Small-Signal Dynamic Stability Enhancement Of A

DC-Segmented AC Power System

by

Sahar Pirooz Azad

A thesis submitted in conformity with the requirements

for the degree of Doctor of Philosophy
Graduate Department of Electrical and Computer Engineering
University of Toronto

Copyright 2014 by Sahar Pirooz Azad

 Abstract
Small-Signal Dynamic Stability Enhancement Of A DC-Segmented AC Power System

Sahar Pirooz Azad

Doctor of Philosophy
Graduate Department of Electrical and Computer Engineering
University of Toronto
2014

This thesis proposes a control strategy for small-signal dynamic stability enhancement of

a DC-segmented AC power system. This control strategy provides four control schemes
based on HVDC supplementary control or modification of the operational condition of
the HVDC control system to improve the system stability by (i) damping the oscilla-
tions within a segment using supplementary current control of a line-commutated HVDC

link, based on the model predictive control (MPC) method (control scheme 1), (ii) min-
imizing the propagation of dynamics among the segments based on a coordinated linear
quadratic Gaussian (LQG)-based supplementary control (control scheme 2), (iii) selec-
tively distributing the oscillations among the segments based on a coordinated LQG-
based supplementary control (control scheme 3) and (iv) changing the set-points of the

HVDC control system in the direction determined based on the sensitivities of the Hopf
stability margin to the HVDC links set-points (control scheme 4). Depending on the
system characteristics, one or more of the proposed control schemes may be effective for
mitigating the system oscillations.

Study results show that (i) control scheme 1 leads to damped low-frequency oscil-
lations and provides fast recovery times after faults, (ii) under control scheme 2, each
segment in a DC-segmented system can experience major disturbances without causing
adjacent segments to experience the disturbances with the same degree of severity, (iii)

control scheme 3 enables the controlled propagation of the oscillations among segments
and damps out the oscillatory dynamics in the faulted segment, and (iv) control scheme
4 improves the stability margin for Hopf bifurcations caused by various events.

ii
 Since power system software tools exhibit limitations for advanced control design,
this thesis also presents a methodology based on MATLAB/Simulink software to (i)
systematically construct the nonlinear differential-algebraic model of an AC-DC system,
and (ii) automatically extract a linearized state space model of the system for the design of

the proposed control schemes. The nonlinear model also serves as a platform for the time-
domain simulation of power system dynamics. The accuracy of the MATLAB/Simulink-
based AC-DC power system model and time-domain simulation platform is validated by
comparison against PSS/E.

iii
 Dedication

To my dear parents, Alireza and Marzeyeh, whom I will always be indebted to.

iv
 To my loving husband
Kasra
who made it all possible.

v
 Acknowledgements
First and foremost, I would like to express my sincere gratitude to my supervisors, Profes-
sor Reza Iravani and Professor Zeb Tate. My first debt of gratitude must go to Professor
Reza Iravani for his deep insight, wisdom, invaluable guidance, advice and limitless sup-
port during the development of this thesis. His patience and understanding has been an
inspiration during my graduate studies.
My deepest gratitude and thanks also go to Professor Zeb Tate who generously dedi-
cated his time and energy to long discussions and without whose help, guidance, encour-
agement and patience, this dissertation would have never been possible.
I also want to express my thanks for the comments and suggestions provided by the
thesis committee members Professor Alexander Prodic, Professor Joshua Taylor, and
Professor Peter Lehn.
I also acknowledge the generous financial support I received from the University of
Toronto, Professor Iravani and Professor Tate.
Finally, I would also like to thank the members of the Energy Systems Group for
valuable advice and instructive discussions.

vi
Contents

1 Introduction 1
1.1 Large-Scale Integration of HVDC Transmission in AC Power Systems . . 1
1.1.1 Embedded HVDC System in an AC Grid . . . . . . . . . . . . . . 2
1.1.2 HVDC Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.3 AC Grid Segmentation . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.4 Mitigation of Dynamic Oscillatory Modes of an AC-DC System . 6
1.2 Statement of the Problem and Thesis Objectives . . . . . . . . . . . . . . 7
1.3 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Thesis Layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 Small-Signal Dynamic Model Development of AC-DC Systems Based

on Computer-Assisted Linearization of AC-DC Systems 10
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Low-Frequency Dynamic Model of the AC System . . . . . . . . . . . . . 12
2.2.1 Turbine-Generator (T-G) Unit Model . . . . . . . . . . . . . . . . 12
2.2.2 Excitation System Model . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.3 Governor System Model . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.4 AC Network Model . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.5 Multi-Machine System Model . . . . . . . . . . . . . . . . . . . . 14
2.3 Low-Frequency Model of the DC System . . . . . . . . . . . . . . . . . . 16
2.3.1 AC to DC Conversion Model . . . . . . . . . . . . . . . . . . . . . 17
2.3.2 DC System Controller Models . . . . . . . . . . . . . . . . . . . . 17
2.3.3 DC Transmission Line Model . . . . . . . . . . . . . . . . . . . . 17
2.3.4 Converter Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.5 Overall DC System Model . . . . . . . . . . . . . . . . . . . . . . 19
2.4 Overall Model of the Multi-Machine AC-DC System . . . . . . . . . . . . 21
2.5 Implementation of the AC-DC System Model in MATLAB/Simulink . . . 22
2.6 Validation of the MATLAB/Simulink-Based AC-DC Model . . . . . . . . 24

vii
 2.6.1 IEEE 14-Bus 1-Segment System . . . . . . . . . . . . . . . . . . . 24
2.6.2 Validation of the Nonlinear AC-DC Model . . . . . . . . . . . . . 26
2.6.3 Validation of the Linearized Dynamic Model . . . . . . . . . . . . 28
2.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3 HVDC Local Supplementary Control (LSC) for Small-Signal Stability

Enhancement 31
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 MPC and LQG Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.1 Linear Quadratic Gaussian (LQG) Control . . . . . . . . . . . . . 32
3.2.2 Model Predictive Controller (MPC) . . . . . . . . . . . . . . . . . 33
3.3 HVDC LSC Based on Optimal Control Theory . . . . . . . . . . . . . . . 36
3.3.1 LSC Based on LQG Method . . . . . . . . . . . . . . . . . . . . . 38
3.3.2 LSC Based on MPC Method . . . . . . . . . . . . . . . . . . . . . 41
3.4 Application of the LSC for Small-Signal Dynamic Stability Enhancement 42
3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

4 Mitigation of Oscillations by Control of the Propagation of Oscillatory

Modes 47
4.1 Global Supplementary Control (GSC) Based on Optimal Control Theory 48
4.2 Study Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3 Study Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.3.1 Dynamics of the Fully-DC-Segmented and Partially-DC-Segmented
Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.2 GSC1 in the Fully-DC-Segmented System . . . . . . . . . . . . . 53
4.3.2.1 Case 1: GSC1 with Current Order Modulation . . . . . 54
4.3.2.2 Case 2: GSC1 with Voltage Reference Modulation . . . . 54
4.3.2.3 Case 3: GSC1 with Current Order and Voltage Reference
Modulation . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.3.3 GSC2 in the Fully-DC-Segmented System . . . . . . . . . . . . . 55
4.3.4 Performance Indices and Sensitivity Analyses of the GSC1 and GSC2 57
4.3.4.1 Sensitivity to the Fault Location . . . . . . . . . . . . . 59
4.3.4.2 Sensitivity to the Operating Point . . . . . . . . . . . . 62
4.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63

5 HVDC Operating-Point Adjustment 64

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

viii
 5.2 Hopf Sensitivity Calculations . . . . . . . . . . . . . . . . . . . . . . . . 65
5.3 Optimization Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.4 Study Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.5 Study Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.5.1 Case Study on the 2-Segment System . . . . . . . . . . . . . . . . 71
5.5.2 Case Studies on the 3-Segment System . . . . . . . . . . . . . . . 73
5.5.2.1 Case 1: Load Variations . . . . . . . . . . . . . . . . . . 74
5.5.2.2 Case 2: Line Impedance Change . . . . . . . . . . . . . 75
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

6 Conclusions 78
6.1 Thesis Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.2 Thesis Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.3 Thesis Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6.4 Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

A MPC Optimization Procedure 82

B System Differential-Algebraic Equations (DAEs) 84

C Calculating the Sensitivity of Stability Margin with Respect to Param-

eters 86

Bibliography 88

ix
List of Tables

2.1 Governor parameters of the T-G units . . . . . . . . . . . . . . . . . . . 26

3.1 Eigenvalues of the linearized WSCC system . . . . . . . . . . . . . . . . 38

3.2 Eigenvalues of the linearized IEEE 14-bus 1-segment system . . . . . . . 38

4.1 Cost function coefficients of the T-G units of the 3-segment system . . . 51

5.1 Cost function coefficients of the T-G units of the 2-segment system . . . 70

x
List of Figures

1.1 The integration of HVDC technology in an AC system . . . . . . . . . . 3

2.1 Block diagram of the AC system model . . . . . . . . . . . . . . . . . . . 16

2.2 Block diagram of the DC system model . . . . . . . . . . . . . . . . . . . 16
2.3 HVDC control system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4 DC transmission line model . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.5 Injection model for HVDC converter stations . . . . . . . . . . . . . . . . 19
2.6 Signal flow of the AC-DC system with supplementary controllers and set-
point adjustment unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.7 Schematic one-line diagram of the IEEE 14-bus system . . . . . . . . . . 25
2.8 Schematic one-line diagram of the IEEE 14-bus 1-segment system . . . . 25
2.9 Voltage magnitude, angle, active and reactive power of machine C1 on bus
3 due to the L-L-L-G fault . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.10 Active and reactive power flow changes of line 6 between buses 3 and 4
due to the L-L-L-G fault . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.11 Voltage magnitude, angle, active and reactive power of machine C3 on bus
8 due to line 3 tripping and reclosure . . . . . . . . . . . . . . . . . . . . 27
2.12 Active and reactive power flow changes of line 6 due to line 3 tripping and
reclosure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.13 Voltages of buses 1 and 3 due to 5% step change in the voltage reference
of G2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.14 DC current and inverter bus voltage dynamics due to 5% step change in
the voltage reference of G2 . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.1 Block diagram of the LQG controller with a limiter . . . . . . . . . . . . 33

3.2 Block diagram of the MPC . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.3 Schematic one-line diagram of the WSCC system . . . . . . . . . . . . . 37
3.4 Block diagram of the MPC-based HVDC supplementary controller . . . . 38

xi
3.5 The control performance and computation time for different values of p
(WSCC system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.6 The single-sided amplitude spectrum of bus 4 voltage angle (IEEE 14-bus
1-segment system) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.7 The single-sided amplitude spectrum of bus 7 voltage angle (WSCC system) 40
3.8 IEEE 14-bus 1-segment system dominant inter-area mode . . . . . . . . . 41
3.9 The spectrogram of bus 4 voltage angle deviations from the steady-state
value for the IEEE 14-bus 1-segment system (MPC in service) . . . . . . 43
3.10 The spectrogram of bus 4 voltage angle deviations from the steady-state
value for the IEEE 14-bus 1-segment system (LQG controller in service) . 44
3.11 Control signal for the cases with LQG and MPC after an L-L-L-G fault
on bus 4 (IEEE 14-bus 1-segment system) . . . . . . . . . . . . . . . . . 45
3.12 WSCC system dominant inter-area mode . . . . . . . . . . . . . . . . . . 46

4.1 Structure of the global supplementary control (GSC) . . . . . . . . . . . 49

4.2 Schematic one-line diagram of the fully-DC-segmented test system . . . . 50
4.3 Active power deviations of T-G unit C3 in each segment, due to an L-L-L-
G fault on bus 10 of segment 1, for the partially- and fully-DC-segmented
systems (with no supplementary controller). . . . . . . . . . . . . . . . . 52
4.4 Active power deviations of AC transmission line 2 in each segment, due to
an L-L-L-G fault on bus 10 of segment 1, with and without GSC1 in service. 53
4.5 Active power deviations of AC transmission line 2, due to an L-L-L-G fault
on bus 10 of segment 1, with GSC2 either enabled or disabled . . . . . . 56
4.6 Active power deviation of T-G unit G1, due to an L-L-L-G fault on bus
10 of segment 3 of the high inertia case, with GSC1, GSC2, and no GSC
enabled . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.7 Performance of GSC1 and GSC2 as the fault location is varied . . . . . . 58
4.8 Performance of GSC1 and GSC2 as the fault location is varied (high inertia
case) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.9 Active power transmitted on the HVDC lines due to a fault on bus 10 of
segment 3 in the high inertia case, with GSC1 or GSC2 enabled . . . . . 61
4.10 Dynamic performance of GSC1 and GSC2 as the HVDC1 operating point
is varied from 70% to 100% of its rated value, in steps of 5%. . . . . . . . 62

