562
0.6333 + j 0.1239 0.6333 -j 0.1239
0.9147 + j 0.0547 0.9147 -j 0.0547
0.8479 1.0000
55
0 5 10
1
0.5
0
0.5
1
x 10
6
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 5 10
5
0
5
10
x 10
7
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
0 5 10
1
0.5
0
0.5
1
x 10
6
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 5 10
1
0.5
0
0.5
1
x 10
6
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Figure 4.1: Open and closed loop responses using centralized periodic output feedback
controller(4 machine and 10 bus system)
4.4 Simulation with nonlinear model
A SIMULINK based block diagram including all the nonlinear blocks is generated [47].
The slip of the machine is taken as output. The output slip signal with gain K and a
limiter is added to V
ref
signal. This is used to damp out the small signal disturbances
via modulating the generator excitation. The output must be limited to prevent the PSS
acting to counter action of AVR. The disturbance considered is a self clearing fault at
particular bus (Bus No.3) cleared after 0.1 second. The limits of PSS output are taken
as 0.1.
Simulation results of dierent generators are shown in Fig. 4.1 without controller and
with controller and the same with fault are shown in Fig. 4.2. As shown in plots, the
proposed controller is able to damp out the oscillations in 8 to 10 seconds after clearing
the fault. The slips are computed relative to the centre of inertia which is dened as
S
coi
=

n
i=1
H
i
.S
i

n
i=1
H
i
S
i
(4.4)
and plotted with time.
56
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller with controller
with controller
with controller
without controller without controller
without controller
without controller
Figure 4.2: Open and closed loop responses with fault at bus 3 using centralized periodic
output feedback controller(4 machine and 10 bus system)
4.5 Decentralized periodic output feedback technique
for multimodel system
For multimodel representation of a plant, it is necessary to design controller which will
robustly stabilize the multimodel system. Multimodel representation of plants can arise in
several ways. When a nonlinear system has to be stabilized at dierent operating points,
linear models are sought to be obtained at those operating points. Even for parametric
uncertain linear systems, dierent linear models can be obtained for the extreme points
of the parameters. These models are used for stabilization of the uncertain system.
Let us consider a family of plant S = {A
i,
B
i,
C
i
} ,dened by
.
x= A
i
x +B
i
u, (4.5)
y = C
i
x, i = 1, ......M. (4.6)
The discrete time invariant systems with sampling interval sec. can be represented
as
x(k + 1) =
i
x(k) +
i
u(k), (4.7)
y(k) = C
i
x(k).
57
The adjoint or the dual for the above systems would be
x(k + 1) =
T
i
x(k) +C
T
i
u(k), (4.8)
y(k) =
T
i
x(k).
Let u(k) = S
i
x be a stabilizing control for the system in Eqn.(4.8). Thus the closed
loop system (
T
i
+ C
T
i
S
i
) becomes stable. But the eigenvalues of (
T
i
+ C
T
i
S
i
) and
_

T
i
+C
T
i
S
i
_
T
are the same. So
_

i
+S
T
i
C
i
_
will also be stable.
Thus S
T
i
G
i
is the output injection gain for the system in Eqn.(4.7). If the LMI
constraints given in Eqns.(3.5- 3.6) are solved using above G
i
, the robust periodic output
feedback gain matrix may becomes full [36]. This results in the control input of each
machine being a function of outputs of all machines. Due to the geographically distributed
nature of power system and lack of communication system (unavoidable delays), the
decentralized robust control scheme may be more feasible and useful than the centralized
control scheme. Decentralized robust periodic output feedback control can be achieved
by making the o diagonal elements of K
0
, K
1
, , K

N1
matrices zero.
So the structure of K
i
( i = 0, , N-1) matrices for 4 machine and 10 bus power
system (machine model 1.1) are assumed as
K
i
=
_

_
k
i11
0 0 0
0 k
i22
0 0
0 0 k
i33
0
0 0 0 k
i44
_

_
, i = 0, , N 1. (4.9)
and for 10 machine and 39 bus power system (machine model 1.0) as
K
i
=
_

_
k
i11
0 0 0 0 0 0 0 0 0
0 k
i22
0 0 0 0 0 0 0 0
0 0 k
i33
0 0 0 0 0 0 0
0 0 0 k
i44
0 0 0 0 0 0
0 0 0 0 k
i55
0 0 0 0 0
0 0 0 0 0 k
i66
0 0 0 0
0 0 0 0 0 0 k
i77
0 0 0
0 0 0 0 0 0 0 k
i88
0 0
0 0 0 0 0 0 0 0 k
i99
0
0 0 0 0 0 0 0 0 0 k
i1010
_

_
, i = 0, , N 1.
(4.10)
58
With this structure of K
i
, the problem can be formulated in the framework of Linear
Matrix Inequalities using Eqns.(3.5-3.6) and the desired matrices can be obtained. Now
it is evident that the control input of each machine is a function of the output of that
machine only and this makes the periodic output feedback technique based power system
stabilizer design a decentralized one. In decentralized PSS, to activate the proposed
controller at same instant, proper synchronization signal is required to be sent to all the
machines.
4.6 Robust decentralized PSS design for
multimachine power system by periodic output
feedback technique
4.6.1 Case study II: 4 machine 10 bus system
The sixteen models of multimachine system are considered for designing robust fast out-
put sampling feedback controller using LMI approach of MATLAB. The controller gains
are applied to simulate a nonlinear model of multimachine system dynamics at dierent
operating points.
The sixteen models of 4 machine 10 bus system with dierent generating power, dier-
ent bus structure and dierent load distribution (60 to 100 percent variations from peak
loads) are obtained using machine model 1.1 as given in Table 4.5:
These models are considered for designing periodic output feedback controller using LMI
approach of MATLAB. The single line diagram of the system is shown in Fig. 2.5. The
machine data, line data and load ow data are given in Table 4.1 to Table 4.3 [47].
The above models multimachine power system are modeled using SIMULINK Toolbox
of MATLAB and the linear state space models are obtained for the same. Then discrete
models are obtained for sampling time = 0.1 sec.
Using the method discussed in Section 4.5 stabilizing output injection gain matrices
G
i
(20x4) is obtained using DLQR theory[41, 52, 53].
This periodic output feedback gain can be obtained which approximately realizes the
designed G for all models of the family. It has been shown in [38], that the controllability
of individual model generically implies the controllability of the augmented system. For
59
Table 4.5: Model parameter variations: 4 machine and 10 bus system
S.No. Active Power Load Bus Structure
(P
G
) (P
L
)
Model 1 [7 7 7.2 7] [9.59, 10.75] as per line data
Model 2 [7 7 7.2 7] [9.59, 11.25] one line 9-10 disconnected
Model 3 [7.2 7.1 7.0 6.9] [10.09, 11.25] as per line data
Model 4 [7.2 7.1 7.0 6.9] [10.09, 11.75] one line 10-7 disconnected
Model 5 [7 7 7.2 7] [10.59,,11.75] as per line data
Model 6 [7.1 6.9 7.5 6.5] [10.59, 12.25] as per line data
Model 7 [7.1 6.9 7.5 6.5] [11.09, 12.25] one line 5-9 disconnected
Model 8 [5 8 6.2 8] [11.09, 12.75] as per line data
Model 9 [7 7 7.2 7] [11.59, 12.75] one line 5-6 disconnected
Model 10 [7 7 7.2 7] [11.59, 13.25] one line 7-10 disconnected
Model 11 [7.2 7.1 7.0 6.9] [11.59, 13.75] one line 7-8 disconnected
Model 12 [7.2 7.1 7.0 6.9] [11.59, 14.25] one line 5-9 disconnected
Model 13 [7 7 7.2 7] [11.59, 14.75] as per line data
Model 14 [7.1 6.9 7.5 6.5] [11.59, 15.25] one line 10-7 disconnected
Model 15 [7.1 6.9 7.5 6.5] [11.59, 15.75] as per line data
Model 16 [5 8 6.2 8] [11.19, 15.75] one line 7-8 disconnected
60
the application mentioned in the chapter, actually the minimum value of

N would have
been 80 (because

N controllability index of the augmented system/number of inputs or
outputs) and that would result 80 gain matrices of 4x4 each. But, from the practical view
point, computation of gains in this way may not be useful because large gains turned out
by this method may magnify the noise and also implementation would be complex. To
reduce this eect of gain, we relax the condition such that the Eqn.(3.4) satises exactly
and also include a constraint on the gain.
The stabilization of 4 machine and 10 bus system at sixteen dierent conditions with
dierent generating power, dierent bus structure and dierent load distribution leading
to sixteen models is done as follows:
Consider the 4 machine and 10 bus system at one condition leading to one 20
th
order
model. An output injection gain matrix can be obtained using DLQR theory. If a periodic
output feedback gain has to be obtained for this system only, then minimum 5 gain
matrices (K
0
, K
1
, K
2
, K
3
, K
4
) each of 4x4 dimension are needed which will exactly realize
the output injection gain computed for this system. The equation (2.53) has solution if
N controllability index of this system/number of inputs or outputs. For one system
controllability index is 20. So minimum 5 gain sequences are needed.
Here as we are dealing with robust stabilization, we have to nd a K which will satisfy

i
K = G
i
, (i = 1, , 16), all these equations. However, if we restrict our gain sequence
to 5, there would not exist a common K which would satisfy
i
K = G
i
for i = 1, , 16.
Thats why this condition is relaxed and a robust K has been sought to be obtained by
satisfying the equation
i
K = G
i
approximately.
Using LMI approach, Eqns.(3.5-3.6) are solved for dierent values of to nd the
robust decentralized gain matrix K. The robust decentralized periodic output feedback
gain matrix K (20x4) is obtained as given below:
K
0
=
_

_
0.5662 0 0 0
0 0.5662 0 0
0 0 0.1409 0
0 0 0 0.1409
_

_
, K
1
=
_

_
2.0876 0 0 0
0 2.0876 0 0
0 0 0.6453 0
0 0 0 0.6453
_

_
,
K
2
=
_

_
3.0503 0 0 0
0 3.0503 0 0
0 0 1.1563 0
0 0 0 1.1563
_

_
, K
3
=
_

_
2.2099 0 0 0
0 2.2099 0 0
0 0 1.0343 0
0 0 0 1.0343
_

_
,
61
Table 4.6: Machine data: 10 machine and 39 bus system
Gen# R
a
X
d
X

d
X
q
X

q
H T

d0
T

q0
T
c
D K
E
T
E
1 0.0 0.2950 0.0647 0.282 0.0647 30.3 6.56 1.5 0.01 0.0 25 0.025
2 0.0 0.0200 0.0060 0.019 0.0060 500.0 6.0 0.7 0.01 0.0 25 0.025
3 0.0 0.2495 0.0531 0.237 0.0531 35.8 5.7 1.5 0.01 0.0 25 0.025
4 0.0 0.3300 0.0660 0.310 0.0660 26.0 5.4 0.44 0.01 0.0 25 0.025
5 0.0 0.2620 0.0436 0.258 0.0436 28.6 5.69 1.5 0.01 0.0 25 0.025
6 0.0 0.2540 0.0500 0.241 0.0500 34.8 7.3 0.4 0.01 0.0 25 0.025
7 0.0 0.2950 0.049 0.292 0.049 26.4 5.66 1.5 0.01 0.0 25 0.025
8 0.0 0.2900 0.057 0.280 0.057 24.3 6.7 0.41 0.01 0.0 25 0.025
9 0.0 0.2106 0.057 0.205 0.057 34.5 4.79 1.96 0.01 0.0 25 0.025
10 0.0 0.2000 0.004 0.196 0.004 42.0 5.7 0.5 0.01 0.0 25 0.025
K
4
=
_

_
0.6248 0 0 0
0 0.6248 0 0
0 0 0.3387 0
0 0 0 0.3387
_

_
.
The closed loop responses with this robust decentralized gain K for all the linearized
models are satisfactory and able to stabilize the outputs. The eigen values of (
N
+KC)
are found to be within the unit circle for all models.
4.6.2 Case study III: 10 machine 39 bus system
The sixteen models of 10 machine 39 bus system with dierent generating power, dierent
bus structure and dierent load distribution (60 to 100 percent variations from peak
loads) are obtained using machine model 1.0. The 10 machines, 39 bus power system
data are considered for designing periodic output feedback controller using LMI approach
of MATLAB. The single line diagram of the system is shown in Fig.2.6.
The machine data, line data and load ow data data are given in Table 4.6 to Table
4.11 [47].
The above models multimachine power system are modeled using SIMULINK Toolbox
of MATLAB and the linear state space models are obtained for the same. Then discrete
models are obtained for sampling time = 0.1 sec.
Using the method discussed in Section 4.5 stabilizing output injection gain matrices
62
Table 4.7: Load ow data I: 10 machine and 39 bus system
Bus. V P
G
Q
G
P
L
Q
L
No. (pu) (degrees) (pu) (pu) (pu) (pu)
1 0.98200 0.00000 5.519816 1.618098 0.0920 0.04600
2 1.03000 -10.96807 10.00000 2.262277 11.0400 2.5000
3 0.98310 2.340665 6.5000 1.660362 0.00000 0.00000
4 1.01230 3.166466 5.08000 1.550982 0.00000 0.00000
5 0.99720 4.189765 6.32000 0.8379978 0.00000 0.00000
6 1.04930 5.198208 6.50000 2.810341 0.00000 0.00000
7 1.06350 7.991468 5.60000 2.296622 0.00000 0.00000
8 1.02780 1.842515 5.40000 0.2757274 0.00000 0.00000
9 1.02650 7.544687 8.30000 0.5969476 0.00000 0.00000
10 1.04750 -4.006471 2.50000 1.83865 0.00000 0.00000
11 1.03501 -9.319072 0.00000 0.00000 0.00000 0.00000
12 1.01664 -6.441693 0.00000 0.00000 0.00000 0.00000
13 0.98552 -9.439605 0.00000 0.00000 3.22000 0.02400
14 0.94983 -10.37121 0.00000 0.00000 5.00000 1.84000
15 0.95056 -9.118281 0.00000 0.00000 0.00000 0.00000
16 0.95205 -8.345428 0.00000 0.00000 0.00000 0.00000
17 0.94408 -10.80150 0.00000 0.00000 2.33800 0.84000
18 .944796 -11.36434 0.00000 0.00000 5.22000 1.76000
19 1.00709 -11.18251 0.00000 0.00000 0.00000 0.00000
20 0.95848 -5.588706 0.00000 0.00000 0.00000 0.00000
21 0.98490 -4.339794 0.00000 0.00000 2.74000 1.1500
22 1.01487 0.1908752 0.00000 0.00000 0.00000 0.00000
23 1.01216 -0.081575 0.00000 0.00000 2.74500 0.84660
24 0.97347 -6.801041 0.00000 0.00000 3.08600 0.92200
25 1.02567 -4.973582 0.00000 0.00000 2.24000 0.47200
63
Table 4.8: Load ow data II: 10 machine and 39 bus system
Bus. V P
G
Q
G
P
L
Q
L
No. (pu) (degrees) (pu) (pu) (pu) (pu)
26 1.01236 -6.207839 0.00000 0.00000 1.39000 0.17000
27 0.99181 -8.328811 0.00000 0.00000 2.81000 0.75500
28 1.01636 -2.467062 0.00000 0.00000 2.06000 0.27600
29 1.01874 0.4586534 0.00000 0.00000 2.83500 0.26900
30 0.98423 -2.018678 0.00000 0.00000 6.28000 1.03000
31 0.95493 -6.528843 0.00000 0.00000 0.00000 0.00000
32 0.93491 -6.512507 0.00000 0.00000 0.07500 0.88000
33 0.95597 -6.377243 0.00000 0.00000 0.00000 0.00000
34 0.95464 -8.215876 0.00000 0.00000 0.00000 0.00000
35 0.95675 -8.534196 0.00000 0.00000 3.20000 1.53000
36 0.97367 -6.891177 0.00000 0.00000 3.29400 0.32300
37 0.98141 -8.100629 0.00000 0.00000 0.00000 0.00000
38 0.98149 -9.085413 0.00000 0.00000 1.58000 0.30000
39 0.98489 -1.018596 0.00000 0.00000 0.00000 0.00000
Table 4.9: Transformer data: 10 machine and 39 bus system
Bus No. R
T
X
T
Tap
From To
39 30 0.0007 0.0138 1.0
39 5 0.0007 0.0142 1.0
32 33 0.0016 0.0435 1.0
32 31 0.0016 0.0435 1.0
30 4 0.0009 0.0180 1.0
29 9 0.0008 0.0156 1.0
25 8 0.0006 0.0232 1.0
23 7 0.0005 0.0272 1.0
22 6 0.0000 0.0143 1.0
20 3 0.0000 0.0200 1.0
16 1 0.0000 0.0250 1.0
12 10 0.0000 0.0181 1.0
64
Table 4.10: Line data I: 10 machine and 39 bus system
Bus No. R
L
X
L
B
c
From To
15 16 0.0002 0.0026 0.0434
14 34 0.0008 0.0129 0.1382
14 15 0.0008 0.0128 0.1342
13 38 0.0011 0.0133 0.2138
13 14 0.0013 0.0213 0.2214
12 25 0.0070 0.0086 0.1460
12 13 0.0013 0.0151 0.2572
11 12 0.0035 0.0411 0.6987
11 2 0.0010 0.0250 0.7500
G
i
(40x10) is obtained using DLQR theory[41, 52, 53].
As in the case of 4 machine 10 bus system, here also the minimum value of

N would
have been 64 and that would result 64 gain matrices of 10x10 each. If a periodic output
feedback gain has to be obtained for single system only, then minimum 4 gain matrices
(K
0
, K
1
, K
2
, K
3
) each of 10x10 dimension are needed as controllability index of single
system is 40. So if we restrict our gain sequence to 4, there would not exist a common
K which would satisfy
i
K = G
i
for i = 1....16. Thats why this condition is relaxed
and a robust K has been sought to be obtained by satisfying the equation
i
K = G
i
approximately. Here gain sequences of K are chosen 10 (K
0
, K
1
, K
2
, K
3
, K
4
, K
5
, K
6
,
K
7
,K
8
, K
9
).
Using LMI approach, Eqns.(3.5-3.6) are solved for dierent values of to nd the
robust decentralized gain matrix K. The robust decentralized periodic output feedback
gain matrix K (100x10) is obtained as given as
K
0
=Diag[-0.195 -17.249 -0.410 -6.571 -0.744 -0.238 -0.280 -0.319 -0.344 -0.070],
K
1
=Diag[-0.185 -12.094 -0.371 -5.428 -0.667 -0.186 -0.219 -0.249 -0.269 -0.054],
K
2
=Diag[-0.175 -7.4385 -0.331 -4.289 -0.588 -0.134 -0.157 -0.179 -0.193 -0.039],
K
3
=Diag[-0.164 -3.284 -0.291 -3.154 -0.508 -0.081 -0.096 -0.109 -0.118 -0.024],
K
4
=Diag[-0.153 0.362 -0.252 -2.023 -0.426 -0.029 -0.034 -0.038 -0.041 -0.008],
65
Table 4.11: Line data II: 10 machine and 39 bus system
Bus No. R
L
X
L
B
c
From To
37 27 0.0013 0.0173 0.3216
37 38 0.0007 0.0082 0.1319
36 24 0.0003 0.0059 0.0680
36 21 0.0008 0.0135 0.2548
36 39 0.0016 0.0195 0.3040
36 37 0.0007 0.0089 0.1342
35 36 0.0009 0.0094 0.1710
34 35 0.0018 0.0217 0.3660
33 34 0.0009 0.0101 0.1723
28 29 0.0014 0.0151 0.2490
26 29 0.0057 0.0625 1.0290
26 28 0.0043 0.0474 0.7802
26 27 0.0014 0.0147 0.2396
25 26 0.0032 0.0323 0.5130
23 24 0.0022 0.0350 0.3610
22 23 0.0006 0.0096 0.1846
21 22 0.0008 0.0135 0.2548
20 33 0.0004 0.0043 0.0729
20 31 0.0004 0.0043 0.0729
19 2 0.0010 0.0250 1.2000
18 19 0.0023 0.0363 0.3804
17 18 0.0004 0.0046 0.0780
16 31 0.0007 0.0082 0.1389
16 17 0.0006 0.0092 0.1130
15 18 0.0008 0.0112 0.1476
66
K
5
=Diag[-0.141 3.498 -0.212 -0.896 -0.343 0.0239 0.0281 0.0320 0.034 0.007],
K
6
=Diag[-0.128 6.119 -0.173 0.226 -0.258 0.0771 0.0905 0.103 0.111 0.022],
K
7
=Diag[-0.115 8.221 -0.134 1.345 -0.172 0.130 0.153 0.174 0.188 0.038],
K
8
=Diag[-0.101 9.799 -0.095 2.460 -0.083 0.184 0.216 0.246 0.265 0.054],
K
9
=Diag[-0.087 10.850 -0.056 3.571 0.0062 0.237 0.279 0.318 0.343 0.069].
The closed loop responses with this robust decentralized gain K for all the linearized
models are satisfactory and able to stabilize the outputs. The eigen values of (
N
+KC)
are found to be within the unit circle.
4.7 Simulation with nonlinear model
In a similar way as discussed in Section 4.4, a SIMULINK based block diagram including
all the nonlinear blocks is generated [47] and the simulations are carried out for 4 machine
10 bus system and 10 machine and 39 bus system.
4.7.1 4 Machine 10 bus system
In 4 machine 10 bus system, the disturbance considered is a self clearing fault at dierent
bus as shown in Table 4.12 cleared after 0.1 second. The limits of PSS output are taken as
0.1. In decentralized PSS, to activate the proposed controller at the same instant, proper
synchronization signal is required to be sent to all machines. All PSSs are to be applied
simultaneously to the respective machine. If there is some delay in synchronization,
the proposed controller can still damp out the oscillations but may give deteriorating
performance as shown in Fig. 4.11 to Fig. 4.18.
The location of fault considered for dierent models are as given in Table 4.12
Simulation results of dierent generators with fault are shown in Fig. 4.3 to Fig. 4.10
without controller and with controller and the same with fault and successive delayed
output of 0.1 second at all generators are shown in Fig. 4.11 to Fig. 4.18. As shown in
the plots, the proposed controller is able to damp out the oscillations in 6 to 8 seconds
after clearing the fault.
67
Table 4.12: Location of fault: 4 machine and 10 bus system
S.No. Model Fault at Bus
1 Model 1 Bus 1
2 Model 2 Bus 2
3 Model 3 Bus 5
4 Model 4 Bus 6
5 Model 5 Bus 7
6 Model 6 Bus 8
7 Model 7 Bus 9
8 Model 8 Bus 10
9 Model 9 Bus 3
10 Model 10 Bus 4
11 Model 11 Bus 5
12 Model 12 Bus 6
13 Model 13 Bus 7
14 Model 14 Bus 8
15 Model 15 Bus 9
16 Model 16 Bus 10
68
0 1 2 3 4 5 6 7 8 9 10
0.03
0.02
0.01
0
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
0.03
0.02
0.01
0
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 1: fault at bus 1
Model 2: fault at bus 2
Figure 4.3: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
5
0
5
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 3: fault at bus 5
Model 4: fault at bus 6
Figure 4.4: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
69
0 1 2 3 4 5 6 7 8 9 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 5: fault at bus 7
Model 6: fault at bus 8
Figure 4.5: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
6
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
5
0
5
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 7: fault at bus 9
Model 8: fault at bus 10
Figure 4.6: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
70
0 1 2 3 4 5 6 7 8 9 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 9: fault at bus 3
Model 10: fault at bus 4
Figure 4.7: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
6
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 11: fault at bus 5
Model12: fault at bus 6
Figure 4.8: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
71
0 1 2 3 4 5 6 7 8 9 10
5
0
5
10
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 13: fault at bus 7
Model 14: fault at bus 8
Figure 4.9: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
6
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
5
0
5
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 15: fault at bus 9
Model 16: fault at bus 10
Figure 4.10: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(4 machine and 10 bus system)
72
0 1 2 3 4 5 6 7 8 9 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 1: fault at bus 1
Model 2: fault at bus 2
Figure 4.11: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 3: fault at bus 5
Model 4: fault at bus 6
Figure 4.12: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback controller(4 machine and 10 bus system)
73
0 1 2 3 4 5 6 7 8 9 10
5
0
5
10
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 5: fault at bus 7
Model 6: fault at bus 8
Figure 4.13: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
2
1
0
1
2
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 7: fault at bus 9
Model 8: fault at bus 10
Figure 4.14: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10 bus system)
74
0 1 2 3 4 5 6 7 8 9 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
0.03
0.02
0.01
0
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 9: fault at bus 3
Model 10: fault at bus 4
Figure 4.15: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
6
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 11: fault at bus 5
Model 12: fault at bus 6
Figure 4.16: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10 bus system)
75
0 1 2 3 4 5 6 7 8 9 10
5
0
5
10
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 13: fault at bus 7
Model 14: fault at bus 8
Figure 4.17: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback Controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
6
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
5
0
5
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 16: fault at bus 10
Model 15: fault at bus 9
Figure 4.18: Open and closed loop responses with fault and synchronizing delay using
robust decentralized periodic output feedback controller(4 machine and 10 bus system)
76
4.7.2 10 Machine 39 bus system
In this case also the disturbance considered is a self clearing fault at dierent buses as
shown in Table 4.13 cleared after 0.1 second. The limits of PSS output are taken as 0.1.
The location of fault considered for dierent models is given in Table 4.13:
Table 4.13: Location of fault: 10 machine and 39 bus system
S.No. Model Fault at Bus
1 Model 1 Bus16
2 Model 2 Bus13
3 Model 3 Bus 9
4 Model 4 Bus 7
5 Model 5 Bus 17
6 Model 6 Bus19
7 Model 7 Bus 21
8 Model 8 Bus 23
9 Model 9 Bus 32
10 Model 10 Bus 34
11 Model 11 Bus 35
12 Model 12 Bus 26
13 Model 13 Bus 25
14 Model 14 Bus 27
15 Model 15 Bus 29
16 Model 16 Bus 31
Simulation results of dierent generators are shown in Fig. 4.19 to Fig. 4.26 without
controller and with controller with fault at particular bus. As shown in plots, the proposed
controller is able to damp out the oscillations in 8 to 12 seconds after clearing the fault.
77
0 5 10 15
0.06
0.04
0.02
0
0.02
0.04
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller
Model 2: fault at bus 13
Model 1: fault at bus 16
Figure 4.19: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(10 machine and 39 bus system)
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 3: fault at bus 9
Model 4: fault at bus 7
Figure 4.20: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(10 machine and 39 bus system)
78
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 5: fault at bus 17
Model 6: fault at bus 19
Figure 4.21: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(10 machine and 39 bus system)
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 7
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 8
without controller
with controller
Model 7: fault at bus 21
Model 8: fault at bus 23
Figure 4.22: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(10 machine and 39 bus system)
79
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 9
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 10
without controller
with controller
Model 10: fault at bus 34
Model 9: fault at bus 32
Figure 4.23: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(10 machine and 39 bus system)
0 5 10 15
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller Model 12: fault at bus 26
Model 11: fault at bus 35
Figure 4.24: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(10 machine and 39 bus system)
80
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 13: fault at bus 25
Model 14: fault at bus 27
Figure 4.25: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller(10 machine and 39 bus system)
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 15: fault at bus 29
Model 16: fault at bus 31
Figure 4.26: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller (10 machine and 39 bus system)
81
4.8 Fast output sampling feedback technique
Consider a linear continuous time-invariant 4 machine 10 bus, MIMO system represented
by
x = Ax +Bu,
y = Cx.
(4.11)
A discrete time invariant system with sampling interval sec. can be represented as
x(k + 1) =