5.1 Optimization flowchart . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

5.2 Schematic one-line diagram of the 2-segment system . . . . . . . . . . . . 70
5.3 Schematic one-line diagram of the fully-DC-segmented test system . . . . 71

xii
5.4 Real part of the closest eigenvalue to the imaginary axis versus the flows
of HVDC links . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.5 Eigenvalues plot of the system corresponding to the two operating points
obtained from the base OPF and optimization problem . . . . . . . . . . 73
5.6 Generation cost and σ versus the optimization step size . . . . . . . . . . 73
5.7 σ versus the optimization step size . . . . . . . . . . . . . . . . . . . . . 74
5.8 Eigenvalues plot of the system corresponding to three operating points
obtained from the optimization problem, and base OPF solution after and
before load variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.9 Loci of the eigenvalues associated with the least damped modes due to
changes in line impedances. The dashed line indicates the change in oper-
ating conditions determined via the method detailed in Fig. 5.1 . . . . . 76

xiii
Nomenclature

αord , βord rectifier and inverter firing angles

Ī, V̄ vectors of current and voltage phasors of each bus in the DQ refer-
ence frame

Īinj , V̄inj vectors of current and voltage phasors at the injection buses

V̄gen vector of voltage phasors at the generator buses

V̄t terminal voltage phasor in the dq reference frame

, ν process and measurement noises

δ rotor angle with respect to a synchronous reference frame

ωs , ω stator and rotor angular frequencies

Ψ stator flux linkage

Ψ1d , Ψ2q direct and quadrature axis damper winding flux linkages

d , q direct and quadrature axis elements

conv inverter/rectifier quantities

xac , uac , yac AC system state vector, input vector and algebraic variables

xdc , udc , ydc DC system state vector, input vector and algebraic variables

YN network admittance matrix

yb , ub output and input vectors of the DC system block

Yred reduced network admittance matrix

CDC , LDC , RDC HVDC line parameters

xiv
D damping constant

Econv , γconv equivalent generator bus voltage magnitude and angle at the con-
verter internal bus

Ef d field voltage

H rotor inertia constant

I, V stator current and voltage

Idref , Vdref rectifier current order and inverter voltage reference

Iconv , Vconv converter current and voltage

KA , TA regulator gain and time constant

KE , TE , SE exciter gain, time constant and saturation

KF , TF rate feedback gain and time constant

KIIcon , KPIcon HVDC rectifier current controller integral and proportional gains

KIV con , KPV con HVDC inverter voltage controller integral and proportional gains

n converter transformer turns ratio

ninj , npas number of buses with and without current injection

nT G , ndc number of T-G units and HVDC links

Pconv , Qconv converter injected active and reactive powers

Tm , Te mechanical and electrical torques

T1 , T3 , R governor time constants and droop

 
Tqo , Tqo damper winding transient and subtransient time constants

Vref , Pref terminal voltage and mechanical power references

Vacconv , θacconv AC voltage magnitude and angle at the converter bus

 
X, X , X leakage, transient, and subtransient reactances

XC commutation reactance

xv
XT converter transformer reactance

Xls , Rs stator leakage reactance and resistance

x, u perturbations of the system state variables and inputs around the
given operating point
 
Tdo , Tdo field winding transient and subtransient time constants

CC constant current

CV constant voltage

GSC global supplementary control

LSC local supplementary control

MPC model predictive control

WSCC Western Systems Coordinating Council

xvi
Chapter 1

Introduction

1.1 Large-Scale Integration of HVDC Transmission

in AC Power Systems
Power system stability is defined as the ability of the power system to remain in a state
of equilibrium under normal conditions and regain a state of equilibrium after being sub-
jected to disturbances [1]. Different forms of power system instabilities, e.g., rotor angle
instability and voltage instability, have been comprehensively explored in the technical
literature. Rotor angle stability requires the rotors of all interconnected synchronous
machines to be in synchronism. Perturbing the system equilibrium leads to an accelera-
tion or deceleration of the machines’ rotors and may lead to loss of synchronism. Rotor
angle stability phenomena are categorized as small-signal dynamic stability and transient
stability. In this thesis the focus is on small-signal rotor angle stability.
Small-signal rotor angle stability is fundamental to the safe operation of the power
system [1], and enables the power system to maintain synchronism under small distur-
bances. The disturbances are considered to be sufficiently small such that the linearized
system model can be used for stability analyses. Small-signal instability is due to lack
of sufficient damping of oscillations. Small-signal oscillations appears in the form of lo-
cal modes, inter-area modes, control modes and torsional modes [1]. Local modes are
associated with the swing of one or a group of generators against the rest of the power
system. Inter-area modes are associated with the swing of a group of generators in one
part of the power system against other aggregates of generators in other parts. The
inter-area oscillations are often experienced over a large part of the power system and
local oscillations usually appear in only a small part of the system [1].
One incident of the inter-area oscillation problem was the oscillations experienced

1
Chapter 1. Introduction 2

on Aug. 14, 2003, where Ontario and much of the northeastern U.S. were subjected
to the largest blackout in North America’s history [2]. More than 263 power plants
tripped offline in Canada and the U.S., leaving 50 million people without power for up
to nine hours. However, Quebec was not affected by the blackout, because its major
interconnections are the high voltage direct current (HVDC) transmission lines.
The classical application of HVDC system is the transmission of bulk power over long
distances due to the lower overall transmission cost and losses as compared with the AC
transmission lines [3]. Furthermore, the amount of transmitted power on HVDC lines
and the transmission distance are not limited by stability constraints. The constraints
associated with stability problems or control strategies are removed by interconnecting
systems via HVDC lines [3]. HVDC system provides some degree of buffering against
cascading failures in the grid and compared to the conventional AC transmission sys-
tem has a higher degree of controllability for the operation of power systems [3]. An
HVDC link connected between two AC systems operates regardless of the voltage and
frequency conditions of the two systems. Therefore, it provides an independent control
for transmitting power between systems. The same applies for an HVDC link within one
AC system. HVDC technology can resolve a large number of existing AC power system
steady-state and dynamic instability issues and improve the security of the system. The
integration of HVDC technology in an AC system can be achieved by

connecting AC buses of a single AC grid through HVDC links,

embedding an HVDC grid in an AC system, and

DC segmentation.

In the first configuration, known as an embedded HVDC system in an AC grid, one

or multiple point-to-point (PTP) HVDC links connect a set of AC system buses in a
single AC grid. In the second configuration, known as the HVDC grid, several AC buses
are interconnected through converter stations which share a common DC transmission
system [3]. In the third configuration, the AC grid is decomposed into smaller AC
segments connected through DC links (and potentially weak AC lines) and the main
power corridors among the segments are the DC links.

1.1.1 Embedded HVDC System in an AC Grid

In this configuration, PTP HVDC links are used for the bulk transmission of electrical
power in a single AC grid. Fig. 1.1 (a) shows a schematic diagram of embedded HVDC
links in a single AC grid. The longest embedded HVDC link in the world is currently
Chapter 1. Introduction 3

Embedded HVDC MTDC

Converter 3 Converter N-1

~ ~ ~ ~
= = = =
Converter 1 Converter 2
HVDC Transmission Line

= = =
~ ~ ~ ~
= Converter 1 Converter 2 Converter N
Converter N-1

AC System AC System
=
~
Converter N

(a) (b)
Meshed DC Grid DC-Segmented AC System
AC Segment 1 AC Segment 2
HVDC12

~ ~= =~
=
=
= ~

=~
HVDC12 ~
~ =
Converter 2
Converter 1

HVDC23
HVDC13 HVDC2N HVDC13
HVDC2N

AC System
=
~
=
=~

Converter 3 Converter N
~
=
~ =~ AC Segment 3
AC Segment N
HVDC3N HVDC3N
~= =~
(c) (d)

Figure 1.1: The integration of HVDC technology in an AC system: (a) Embedded HVDC
system in an AC grid, (b) MTDC, (c) Meshed DC grid, and (d) DC segmentation.

the 2071 km, ±800 kV, 6400 MW link in the People’s Republic of China [4]. The longest
transmission link in the world, 2375 km, will be the Rio Madeira link in Brazil, which is
scheduled for completion in 2013 [4].
Chapter 1. Introduction 4

1.1.2 HVDC Grid

Another approach for the integration of HVDC transmission in an AC grid is embed-
ding an HVDC grid in the AC system. The meshed HVDC grid includes multiple DC
converters which are interconnected by a meshed DC transmission network. In this con-
figuration, if one HVDC line is lost, another line supports the partially isolated node [5].
This configuration should contain at least one loop and can be realized only based on self-
commutated voltage-sourced converter (VSC) technology. Fig. 1.1 (c) shows a schematic
diagram of a meshed HVDC grid. Multi-terminal HVDC (MTDC) configuration is a
special type of the HVDC grid, where both line-commutated current-sourced converter
(LCC) and VSC technologies can be utilized [3]. Fig. 1.1 (b) shows a schematic diagram
of an MTDC grid.
The HVDC grid offers fast response time, reduces congestion, improves system dy-
namic stability and performance under disturbances, provides flexibility in power flow
control and facilitates the integration of renewable energy sources [3, 6–9]. In the Euro-
pean power system, where massive renewable energy sources in offshore or remote loca-
tions are utilized, this configuration is a solution that will integrate substantial amount of
renewable energy to the grid [10]. Besides, effective oscillation damping can be provided
by the independent active and reactive power modulations and transmission bottlenecks
can be addressed via the fast power flow controllability of the VSC-HVDC system, oscil-
lation damping and dynamic voltage support. In comparison to a multiple point-to-point
HVDC configuration, the MTDC configuration has less number of converter stations and
lower transmission loss. Currently, for MTDC applications, only LCC-based HVDC has
been used. Most notable lines are the one from Hydro-Quebec to New England and
the one between Sardinia, Corsica and Italy (SACOI). In comparison to an LCC-based
MTDC system, where the function of each converter station (rectifier or inverter) is of-
ten fixed, the VSC technology of the DC grid provides bi-directional power transfer by
varying the current direction and facilitates the realization of DC grids of more than a
few terminals [7].

1.1.3 AC Grid Segmentation

One approach to mitigate and/or geographically localize oscillatory modes of an AC
system to enhance the system stability is to partition the AC system into segments
by back-to-back (BTB) and/or PTP HVDC links [11, 12], based on VSC and/or LCC
technologies as shown in Fig. 1.1 (d). The segments can also have AC lines connections;
however, the AC lines should not constitute major power corridors.
Chapter 1. Introduction 5

Segmentation of the AC grid can be achieved through the conversion of AC links to

BTB or PTP DC links [12, 13], and/or the installation of new HVDC links. The size of
each AC segment is based on a trade-off between the converter cost, potential gain in
reliability and power transfer capability enhancement, geographical/political boundaries,
and the system operational characteristics and requirements [14]. The boundaries of the
AC segments are determined according to

congested areas that need more transfer capability,

locations where longer HVDC lines can be formed from existing AC lines,

locations that require the least back-to-back MVA, and

locations where HVDC links can replace stability-limited AC links [15].

Some of the key benefits of the DC segmentation are [11, 12, 15]:

minimizing cascading outages and widespread blackouts,

confining system collapse to the faulted segment,

increasing power transfer capability among segments,

providing significant local oversight on the grid,

potentially easier grid expansion planning and investment decisions,

state estimators performance enhancement,

feasibility of intelligent/self-healing grid planning,

more effective wide area measurement system (WAMS)-based applications,

controllability of the inter-area power flows,

reducing operational complexity and uncertainty, and

increasing system resiliency to natural and man-made disturbances.

The work reported in [11] is the first systematic assessment of the DC segmentation
concept and its potential advantages. In [11], the reliability and transfer capability im-
provement of the Eastern Interconnection (EI) of North America with segmentation is
examined by comparing the dynamic performance of the system before and after segmen-
tation. In [12], the response of an AC test system with and without segmentation to two
Chapter 1. Introduction 6

typical disturbances, generation loss and line-trip, are studied. The study shows that
segmentation can prevent cascading outages and reduces fault impacts on neighboring
segments. Decomposing the AC grid into smaller segments does not eliminate all the
stability problems associated with large AC grids. However, conceptually, DC segmen-
tation of the interconnected AC system results in more localized and more manageable
problems [15].