x(k) +

u(k), (4.12)
y(k) = Cx(k).
Let u(k) = Fx(k) be a stabilizing control for the system in Eqn.(4.12). Thus closed
loop system(4.12) (

F) becomes stable and has no eigenvalues at the origin.


Using this F, LMI constraints given in Eqns.(2.65-2.66) are solved for dierent values
of to nd the centralized gain matrix.
4.9 Case study IV: Centralized PSS design for
multimachine power system (4 machine 10 bus)
by fast output sampling feedback technique
As discussed in Section 4.3, the discrete model is obtained for sampling time = 0.1 sec.
Using the method discussed in Section 4.8 stabilizing state feedback gain matrix F
(16 x 4) is obtained.
Using LMI approach, Eqns.(2.65-2.66) are solved for dierent values
1
,
2
and
3
to
nd the centralized gain matrix, L. The centralized fast output sampling feedback gain
matrix L (4x16) is obtained as
82
L=
_

_
111.48 22.3578 3.52790 6.40480
92.5191 98.652 4.84100 8.19710
2.63120 5.15920 113.16 20.3053
4.37680 7.77480 92.2779 98.785
19.1243 81.6692 0.92200 1.66980
79.1386 82.3857 2.00450 3.40070
0.69300 1.31650 19.2098 82.0006
1.70880 2.98040 79.6398 83.3067
115.973 99.6258 0.1390 0.4040
46.9279 199.064 0.37720 0.62510
0.40590 0.48550 117.521 102.117
0.82920 1.35960 48.4882 201.774
9.8230 196.62 3.8357 6.8838
215.03 163.23 6.6951 11.363
3.3275 6.2563 9.8558 197.46
6.4560 11.329 216.91 166.78
_

_
T
The closed loop responses with this centralized gain L for the linearized model are
satisfactory and able to stabilize the outputs. The eigen values of (
N
+LC) are found
to be within the unit circle and given in Table 4.14.
Table 4.14: Eigen values of the closed loop system (
N
+ LC)
0.9888 + 0.0906i 0.9888 - 0.0906i
0.9883 +0.0900i 0.9883 - 0.0900i
0.9077 + 0.0661i 0.9077 - 0.0661i
0.9072 +0.0623 0.9072 - 0.0623
0.9926 + 0.0177i 0.9926 - 0.0177i
0.9058 0.9058
0.9977 0.9858
0.5497 0.5754
83
0 5 10
1
0.5
0
0.5
1
x 10
6
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
0 5 10
5
0
5
10
x 10
7
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
0 5 10
1
0.5
0
0.5
1
x 10
6
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
0 5 10
1
0.5
0
0.5
1
x 10
6
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
with controller
with controller
with controller
without controller
without controller
without controller
without controller
Figure 4.27: Open and closed loop responses using centralized fast output sampling feed-
back controller(4 machine and 10 bus system)
4.10 Simulation with nonlinear model
A SIMULINK based block diagram including all the nonlinear blocks is generated using
machine model 1.0 [47]. The slip of the machine is taken as output. The output slip signal
with centralized gain L and a limiter is added to V
ref
signal. This is used to damp out
the small signal disturbances via modulating the generator excitation. The output must
be limited to prevent the PSS acting to counter action of AVR. Dierent operating points
are taken as the dierent models. The disturbance considered is a self clearing fault at
particular bus (Bus No.3) cleared after 0.1 second. The limits of PSS output are taken
as 0.1.
Simulation results of dierent generators are shown in Fig. 4.27 for open and closed
loop system. The simulation are also carried out with fault and shown in Fig. 4.28. As
shown in plots, the proposed controller is able to damp out the oscillations in 5 to 6
seconds after clearing the fault.
84
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
0 5 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
with controller
with controller
with controller
without controller
without controller
without controller
without controller
Figure 4.28: Open and closed loop responses with fault at bus 3 using centralized fast
output sampling feedback controller(4 machine and 10 bus system)
4.11 Decentralized fast output sampling feedback
technique for multimodel
Let us consider a family of plant S = {A
i,
B
i,
C
i
} ,dened by
.
x= A
i
x +B
i
u, (4.13)
y = C
i
x, i = 1, ......M. (4.14)
The discrete time invariant systems with sampling interval sec. can be represented
as
x(k + 1) =
i
x(k) +
i
u(k), (4.15)
y(k) = C
i
x(k).
Let u(k) = Fx be a stabilizing control for the reduced order system in Eqn.(4.15).
Thus closed loop reduced system(4.15) (
i
+
i
F) becomes stable and has no eigenvalues
at the origin.
Using this F, If the LMI constraints given in Eqns.(3.17-3.18) are solved, the robust
fast output sampling feedback gain matrix may become full[36]. This results in the control
input of each machine being a function of outputs of all machines. As robust decentralized
85
control scheme may be more feasible than the centralized control scheme, robust decen-
tralized fast output sampling feedback control is obtained by making the o diagonal
elements of L
0
, L
1
, , L
N1
matrices zero.
So the structure of L
i
( i = 0, , N-1) matrices for 4 machine and 10 bus power
system (machine model 1.0) are assumed as
L
i
=
_

_
l
i11
0 0 0
0 l
i22
0 0
0 0 l
i33
0
0 0 0 l
i44
_

_
, i = 0, , N 1. (4.16)
and for 10 machine and 39 bus power system (machine model 1.0) as
L
i
=
_

_
l
i11
0 0 0 0 0 0 0 0 0
0 l
i22
0 0 0 0 0 0 0 0
0 0 l
i33
0 0 0 0 0 0 0
0 0 0 l
i44
0 0 0 0 0 0
0 0 0 0 l
i55
0 0 0 0 0
0 0 0 0 0 l
i66
0 0 0 0
0 0 0 0 0 0 l
i77
0 0 0
0 0 0 0 0 0 0 l
i88
0 0
0 0 0 0 0 0 0 0 l
i99
0
0 0 0 0 0 0 0 0 0 l
i1010
_

_
, i = 0, , N 1. (4.17)
With this structure of L
i
, the problem can be formulated in the framework of Linear
Matrix Inequalities using Eqns.(3.17-3.18) and the desired matrices can be obtained. Now
it is evident that the control input of each machine is a function of that machine only and
this makes the fast output sampling based power system stabilizer design a decentralized
one.
86
4.12 Robust decentralized PSS design for
multimachine power system by fast output
sampling feedback technique
4.12.1 Case study V: 4 machine 10 Bus system
As discussed in Section 4.6.1 using machine model 1.0, the discrete models are obtained
for sampling time = 0.1 sec.
Using the method discussed in Section 4.11, a common stabilizing state feedback gain
matrix F (4 x16) is obtained.
This fast output sampling feedback gain can be obtained which approximately realizes
the designed F for all models of the family. It has been shown in [38], that the observability
of individual models generically implies the observability of the augmented system. For
the application mentioned in the chapter, actually the minimum value of

N would have
been 64 (because

N observability index of the augmented system/number of inputs or
outputs) and that would result 64 gain matrices of 4x4 each. However, in this method, the
elements of gain matrices may turn out to be high which may magnify the noise and also
implementation would be complex. To reduce this eect of gain, we relax the condition
such that the Eqn. (3.16) satises exactly and include a constraint on the gain.
The stabilization of 4 machine and 10 bus system at sixteen dierent conditions with
dierent generating power, dierent bus structure and dierent load distribution leading
to eight models is done as follows:
Consider the 4 machine and 10 bus system at one condition leading to one 16
th
or-
der model. A state feedback gain matrix can be obtained using LMI theory. If a fast
output sampling feedback gain has to be obtained for this system only, then minimum 4
gain matrices (L
0
,L
1
,L
2
,L
3
) each of 4x4 dimension are needed which will exactly realize
the state feedback gain computed for the system. The equation (3.15) has solution if
N observability index of this system/number of inputs or outputs. For one system
controllability index is 16. So minimum 4 gain sequences are needed.
Here as we are dealing with robust stabilization, we have to nd a L which will satisfy
LC
i
= F, (i = 1, , 16). However, if we restrict our gain sequence to 4, there would
not exist a common L which would satisfy LC
i
= F for i = 1, , 16. Thats why this
condition is relaxed and a robust L has been sought to be obtained by satisfying the
equation LC
i
= F approximately.
87
Using LMI approach, Eqns.(3.17-3.18) are solved for dierent values
1
,
2
and
3
to nd the robust decentralized gain matrix, L. The robust decentralized fast output
sampling feedback gain matrix L (4x16) is obtained as given below:
L
0
=
_

_
26.6968 0 0 0
0 26.6968 0 0
0 0 26.9161 0
0 0 0 26.9161
_

_
,L
1
=
_

_
9.7679 0 0 0
0 9.7679 0 0
0 0 10.3681 0
0 0 0 10.3681
_

_
,
L
2
=
_

_
0.9861 0 0 0
0 0.9861 0 0
0 0 2.1116 0
0 0 0 2.1116
_

_
,L
3
=
_

_
2.3242 0 0 0
0 2.3242 0 0
0 0 0.7504 0
0 0 0 0.7504
_

_
.
The closed loop responses with this robust decentralized gain L for all the linearized
models are satisfactory and are able to stabilize the outputs. The eigen values of (
N
+
LC) are found to be within the unit circle.
4.12.2 Case study VI: 10 machine 39 bus system
As discussed in Section 4.6.2, the discrete models are obtained for sampling time = 0.1
sec.
Using the method discussed in section 4.11, a common stabilizing state feedback gain
matrix F (10 x40) is obtained.
As in the case of 4 machine 10 bus system, here also the minimum value of