1.1.4 Mitigation of Dynamic Oscillatory Modes of an AC-DC

System
Two general methods of improving power system small-signal stability are the installation
of new devices and improving the control of existing devices. In particular, the following
methods have been investigated [16]:

installing new infrastructure, such as adding new transmission lines to the power
system or installing new generation capacity,

installing flexible AC transmission system (FACTS) controllers, e.g. unified power

flow controllers (UPFCs) [17–19], thyristor controlled phase shifting transform-
ers [17], thyristor controlled series capacitors (TCSCs) [17,20,21], static VAR com-
pensators (SVCs) [22, 23], static phase shifters (SPSs) [24] and static synchronous
compensators (STATCOMs) [25],

modifying the control scheme by adopting power system stabilizers (PSSs) [26–31],
and

modifying the control scheme by adopting HVDC modulation and set-point adjust-
ment [29, 32].

Improving the HVDC control scheme, i.e., adopting HVDC modulation techniques
and adjusting the set-point of the HVDC links, is an attractive alternative to additional
infrastructure installation. Reported studies have shown that the stability of power
systems can be improved by the judicious control of the existing HVDC connections [33].
In this thesis, we focus on the HVDC power modulation to mitigate oscillatory dynamics
and we refer to this controller as the supplementary controller (SC). Furthermore, a set-
point tuning method is proposed to increase the power system stability margin (a metric
for the stability of the closed loop system). A higher stability margin corresponds to a
more stable system with smaller amplitude oscillatory transients.
Chapter 1. Introduction 7

1.2 Statement of the Problem and Thesis Objectives

Although there have been studies on the conceptual segmentation of the AC grid with
HVDC links, small-signal stability enhancement in such a configuration has not been
investigated. The objective of this thesis is to enhance the small-signal stability of a
DC-segmented AC power system. Four control schemes have been proposed to achieve
the main objective of this thesis. Control schemes 1-4 improve the system stability by
damping the oscillations within the segment, minimizing the propagation of the oscilla-
tions to the neighboring segments, distributing the oscillations among the segments and
changing the set-points of the HVDC control system, respectively. Depending on the
system characteristics, one or more of the proposed control schemes may be effective for
the mitigation of system oscillations. Improving the system stability by modifying the
control scheme is appealing since it is cost effective in comparison to the installation of
additional equipment, and can be implemented by modifying the existing control schemes
instead of installing physical power apparatus.
To achieve the main objective of this thesis, this research work focuses on:

Developing a small-signal dynamic model based on computer-assisted linearization

of AC-DC systems for control design, time-domain simulation and systematic per-
formance evaluation of the control scheme (The developed model can be used for
various control designs and its application is not limited to the controllers designed
in this thesis).

Developing a HVDC local supplementary control (LSC) scheme, using the small-
signal dynamic model of the system, to damp inter-area oscillations in the AC-DC
system.

Developing a HVDC global supplementary control (GSC) scheme, using the small-
signal dynamic model of the system, to improve the system stability by distributing
the oscillations among the segments.

Developing a GSC scheme, using the small-signal dynamic model of the system, to
improve the system stability by minimizing the propagation of the oscillations to
the other segments.

Developing an operating point tuning scheme to adjust the HVDC control set-
points, to enhance the small-signal stability of an AC-DC system.

Each of these milestones has been achieved based on HVDC supplementary control or
modification of the operational condition of the HVDC control system.
Chapter 1. Introduction 8

1.3 Methodology
In order to achieve the aforementioned thesis objective, the following methodology is
employed:

Systematically construct the nonlinear differential-algebraic model of an AC-DC

system,

Automatically extract a linearized state space model of the nonlinear system (using
the developed MATLAB/Simulink-based platform),

Design linear control schemes for the linearized system model (using the developed
MATLAB/Simulink-based platform) and design a control scheme based on the
sensitivity of the system stability margin with respect to the parameter space, and

Perform time-domain simulation: MATLAB/Simulink environment is used to eval-

uate the agreement between the corresponding dynamic responses of the automati-
cally generated linearized model and the nonlinear model to small disturbances and
validate the accuracy of the linearized model. The performance of the proposed
control schemes and the dynamic behaviour of the system, including the proposed
controllers, under various faults and disturbances are investigated through time-
domain simulations in the developed MATLAB/Simulink-based platform.

1.4 Thesis Layout

The rest of this thesis is organized as follows:

chapter 2 introduces a modeling approach and integration procedure for developing

AC-DC power system models in the MATLAB/Simulink environment and demon-
strates the effectiveness of the automatic linearization provided by Simulink. In the
following chapters, the developed AC-DC model will be used to design the control
schemes to damp oscillations.

chapter 3 introduces, formulates and evaluates an approach for damping the os-
cillations of power systems based on supplementary current control of an LCC
HVDC link within a segment. The proposed control is based on the model pre-
dictive control (MPC) method. This chapter presents the MPC design procedure
and evaluates the performance of a discrete-time MPC-based HVDC supplemen-
tary controller for mitigating oscillatory modes of two study systems. This chapter
Chapter 1. Introduction 9

also compares the damping effect of the designed MPC-based controller with that
of a linear quadratic Gaussian (LQG) controller for the same test systems.

chapter 4 introduces two control schemes based on the supplementary control of

inter-segment HVDC links, to (i) confine the oscillatory dynamics initiated in a
segment within that segment and minimize their propagation to other segments,
or (ii) distribute the oscillatory modes among the segments. An LQG control
method is adopted to design the HVDC supplementary controls. Each HVDC
supplementary control provides simultaneous modulation of the current order and
voltage reference of the corresponding rectifier and inverter stations. Performances
of both GSC schemes are evaluated and compared. This chapter also introduces a
sensitivity measure to evaluate the performance of each design to the variations in
the system parameters and operating point.

chapter 5 presents a control scheme based on the adjustment of the set-point value
of the HVDC lines to improve the stability margin and control Hopf bifurcations
(at a Hopf point, a complex pair of eigenvalues of the linearized system crosses the
imaginary axis) caused by gradual variation of the parameters such as loads. In this
chapter, local bifurcation theory and computation of the sensitivity of the stability
margin and the real part of the critical eigenvalues with respect to the parameter
space are discussed. The optimization problem together with a brief description of
modeling the system for this type of optimization are also presented. The proposed
control scheme is evaluated on two test systems for several Hopf bifurcations caused
by a variety of events such as load and line impedance variations.

chapter 6 summarizes the contributions of the thesis, presents its conclusions, and
recommends future research directions.
Chapter 2

Small-Signal Dynamic Model

Development Based on
Computer-Assisted Linearization

2.1 Introduction

Systematic design of a controller, e.g., a linear HVDC supplementary controller, requires

a differential-algebraic model of the system, which includes both DC and AC components.
Linearization of the models is one of the requirements for the control design process, since
most of the controllers applied to power systems are linear controllers. The characteristics
of the nonlinear system to be controlled vary due to topological changes, load/generation
variations, etc. These changes must be taken into account in the linearized model. There-
fore, to achieve the desired performance in the system, the linearization process often has
to be repeated each time the system is subjected to a change. Thus, automation of the
linearization process is highly desirable and makes the controller design process faster
and more accurate.
The design of a linear controller requires (i) a mathematical description of the nonlin-
ear AC-DC system model, (ii) steady-state solution of the system to obtain the operating
point about which the linearized model is developed, and (iii) linearization of the non-
linear model. The automation of steps (ii) and (iii) is essential for the practical design of
controllers, given that the operating point and parameters of the system vary and, with
each change, the linearization process has to be repeated.
There exists a host of software packages, e.g., PSS/E, DIGSILENT, and PST, for the
analysis of low-frequency dynamics and steady-state response of power systems. There

10
Chapter 2. Small-Signal Dynamic Model Development 11

are three main limitations in using the existing software tools to design controllers for
AC-DC systems: (i) lack of provisions to provide a full nonlinear state-space model
of the AC-DC system, (ii) inability to automatically generate a state-space linearized
model, and (iii) limited component modeling capability. For example, two of the popular
commercial packages for power system analysis, DIGSILENT and PSS/E, provide some
limited information of the linearized model to the user, but do not provide the full linear
model needed in controller design. The NEVA-Eigenvalue and Modal Analysis module
of the PSS/E software only provide eigenvalues and eigenvectors, and DIGSILENT only
provides the eigenvalues, eigenvectors, controllability, observability and participation fac-
tors for each state variable. Although these eigenvalue analyses are likely based on an
internal linearized model, the full linearized state space model is not available to the user.
Instead, only data extracted from that model is provided to the user, whereas for control
design purposes, the full linearized state space model is required.
The Power System Toolbox (PST) consists of a set of coordinated MATLAB m-files
that model the power system components for power flow and stability studies, provides
a full linearized model of the AC-DC system [34] and does not have the above mentioned
limitation. However, PST has a limited set of models (e.g., the dynamics of the HVDC
controllers and the DC transmission line are not included in this software) and is not
being actively developed. In addition, custom models cannot be automatically linearized
in this software.
As an alternative to specialized power system software, MATLAB/Simulink [35] has
been successfully used for time-domain simulation of AC-DC power systems and offers
specific advantages when compared with standard power system analysis packages. For
example, once a model is built in Simulink, one can obtain alternative representations
of the system (e.g., state-space models or transfer functions), using the automated tools
that are parts of the Simulink software [35]. Furthermore, the block structure of this
software enables a controller designer to construct new device and controller models using
Simulink’s extensive library of standard control blocks and obtain the new linearized
model automatically. Given its capabilities, this platform is a suitable environment for
controller design.
The objective of this chapter is to present a nonlinear state-space model of an in-
terconnected AC-DC power system and describe the approach to define the model in
Simulink. Particular emphasis is placed on describing the approach to combine the dif-
ferential and algebraic models of the DC and AC subsystems and to enable automatic
linearization for control design. Furthermore, the accuracy of the automatic linearization
provided by Simulink is demonstrated.
Chapter 2. Small-Signal Dynamic Model Development 12

2.2 Low-Frequency Dynamic Model of the AC Sys-

tem
The balanced AC systems under consideration include the turbine-generator (T-G) units
and their controls, AC transmission lines, and transformers. The detailed model of each
of these components is given in [36], and a brief summary of each is provided below.

2.2.1 Turbine-Generator (T-G) Unit Model

The electromechanical system of each T-G unit is assumed to include a synchronous

machine (SM) and a turbine system. The electrical system of each SM is represented by
one field winding and one damper winding on the rotor d-axis and two damper windings
on the rotor q-axis. The stator dynamics of each SM are neglected and it is assumed that
the SM magnetic circuit is linear. The model of each machine rotor electrical system is
expressed in a d-q reference frame which rotates at the corresponding rotor speed. The
stator circuitry of the ith SM is represented by the following algebraic equations [36]

ωi
Rsi Idi + Ψqi + Vdi = 0, (2.1)
ωs

ωi
Rsi Iqi − Ψdi + Vqi = 0, (2.2)
ωs
where   
 (Xdi − Xlsi )  (Xdi − Xdi )
Ψdi = −Xd Idi +  Eqi +  Ψ1di , (2.3)
(Xdi − Xlsi ) (Xdi − Xlsi )
  
 (Xqi − Xlsi )  (Xqi − Xqi )
Ψqi = −Xqi Iqi −  Edi +  Ψ2qi . (2.4)
(Xqi − Xlsi ) (Xqi − Xlsi )
The dynamics of the rotor electrical system of the ith SM in a dqo frame rotating at the
rotor speed (zero sequence is neglected) are given by



dEqi 
 X − Xdi
 
 
 
Tdoi = Ef di −(Xdi −Xdi ) Idi − di 2 (Ψ 1di +(X di −X lsi )I di −Eqi ) −Eqi , (2.5)
dt (Xdi − Xlsi )

 dΨ1di  
Tdoi = −Ψ1di + Eqi − (Xdi − Xlsi )Idi , (2.6)
dt


dEdi 
  
Xqi − Xqi    
 
Tqoi = (Xqi − Xqi ) Iqi −  Ψ2qi + (Xqi − Xlsi )Iqi + Edi − Edi , (2.7)
dt (Xqi − Xlsi ) 2
Chapter 2. Small-Signal Dynamic Model Development 13

 dΨ2qi  
Tqoi = −Ψ2qi − Edi − (Xqi − Xlsi )Iqi . (2.8)
dt
The mechanical system of the ith T-G unit is represented by an equivalent rigid mass
and its dynamics are given as [36]

dδi
= ωi − ωs , (2.9)
dt

2Hi dωi ωs  
= Tmi − Di (ωi − ωs ) − 2
Vdi Idi + Vqi Iqi + Rsi (Idi 2
+ Iqi ) . (2.10)
ωs dt ωi

2.2.2 Excitation System Model

It is assumed that each SM is equipped with an IEEE Type-I exciter system [36] and
dynamically described by

dEf di
TEi = − (KEi + SE (Ef di )) Ef di + VRi , (2.11)
dt

dRf i KF i
TF i = −Rf i + Ef di , (2.12)
dt TF i

dVRi KAi KF i
TAi = −VRi + KAi Rf i − Ef di + KAi (Vref i − |V¯ti |). (2.13)
dt TF i
If VRi reaches its limit, it is set to that limit in (2.11).