N would
have been 64 and that would result 64 gain matrices of 10x10 each. If a fast output
sampling feedback gain has to be obtained for single system only, then minimum 4 gain
matrices (L
0
,L
1
,L
2
,L
3
) each of 10x10 dimension are needed. For one system controllability
index is 40. So minimum 4 gain sequences are needed.
So if we restrict our gain sequence to 4, there would not exist a common L which
would satisfy LC
i
= F, for i = 1....16. Thats why this condition is relaxed and a robust
L has been sought to be obtained by satisfying the equation LC
i
= F, approximately.
Here the gain sequences of L are chosen 10 (L
0
, L
1
, L
2
, L
3
, L
4
, L
5
, L
6
, L
7
,L
8
, L
9
).
Using LMI approach, Eqns.(3.17-3.18) are solved for dierent values
1
,
2
and
3
to nd the robust decentralized gain matrix, L. The robust decentralized fast output
sampling feedback gain matrix L (10x100) is obtained as given as
L
0
=Diag[40.142 41.930 -50.524 4.348 -12.844 -37.016 -27.653 -26.207 -20.188 -6.320],
88
L
1
=Diag[-128.082 -110.360 178.522 -14.590 31.377 32.804 23.907 21.7127 16.537 6.332],
L
2
=Diag[50.600 -4.882 -107.846 5.179 11.353 49.192 37.120 35.756 27.655 8.044],
L
3
=Diag[112.503 95.188 -156.148 15.093 -49.264 31.637 24.748 25.185 19.761 4.104],
L
4
=Diag[32.968 102.699 12.297 5.717-36.278 -2.694 -1.064 0.450 0.678 -1.763],
L
5
=Diag[-71.169 25.198 162.766 -10.115 39.128 39.128 -28.605 -27.168 -20.955 -6.882],
L
6
=Diag[-95.293 -82.825 126.778 -15.241 74.489 -60.330 -46.447 -46.042 -35.962 -9.277],
L
7
=Diag[-17.624 -147.119 -78.670 -2.777 -3.083 -55.251 -43.367 -44.227 -34.812 -7.446],
L
8
=Diag[77.082 -88.350 -220.456 14.279 -113.798 -9.608 -8.283 -9.5576 -7.741 -0.203],
L
9
=Diag[-0.509 170.770 134.634 -1.218 60.736 89.621 69.784 70.257 55.161 13.427].
The closed loop responses with this robust decentralized gain L for all the linearized
models are satisfactory and are able to stabilize the outputs. The eigen values of (
N
+
LC) are found to be within the unit circle.
4.13 Simulation with nonlinear model
A SIMULINK based block diagram including all the nonlinear blocks is generated using
machine model 1.0[47]. The output slip signal with robust decentralized gain L and a
limiter is added to V
ref
signal. The output must be limited to prevent the PSS acting to
counter action of AVR. Dierent operating points are taken as the dierent models.
4.13.1 4 Machine 10 bus system
In 4 machine 10 bus system,the disturbance considered is a self clearing fault at dierent
bus cleared after 0.1 second. The limits of PSS output are taken as 0.1. The location
of fault considered for dierent models are given in Table 4.12. In decentralized PSS, to
activate the proposed controller at same instant, proper synchronization signal is required
to be sent to all the machines. All PSSs are to be applied simultaneously to the respective
machines.
Simulation results of dierent generators with fault are shown in Fig.4.33 to Fig. 4.36
without controller and with controller. As shown in plots, the proposed controller is able
to damp out the oscillations in 4 to 5 seconds after clearing the fault.
89
0 1 2 3 4 5 6 7 8 9 10
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
15
10
5
0
5
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 1: fault at bus 1
Model 2: fault at bus 2
Figure 4.29: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
0.06
0.04
0.02
0
0.02
0.04
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 3: fault at bus 5
Model 4: fault at bus 6
Figure 4.30: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
90
0 1 2 3 4 5 6 7 8 9 10
5
0
5
10
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 5: fault at bus 7
Model 6: fault at bus 8
Figure 4.31: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
6
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 7: fault at bus 9
Model 8: fault at bus 10
Figure 4.32: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
91
0 1 2 3 4 5 6 7 8 9 10
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
0.03
0.02
0.01
0
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 9: fault at bus 3
Model 10: fault at bus 4
Figure 4.33: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
0.04
0.02
0
0.02
0.04
0.06
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 11: fault at bus 5
Model 12: fault at bus 6
Figure 4.34: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
92
0 1 2 3 4 5 6 7 8 9 10
5
0
5
10
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
with controller
without controller
0 1 2 3 4 5 6 7 8 9 10
4
2
0
2
4
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
with controller
without controller
Model 13: fault at bus 7
Model 14: fault at bus 8
Figure 4.35: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
0 1 2 3 4 5 6 7 8 9 10
5
0
5
10
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
with controller
with controller
0 1 2 3 4 5 6 7 8 9 10
10
5
0
5
x 10
3
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
with controller
without controller
Model 15: fault at bus 9
Model 16: fault at bus 10
Figure 4.36: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(4 machine and 10 bus system)
93
0 5 10 15
0.1
0.05
0
0.05
0.1
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller
Model 1: fault at bus 16
Model 2: fault at bus 13
Figure 4.37: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
4.13.2 10 Machine 39 bus system
For 10 machine 39 bus system, the disturbance considered is a self clearing fault at dierent
bus cleared after 0.1 second. The limits of PSS output are taken as 0.1. The location
of fault considered for dierent models are given in Table 4.13.
Simulation results of dierent generators with fault are shown in Fig. 4.37 to Fig. 4.44
without controller and with controller. As shown in plots, the proposed controller is able
to damp out the oscillations in 7 to 10 seconds after clearing the fault.
4.14 Concluding remarks
In this chapter, 4 machine 10 bus system is considered for the design of centralized power
system stabilizers. The slip signal is taken as output. The centralized power system sta-
bilizers are designed using periodic output feedback and fast output sampling feedback
techniques and applied to the linear as well as nonlinear system. The simulations show
that these output feedback control techniques can be used to design power system sta-
bilizer for 4 machine and 10 bus power system. These methods are more general than
static output feedback and also leads to single controller for various operating points of 4
machine and 10 bus system.
94
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 3: fault at bus 9
Model 4: fault at bus 7
Figure 4.38: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
0 5 10 15
0.03
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.03
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 6: fault at bus 19
Model 5: fault at bus 17
Figure 4.39: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
95
0 5 10 15
0.03
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 7
without controller
with controller
0 5 10 15
0.03
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 8
without controller
with controller
Model 7: fault at bus 21
Model 8: fault at bus 23
Figure 4.40: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 9
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 10
without controller
with controller
Model 9: fault at bus 32
Model 10: fault at bus 34
Figure 4.41: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
96
0 5 10 15
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller
Model 11: fault at bus 35
Model 12: fault at bus 26
Figure 4.42: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 13: fault at bus 25
Model 14: fault at bus 27
Figure 4.43: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
97
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 15: fault at bus 29
Model 16: fault at bus 31
Figure 4.44: Open and closed loop responses with fault using decentralized fast output
sampling feedback controller(10 machine and 39 bus system)
The chapter also deals with the design of decentralized power system stabilizers for the
multimachine power system viz 4 machine 10 bus and 10 machine 39 bus systems. The
design schemes of robust power system stabilizer for both multimachine power systems
using decentralized periodic output feedback and fast output sampling feedback have
been developed. It is found that the designed robust decentralized controllers provide
good damping enhancement for various operating points of both cases.
The proposed methods result satisfactory response behaviour to damp out the oscilla-
tions. The robust decentralized control can be applied simultaneously to all the machines.
The input applied to each machine is a function of output of the respective machine only.
Thus the applied control schemes are robust and decentralized in nature, whereas the
convectional power system stabilizers are designed sequentially. The conventional power
system stabilizer is dynamic in nature and is required to be tuned according to power char-
acteristics, whereas, the proposed controllers gains are static in nature and one controller
structure is able to damp out the oscillations for all models.
The centralized and robust decentralized fast output sampling feedback methods pro-
vide satisfactory damping enhancement for various operating points of multimachine sys-
tems compared to centralized and robust decentralized periodic output feedback method.
The control input required for these models are of smaller magnitudes in fast output
98
sampling feedback method compared to periodic output feedback methods.
As shown in plots, the centralized fast output sampling feedback controller requires
to damp out the oscillations in 5 to 6 seconds after clearing the fault whereas, centralized
periodic output feedback controller is able to damp out the oscillations in 8 to 10 seconds.
Also shown in plots, the decentralized fast output sampling feedback controller requires
to damp out the oscillations in 4 to 5 seconds after clearing the fault whereas, decentralized
periodic output feedback controller is able to damp out the oscillations in 6 to 8 seconds
for 4 machine with 10 bus power system.
Further the decentralized fast output sampling feedback controller requires to damp
out the oscillations in 7 to 10 seconds after clearing the fault whereas, decentralized
proposed periodic output feedback controller is able to damp out the oscillations in 8 to
12 seconds for 10 machine with 39 bus power system.
The selection of faults at dierent buses are considered randomly. The nonlinear
simulations with proposed controllers for all generators using these fault conditions give
satisfactory results. some of the results are presented in respective case studies.
99
.
100
Chapter 5
Design of PSS for Multimachine
Power System via Reduced order
Model
5.1 Introduction
For many processes (like large power system, chemical plants and nuclear reactors) the
order of the state matrix may be quite large. It would be dicult to work with these
large systems in their original form. In such cases, it is common to study the process by
approximating it to a simpler model. These mathematical models correspond to approxi-
mating a system by its dominant pole-zeros in the complex plane. They generally require
empirical determination of the system parameters. Many dierent methods have been de-
veloped to accomplish the purpose by estimating the dominant part of the large system
and nding a simpler (or reduced order) system representation that has its behavior akin
to the original system. This chapter is organized as follows.
In Section 5.2, a brief review on system reduction method based on dominant eigen-
value retention is presented. In Section 5.3, decentralized periodic output feedback tech-
nique via reduced order model for multimodel system is discussed. In Section 5.4, the
design of robust decentralized PSS for multimachine power system (10 machine 39 bus
system) using periodic output feedback technique via reduced order model is presented
whereas Section 5.5 contains the simulations with nonlinear model using proposed periodic
output feedback controller. In Section 5.6, decentralized fast output sampling feedback
technique via reduced order model for multimodel system is discussed. In Section 5.7, the
101
design of robust decentralized PSS for multimachine power system (10 machine 39 bus
system) using fast output sampling feedback technique via reduced order model is pre-
sented, whereas Section 5.8 contains the simulations with nonlinear model using proposed
fast output sampling feedback controller followed by concluding remarks.
5.2 Model order reduction based on dominant modes
retention
It is usually possible to describe the dynamics of physical systems by a number of simul-
taneous linear dierential equations with constant coecients.
x = Ax +Bu, where A = n n matrix (5.1)
y = Cx. (5.2)
The simulation and design of controllers become very cumbersome if the order of the
system goes high. One way to overcome this diculty is to develop a reduced model of the
higher order system. One of the well known technique is based on dominant eigenvalue
retention. This is explained in the following section.
5.2.1 Davison technique
A system of higher order can be numerically approximated to one of smaller order by
Davisons method [45]. The method suggests that a large (n n) system can be reduced
to a simpler (r r) model (r n) by considering the eects of the r most dominant
(dominant in the sense of being closest to instability) eigenvalues alone. The principle of
the method is to neglect eigenvalues of the original system that are farthest from the origin
and retain only dominant eigenvalues and hence dominant time constants of the original
system in the reduced model. This implies that the overall behavior of the approximate
system will be very similar to that of the original system since the contribution of the
unretained eigenvalues to the system response are important only at the beginning of the
response, whereas the eigenvalues retained are important throughout the whole of the
response. This method is briey explained in the following.
For the system represented Eqns.(5.1-5.2), consider the linear transformation
x = Pz, (5.3)
102
which transforms the model Eqn.(5.1-5.2) into the following form:
z =

Az +

Bu, where

A = r r matrix (5.4)
y =

Cz, (5.5)
and

A = P
1
AP, (5.6)

B = P
1
B, and

C = CP;

A is in the diagonal form as

A = diag [
1
,
2
, . . . ,
n
] , (5.7)
and
Re(
1
) Re(
2
) Re(
n
). (5.8)
Further assume that r eigenvalues are only dominant; i.e., the order of the reduced
model is r, and partition the model in Eqn.(5.4-5.5) as
z
1
=

A
1
z
1
+

B
1
u, (5.9)
z
2
=

A
2
z
2
+

B
2
u, (5.10)
and
y =

C
1
z
1
+

C
2
z
2
, (5.11)
where

A
1
= diag [
1
,
2
, . . . ,
r
] , (5.12)

A
2
= diag [
r+1
,
r+2
, . . . ,
n
] , (5.13)

B
1
= first r rows of

B, (5.14)
and

B
2
= remaining (n r) rows of

B, (5.15)
and are, respectively, r r, (n r) (n r), r m, and (n r) m matrices obtained
by partitioning of

A and

B suitably. In Eqns.(5.9-5.10), the order of z
1
is r and that of
z
2
is (n r).
103
Now, because the contribution of the modes represented by the eigenvalues
r+1
,
r+2
,. . .,
n
is not signicant, it may be assumed that z
2
=0, whereby we have from Eqn.(5.3)
_
_
x
1
x
2
_
_
=
_
_
P
11
P
21
_
_
z
1
, (5.16)
where P
11
and P
21
are, respectively, r r and (n r)r submatrices obtained by par-
titioning of P, and z
1
and z
2
are, respectively, r and n r dimensional state vectors
corresponding to the original state variables. It follows from Eqn.(5.16) that
z
1
= P
1
11
x
1
, (5.17)
with which the model in Eqn.(5.9) can be transformed to
x
1
= P
11

A
1
P
1
11
x
1
+P
11

B
1
u = A
r
x
1
+B
r
u, (5.18)
and
y =

C
1
P
1
11
x
1
= C
r
x
1
. (5.19)
Moreover, from Eqns.(5.16-5.17), we have
x
2
= P
21
P
1
11
x
1
. (5.20)
Thus, the original n
th
order model represented by Eqns.(5.1-5.2) is reduced to an r
th
order model given by Eqns.(5.18-5.19). The state variables of the approximate model are
the same as the rst r state variables of the original higher-order model. The remaining
state variables are given in terms of the rst r state variables by Eqn.(5.20).
5.3 Decentralized periodic output feedback control
for multimodel system via reduced order model
Let us consider a family of plant S = {A
i,
B
i,
C
i
}, dened by
.
x= A
i
x +B
i
u, (5.21)
y = C
i
x, i = 1, ......M. (5.22)
The discrete time invariant systems with sampling interval sec. can be represented
as
x(k + 1) =
i
x(k) +
i
u(k), (5.23)
y(k) = C
i
x(k).
104
The adjoint or the dual for the above systems would be
x(k + 1) =
T
i
x(k) +C
T
i
u(k), (5.24)
y(k) =
T
i
x(k).
There exists a transformation V
i
, such that,
x = V
i
z, (5.25)
transforms the above system into the following block modal form
z(k + 1) =

i
z(k) +

C
i
u(k), (5.26)
y(k) =

i
z(k),
where,

i
=
_
_

1i
0
0
2i
_
_
,

C
i
=
_
_
C
1i
C
2i
_
_
, (5.27)

i
=
_

1i

2i
_
,
and the eigenvalues are arranged in their order of dominance.
We now extract an r
th
order model, retaining the r dominant eigenvalues, by truncating
the above systems. Using Eqn.(5.26-5.27), we get,
z
r
(k + 1) =
1i
z
r
(k) +C
1i
u(k), (5.28)
y
r
(k) =
1i
z
r
(k).
Let u(k) = S
ri
z
r
be a stabilizing control for the reduced order model in Eqn. (5.28).
Thus the closed loop reduced model (
1i
+C
1i
S
ri
) becomes stable. Now,
z
r
=
_
I
r
: 0
r(nr)

z, (5.29)
=
_
I
r
: 0
r(nr)

V
1
i
x.
Therefore, we get,
u(k) = S
ri
_
I
r
: 0
r(nr)

V
1
i
x, (5.30)
S
i
x,
105
which makes the closed loop systems
_

T
i
+C
T
i
S
i
_
stable. But the eigenvalues of (
T
i
+
C
T
i
S
i
) and
_

T
i
+C
T
i
S
i
_
T
are the same. So
_

i
+S
T
i
C
i
_
will also be stable. Thus
S
T
i
G
i
is the output injection gain for the system in Eqn.(5.23).
If the LMI constraints given in Eqns.(3.5-3.6) are solved using the above G
i
, the robust
periodic output feedback gain matrix may become full [36]. This results in the control
input of each machine being a function of outputs of all machines. But, as discussed in
Chapter 4 that the decentralized robust control scheme may be more feasible and useful
than the centralized control scheme and the same can be achieved by making the o
diagonal elements of K
0
, K
1
, , K

N1
matrices zero. The structure of K
i
( i = 0, ,
N-1) matrices for 10 machine and 39 bus power system are assumed as given in Eqn.(4.10).
5.4 Case study I: Robust decentralized PSS design
for multimachine power system by periodic out-
put feedback technique via reduced order model
The sixteen models of multimachine system are considered for designing robust periodic
output feedback controller via reduced order model using LMI approach of MATLAB.
Then controller gains are applied to simulate a nonlinear model of multimachine system
dynamics at dierent operating points.
The sixteen models of 10 machine 39 bus power system with dierent generating power,
dierent bus structure and dierent load distribution (60 to 100 percent variations from
peak loads) are obtained. The single line diagram of the system is shown in Fig.2.6.
The machine data, line data and load ow data are given in Table 4.6 to Table 4.11
[47].
The above models of the multimachine power system are modeled using SIMULINK
Toolbox of MATLAB and the linear state space models are obtained for the same. Then
discrete models are obtained for sampling time = 0.1 sec. The reduced order models of
the order of 20 are computed from the adjoint discrete models as discussed in Section 5.3.
Using the method discussed in Section 5.3 stabilizing gain matrices S
ri
(10x20) can
be obtained for reduced order model using DLQR theory[41, 52, 53]. Using aggregation
techniques, the output injection gain G
i
(40x10) can be calculated for the higher order
(actual) models.
The periodic output feedback gain can be obtained which approximately realizes the
106
designed G
i
for all the models of the family.
Consider the described system at one operating condition leading to one 40
th
order
model. As discussed above, an output injection gain matrix can be obtained for this
system via reduced order model. If a periodic output feedback gain has to be obtained for
this system only, then minimum 4 gain matrices (K
0
, K
1
, K
2
, K
3
) each of 10x10 dimension
are needed, which will exactly realize the output injection gain computed for this system.
The equation (2.53) has a solution if N controllability index of this system/number
of inputs or outputs. For one system controllability index is 40. So minimum 4 gain
sequences are needed.
Here as we are dealing with robust stabilization, we have to nd a K which will satisfy

i
K = G
i
, (i = 1, , 16), all these equations. So, if we restrict our gain sequence to
4, there would not exist a common K which would satisfy
i
K = G
i
for i = 1, , 16.
Thats why this condition is relaxed and a robust K has been sought to be obtained by
satisfying the equation
i
K = G
i
approximately. Here gain sequences of K are chosen 10
(K
0
, K
1
, K
2
, K
3
, K
4
, K
5
, K
6
, K
7
,K
8
, K
9
).
Using the G
i
, LMI constraints given in Eqns.(3.5 - 3.6) are solved for dierent values
of to nd the robust decentralized gain matrix K for the actual models. The robust
decentralized periodic output feedback gain matrix K (100x10) via reduced order model
is obtained as given as
K
0
=Diag[0.0499 0.0419 0.0327 0.0224 0.0109 -0.0017 -0.0155 -0.0304 -0.0465 -0.0637],
K
1
=Diag[-0.0146 -0.0147 -0.0136 -0.0112 -0.0076 -0.0026 0.0036 0.0111 0.0199 0.0299],
K
2
=Diag[0.0439 0.0072 -0.0193 -0.0358 -0.0422 -0.0385 -0.0247 -0.0008 0.0331 0.0772],
K
3
=Diag[0.0541 -0.0067 -0.0492 -0.0734 -0.0793 -0.0669 -0.0362 0.0128 0.0801 0.1657],
K
4
=Diag[-0.0675 -0.0373 -0.0127 0.006 0.0192 0.0266 0.0283 0.0242 0.0144 -0.0011],
K
5
=Diag 1.0e-003 *[0.1340 0.1345 0.1350 0.1356 0.1361 0.1367 0.1372 0.1378 0.1383
0.1389],
K
6
=Diag[0.0499 0.0498 0.0498 0.0498 0.0497 0.0497 0.0497 0.0496 0.0496 0.0496],
K
7
=Diag 1.0e-003 *[0.7035 0.7063 0.7091 0.7120 0.7148 0.7177 0.7206 0.7234 0.7263
0.7293],
K
8
=Diag[0.0658 0.0658 0.0657 0.0657 0.0657 0.0656 0.0656 0.0656 0.0655 0.0655],
107
K
9
=Diag[0.1657 0.1657 0.1657 0.1657 0.1657 0.1657 0.1657 0.1657 0.1657 0.1657].
The closed loop responses with this robust decentralized gain K via reduced order
model for all the linearized models are satisfactory and are able to stabilize the outputs.
The eigen values of (
N
+KC) are found to be within the unit circle.
5.5 Simulation with nonlinear model
A SIMULINK based block diagram including all the nonlinear blocks is generated [47].
The output slip signal with robust decentralized gain K and a limiter is added to V
ref
signal. This is used to damp out the small signal disturbances via modulating the gener-
ator excitation. The output must be limited to prevent the PSS acting to counter action
of AVR. Dierent operating points are taken as the dierent models. The disturbance
considered is a self clearing fault at dierent buses cleared after 0.1 second. The limits of
PSS output are taken as 0.1. The location of fault considered for dierent models are
given in Table 4.13.
Simulation results of dierent generators are shown in Fig. 5.1 to Fig. 5.8 without
controller and with controller and with fault at particular bus. As shown in plots, the
proposed controller is able to damp out the oscillations in 12 to 15 seconds after clearing
the fault.
5.6 Decentralized fast output sampling feedback
technique via reduced order model for multimodel
system
Let us consider a family of plant S = {A
i,
B
i,
C
i
}, dened by
.
x= A
i
x +B
i
u, (5.31)
y = C
i
x, i = 1, , M. (5.32)
The discrete time invariant systems with sampling interval sec. can be represented
as
x(k + 1) =
i
x(k) +
i
u(k), (5.33)
y(k) = C
i
x(k).
108
0 5 10 15
0.1
0.05
0
0.05
0.1
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller
Model 1: fault at bus 16
Model 2: fault at bus 13
Figure 5.1: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.04
0.02
0
0.02
0.04
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 3: fault at bus 9
Model 4: fault at bus 7
Figure 5.2: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
109
0 5 10 15
0.04
0.02
0
0.02
0.04
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.04
0.02
0
0.02
0.04
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 5: fault at bus 17
Model 6: fault at bus 19
Figure 5.3: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 7
without controller
with controller
0 5 10 15
0.03
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 8
without controller
with controller
Model 7: fault at bus 21
Model 8: fault at bus 23
Figure 5.4: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
110
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 9
without controller
with controller
0 5 10 15
0.03
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 10
without controller
with controller
Model 9: fault at bus 32
Model 10: fault at bus 34
Figure 5.5: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
0 5 10 15
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller
Model 1: fault at bus 1
Model 11: fault at bus 35
Model 12: fault at bus 26
Figure 5.6: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
111
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 14: fault at bus 27
Model 13: fault at bus 25
Figure 5.7: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 15: fault at bus 29
Model 16: fault at bus 31
Figure 5.8: Open and closed loop responses with fault using robust decentralized periodic
output feedback controller via reduced order model
112
There exists a transformation V
i
, such that,
x = V
i
z, (5.34)
transforms the above system into the following block modal form
z(k + 1) =

i
z(k) +

i
u(k), (5.35)
y(k) =

C
i
z(k),
where,

i
=
_
_

1i
0
0
2i
_
_
,

i
=
_
_

1i

2i
_
_
, (5.36)

C
i
=
_
C
1i
C
2i
_
,
and the eigenvalues are arranged in their order of dominance.
We now extract an rth order model, retaining the r dominant eigenvalues, by trun-
cating the above systems. Using Eqns.(5.35-5.36), we get,
z
r
(k + 1) =
1i
z
r
(k) +
1i
u(k), (5.37)
y(k) = C
1i
z
r
(k).
Let u(k) = S
r
z
r
be a stabilizing control for the reduced order model in Eqn.(5.37).
Thus the closed loop reduced model (
1i
+
1i
S
r
) becomes stable. Now,
z
r
=
_
I
r
: 0
r(nr)

z, (5.38)
=
_
I
r
: 0
r(nr)

V
1
x.
Therefore, we get,
u(k) = S
r
_
I
r
: 0
r(nr)