2.2.3 Governor System Model

Each T-G unit is equipped with a TGOV1 type governor [36] which is dynamically
described by
dTsvi  ωi Pref i 1 
= 1/T1i − Tsvi + − (ωi − ωs ) , (2.14)
dt ωs Ri Ri
dTmi  
= 1/T3i − Tmi + Tsvi . (2.15)
dt
Chapter 2. Small-Signal Dynamic Model Development 14

2.2.4 AC Network Model

The AC network is composed of AC transmission lines and transformers. Since the

objective of this study is to investigate low-frequency dynamics, e.g., 0.1 to 2 Hz, the
AC transmission network is represented by the positive sequence algebraic equations
associated with the network nodal equations

Ī = YN V̄, (2.16)

where Ī and V̄ are expressed in the system global DQ reference frame at the high voltage
bus of T-G unit 1 [36] with components I¯i = IDi + jIQi and V̄i = VDi + jVQi , respectively.
Power system loads are modeled as constant impedances and included in the YN matrix.

2.2.5 Multi-Machine System Model

The dynamic model of each T-G unit, (2.1)-(2.2) and (2.5)-(2.15) is transformed to the
system global DQ reference frame and arranged in a state-space form. These equations
constitute the following set of nonlinear differential-algebraic equations (DAEs)

ẋac = fac (xac , uac , yac ), (2.17)

gac (xac , yac ) = 0, (2.18)

where  T
xac = xT1ac xT2ac . . . xTnT Gac , (2.19)
 
xiac = [Eqi Edi Ψ1di Ψ2qi δi ωi Ef di Rf i VRi Tsvi Tmi ]T , (2.20)
 T
uac = uT1ac uT2ac . . . uTnT Gac , (2.21)
 T
uiac = Pref i ωs Vref i , (2.22)
 T
yac T T T
= y1ac y2ac . . . ynT G ac , (2.23)
 T
yiac = VDi VQi Vdi Vqi IDi IQi Idi Iqi , (2.24)
 T
fac = fT1ac fT2ac . . . fTnT Gac , (2.25)
Chapter 2. Small-Signal Dynamic Model Development 15

⎡       ⎤
  X −X   
1/Tdoi Ef di − (Xdi − Xdi ) Idi − di di 2 (Ψ1di + (Xdi − Xlsi )Idi − Eqi ) − Eqi
⎢   (Xdi −Xlsi )
  ⎥
⎢ Xqi −Xqi 
 
 
⎥
⎢   
1/Tqoi (Xqi − Xqi ) Iqi − (X  −X )2 Ψ2qi + (Xqi − Xlsi )Iqi + Edi − Edi

⎥
⎢  qi lsi  ⎥
⎢ ⎥
⎢   
1/Tdoi − Ψ1di + Eqi − (Xdi − Xlsi )Idi ⎥
⎢ ⎥
⎢   ⎥
⎢   
1/Tqoi − Ψ2qi − Edi − (Xqi − Xlsi )Iqi ⎥
⎢ ⎥
⎢ ⎥
⎢ ωi − ωs ⎥
⎢ ⎥
⎢     ⎥
fiac =⎢
⎢
ωs
Tmi − Di (ωi − ωs ) − ωi Vdi Idi + Vqi Iqi + Rsi (Idi + Iqi )
ωs 2 2 ⎥
⎥
⎢ 2Hi
  ⎥
⎢ ⎥
⎢ 1/TEi − (KEi + SE (Ef di )) Ef di + VRi ⎥
⎢   ⎥
⎢ ⎥
⎢ 1/TF i − Rf i + K Fi
E f di ⎥
⎢  TF i
 ⎥
⎢ ⎥
⎢ 1/TAi − VRi + KAi Rf i − TF i Ef di + KAi (Vref i − |V¯ti |)
KAi KF i
⎥
⎢   ⎥
⎢ ⎥
⎢ 1/T1i − Tsvi ωωsi + R
Pref i
− 1
(ω − ω ) ⎥
⎣  i Ri

i s
⎦
1/T3i − Tmi + Tsvi .
(2.26)
In (2.18), gac (xac , yac ) = 0 represents the algebraic equations of the stator circuitry of
the SMs (gSMac ), (2.28)-(2.29), network nodal equations (gN Eac ), (2.16), and coordinate
transformation equations (gCOac ), (2.30)-(2.33).
 T
gac = gTSMac gTN Eac gTCOac , (2.27)

 T
gSMac = gT1ac gT2ac . . . gTnT G ac , (2.28)
 ωi

Rsi Idi + Ψ
ωs qi
+ Vdi
giac = . (2.29)
Rsi Iqi − ωi
Ψ
ωs di
+ Vqi
Coordinate transformation algebraic equations gCOac include (2.30)-(2.33) for all the gen-
erator buses.
VDi − Vdi sin δi − Vqi cos δi = 0, (2.30)

VQi − Vqi sin δi + Vdi cos δi = 0, (2.31)

Idi − IDi sin δi + IQi cos δi = 0, (2.32)

Iqi − IQi sin δi − IDi cos δi = 0. (2.33)

The complete differential-algebraic model of the multi-machine AC system including the

T-G units, coordinate transformation and AC network algebraic equations (2.17)-(2.18)
are combined based on the block diagram of Fig. 2.1.
Chapter 2. Small-Signal Dynamic Model Development 16

‫ݏ݀ܽ݋ܮ‬
ܸത௜ ܸത௝
Network Model To Other Machines
‫ܫ‬௜ҧ ‫ܫ‬௝ҧ

ܸ௥௘௙௜ Exciter # i Governor # i ܲ௥௘௙௜

Model Model

‫ܫ‬஽௜ ǡ ‫ܫ‬ொ௜ ‫ܧ‬௙ௗ௜ ܶ௠௜

ܸௗ௜ ǡ ܸ௤௜
߱௜
T-G Unit # i
DQ\dq Transform
Model
ܸ஽௜ ǡ ܸொ௜ ‫ܫ‬ௗ௜ ǡ ‫ܫ‬௤௜
ߜ௜

Figure 2.1: Block diagram of the AC system model

DC
AC to DC Rectifier
Line
Conversion Model
Model
Model

Controller Model

PI
Controller
(Rectifier) Inverter
Model
PI
Controller
(Inverter)

Figure 2.2: Block diagram of the DC system model

2.3 Low-Frequency Model of the DC System

The model of a DC system, to be integrated with the AC network model, includes
models of four components, i.e., the HVDC controller, DC line, AC to DC conversion
block and converter stations, Fig. 2.2. The models of these components are described in
the following sections.
Chapter 2. Small-Signal Dynamic Model Development 17

2.3.1 AC to DC Conversion Model

This block relates the corresponding AC and DC quantities based on the steady-state
voltage-current relationship of the HVDC rectifier (rec) and inverter (inv) buses, i.e.,
√
3 2 3Xci
Vdreci = nVacreci cos αordi − Idreci , (2.34)
π π

√
3 2 3Xci
Vdinvi = nVacinvi cos βordi + Idinvi . (2.35)
π π

2.3.2 DC System Controller Models

The HVDC control system consists of the HVDC main and supplementary controllers [1].
Supplementary controllers are used to enhance the AC system dynamic performance. In
this thesis, the supplementary controllers will be used to improve the damping of AC
system oscillations and will be further explained in chapter 3.
In the HVDC main controller, the voltage and current regulation responsibilities are
assigned to separate stations. Under normal operation, the rectifier station controls the
corresponding DC side current (constant current mode (CC)) and the inverter station
regulates the corresponding DC side voltage (constant voltage mode (CV)), e.g., using
proportional-integral (PI) controllers. The rectifier regulates the current by adjusting the
valve firing angle, αord , and the inverter regulates the voltage by adjusting the inverter
firing angle, βord .
The block diagram of the HVDC control system is shown in Fig. 2.3. In Fig. 2.3,
the CC and CV modes of the main HVDC controller and also HVDC supplementary
controllers are depicted. The supplementary controllers modulate the current order and
voltage reference and their outputs are added to the input of the main HVDC controllers.

2.3.3 DC Transmission Line Model

The DC line is represented by a T-equivalent shown in Fig. 2.4, and dynamically modeled
as
dIdreci 1
= (Vdreci − Vci − Rdci Idreci ), (2.36)
dt Ldci
dIdinvi 1
= (−V dinvi + Vci − Rdci Idinvi ), (2.37)
dt Ldci
Chapter 2. Small-Signal Dynamic Model Development 18

measured
measured
d
d inv d
rec PI ord
Limits +
PI ȕ Controller +
d
ord + Controller Limits
+
Supplementary
Supplementary Controller
Controller

Modulation
Modulation Signal Signal

Figure 2.3: HVDC control system

Idrec Id
inv
Vc
Vd + Ldc + Ldc + Vd
rec Rdc Rdc inv
- - -
Figure 2.4: DC transmission line model

dVci 1
= (Id − Idinvi ). (2.38)
dt Cdci reci

2.3.4 Converter Model

The injection modeling approach [37] is used to represent each converter station. Each
converter station is represented by a dynamic equivalent generator behind the reactance of
the corresponding converter transformer, Fig. 2.5. The voltage and angle at the equivalent
generator buses are adjusted to generate the desired active and reactive power flows as

Econvi Vacconvi sin(γconvi − θacconvi )

Pconvi = , (2.39)
XTi

2
Econvi − Econvi Vacconvi cos(γconvi − θacconvi )
Qconvi = . (2.40)
XTi
Chapter 2. Small-Signal Dynamic Model Development 19

ܸ௔௖ ௥௘௖ ‫ߠס‬௔௖ ௥௘௖ ܸ௔௖ ௜௡௩ ‫ߠס‬௔௖ ௜௡௩

்ܺ ்ܺ

‫ܧ‬௥௘௖ ‫ߛס‬௥௘௖ ‫ܧ‬௜௡௩ ‫ߛס‬௜௡௩

ܲ௥௘௖ ൅ ݆ܳ௥௘௖
ܲ௜௡௩ ൅ ݆ܳ௜௡௩

‫ݎ݋ݐܽݎ݁݊݁ܩݐ݈݊݁ܽݒ݅ݑݍܧ‬ ‫ݎ݋ݐܽݎ݁݊݁ܩݐ݈݊݁ܽݒ݅ݑݍܧ‬

Figure 2.5: Injection model for HVDC converter stations

Therefore, the voltage and angle at the equivalent generator buses are

Econvi = a2convi + b2convi , (2.41)

 
−1 aconvi
γconvi = θacconvi + tan , (2.42)
bconvi
where
XT Pconvi
aconvi = , (2.43)
Vacconvi
   
1 XT2i Pconvi
2
bconvi = Vacconvi + Vac2 convi − 4 − XTi Qconvi , (2.44)
2 Vac2 convi

Preci = −Vdreci Idreci , Qreci = Preci tan ϕreci , (2.45)

X c Id
ϕreci = cos−1 (cos αordi − √ i reci ), (2.46)
2nVacreci

Pinvi = Vdinvi Idinvi , Qinvi = Pinvi tan ϕinvi , (2.47)

X c Id
ϕinvi = cos−1 (cos βordi + √ i invi ). (2.48)
2nVacinvi

2.3.5 Overall DC System Model

The HVDC link equations are arranged in a state-space form and constitute a set of
nonlinear differential and output equations which represent the HVDC system block as
Chapter 2. Small-Signal Dynamic Model Development 20

given by
ẋdc = fdc (xdc , ub ), (2.49)

yb = hdc (xdc , ub ) (2.50)

where  T
xdc = xT1dc xT2dc . . . xTndcdc , (2.51)
 T
xidc = Vci Idreci Idinvi Iconi Vconi , (2.52)
 T
ub = uTb1 uTb2 . . . uTbndc , (2.53)
 T
ubi = θacreci θacinvi Vacreci Vacinvi Vdref i Idref i , (2.54)
 T
T T T
yb = yb1 yb2 . . . ybndc , (2.55)
 T
ybi = γreci γinvi Ereci Einvi . (2.56)

Icon and Vcon are the states for the PI controllers used to regulate the current and voltage.
The DAEs of the DC system are