V
1
i
x, (5.39)
S
i
x,
which makes the closed loop systems (
i
+
i
S
i
) stable and has no eigenvalues at the
origin.
Thus S
i
F
i
are the stabilizing state feedback gains for the system in Eqn.(5.33).
Using these F
i
, the following inequalities are solved
L <
1
,
113
LD
0i
F
i

i
<
2i
, i = 1......M,
LC
i
F
i
<
3i.
(5.40)
The controller obtained from the above equation will give desired behavior, but might
require excessive control action. To reduce this eect we relax the condition that L exactly
satisfy the above linear equation and include a constraint on the gain L. Thus we arrive
at the following in equations
_
_

2
1
I L
L
T
I
_
_
< 0,
_
_

2
2i
I (LD
0i
F
i

i
)
(LD
0i
F
i

i
)
T
I
_
_
< 0,
_
_

2
3i
I (LC
i
F
i
)
(LC
i
F
i
)
T
I
_
_
< 0. (5.41)
If the LMI constraints given in Eqns.(5.40-5.41) are solved using the above F
i
, the
robust fast output sampling feedback gain matrix may become full [37]. This results in
the control input of each machine being a function of outputs of all machines. To obtain
the robust decentralized fast output sampling feedback control o diagonal elements of
L
0
, L
1
, , L
N1
matrices are made zero. The structure of L
i
( i = 0, , N-1) matrices
for 10 machine 39 bus power system are assumed as given in Eqn.(4.17).
5.7 Case study II: Robust decentralized PSS design
for multimachine power system by fast output
sampling feedback technique via reduced order
model
As discussed in Section 5.4, the discrete models are obtained for sampling time = 0.1
sec. The reduced order models are computed from the discrete models as discussed in
Section 5.6.
Using the method discussed in Section 5.6 common stabilizing gain matrix S
r
(10 x20)
via reduced order model can be obtained using LMI Toolbox [52], [53]. Using aggregation
114
techniques, the state feedback gain F
i
(10 x40) can be calculated for the higher order
(actual) models.
This fast output sampling feedback gain can be obtained which approximately realizes
the designed F
i
.
Here, also consider one condition leading to one 40
th
order model. A state feedback
gain matrix can be obtained for this system via reduced order. If a fast output sampling
feedback gain has to be obtained for this system only, then minimum 4 gain matrices
(L
0
, L
1
, L
2
, L
3
) each of 10x10 dimension are needed, which will exactly realize the state
feedback gain computed for this system. The equation (3.15) has a solution if N observ-
ability index of this system/number of inputs or outputs. For one system controllability
index is 40. So minimum 4 gain sequences are needed.
Here as we are dealing with robust stabilization, we have to nd a L which will satisfy
LC
i
= F
i
, (i = 1, , 16), all these equations. So, if we restrict our gain sequence to 4,
there would not exist a common L which would satisfy LC
i
= F, for i = 1, , 16. Thats
why this condition is relaxed and a robust L has to be obtained by satisfying the equation
LC
i
= F
i
, approximately. Here gain sequences of L are chosen 10 (L
0
, L
1
, L
2
, L
3
, L
4
, L
5
,
L
6
, L
7
,L
8
, L
9
).
Using the F
i
, LMI constraints given in Eqns.(3.17-3.18) are solved for dierent values
of
1
,
2
and
3
to nd the robust decentralized gain matrix L for the actual models. The
robust decentralized fast output sampling feedback gain matrix L (10x100) via reduced
order model is obtained as given as
L
0
=Diag[ -7.241 -1.562 2.229 4.368 5.063 4.500 2.846 0.251 -3.153 -7.248],
L
1
=Diag[-0.0613 -0.0046 0.0324 0.0521 0.0568 0.0485 0.0290 0.0001 -0.0370 -0.0808],
L
2
=Diag[0.7397 0.0010 -0.3605 -0.4526 -0.3690 -0.1907 0.0121 0.1794 0.2600 0.2105],
L
3
=Diag[-0.3177 0.4268 0.5274 0.2497 -0.1791 -0.5660 -0.7483 -0.5903 0.0202 1.1747],
L
4
=Diag[1.2500 0.1554 -0.4950 -0.7819 -0.7845 -0.5805 -0.2458 0.1453 0.5210 0.8117],
L
5
=Diag[-0.9811 -0.1696 0.3381 0.5947 0.6473 0.5378 0.3034 -0.0229 -0.4116 -0.8370],
L
6
=Diag[-0.5786 -0.0648 0.2333 0.3603 0.3572 0.2621 0.1104 -0.0655 -0.2355 -0.3719],
L
7
=Diag[0.5883 0.1471 -0.1616 -0.3478 -0.4207 -0.3890 -0.2604 -0.0425 0.2580 0.6346],
L
8
=Diag[-0.0017 -0.0013 -0.0010 -0.0007 -0.0005 -0.0002 -0.0000 0.0002 0.0004 0.0006],
115
0 5 10 15
0.06
0.04
0.02
0
0.02
0.04
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller
Model 1: fault at bus 16
Model 2: fault at bus 13
Figure 5.9: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
L
9
=Diag[-0.0120 -0.0090 -0.0060 -0.0030 -0.0000 0.0029 0.0059 0.0089 0.0118 0.0148].
The closed loop responses with this robust decentralized gain L via reduced order
model for all the linearized models are satisfactory and able to stabilize the outputs. The
eigen values of (
N
+ LC) are found to be within the unit circle.
5.8 Simulation with nonlinear model
As in the case of decentralized periodic output feedback control via reduced order model,
here also nonlinear simulations are carried out in the similar way.
Simulation results of dierent generators with fault are shown in Fig. 5.9 to Fig. 5.16
without controller and with controller. As shown in plots, the proposed controller is able
to damp out the oscillations in 8 to 12 seconds after clearing the fault.
5.9 Concluding remarks
In this chapter, 10 machine and 39 bus power system is considered for the design of
decentralized power system stabilizers via reduced order model. The design schemes
of robust power system stabilizers for multimachine power systems using decentralized
116
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 3: fault at bus 9
Model 4: fault at bus 7
Figure 5.10: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 5: fault at bus 17
Model 6: fault at bus 19
Figure 5.11: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
117
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 7
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 8
without controller
with controller
Model 8: fault at bus 23
Model 7: fault at bus 21
Figure 5.12: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 9
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 10
without controller
with controller
Model 9: fault at bus 32
Model 10: fault at bus 34
Figure 5.13: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
118
0 5 10 15
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 1
without controller
with controller
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 2
without controller
with controller
Model 11: fault at bus 35
Model 12: fault at bus 26
Figure 5.14: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
0 5 10 15
0.015
0.01
0.005
0
0.005
0.01
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 3
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 4
without controller
with controller
Model 13: fault at bus 25
Model 14: fault at bus 27
Figure 5.15: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
119
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 5
without controller
with controller
0 5 10 15
0.02
0.01
0
0.01
0.02
Time in seconds
R
e
l
a
t
i
v
e

S
l
i
p
GENERATOR 6
without controller
with controller
Model 15: fault at bus 29
Model 16: fault at bus 31
Figure 5.16: Open and closed loop responses with fault using Decentralized Fast Output
Sampling Feedback Controller via reduced order model
periodic output feedback and fast output sampling feedback via reduced order model
have been developed. It is found that the designed robust decentralized controllers via
reduced order model provide good damping enhancement for various operating points
of multimachine power systems. The proposed methods result in satisfactory response
behaviour to damp out the oscillations. These methods also lead to robust decentralized
control for multimachine power system. The robust decentralized control via reduced
order model can be applied simultaneously to all the machines. The input applied to
each machine is a function of output of the respective machine only. Thus the applied
control scheme is robust and decentralized in nature, whereas, the conventional power
system stabilizers are designed sequentially. The conventional power system stabilizer is
dynamic in nature and is required to be tuned according to power characteristics whereas,
the proposed controller gains are static in nature and one controller structure is able to
damp out the oscillations for all models.
The robust decentralized fast output sampling feedback method via reduced order
model provides better damping enhancement for various operating points of multimachine
systems as compared to robust decentralized periodic output feedback method via reduced
order model. The control input required for these models are of smaller magnitudes in
fast output sampling feedback method via reduced order model as compared to periodic
120
output feedback method via reduced order model.
As shown in plots, the decentralized fast output sampling feedback controller via
reduced order model requires to damp out the oscillations in 8 to 12 seconds after clearing
the fault whereas, decentralized periodic output feedback controller via reduced order
model is able to damp out the oscillations in 12 to 15 seconds. The decentralized robust
power system stabilizers design via reduced order model given in this chapter are just
illustration of techniques. This technique will be more useful when it has to be applied
to larger power system where it is very hard to compute the state feedback gains. It has
been observed from the simulations results that the both robust decentralized methods
via reduced order model provide enough damping enhancement for various operating
points of multimachine systems, but take little more time to damp out the oscillations in
camparison to the full order decentralized robust methods discussed in Chapter 4.
The selection of faults at dierent buses are considered randomly. The nonlinear
simulations with proposed controllers for all generators using these fault conditions give
also satisfactory results. some of the results are presented in respective case studies.
121
.
122
Chapter 6
Conclusions and Future work
This thesis attempts to use the powerful properties of a periodic output feedback method
and fast output sampling feedback technique and proposes design method of robust de-
centralized power system stabilizers (PSS). One of the primary requirements of a good
decentralized method is that the resulting PSS should be robust enough to wide variations
in system parameters, while also being computationally manageable. In this respect, the
proposed periodic output feedback and fast output sampling feedback methods provide
satisfactory results.
The work presented here deals with the design of robust PSS for SMIB and central-
ized and robust decentralized PSS for multimachine system. Periodic output feedback
and fast output sampling feedback methods have been used in the proposed PSS tuning
approaches.
The thesis also examines design approaches for robust decentralized PSS for multi-
machine system using reduced order model. Periodic output feedback and fast output
sampling feedback methods have been also used in the proposed PSS tuning approaches.
Investigations reveal that the classical approach does provide satisfactory performances
for operating conditions up to the nominal but gives deteriorating responses when the
load increases. Moreover, the classically tuned PSS fails to stabilize the system at some
operating conditions.
These methods are more general than static output feedback and the control input
required for these models are of smaller magnitudes. It is found that the designed robust
controllers provide good damping enhancement for various operating points of single ma-
chine connected to an innite bus system and multimachine system. The conventional
power system stabilizer is dynamic in nature and is required to be tuned according to the
power characteristics. whereas, the proposed controllers gains are static in nature and
123
one controller structure is able to damp out the oscillations for all models.
These periodic output feedback and fast output sampling feedback methods also lead
to one centralized and robust decentralized controls for multimachine power system. The
robust decentralized periodic output feedback and fast output sampling feedback con-
trollers can be applied simultaneously to all the machines. The input applied to each
machine is a function of the output of that respective machine only. Thus the applied
control schemes are robust and decentralized in nature, whereas, the conventional power
system stabilizers are designed sequentially.
In multimachine power system the order of the states matrix may be quite large. It
would be dicult to work with these large systems in their original form. In such cases, it
is common to study the system by approximating it to a simpler model. Designed robust
decentralized periodic output feedback and fast output sampling feedback controllers via
reduced order model provide good damping enhancement for various operating points of
multimachine power systems.
6.1 Salient features of the present work
The thesis provides a broad-ranging overview of the research work carried out in the
area of PSS tuning over the past three decades and brings out the main research
issues that have been addressed.
The thesis provides a detailed description of the development of the system math-
ematical models, both for single machine innite bus as well as the multimachine
system under small perturbations. These models are generic enough and can be
applied to large sized power systems.
The thesis provides an exhaustive analysis of periodic output feedback and fast
output sampling feedback control methods applicable to single machine and mul-
timachine power system for designing robust PSS and brings out a comparison of
performances achieved by systems having PSS designed using these methods.
The thesis proposes a novel approach to design decentralized PSS using periodic
output feedback and fast output sampling feedback control methods. The main
feature of this method is that the robust decentralized control can be applied simul-
taneously to all the machines. The input applied to each machine is a function of
output of the respective machine only. Thus the applied control schemes are robust
124
and decentralized in nature whereas the conventional power system stabilizers are
designed sequentially.
The thesis develops a novel design on robust decentralized periodic output feedback
and fast output sampling feedback controllers via reduced order model which provide
good damping enhancement for various operating points of multimachine power
systems. The proposed methods result in satisfactory response behaviour to damp
out the oscillations.
6.2 Scope of future work in this area
Tuning of PSS for large interconnected power systems has been a challenging prob-
lem for power engineers and though a lot of work has been reported in this area,
several issues remain unresolved. Based on the work reported in this thesis, we
briey layout some of the issues that need to be addressed within the same frame-
work discussed here, in order to gain an exhaustive understanding of the problem
of PSS tuning and its characteristics.
The system investigated has been limited up to a ten generator, thirty nine bus
system. It would be desirable to examine periodic output feedback and fast output
sampling feedback methods based PSS tuning for larger and more realistic systems.
Based on the experience accumulated during simulations and due to the development
of both the system model and the periodic output feedback and fast output sampling
feedback methods in a generic manner, the extension of the work could be done
without diculties.
As mentioned in one of the earlier chapters, siting of PSS is an important issue, more
so, when the system size increases considerably. It is thus important to examine the
periodic output feedback and fast output sampling feedback methods based PSS
tuning method while incorporating the PSS siting issues.
The systems considered in the thesis assume that the loads are constant impedance
loads. It would be of interest to the designer to understand how the dynamics of the
system will be aected by the load dependence on voltage and consequently, how
the optimal PSS parameters will be aected.
125
The powerful properties of periodic output feedback and fast output sampling feed-
back methods based optimization can be further exploited to examine various other
controller structures and determine their globally optimal settings.
Test and implement periodic output feedback and fast output sampling feedback
strategies to various other power system phenomena.
126
Bibliography
[1] E.V.Larsen and D.A. Swann, Applying power system stabilizers part- I: general
concepts, IEEE Trans. on Power Apparatus and Systems, vol. PAS-100, No. 6, pp.
3017-3024, June 1981.
[2] E.V.Larsen and D.A. Swann, Applying power system stabilizers part- II: perfor-
mance objective and tuning concepts, IEEE Trans. on Power Apparatus and Sys-
tems, vol. PAS-100, No. 6, pp. 3025-3033, June 1981.
[3] P. Kundur, M. Klein, G. J. Rogers and M. S. Zywno, Application of power sys-
tem stabilizers for enhancement of overall system stability, IEEE Trans. on Power
Systems, vol. 4, pp. 614-626, May 1989.
[4] F.R.Schlief and J.H. White, Damping for the North-west-southwest tie line oscilla-
tions - a analog study, IEEE Trans. on Power Apparatus and Systems, vol. PAS-85,
pp. 1239-1246, December 1966.
[5] F.R.Schlief and R.R. Angell, Damping of system oscillations with a hydrogenerating
unit, IEEE Trans. on Power Apparatus and Systems, vol. PAS-86, pp. 438-442, April
1967.
[6] C. Concordia and F.P. de mello, Concepts of synchronous machine stability as
aected by excitation Control, IEEE Trans. on Power Apparatus and Systems, vol.
PAS-88, pp. 316-329, April 1969.
[7] F.R.Schlief, H.D. Hunkins, E.E. Hattan and W. B. Gish, Control of rotating exciters
for power system damping: pilot applications and experience, IEEE Trans. on
Power Apparatus and Systems, vol. PAS-88, pp. 1259-1266, August 1969.
[8] E. J. Warchol, F.R.Schlief, W. B. Gish and J.R. Church, Alinement and modeling of
Hanford excitation control for system damping, IEEE Trans. on Power Apparatus
and Systems, vol. PAS-90, pp. 714-724, March/April 1971.
127
[9] A. D. Gerhart, T. Hillesland, Jr., J.F. Luini and M. L. Rockeld, Power system
stabilizer eld testing and digital simulation, IEEE Trans. on Power Apparatus
and Systems, vol. PAS-90, pp. 2095-2100, September/October 1971.
[10] F.P. DeMello, P.J. Nolan, T.F. Laskowski and J.M. Undrill, Co-ordinated applica-
tion of stabilizers in multi-machine power systems, IEEE Trans. on Power Appa-
ratus and Systems, vol. PAS-99, pp. 892-901, May/June80.
[11] O.H.Abdalla, S.A.Hassan and N.T.Tweig, Co-ordinated stabilization of a multi-
machine power system, IEEE Trans. on Power Apparatus and Systems, vol.PAS-
103, pp. 483-494, Mar.84.
[12] A. Doi and S. Abe, Co-ordinated synthesis of power system stabilizers in multima-
chine power systems, IEEE Trans. on Power Apparatus and Systems, vol.PAS-103,
pp.1473-1479, June84.
[13] C.M. Lim and S. Elangovan, Design of stabilizers in multi-machine power systems,
Proceedings of IEE, Part-C, vol.132, pp. 146-153, May85.
[14] C.M. Lim and S.Elangovan, New approach to power system stabilizer design, Elec-
tric Power Systems Research, vol.8, pp. 285-292, 1985.
[15] P. Kundur, D. C. Lee and H.M. Zein El-din, Power system stabilizers for ther-
mal units: analytical techniques and on-site validation, Paper F80-227-9 presented
IEEE PES Winter Meeting, New York, February 1980.
[16] R.J. Fleming, M.A. Mohan and K. Parvatisam, Selection of parameters of stabilizers
in multimachine power systems, IEEE Trans. on Power Apparatus and Systems, vol.
PAS-100, pp. 2329-2333, 1981.
[17] K. E. Bollinger, A. Laha, R. Hamilton and T. Harras, Power system stabilizer
design using root locus methods, IEEE Trans. on Power Apparatus and Systems,
vol. PAS-94, pp. 1484-1488, September/October 1975.
[18] Y.N. Yu and Q.H. Li, Pole placement power system stabilizers design of an unstable
nine-machine system, IEEE Trans. on Power Systems, vol. PWRS-5, pp. 353-358,
1990.
[19] Fathi A. Saleh and Magdi S. Mahmoud, Design of power system stabilizer using
reduced-order methods, Electric Power System Research, vol. 33, pp. 219-226, 1995.
128
[20] Y.N. Yu and C. Siggers, Stabilization and optimal control signal for a power system,
IEEE Trans. on Power Apparatus and Systems, vol. PAS-90, pp. 1469-1481, 1971.
[21] H.A.M. Moussa and Y.N. Yu,Optimal power system stabilization through excitation
and/or governor control, IEEE Trans. on Power Apparatus and Systems, vol.PAS-
91, pp. 1166-1174, May/June72.
[22] B. Kumar and E.F. Richards, An optimal eigenvalue assignment for improved dy-
namic stability in power system, IEEE Trans. on Power Apparatus and Systems,
vol. PAS-101, pp. 1570-1577, 1982.
[23] Y.N.Yu, Vong, K. Vongsuriya and L.N. Vedman, Application of an optimal control
theory to a power system, IEEE Trans. on Power Apparatus and Systems, vol.
PAS-89, pp. 55-62, 1970.
[24] K.R. Padiyar, S.S. Prabhu, M.A. Pai and K. Gomathi, Design of stabilizers by pole
assignment with output feedback, Electric Power Systems Research, vol.2, No.3, pp.
140-145, July1980.
[25] S. Chen and O. P. Malik, H

based power system stabilizer design, IEE Proceed-


ings pt. C, vol. 142, No. 2, March 1995, pp. 179-184.
[26] T.C. Yang,Applying H

optimisation method to power system stabilizer design


Part 1: single-machine innite-bus system, Electric Power Systems Research, vol.19,
No.1, pp. 29-35, 1995.
[27] T.C. Yang, Applying H