ẋdc = fdc (xdc , udc , ydc ), (2.57)

gdc (xdc , ydc ) = 0, (2.58)

where  T
udc = uT1dc uT2dc . . . uTndcdc , (2.59)
 T
uidc = Vdref i Idref i , (2.60)
 T
T T T
ydc = y1dc y2dc . . . yndcdc , (2.61)

yidc = [Vdiinv Vdirec αordi βordi γreci γinvi Ereci Einvi Pinvi Preci Qinvi Qreci ]T , (2.62)
 T
fdc = f1Tdc f2Tdc . . . fnTdcdc , (2.63)
Chapter 2. Small-Signal Dynamic Model Development 21

⎡ ⎤
1
(Idreci − Idinvi )
⎢ Cdci
⎥
⎢ 1 (V − − ⎥
⎢ Ldci reci
d V ci Rdc I d ) ⎥
⎢
i reci
⎥
fdci ⎢
= ⎢ L (−V dinvi + Vci − Rdci Idinvi )⎥
1
⎥. (2.64)
⎢ dci ⎥
⎢ KIIcon (Idreci − Idref i ) ⎥
⎣ ⎦
KIV con (Vdinvi − Vdref i )

In (2.58), gdc (xdc , ydc ) = 0 represents the algebraic equations of the AC to DC conver-
sion block (gCVdc ), (2.66)-(2.67), network nodal equations (gN Edc ), and converter model
equations (gCOdc ). Converter model equations, gCOdc , include (2.39)-(2.40), (2.45),(2.47)
for all the converter stations.
 T
gdc = gTCVdc gTN Edc gTCOdc , (2.65)

 T
gCVdc T T T
= g1dc g2dc . . . gndcdc , (2.66)
⎡ √
3Xci
⎤
−Vdreci + 3 2
nVacreci cos αordi − Idreci
⎢ π
√
π
⎥
⎢−Vd + 3 2
nVacinvi cos βordi
+
3Xci
Idinvi ⎥
⎢ invi π π ⎥
gidc =⎢ ⎥. (2.67)
⎢ −αord + KP (Id − Id ) + Iconi ⎥
⎣ i Icon reci ref i ⎦
−βordi + KPV con (Vdinvi − Vdref i ) + Vconi

2.4 Overall Model of the Multi-Machine AC-DC Sys-

tem
The overall model of the multi-machine AC-DC system is constructed from (2.17)-(2.18)
and (2.57)-(2.58) as a set of DAEs:
 
ẋac
= f(xac , xdc , ydc , yac , udc , uac ), (2.68)
ẋdc
⎡ ⎤
gSMac
⎢ ⎥
⎢ ⎥
⎢ gCOac ⎥
⎢ ⎥
0=⎢
⎢ gCVdc ⎥,
⎥ (2.69)
⎢ ⎥
⎢ gCOdc ⎥
⎣ ⎦
Īinj − Yred V̄inj
Chapter 2. Small-Signal Dynamic Model Development 22

where  T
V̄inj = V̄gen Erec ∠γrec Einv ∠γinv . (2.70)

To calculate Yred , the network admittance matrix YN is partitioned into four matrices
as
ninj npas
     
ninj Īinj ninj YA YB V̄inj
= , (2.71)
npas 0 npas YC YD V̄pas

where V̄pas is the vector of voltage phasor at the buses without current injection. Injection
buses include the generator and HVDC buses. Since there are no current injections at
the npas network buses, these buses can be eliminated from (2.71). Thus
 −1

Īinj = YA − YB YD YC V̄inj = Yred V̄inj . (2.72)

At a given operating point, linearization of the model and elimination of the algebraic
variables results in the following linearized model of the AC-DC system:

ẋ = Ax + Bu, (2.73)

where A and B are block structured matrices, with nT G blocks associated with the T-G
units and ndc blocks associated with the HVDC links.

2.5 Implementation of the AC-DC System Model in

MATLAB/Simulink
This section describes an implementation of the AC-DC system model, equations (2.68)-
(2.69), in the MATLAB/Simulink environment for time-domain simulation and auto-
matic generation of the linearized system state-space model.
The general signal flow of the AC-DC system model, including a DC-link supplemen-
tary controller, is shown in Fig. 2.6. Details of the HVDC supplementary controller will
be discussed in chapter 3. Each block is constructed in the Simulink package. A challenge
in the implementation of the system model within the MATLAB/Simulink environment
is handling algebraic loops. An algebraic loop is formed when a signal loop with only
direct feedthrough blocks exists, i.e., the output of the block is directly controlled by
the input and no state variable exists in the block [35]. To overcome this problem, the
outputs of the Network Equations block can be supplied to other blocks through Delay
Chapter 2. Small-Signal Dynamic Model Development 23

‫ܫ‬ଵҧ 
ܸത௧ଵ  ‫ܫ‬ଶҧ 
T-G1

Network Equations

ǥ
ܲ௥௘௙ଵ  ܸത௧ଶ 
ܸ௥௘௙ଵ  ‫ܫ‬௜ҧ 

ǥ
ܸത௧௜  ܸ௔௖ ௜௡௩
ܸ௔௖ ௥௘௖
T-G2 ߠ௔௖ ௜௡௩
ܲ௥௘௙ଶ 
ߠ௔௖ ௥௘௖

‫ܧ‬௜௡௩೔ ‫ߛס‬௜௡௩೔ 
ܸ௥௘௙ଶ 
ǥ


HVDC
‫ܧ‬௥௘௖ ௜ ‫ߛס‬௥௘௖೔ 
System ‫ܫ‬ௗ೚ೝ೏
#i є Modulation
T-G i ‫ܫ‬ௗೝ೐೑ Signal Set-Point
ܲ௥௘௙௜  Supplementary
Adjustment
ܸ௥௘௙௜  ܸௗೝ೐೑ Controller
Unit
є
ܸௗ೚ೝ೏

Figure 2.6: Signal flow of the AC-DC system with supplementary controllers and set-
point adjustment unit

Blocks or first-order filters. By choosing a sufficiently small time-constant for the filter
(e.g., 0.0005 s), the algebraic loop can be avoided without affecting the dynamics of
interest.
Simulation of the AC-DC system dynamic behaviour requires the coordinated solu-
tion of the system’s DAEs. At every integration step of the differential equations, the
DC system, with respect to each converter station, is represented by a Thevenin equiv-
alent [37] and the AC system is solved. Next, the terminal voltages of the generators
and the equivalent generators are substituted in the network equations and the generator
terminal currents are calculated. The voltage phasor of the converter buses are then used
to calculate the voltage phasors of the equivalent generators at the converters’ internal
buses (buses connected to the equivalent generators). Each phasor voltage and the corre-
sponding converters transformer reactance form the DC side Thevenin equivalent in the
subsequent iteration.
One major advantage of implementing the AC-DC system model in the MATLAB/
Simulink environment is the capability to automatically extract the linear state-space
model of Simulink-implemented systems given an operating point, system states and
Chapter 2. Small-Signal Dynamic Model Development 24

inputs. The automated linearization process enables a control design that takes into
account a wide range of parameters and system conditions. The “linmod” command of
the MATLAB environment is used to extract the continuous linear state-space model of
the system around a given operating point. To linearize the system model, the inputs
and outputs of the system are specified in the Simulink block diagram and are used
to compute the system linear model by linearizing each block individually. The default
algorithm uses the Padé approximation to linearize the delay blocks and preprogrammed,
analytic Jacobians for linearizing the remaining blocks. This approach results in a more
accurate linearized model compared to those obtained from numerical perturbation of
the inputs and states [35].
The operating point at which the system is linearized is determined by solving the
DC and AC system power flow equations using the parameters specified in the Simulink
model. A standard iterative solver of the AC and DC power flow equations can be used,
e.g., as detailed in [1].

2.6 Validation of the MATLAB/Simulink-Based AC-

DC Model

In this section, validation of the accuracy of the nonlinear and linearized AC-DC system
models is investigated. The IEEE 14-bus 1-segment system is developed for this investi-
gation. In the plots shown, per unit quantities are based on the corresponding machine
ratings.

2.6.1 IEEE 14-Bus 1-Segment System

Fig. 2.7 shows a schematic diagram of the IEEE 14-bus system [38] which includes five
T-G units and twenty AC transmission lines. Based on the system of Fig. 2.7, the IEEE
14-bus 1-segment system is constructed. Fig. 2.8 shows the schematic one-line diagram
of the 1-segment system which is the same as Fig. 2.7, except an LCC HVDC link is
connected between bus 2 (rectifier station) and bus 4 (inverter station). The HVDC link
and control parameters are given in [37]. The governor parameters of the T-G units are
given in Table 2.1.
Chapter 2. Small-Signal Dynamic Model Development 25

13
14
12 19 20
17

11
G1 13 10

9
16
Slack 1
12 11
18
Bus 9 15 7

2 C3
6 14
10 8 8
C2
1 7 4
5

5
4
2
6

3
G2
3

C1

Figure 2.7: Schematic one-line diagram of the IEEE 14-bus system

13
14
12 19 20
17

11
G1 13 10

9
16
Slack 1
12 11
18
Bus 9 15 7

2 C3
6 14
10 8 8
C2
1 7 4
5

I1
R1 PDC1
5
4
2
6

3
G2
3

C1

Figure 2.8: Schematic one-line diagram of the IEEE 14-bus 1-segment system
Chapter 2. Small-Signal Dynamic Model Development 26

Table 2.1: Governor parameters of the T-G units

i T1i T2i T3i Ri

1 0.02 0 0.5 0.05

2 0.02 0 0.5 0.05

3 0.2 0 5.0 0.5

4 0.2 0 5.0 0.5

5 0.2 0 5.0 0.5

(a)
1.05
1
|V|(pu)

0.95 Simulink
0.9 PSS/E
0.85
2 3 4 5 6 7 8 9 10
(b) t(s)
−10
∠V (deg)

−15

2 3 4 5 6 7 8 9 10
(c) t(s)

0.2
P G (pu)

−0.2
2 3 4 5 6 7 8 9 10
t(s)
(d)
1
Q (pu)

0.5
G

2 3 4 5 6 7 8 9 10
t(s)

Figure 2.9: Voltage magnitude, angle, active and reactive power of machine C1 on bus 3
due to the L-L-L-G fault

2.6.2 Validation of the Nonlinear AC-DC Model

To demonstrate the accuracy of the time-domain simulation studies based on the devel-
oped MATLAB/Simulink nonlinear dynamic model, the dynamic behaviour of the IEEE
14-bus 1-segment test system, Fig. 2.8, is investigated and compared against simulation
Chapter 2. Small-Signal Dynamic Model Development 27

(a)

−0.05 Simulink
−0.1 PSS/E

P 6 (pu)
−0.15
−0.2
−0.25

0 2 4 6 8 10
t(s)
(b)
0.8
0.6
Q 6 (pu)

0.4
0.2
0
0 2 4 6 8 10
t(s)

Figure 2.10: Active and reactive power flow changes of line 6 between buses 3 and 4 due
to the L-L-L-G fault
(a)
1.02
|V|(pu)

1 Simulink
PSS/E
0.98
19.5 20 20.5 21 21.5 22
t(s)
(b)
-10
‘V (deg)

-15

-20
19.5 20 20.5 21 21.5 22
(c) t(s)
0.5
PG (pu)

-0.5
19.5 20 20.5 21 21.5 22
(d) t(s)
QG (pu)

0.4

19.5 20 20.5 21 21.5 22

t(s)

Figure 2.11: Voltage magnitude, angle, active and reactive power of machine C3 on bus
8 due to line 3 tripping and reclosure
Chapter 2. Small-Signal Dynamic Model Development 28

(a)
0

P 6 (pu)
−0.5 Simulink
PSS/E

−1
19.5 20 20.5 21 21.5 22
(b) t(s)
0.4

0.3
Q 6 (pu)

0.2

0.1
19.5 20 20.5 21 21.5 22
t(s)

Figure 2.12: Active and reactive power flow changes of line 6 due to line 3 tripping and
reclosure

results obtained from PSS/E. The generators, exciters, and DC line models used in the
PSS/E simulations are GENROU, IEEET1 and CDC4T, respectively. In the first case
study, the test system is subjected to an L-L-L-G fault on bus 13 at time t = 2 s that
self-clears after 5 cycles. Figs. 2.9-2.10 show the system’s dynamic response to this fault.
In the second case study, line 3 between buses 2 and 3 is opened by the line circuit
breakers at time t = 20 s and successfully reclosed after 5 cycles. Figs. 2.11-2.12 show
the system’s dynamic response to this disturbance.
Figs. 2.9-2.10 and Figs. 2.11-2.12 show close agreements between the corresponding
systems dynamic responses using MATLAB/Simulink and PSS/E. The small discrepan-
cies are likely due to a difference in the DC link model (the CDC4T model used in the
PSS/E simulations does not represent the DC line and DC controller dynamics [39]).