optimisation method to power system stabilizer design


Part 1: multi-machine power system, Electric Power Systems Research, vol.19,
No.1, pp. 37-43, 1995.
[28] A. Soos and O. P. Malik, An H
2
optimal adaptive power system stabilizer , IEEE
Trans. on Energy Conversion, vol.EC-17, pp. 143-149, March2002.
[29] P. Shrikant Rao and I. Sen, Robust pole placement stabilizer design using linear
matrix inequalities, IEEE Trans. on Power Systems, vol. 15, No. 1, pp. 3035-3046,
February 2000.
[30] V.L. Syrmos, C. Abdallah, P. Dorato and K. Grigoriadis, Static output feedback: a
survey, Automatica, vol. 33, No. 6, pp. 577-590, 1996.
129
[31] V.L. Syrmos, C. adallah and P. Dorato, Static output feedback: A survey , Pro-
ceedings of the 33
rd
IEEE Conference on Decision and control, Lake Buena, Vasta
FL(USA), vol.VM-13 1:30, pp. 837-842, December1994.
[32] A. Astol and P. Colaneri, Static output feedback stabilization of linear and non-
linear systems, Proceedings of the 33
rd
IEEE Conference on Decision and control,
Sydney, Australia, pp. 2920-2925, December2000.
[33] A.B.Chammas and C.T.Leondes, Pole assignment by piecewise constant output
feedback, International Journal of Control, vol. 29, pp. 31-38,1979.
[34] T.L.Huang, S.C. Chen, T.Y.Hyvang and W.T.Yang, Power system output feedback
stabilizer design via optimal sub-eigen structure assignment, Electric Power System
Research, vol.21, pp. 107-114, 1991.
[35] T.L.Huang, S.C. Chen, T.Y.Hwang and W.T.Yang, Power system output feedback
stabilizer design via optimal subeigenstructure assignment, IEEE Trans. on Power
Systems, vol. PWRS-6,No.3, pp. 1035-1041, 1991.
[36] Rajeev Gupta, B. Bandyopadhyay and A.M.Kulkarni, Design of power system sta-
bilizer for multimachine power system using periodic output feedback technique,
International conference on Quality, Reliability and Control, ICQRC2001, Mumbai,
India, pp. C45-1-8, December 2001.
[37] Rajeev Gupta, B. Bandyopadhyay and A.M.Kulkarni, Design of power system stabi-
lizer for multimachine power system using fast output sampling feedback technique,
4
th
Asian Control conference, ASCC2002, Singapore, pp. 1916-1921, September
2002.
[38] H.Werner and K.Furuta, Simultaneous stabilization based on output measurement,
Kybernetika, vol. 31, pp. 395-411, 1995.
[39] H. Werner, Fast output sampling regulators with integral action, Proceedings of
5th European Control Conference, Karlsruhe, Germany, Paper DA-9-3, 1999.
[40] B.M.Patre, B.Bandyopadhyay and H.Werner, Control of discrete two-time scale
system by using piece constant periodic output feedback, System Science, vol. 23,
pp. 23-37, 1997.
130
[41] M. Aldeen and F. Crusca, Multimachine power system stabilizer design based on
new LQR approach, IEE Proceedings of generation, transmission and distribution,
vol. 142, No. 6, pp. 395-411, 1995.
[42] E.D. Tuglie, M.L. Scala, R. Sbrizzai and M.Trovato, Sequential design of a de-
centralized control structure for power system stabilizers, Electric Power Systems
Research, vol.50, No.1, pp. 91-98, 1999.
[43] C.L. Chen and Y.Y. Hsu, An ecient algorithm for the design of decentralized
output feedback power system stabilizer, IEEE Transactions on Power Systems,
vol. PWRS-3, No.3, pp. 999-1004, 1988.
[44] G.E.Boukarim, S. Wang, J.H.Chow, G.C. Taranto and N.Martin, A Comparison of
classical, robust and decentralized control design for multiple power system stabiliz-
ers, IEEE Trans. on Power Systems, vol. PWRS-15,No.4, pp. 1287-1292, 2000.
[45] E. J. Davison, A method for simplifying linear dynamic systems, IEEE Trans. on
Automatic Control, vol.AC-11, pp. 93-101, January 1966.
[46] E.V.Larsen and D.A.Swann, Applying power system stabilizers Part- III: practical
consideration, IEEE Trans. on Power Apparatus and Systems, vol. PAS-100, No. 6,
pp. 3035-3046, June 1981.
[47] K.R. Padiyar, Power System Dynamics Stability and Control, Interline publishing
private Ltd., Bangalore, 1996.
[48] W.G. Heron and R.A. Phillips, Eect of a modern amplidyne voltage regulator
on under excited operation of large turbine generators, AIEE Trans. 71(pt.3), pp.
692-97, 1952.
[49] F.D. DeMello and C.Concordia, Concepts of synchronous machine stability as af-
fected by excitation control, IEEE Trans. on Power Apparatus and Systems, Vol.
PAS-88, pp. 316-329, 1969.
[50] P.Kundur, Power System Stability and Control, McGraw-Hill, Inc., Newyork, 1993.
[51] P.Gahenet, A.Nemirovski, A.J.Laub, and M.Chilali, LMI toolbox for Matlab: user
manual, The Math works Inc., Natick MA, 1995.
[52] Natarajan Narasimhamurti, Using Matlab control system toolbox: student work
book, University of Michigan-Dearbon.
131
[53] Urban Bruneer, Review of classical and modern control theory using Matlab: stu-
dent edition, part 1, Printice Hall, Usa, September 2002.
132
Summary
In the late 1950s most of the new generating units added to the electric utility systems
were equipped with continuously acting voltage regulators. As these units became a larger
percentage of generating capacity, it became apparent that the voltage regulator action
had a detrimental impact upon the dynamical stability (or small scale stability) of the
power system. Oscillations of small magnitude and low frequency often persist for long
periods of time and in some cases it can cause limitations on the power transfer capability.
Power system stabilizers (PSSs) were developed to aid in damping these oscillations via
modulation of the generator excitation. The development has evolved the use of various
tuning techniques and input signals and learning to deal with turbine generator shaft
torsional mode of vibrations
Power System Stabilizers (PSS) are added to excitation systems, to enhance the damp-
ing of electric power system during low frequency oscillations. For large scale power
systems comprising of many interconnected machines, the power system stabilizer (PSS)
parameter tuning is a complex exercise due to the presence of several poorly damped
modes of oscillation. The problem is further complicated by continuous variation in power
system operating conditions. In the simultaneous tuning approach, exhaustive computa-
tional tools are required to obtain optimal parameter settings for the PSS, while in case
of sequential tuning, although the computational burden is lesser, evaluating the tuning
sequence is an additional requirement. There is a further problem of eigenvalue drift.
This thesis attempts to use the powerful properties of a periodic output feedback
method and fast output sampling feedback technique and proposes design method of
robust decentralized power system stabilizers (PSS). One of the primary requirements of
a good decentralized method is that the resulting PSS is robust enough to wide variations
in system parameters, while also being computationally manageable. In this respect, the
proposed periodic output feedback and fast output sampling feedback methods provide
satisfactory results.
In static output feedback techniques, one approach to pole placement problem is to
1
consider the potential of time-varying periodic output feedback. It was shown by Cham-
mas and Leondes that a controllable and observable plant was discrete time pole assignable
by periodically time-varying piecewise constant output feedback. With Fast output sam-
pling approach proposed by Werner and Furuta, it is generically possible to simultaneously
realize a given state feedback gain for a family of linear, observable models. This approach
requires to increase the low rank of the measurement matrix of an associated discretized
system, which can be achieved by sampling the output several times during one input
sampling interval, and constructing the control signal from these output samples. Since
the feedback gains are piecewise constant, their method could be easily implemented and
indicated a new possibility. Such control laws can stabilize a much larger class of systems
than the static output feedback
Based on a recent convex programming algorithm for simultaneous stabilization by
linear state feedback, periodic output feedback and fast output sampling feedback are the
two types of multirate techniques for stabilizing a family of systems, when either a si-
multaneous stabilizing state feedback gain or a simultaneous stabilizing output injection
matrix exist, and complete state information is not available. In the periodic output
feedback technique the rate of applying control input signal is higher than the output
sampling rate, whereas in fast output sampling feedback technique the output signal is
sampled at a faster rate than that of the control input applied. These controllers provide
superior alternative to observer based state feedback control. Nondynamic feedback of
output samples allows a particularly exible design.
This thesis proposes the applications of periodic output feedback and fast output
sampling feedback techniques for designing robust power system stabilizers for a single
machine connected to an innite bus system. The slip signal is taken as output. The
powerful properties of these methods are used to optimize the PSS parameter settings.
The nonlinear model of a machine is linearized at dierent operating points and linear
models are obtained. In periodic output feedback method, for each of these linear models,
an output injection gain is obtained using DLQR technique. A robust periodic output
feedback gain which realizes these output gains is obtained using LMI approach. This
robust periodic output control is applied to the nonlinear model of the single machine
at dierent operating (equilibrium) points. This method succeeds in achieving a robust
tuned PSS parameter set.
In fast output sampling feedback method, for all of these linear models, a single
stabilizing state feedback gain is obtained. A robust fast output sampling feedback gain
2
which realizes this state feedback gain is obtained using LMI approach. This method does
not require the state of the system for feedback and is easily implementable. This robust
fast output sampling control is applied to nonlinear model of a single machine at dierent
operating (equilibrium) points. This method also succeeds in achieving a robust tuned
PSS parameter set and is compared to conventional power system stabilizers.
These methods are further extended for designing centralized and decentralized con-
trollers for multimachine power system. The Nonlinear dierential equations governing
the behavior of a multimachine power system is linearized about a particular generating
power, bus structure and load distributions to obtain a linear model which represents the
small signal oscillatory response of a power system. Variations in the generating power,
bus structure and load distribution of the system result in the variations in the param-
eters of the small signal model. A given range of variations in the generating power,
bus structure and load distribution of a particular system thus generates a set of linear
models each corresponding to one particular generating power, bus structure and load
distribution. Since, at any given instant, the actual plant could correspond to any model
in this set, a robust controller would have to import adequate damping to each one of
this entire set of linear models.
In the centralized controllers for multimachine power system, periodic output feedback
and fast output sampling feedback gain matrices are generally full, that makes control
input as a function of output of all machines. Power system stabilizers for all machines
can be applied simultaneously to all the machines. It is found that the designed controller
provides a good damping enhancement for a multimachine power system.
As mentioned above in centralized control, control input of each machine becomes
a function of outputs of all machines. Also centralized PSSs require transmission of
signal among the generating units. This requirement in itself no longer constitutes a
problem from practical and technical view points. This is due to the rapid advancement
in optical ber communication and their adoption by the power utilities. However, if a
completely decentralized PSS can be found, so that no signicant deterioration in the
system performance is experienced as compared to the state and centralized periodic
output feedback based schemes, then such a scheme would be more advantageous, in terms
of practicability and reliability. In such schemes, not only is the cost of implementation
drastically reduced, but also, the risk of loss of stability due to signal transmission failure
is minimized. Also, due to the geographically distributed nature of power systems, a
decentralized control scheme may be more feasible than a centralized control scheme.
3
In the decentralized power system stabilizer, the control input for each machine should
be a function of the output of that machine alone. This can be achieved by designing
a decentralized PSS using periodic output feedback and fast output sampling feedback
techniques in which the gain matrix should have all o-diagonal terms zero or very small
compare to the diagonal terms. In decentralized PSS, to activate the proposed controller
at same instant, proper synchronization signal is required to be sent to all machines. All
PSSs can be applied simultaneously to the respective machines. So the decentralized
stabilizer design problem can be translated into a problem of diagonal gain matrix design
for multimachine power system. This thesis proposes the design of a decentralized robust
power system stabilizer for multimachine system using periodic output feedback and fast
output sampling feedback techniques.
Design of robust decentralized power system stabilizers for multimachine power system
via reduced order model by using both the multirate output feedback methods have been
also proposed. For large power systems, the order of the state matrix may be quite large. It
would be dicult to work with these complex systems in their original form. In particular
the computation of the output injection gain which is needed to obtain the decentralized
periodic output feedback based power system stabilizer becomes very tedious for large
power system. One of the ways to overcome this diculty is to develop a reduced order
model for large power systems. Then an output injection gain can be computed from
the reduced model of the power system, and using the aggregation techniques, an output
injection gain can be obtained for the higher order (actual) model. The decentralized
periodic output feedback gain which realizes this output injection gain can be obtained
for the actual model.
The decentralized power system stabilizers based on fast output sampling method has
been also designed via reduced order model in this thesis. In this case a state feedback gain
from the reduced model of the power system is obtained and then, using the aggregation
techniques, a state feedback gain is obtained for the higher order (actual) model. The
decentralized fast output sampling feedback gain which realizes this state feedback gain
can be obtained for the actual model.
These methods give very good results for the design of Power System Stabilizers for
single machine and multimachine power system. Investigations reveal that the classical
approach does provide satisfactory performances for operating conditions up to the nom-
inal but gives deteriorating responses when the load increases. Also the classically tuned
PSS fails to stabilize the system at some operating conditions.
4
List of Publications based on this work
Conferences:
1. Rajeev Gupta, B. Bandyopadhyay and A. M. Kulkarni, Design of power system
stabilizer for multimachine power system using periodic output feedback, Interna-
tional Conference on Reliability, Quality and Control, Hotel Le Meredian, Mumbai
(ICQRC-2001), pp. C45-1:8, December 2001.
2. Rajeev Gupta, B. Bandyopadhyay and A. M. Kulkarni, Design of power system sta-
bilizer for multimachine power system using fast output sampling feedback, Asian
Control Conference (ASCC-2002), Singapore, pp. 1916-1921, 25-27 September 2002
3. Rajeev Gupta, B. Bandyopadhyay and A. M. Kulkarni, Design of decentralized
power system stabilizer for multimachine power system using periodic output feed-
back, Seventh International Conference on Control, Automation, Robotics and Vi-
sion (ICCARV-2002) Singapore, pp. 1676-1681, 2-5 December 2002.
4. Rajeev Gupta, B. Bandyopadhyay and A. M. Kulkarni, Design of decentralized
power system stabilizer for multimachine power system using fast output sampling
feedback, National Power System Conference (NPSC-2002), IIT Kharagpur, pp.
362-367, 26-29 December 2002.
Journals:
1. Rajeev Gupta, B. Bandyopadhyay and A. M. Kulkarni, Design of power system
stabilizer for single machine power system using periodic output feedback, IEE
proceedings on Generation, Transmission and Distribution, vol.150, No. 2, pp. 211-
216, March 2003.
2. Rajeev Gupta, B. Bandyopadhyay and A. M. Kulkarni, Design of power system
stabilizer for single machine power system using fast output sampling feedback,
Electric Power System Research, vol.65, pp. 247-257, 2003.
Acknowledgments
It is a pleasant task to express my gratitude to all those near and dear ones who have
accompanied and helped me in my research program.
I would like to express here my sincere gratitude to my supervisors, Prof. B. Bandy-
opadhyay and Prof. Anil M. Kulkarni, whose precious advice and friendly encouragement
made this work go smoothly throughout the period of research. They are the source of
never-ending inspiration for me. I have been extremely lucky to have them as my men-
tors. Many thanks go to my examiner, Professor S.A. Khaparde, for his support and
useful discussions.
I express my gratitude to Prof. P.S.V. Nataraj and Prof. R.N. Banavar for their
constructive suggestions oered to me for the successful completion of my thesis.
I gratefully acknowledge the constructive criticism and valuable inputs that S. Janard-
hanan, G.L. Sharma, Uday Maskar and Shubhanga Aithal have contributed with during
the course of this work.
I am extremely grateful to Kota Engineering College Society, Prof. S.C. Gupta, Prin-
cipal, Engineering College, Kota and Prof. D.K.Parihar, Head, Department of Electrical
Engineering, Engineering College, Kota for sponsoring me for Ph.D. program under Qual-
ity Improvement Program. I am also extremely grateful to Prof. N.C. Bhandari, currently
Principal, Engineering College, Kota and Prof. M.L. Gupta, currently Head, Department
of Electrical Engineering, Engineering College, Kota for their cooperation in administra-
tive and nancial matters during my stay at IIT Bombay, Mumbai.
I am also thankful to my colleagues Dr. K.V.S. Rao and Mr. Ranjan Maheshwari for
their cooperation and making my stay at IIT Bombay, Mumbai troubleless by managing
all the matters at Engineering College, Kota and informing me all needful informations
time to time.
The colleagues at the department in general, and the Systems and Control Group
in particular Rajshekhar, Prakash, Sachin, Kubal, Nagendra Prasad, Jignesh, Arun,
Shankarnarayan and many more created a supportive and friendly working environment.
They made the coee break worth more than a coee during the break.
I owe my deepest gratitude to my parents for their continuous inspiration and moti-
vation throughout my life. I also owe my deepest gratitude to my family, and to my wife
Bharti especially, for their love, understanding and support that gave me the comfort I
needed to fully concentrate on my work. Here, I cant forget my daughters, Rupal and
Disha for their childish amusing activities, after long tiring work at Syscon lab, that made
me always fresh and cheerful.
I am grateful to my sister and her family to give me all types of help in this busy city
of Mumbai and caring me family in my absence. I am also thankful to my in-laws for
their moral boost to me.
I would also like to thank my school teachers from town named Baswa in Rajasthan
who contributed greatly to my formative years and further development. A chain of
gratitude is never-ending-there are so many people who have played signicant role in
my success. Though I am not able to mention their names here, deep in my heart is an
overwhelming feeling of warmth and gratitude for them.
Finally I am thankful to God for inspiring and guiding this humble human creature,
without whose blessings, nothing would have been possible.

Вам также может понравиться