2.6.3 Validation of the Linearized Dynamic Model

To validate the automatically generated linearized model, the dynamic behaviour of the
IEEE 14-bus 1-segment test system due to a 5% step increase in the reference voltage of
generator 2 (Vref 2 ) is simulated using nonlinear and linearized models. The simulation
results from both models are compared in Figs. 2.13-2.14. Close agreement between the
corresponding results of the linearized and nonlinear models verify the accuracy of the
linearized model.
Chapter 2. Small-Signal Dynamic Model Development 29

(a)
1.064

1.062

V (pu)
t3
1.06 Nonlinear System
Linear System
1.058
0 2 4 6 8 10
t(s)
(b)
1.02

1.015
V (pu)
t1

1.01 Nonlinear System

Linear System

0 2 4 6 8 10
t(s)

Figure 2.13: Voltages of buses 1 and 3 due to 5% step change in the voltage reference of
G2

(a)
1.055

1.05
E inv(pu)

Linear System
1.045
Nonlinear System
1.04
0 2 4 6 8 10
t(s)
(b)
0.5002
Linear System
I dinv(pu)

0.5001
Nonlinear System

0.5

0.4999
0 2 4 6 8 10
t(s)

Figure 2.14: DC current and inverter bus voltage dynamics due to 5% step change in the
voltage reference of G2
Chapter 2. Small-Signal Dynamic Model Development 30

2.7 Conclusions
This chapter presents a MATLAB/Simulink-based framework which can be used for
controller design in interconnected AC-DC power systems (This framework will be used
in the following chapters to design controllers). The key feature of this platform, i.e.,
modeling flexibility, accuracy of time-domain simulation, and automatic derivation of
linear models, make it a feasible analytical tool for the linear controller design. The
developed platform also enables for easy inclusion of additional component models, e.g.,
those of FACTS controllers and custom-specified apparatus, and automatic linearization
of the models for controller design.
The MATLAB/Simulink-based AC-DC power system model can also be used as an
independent platform for the time-domain simulation of AC-DC power system transient
phenomena. The accuracy of the simulation results based on the developed model was
validated by comparison to the corresponding simulation results obtained from PSS/E.
The capability to automatically extract an accurate linear model of an AC-DC power
system from the complete differential-algebraic nonlinear model and the accuracy of the
nonlinear model were also demonstrated by comparison of the linear and nonlinear model
responses to small-signal dynamics. Automatic extraction of the linear state space model
enables (i) linear control design and (ii) ease of adjusting the designed controller due to
changes in the system operating point, configuration and parameter values.
Chapter 3

HVDC Local Supplementary

Control (LSC)

3.1 Introduction
Supplementary control of the classical HVDC rectifier station, which provides modulation
to the rectifier current controller, has been long recognized as an effective means for the
mitigation of oscillations of interconnected power systems [33]. The technical literature
reports the following approaches for the design of the HVDC supplementary controller:

Nonlinear feedback linearization method [16]: In this method, system dynamics are
transformed into a linear form using a nonlinear pre-feedback loop, and then, for
the linearized system, another feedback loop is designed. The main advantage of
this approach is that the proposed state feedback linearization does not rely on the
assumption that there is only a small deviation of the states from the equilibrium.
However, it relies on finding the nonlinear pre-feedback loop, which is not always
possible.

Numerical optimization methods: The optimization based methods search in the

parameter space to identify the set of control parameters for the desired perfor-
mance and can utilize nonlinear programming [40, 41], and heuristic methods, e.g.,
neural networks and genetic algorithms [42]. One of the drawbacks of these methods
is that the search in the parameter space is time-consuming.

LQG controller: The linear quadratic regulator (LQR), in which the objective
function is a quadratic function of the state vector and control inputs, has been
widely used in the design of power system controllers [43]. In the LQR design, if all

31
Chapter 3. LSC for Small-Signal Stability Enhancement 32

system states are not measured, an observer is required to estimate the states. The
LQ regulators and the state estimators together form the LQG controller. While
LQG control has been widely studied in the context of power systems [44, 45], it
is not able to incorporate all system constraints, e.g., hard limits on the control
signal.

To overcome the drawbacks of these methods and enhance the damping of inter-area
oscillations in power systems, this thesis proposes an HVDC supplementary controller
based on the MPC methodology. MPC combines a prediction strategy and a control
strategy to hold the system output at a reference value by adjusting the control signal.
The advantage of the MPC compared with other optimal control strategies, e.g., LQG,
is that it adjusts the control signal to achieve the objectives while explicitly respecting
the plant constraints [46, 47].
This chapter describes the MPC-based and LQG-based supplementary controller de-
sign procedures and evaluates the performance of a discrete-time MPC-based HVDC
supplementary controller for mitigating oscillatory modes of two study systems. This
chapter also compares the damping effect of the designed MPC-based controller with
that of an LQG controller for the same test systems. While the proposed MPC-based
controller can be applied to both LCC- and VSC-HVDC systems, this thesis only con-
siders the LCC-HVDC system.

3.2 MPC and LQG Controllers

Optimal control theory determines the control signal of a dynamic system over a time
horizon to minimize a performance index (cost function). This section summarizes two
optimal controllers, the MPC-based and LQG controllers, which are designed and eval-
uated in the reported studies.

3.2.1 Linear Quadratic Gaussian (LQG) Control

A linear dynamic system can be described by

ẋ(t) = Ax(t) + Bu(t) + (t), (3.1)

y(t) = Cx(t) + Du(t) + ν(t), (3.2)

where x(t), y(t), u(t), (t) and ν(t) are the state, output, input, process noise and mea-
surement noise vectors, respectively. The goal of an LQR is to identify a state-feedback
Chapter 3. LSC for Small-Signal Stability Enhancement 33

‫ܝ‬ ‫ܡ‬
‫ܠ‬ሶ ൌ ‫ ܠۯ‬൅ ۰‫ ܝ‬൅ ૓

۹
‫ܠ‬ොሶ ൌ ሺ‫ ۯ‬െ ‫ۺ‬۱ሻ‫ܠ‬ො ൅ ۰‫ ܝ‬൅ ‫ܡۺ‬

Figure 3.1: Block diagram of the LQG controller with a limiter

law u(t) = Kx(t) for the system of (3.1)-(3.2), subject to additive white Gaussian noises,
that minimizes the quadratic cost function

τ
 T 
J(u) = lim x (t)Qx(t) + uT (t)Ru(t) dt, (3.3)
τ →+∞
0

where Q and R are weighting matrices and commonly constructed as diagonal matri-
ces [43]. Entries of Q and R penalize the deviations of the corresponding state variables
from zero and variations of the controller output, respectively. Q and R determine the
trade-off between tracking performance and control effort. The optimal feedback gain,
K, is given by
K = R−1 BT S, (3.4)

where S is the solution of the Riccati equation

AT S + SA − (SB)R−1 (BT S) + Q = 0. (3.5)

LQ-optimal state feedback requires full state measurement. If all the states are not
measured, an observer, e.g., a Kalman filter [48], is required. The combination of the
LQR with a Kalman filter forms the LQG controller as shown in Fig. 3.1.

3.2.2 Model Predictive Controller [49]

The objective of an MPC is to hold the system output at a reference value by adjusting
the control signal. Fig. 3.2 represents the block diagram of the MPC, where, r, u, ,
ν, y and ȳ represent the reference, control, process noise, measurement noise, measured
output, and true value of the output signals, respectively.
Chapter 3. LSC for Small-Signal Stability Enhancement 34

‫ܚ‬ ૅ
‫ܝ‬
‫ܥܲܯ‬ +
݈ܲܽ݊‫ݐ‬
ࣕ + ‫ܡ‬
‫ܡ‬ത 

Figure 3.2: Block diagram of the MPC

An MPC combines a prediction strategy and a control strategy. The main differ-
ence between the MPC and other optimal control strategies, e.g., the LQG approach,
is that it adjusts the control signal to achieve the objectives while respecting the con-
trol constraints. Moreover, the MPC optimization horizon is always finite (versus LQG
optimization, which often has an infinite time horizon). The additional constraints are
typically upper or lower bounds of the control signals or their derivatives. This chapter
adopts discrete-time MPC since it is computationally less demanding as compared with
the continuous-time MPC.
The controller operates in two phases: estimation and optimization. In the for-
mer, the current true outputs (i.e., outputs without measurement noise), ȳ(k), and
states of the system, x(k), are estimated. In the latter, first, the future outputs,
y(k + 1), . . . , y(k + p), where p is the prediction horizon, are estimated as a function
of the control moves using the past and current measurements. Then, based on the val-
ues of the reference signals and constraints over the prediction horizon, M control moves,
u(k), u(k + 1), . . . , u(k + M), where M (1 <= M <= p) is the control horizon, are cal-
culated. p and M are basic tuning parameters for MPC. p is generally set long enough
to capture the steady-state effects of all calculated future control moves [47]. A larger M
improves the performance of the MPC at the expense of additional computations [47].
Finally, the controller move at the current sampling time u(k) is applied to the system.
At the next sampling time, new measurements are obtained and the control signal is
revised. This procedure is repeated at each sampling instant.
At each sampling instant, the MPC estimates the current states x(k) from the avail-
able measurements. If all the states are measured, the state estimation only considers
the effect of noise on the measurements. The discretized model of the system, (3.1)-(3.2),
is
x(k + 1) = Ax(k) + Bu(k) + (k), (3.6)

y(k) = Cx(k) + ν(k), (3.7)

where ν(k) and (k) are the measurement and process noise assumed to be independent
Chapter 3. LSC for Small-Signal Stability Enhancement 35

(of each other), white, and with normal probability distributions. The current outputs
and states of the system are estimated as

x̂(k|k) = x̂(k|k − 1) + L(y(k) − ŷ(k)), (3.8)

ŷ(k) = Cx̂(k|k − 1), (3.9)

where L is the Kalman gain [48]. The main difference between the LQG and MPC
optimization is that the MPC optimization is over a finite horizon and constrained. The
MPC action at time k is obtained by solving the optimization problem


p−1

ny
y
min wi+1,j (yj (k + i + 1|k) (3.10)
Δu(k|k),...,Δu(k+M −1|k)
i=0 j=1


nu
−rj (k + i + 1))2 + Δu
wi,j Δuj (k + i|k)2 ,
j=1

∀i=0,...,p−1, ∀j=1,...,nu
  
ujmin (i) ≤ uj (k + i|k) ≤ ujmax (i),
Δujmin (i) ≤ Δuj (k + i|k) ≤ Δujmax (i), (3.11)
uj (i + k) = uj (i + k − 1) + Δuj (i + k|k),
Δu(k + h|k) = 0, ∀h = M, . . . , p − 1. (3.12)
y Δu
where wi,j is the weight for output j, wi,j is the rate weight for control signal j at i
step ahead from the current step, rj (i) is the set-point at time step i, k is the current
step, ny is the number of outputs, Δuj (k + i|k) is the adjustment of control signal j
at time step k + i based on the measurements at time step k, and nu is the number of
control signals. uj,min , uj,max , Δuj,min and Δuj,max , are the lower and upper bounds on
the control signal and control signal variation, respectively. The control signal applied
to the plant is u(k) = u(k − 1) + Δu(k|k). Complete details of the MPC optimization
are provided in Appendix A.
For set-point tracking purposes, the cost function only consists of the weighted sum
of the deviation of the outputs from their reference values and a weighted sum of the
controller adjustments over the prediction horizon. The weights specify the trade-offs
in the controller design and reflect the importance of the corresponding variables to the
overall performance of the system. If a particular output weight is large, deviations of the
corresponding output dominate the cost function. Increasing the rate weights forces the
controller to make smaller adjustments and degrades set-point tracking. In the reported
Chapter 3. LSC for Small-Signal Stability Enhancement 36

studies, the constraints are assumed to be hard constraints and thus must not be violated.
A quadratic optimization solver, based on the KWIK algorithm [50], is used to solve the
optimization problem.
The communication delay Td associated with the measured signal ym can be consid-
ered in the system by adding one state equation to the plant model based on the time
delay Padé approximation [28, 51] as

dyd 2 Td
= (−yd + ym − ym ), (3.13)
dt Td 2

where yd is the delayed signal associated with ym .

While the LQR requires full state feedback, neither LQG nor MPC requires full
state feedback. These controllers are equipped with observers which provide the state
estimates. A brief description of the prediction process is provided in Appendix A. The
models used for online dynamic security assessment (DSA), [52–54], can be used for the
MPC and LQR design as well.
The study results reported in this chapter and also the performance comparison of
the LQG-based and MPC-based controllers are based on the assumption that the system
model is available. This model can be derived using the same methods applied in the
online DSA. Since the robustness of a generic controller for nonlinear systems, e.g., LQG
or MPC, cannot be guaranteed, further action are required to ensure the desired controller
performance in case of uncertainties in the system parameters, topology, operating point
and load conditions [46, 55–57].

3.3 HVDC LSC Based on Optimal Control Theory

In this section, to investigate the performance of the designed controllers which are used
to damp oscillations within a segment, two study systems (WSCC and IEEE 14-bus 1-
segment) with different characteristics in terms of the integration of the DC system in
the AC system are selected. The former has an HVDC link in parallel with an AC path,
and the latter has an HVDC link embedded in the AC system. Designing a controller
to damp the oscillations for the IEEE 14-bus 1-segment system compared to the WSCC
system is more challenging; since control of the HVDC link in the former configuration
affects more than one AC path. The details of the IEEE 14-bus 1-segment system are
provided in chapter 2.
In this study, the WSCC 9-bus system [58] is augmented with a classical HVDC
link as shown in Fig. 3.3. The WSCC system includes three T-G units and nine AC
Chapter 3. LSC for Small-Signal Stability Enhancement 37

18 KV 230 KV 230 KV 230 KV

Y=j0.0745 Y=j0.1045
Z=0.0085+j0.072 Z=0.0119+j0.1008

G2 G3
163 MW 2 7 9 3 85 MW
8
100 MW
Tap=18/230 I1 35 MVAR
Tap=13.8/230
Z=j0.0586
Z=j0.0625

Z=0.039+j0.17
Y=0

Z=0.032+j0.161
Y=0

Y=j0.153

Y=j0.179
PDC1

R1

5 6
230 KV 230 KV
Y=j0.088 Y=j0.079
Z=0.01+j0.085 Z=0.017+j0.092
90 MW
125 MW
30 MVAR
50 MVAR
4 230 KV
Tap=16.5/230
1 Z=j0.0576
Y=0
Slack 16.5 KV
Bus

G1

Figure 3.3: Schematic one-line diagram of the WSCC system

transmission lines. The rectifier and inverter stations are connected to buses 4 and 7,
respectively. The parameters of the WSCC system, including those of the T-G units
and their IEEE Type-1 exciter systems, are given in [58]. The HVDC link and control
parameters are given in [37].

To mitigate the inter-area oscillatory modes of the WSCC system and the IEEE
14-bus 1-segment system, the HVDC supplementary controller modulates the reference
(set-point) of the corresponding HVDC rectifier current controller, Fig. 3.4. The oper-
ating point of each study system, prior to each case study, is obtained from an AC-DC
power flow solution when the HVDC link transfers 50 MW (WSCC system) and 25 MW
(IEEE 14-bus 1-segment system). The LQG controller and the MPC-based controller,
for each study system, are designed based on the linearization of the system model about
the steady-state operating point. The linearized model of each study system is deter-
mined using the method discussed in chapter 2. The inter-area modes of the linearized
WSCC and IEEE 14-bus 1-segment systems, with frequencies less than 2 Hz and damp-
ing ratios less than 0.5, are presented in Tables 3.1 and 3.2, respectively. The states
with participation factors [1] greater than 0.06 in each mode and their corresponding
participation factors are also given in Tables 3.1 and 3.2. For a complex pair of eigenval-
ues κi = σi ± jωi , the damping ratio ζi and frequency fi are defined as √ −σ 2
i
2
ωi
and 2π ,
σi +ωi
respectively [1].
Chapter 3. LSC for Small-Signal Stability Enhancement 38

Measurement Ȟ
Modulation To the HVDC Main
Noise
Signal Controller
+
y +
MPC-Based +
Measurement Idref
Supplementary
Idord +
r Controller
Reference Signal

Figure 3.4: Block diagram of the MPC-based HVDC supplementary controller

Table 3.1: Eigenvalues of the linearized WSCC system

Inter-Area Frequency Damping Associated States
Mode (Hz) Ratio (Participation Factors)
1 1.385 0.04 δ2 (0.4), ω2 (0.3), ω1 (0.14), δ3 (0.07)

2 0.142 0.29 Ef d2 (0.24), Eq2 (0.17),

Eq1 (0.11), Ef d1 (0.1), Ψ1d2 (0.07)

3.3.1 LSC Based on LQG Method

To prevent ill-conditioning when solving the Riccati equation, it is necessary to adopt

a reduced-order system model to design the LQG controller. The Kalman decomposi-
tion [59] technique is used to perform model order reduction by removing the uncontrol-
lable and unobservable states. The reduced orders of the IEEE 14-bus 1-segment and
WSCC systems are 63 and 41, respectively.
Choosing the weighting matrices is a critical step in designing the LQG controller.
In this chapter, a systematic approach to set the weighting matrices is proposed. In this

Table 3.2: Eigenvalues of the linearized IEEE 14-bus 1-segment system

Inter-Area Frequency Damping Associated States
Mode (Hz) Ratio (Participation Factors)
1 1.26 0.05 ωC2 (0.2), δC2 (0.16), ωC3 (0.13),
δC3 (0.11), Ψ2qC2 (0.08),
ωC1 (0.08), δC1 (0.07)
2 1.37 0.26 δC3 (0.27), δC1 (0.23),
ωC3 (0.09), δ2 (0.07)
3 1.47 0.32 δC1 (0.24), δC2 (0.17),
ωC1 (0.14), ωC3 (0.11)
Chapter 3. LSC for Small-Signal Stability Enhancement 39

0.95

error(pu)
0.9

0.85
1 2 3
10 10 10
p

1.4
computation time(pu)

1.3

1.2

1.1

0.9
0 100 200 300 400 500 600 700 800 900
p

Figure 3.5: The control performance and computation time for different values of p
(WSCC system)

approach, to damp the mode of interest, the weights corresponding to this mode in the
Q matrix are set to non-zero values and the weights corresponding to the rest of the
modes are set to zero. To specify the Q matrix, the participation factors of the states
of the system in the mode of interest are calculated and those with participation factor
greater than 0.06 are given non-zero weights. The weight assigned to each of these states
corresponds to the participation factor of each of the states in the mode which should be
damped.
The inter-area oscillatory modes of interest in the WSCC and the IEEE 14-bus 1-
segment systems are at frequencies of 1.38 Hz and 1.26 Hz, respectively. These modes
have the least damping compared to the rest of the modes. The inputs to the controller
are determined through a modal observability analysis of the network [1]. For the WSCC
system, the voltage angles of buses 2, 4, 5, 6, 7, 8 and the active power flows of lines
1 and 9 exhibit the highest observability of the critical mode and are selected as the
input signals of the controller. For the IEEE 14-bus 1-segment system, voltage angles of
buses 2, 4, 5, 6, 7, 9, 14 and the active power flow of lines 1 and 3 are selected as the
controller inputs, based on the same method used for the WSCC system. To maintain the
robustness feature of the LQR in the LQG design, the weighting matrices of the Kalman
filter are tuned such that the loop transfer function of the system with the LQG controller
approaches the loop transfer function of the system with the LQR controller [60].
To limit the impact of modulation on the network flows, limit transients and avoid
Chapter 3. LSC for Small-Signal Stability Enhancement 40

0.2

0.18 No Control
LQG
0.16 MPC

0.14

0.12
Amplitude

0.1

0.08

0.06

0.04

0.02

0
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
f(Hz)

Figure 3.6: The single-sided amplitude spectrum of bus 4 voltage angle (IEEE 14-bus
1-segment system)

0.7
No Control
LQG
0.6 MPC

0.5
Amplitude

0.4

0.3

0.2

0.1

0
0.5 1 1.5 2
f(Hz)

Figure 3.7: The single-sided amplitude spectrum of bus 7 voltage angle (WSCC system)
Chapter 3. LSC for Small-Signal Stability Enhancement 41

(a)
LQG
1.7 MPC

P 1 (pu)
1.6

1.5
2.5 3 3.5 4 4.5 5
(b) t(s)

P 8 (pu) 0.3

0.25

0.2
2.5 3 3.5 4 4.5 5
t(s)
(c)
−8
−8.5
θ4 (deg)

−9
−9.5
2.5 3 3.5 4 4.5 5
t(s)
(d)
−10
θ (deg)

−12
8

−14
2.5 3 3.5 4 4.5 5

Figure 3.8: IEEE 14-bus 1-segment system dominant inter-area mode: (a) active power
flow of line 1, (b) active power flow of line 8, (c) bus 4 voltage angle and (d) bus 8 voltage
angle

potential voltage fluctuations in other areas, the DC power modulation must be bounded.
The bound on the DC power modulation limits the deviation of the operating point from
the optimal system operating point. A representative value of 10% was chosen to be the
permissible value in this system. This limit would ultimately be decided by the system
operator based on system voltage and security limits. Since the LQG controller does not
take into account the constraints on the inputs and outputs of the system, the control
signal may exceed the limits. Therefore, a limiter is imposed on the controller output.

3.3.2 LSC Based on MPC Method

The MPC-based controller output (modulation signal) is added to the HVDC link ref-
erence current signal and its inputs are the same as those of the LQG controller. If
oscillations in the measurement signals which display the most observability of the criti-
cal mode are damped out, then adequate damping for those modes in the entire system
Chapter 3. LSC for Small-Signal Stability Enhancement 42

is achieved. In this study, the measurement noise is modeled as a white noise with unity
amplitude.
t
To choose the design parameters, a performance index error = 0 Δp(t)T Δp is de-
fined in which Δp is the power deviation vector of all the AC transmission lines from
their steady-state values and t is the simulation time. This performance index repre-
sents the power deviation of all the AC transmission lines from their steady-state values
during the simulation time. The most important design parameter with respect to the
control performance and computation time is the prediction horizon p. Fig. 3.5 shows
the trade-off between the computation time and control performance for different values
obtained with p for the WSCC system (by increasing p the control performance improves
and the computation time increase). In Fig. 3.5, the computation time and the error at
different values of p are divided by those of p=10. To have a fair balance between the
computation time and the performance, for the WSCC system, p, control horizon M,
and control interval (sampling time) are selected to be 50, 5, and 0.02 s, respectively. In
Fig. 3.5, for p greater than 40, the error index becomes almost flat and for the minimum
error index, at p=50 the computation time is at its minimum. The same parameters for
the IEEE 14-bus 1-segment system are 500, 5, and 0.02 s, respectively.

3.4 Application of the LSC for Small-Signal Dynamic

Stability Enhancement
The proposed LSC is applied to two study systems, WSCC and IEEE 14-bus 1-segment.
The objectives of this section are to (i) demonstrate the MPC and LQG application
procedures, (ii) illustrate and verify the effectiveness of the LSC for mitigating inter-
area oscillations and (iii) highlight the superior performance of the MPC-based LSC
as compared with the widely used LQG-based LSC. The reported studies of this section
consider three scenarios: (i) no supplementary controller in service (no control), (ii) LQG
controller operational (LQG), and (iii) MPC operational (MPC). For each case study,
the system under consideration is subject to a 5-cycle, temporary, L-L-L-G fault at the
inverter AC bus (bus 4 of the IEEE 14-bus 1-segment system and bus 7 of the WSCC
system).
Figs. 3.6-3.7 show the frequency spectrum of the voltage angle of the inverter bus
(with respect to the voltage angle of bus 1) obtained from the discrete Fourier transform.
Fig. 3.6 shows that the 1.26 Hz low-frequency oscillatory mode (inter-area mode) of the
IEEE 14-bus 1-segment system is excited by the disturbance. Fig. 3.6 illustrates that the
MPC effectively reduces the amplitude of the dominant mode by a factor of 9 while the
Chapter 3. LSC for Small-Signal Stability Enhancement 43

Oscillation Amplitude
100
1.8
1.6 80

1.4

Frequency (Hz)
1.2 60

1
40
0.8
0.6
20
0.4
0.2
0
5 10 15
Time (s)

Figure 3.9: The spectrogram of bus 4 voltage angle deviations from the steady-state
value for the IEEE 14-bus 1-segment system (MPC in service)

LQG controller reduces the amplitude by a factor of 3.

Fig. 3.7 shows that the MPC reduces the amplitude of the 1.38 Hz dominant inter-
area mode of the WSCC system to one fourth of the amplitude observed with the LQG
controller in service. Both the MPC-based and LQG controller are designed to damp
the dominant oscillatory modes of the system. In doing so, these controllers may slightly
reduce the damping ratios of the other modes. However, the damping impact of the
controllers on the non-dominant modes is insignificant as compared to the impact on
the dominant mode. The time-domain behavior of the IEEE 14-bus 1-segment system,
subject to the fault at t=2 s, is shown in Fig. 3.8. The faster reduction in oscillations
shows the MPC is more effective in damping the dominant inter-area mode as compared
with the LQG controller.
Traditional analyses such as eigenvalue analysis are commonly used to evaluate the
performance of linear systems. Since the closed-loop system of this study is nonlinear,
the transfer function of the closed-loop system does not exist (the transfer function of
the linearized system changes at each operating point) and eigenvalue analysis is not
possible. To address this issue, spectrograms are used to show how all the frequency
components of the system change during the time. Spectrograms simultaneously capture
the time-domain and frequency-domain information of the system signals.
The spectrograms of the voltage angle deviations on bus 4 of the IEEE 14-bus 1-
segment system are depicted in Figs. 3.9-3.10. The spectrogram shows the amplitude
of the time-dependent Fourier transform of the bus 4 voltage angle deviations versus
Chapter 3. LSC for Small-Signal Stability Enhancement 44

Oscillation Amplitude
100
1.8
1.6 80

1.4

Frequency (Hz)
1.2 60

1
40
0.8
0.6
20
0.4
0.2
0
5 Time (s) 10 15

Figure 3.10: The spectrogram of bus 4 voltage angle deviations from the steady-state
value for the IEEE 14-bus 1-segment system (LQG controller in service)

time. The time-dependent Fourier transform is the discrete-time Fourier transform for
a sequence computed by a sliding time window of 1000 points and 50% overlap between
the segments. Figs. 3.9-3.10 show that the MPC-based controller damps the oscillations
faster than the LQG controller. Furthermore, with reference to the modes of interest
specified in Table 3.2, the amplitudes of the frequency components of the selected signal
when the MPC is in service is smaller compared to the case when the LQG controller is in
service, even at early times. For a fault of the same duration applied to every other bus
on the system, with the MPC-based supplementary controller in service, faster damping
is achieved as compared to when the LQG-based supplementary controller is in service.
The reason for the superior performance and fast damping of the MPC-based con-
troller as compared with those of the LQG controller is its ability to optimize the objective
function while respecting the system constraints. The control signals of the MPC-based
and LQG controllers are shown in Fig. 3.11. The L-L-L-G fault at the inverter station
requires a large control signal to achieve a short settling time. In the case of the LQG
controller, since the control signal must be bounded and this constraint is not considered
in the LQG controller optimization, the controller provides a large control signal which
is then constrained to its limits. Therefore, the controller performance deteriorates. If
the parameters are tuned such that the control signal remains between its bounds, the
damping effect diminishes.
In general, in the design of the LQG controller, the weighting matrices should be
selected to retain the control signal between its lower and upper bounds. If the control
Chapter 3. LSC for Small-Signal Stability Enhancement 45

0.1 MPC
LQG
0.08

0.06

0.04

0.02
u(pu) 0

−0.02

−0.04

−0.06

−0.08

−0.1
1 2 3 4 5 6 7 8 9
t(s)

Figure 3.11: Control signal for the cases with LQG and MPC after an L-L-L-G fault on
bus 4 (IEEE 14-bus 1-segment system)

signal exceeds its limits repeatedly, the actuator will be saturated and the controller will
not be effective. If the gains are modified by adjusting the control weighting matrix
such that the control signal remains between its bounds, the controller response time
increases. In order to keep the control signal bounded, a small gain should be selected
for the LQG controller and, since this gain does not change with time, the controller will
have less influence on the system response. However, in the case of MPC, the “gain” is
allowed to vary with time and, when feasible, a larger gain can be applied to damp out
the oscillations faster.
Fig. 3.12 highlights the effect of communication time delay of the remote signals on
the damping effect of the MPC-based supplementary controller. It is assumed that all
signals are delayed 200 ms [51]. Fig. 3.12 indicates that the MPC-based supplementary
controller performance remains satisfactory despite the introduction of delay.

3.5 Conclusions
In this chapter, a supplementary classical HVDC controller based on the model predictive
control (MPC) methodology and linear quadratic Gaussian (LQG) control is proposed,
designed and applied to two study systems, WSCC and IEEE 14-bus 1-segment, to
mitigate the dominant inter-area oscillatory modes. The small-signal dynamic modeling
approach is used to determine the critical inter-area modes of the study systems and a
Chapter 3. LSC for Small-Signal Stability Enhancement 46

(a)

0.1
No Delay
With Delay
0

P8(pu)
−0.1

−0.2

−0.3
1 1.5 2 2.5 3 3.5 4 4.5 5
t(s)
(b)

15

10
θ (deg)

5
8

−5
1 1.5 2 2.5 3 3.5 4 4.5 5
t(s)

Figure 3.12: WSCC system dominant inter-area mode: (a) active power flow of line 8
and (b) bus 8 voltage angle

systematic approach to design the MPC-based and LQG controllers is proposed. The
main feature of the MPC control is its ability to cope with hard constraints on inputs,
outputs, and states. Since these constraints are directly included in the optimization
problem formulation, future constraint violations are prevented. The MPC can address
actuator limitations and enables operation closer to the system constraints which results
in faster system response. Furthermore, while the LQG controller has a static gain,
the online optimization of the MPC permits the gain of the controller to vary with
time, resulting in faster damping. This chapter also reports the simulation results of the
designed MPC-based and LQG-based HVDC supplementary controllers for the mitigation
of inter-area oscillations. Simulation results show that both are beneficial to overall
system dynamic performance and lead to damped low-frequency oscillations under small-
signal disturbances; however, the MPC methodology provides faster recovery times after
faults. The proposed supplementary controller has satisfactory performance even if the
communication delay is accounted for.
Chapter 4

Mitigation of Oscillations by Control

of the Propagation of Oscillatory
Modes Using an HVDC
Supplementary Control Scheme

This chapter presents and investigates a new LCC-HVDC global supplementary control
(GSC) strategy for stabilizing and enhancing the dynamic performance of a large AC
system that is segmented by LCC-HVDC links. The GSC can stabilize the AC system
while either minimizing the propagation of oscillatory dynamics from one segment to
other segments (GSC1) or enabling their controlled transfer from a disturbed segment to
other segments (GSC2).
In a DC-segmented AC grid, the impact of the LCC-HVDC links on the propagation of
oscillatory dynamics initiated within a segment can be manifested through the following
mechanisms:

1. Partial “buffering” of the HVDC links through their main (constant current (CC)
and constant voltage (CV)) controllers [1], which inherently limit the transfer of
dynamics between the AC segments.

2. The use of HVDC supplementary controllers to selectively control the transfer of

oscillatory dynamics among the segments.

The main idea of this chapter is to deploy coordinated supplementary controls of

inter-segment HVDC links to (i) stabilize the system and (ii) achieve one of the following
objectives, based on the system requirements:

47
Chapter 4. Control of the Propagation of Oscillatory Modes 48

(ii-a) Confine the oscillatory dynamics within the initiating segment

(ii-b) Selectively permit the controlled transfer of oscillatory dynamics from the per-
turbed segment to other segments in order to enhance the overall system damping

The controllers used to achieve these two objectives are referred to throughout the remain-
der of the chapter as the global supplementary controllers GSC1 and GSC2, respectively.
To the best of our knowledge, neither scheme has been explored in the technical liter-
ature in the context of DC-segmented systems. Since the measurements (inputs to the
controllers) are received from all the segments, the proposed control schemes are referred
as “global” supplementary control schemes.
Because maintaining system stability is a goal for both of the GSCs, and the weighting
matrices of an LQG controller can be selected such that the closed-loop stability of the
(linearized) system is guaranteed [61], both GSC1 and GSC2 are implemented as LQG
controllers. Both GSC1 and GSC2 provide simultaneous modulation of the current order
and voltage reference of the corresponding rectifier and inverter stations.

4.1 GSC Based on Optimal Control Theory

Fig. 2.3 depicts the CC and CV modes of the main HVDC controller and the supplemen-
tary controllers. The supplementary controllers modulate the current order and voltage
reference. The HVDC main controller, in comparison to an AC link, provides a high de-
gree of buffering of the two AC sides. However, during a disturbance in the AC segments,
the DC power transfer does not remain absolutely constant and is subject to dynamics.
HVDC supplementary control can be used to either mitigate the dynamics or to facilitate
their transfer between AC segments in a controlled manner.
Modulation of the HVDC link power transfer by a supplementary controller has been
used to mitigate the oscillatory modes of AC systems with embedded HVDC links, e.g.,
in the WECC system [32]. The conventional strategy for HVDC supplementary control
is to modulate the rectifier current order and not use the inverter voltage reference
modulation in the supplementary control process. However, coordinated DC line power
flow modulation through both the rectifier and inverter stations can (i) enhance the
modulation effect and (ii) extend its application to cases where contribution from either
or both stations is required.
This chapter evaluates the performance of the HVDC supplementary controller to:
(1) stabilize the system and (2) minimize the propagation of oscillations from the per-
turbed segment to other segments or, in a controlled manner, deliberately transfer the
Chapter 4. Control of the Propagation of Oscillatory Modes 49

Global Supplementary Controller (GSC)

Segment 2
HVDC12
~= =~
Segment 1

~= =~

HVDC13 HVDC23

Segment 3
~=

~=

Figure 4.1: Structure of the global supplementary control (GSC)

oscillations, from the perturbed segment to other segments based on (i) rectifier cur-
rent order modulation, (ii) inverter voltage reference modulation, and (iii) simultaneous
rectifier current order and inverter voltage reference modulation. The inverter voltage
is usually not modulated to minimize the possibility of commutation failures. To avoid
commutation failure, the reported studies in this chapter limit the voltage reference mod-
ulation to 3% of its rated value. Provisions are made for the inverter constant extinction
angle control to override the voltage control whenever the minimum extinction angle is
reached.
Fig. 4.1 illustrates the general structure of the GSC. The coordination center receives
signals from all converter stations and converter buses as inputs and uses this informa-
tion to coordinate the supplementary control of all HVDC links. Two supplementary
controllers, which can be applied to the main controllers of the rectifier and/or inverter
stations, are considered in this chapter. The main difference between the two controllers
is in how perturbations in the inter-segment DC line flows are penalized relative to os-
cillations in the other system states. The first controller, GSC1, attempts to reduce
perturbations in the HVDC line flows by penalizing oscillations in the DC line states.
The second controller, GSC2, does not specifically target the HVDC line states; instead,
it focuses on limiting the variation in a more complete set of system states. Which con-
troller is more appropriate for use on a given system depends on the overall operating
Chapter 4. Control of the Propagation of Oscillatory Modes 50

DC2
19 20 19 20 19 20
17 17 17
13 13 13

12 11 18 16 12 11 18 16 16
12 11 18
9 15 9 15 15
2 9
2 2

10 8 14 10 8 14 8 14
10
1 1 1
7 7 7

5 4 6 5 4 6 5 4 6

3 3 3
DC1

DC3

Figure 4.2: Schematic one-line diagram of the fully-DC-segmented test system

goals. If limiting DC line flow variation is of paramount importance—e.g., to ensure that

segments remain isolated as much as possible during disturbances—then GSC1 is the
appropriate controller to use. On the other hand, GSC2 will tend to use the full range of
control capabilities—including intentional modulation of the inter-segment line flows—to
stabilize the entire system as quickly as possible. For systems where the fast damping of
all AC segments, rather than the isolation of all segments, is of paramount importance,
GSC2 will tend to be a better choice than GSC1.

4.2 Study Systems

To study the DC segmentation concept and validate the accuracy of the proposed control
schemes for the inter-area HVDC links, the IEEE 14-bus 3-segment system (fully-DC-
segmented and partially-DC-segmented configurations) is developed.
The fully-DC-segmented configuration, Fig. 4.2, is composed of three identical seg-
ments where each one is a copy of the IEEE 14-bus system. In the 3-segment system,
the three segments are connected by three LCC-HVDC links, i.e., HVDC1, HVDC2,
and HVDC3. The parameters of the HVDC system are given in [37]. The coefficients
of the cost functions for the T-G units of the 3-segment system are given in Table 4.1.
The difference between the cost function of the corresponding T-G units from the three
segments introduces an imbalance between the conventional optimal power flow (OPF)
solution of the segments to include HVDC transactions in the 3-segment study case.
For the sake of comparison, a partially-DC-segmented configuration is also investi-
gated, in which HVDC2 and HVDC3 of the fully-DC-segmented system are replaced by
AC lines that result in the same inter-segment power transfers under the equivalent sys-
Chapter 4. Control of the Propagation of Oscillatory Modes 51

Table 4.1: Cost function coefficients of the T-G units of the 3-segment system

Segment 1 Segment 2 Segment 3

i ci2 ci1 ci0 i ci2 ci1 ci0 i ci2 ci1 ci0

1 0.046 21.6 0 1 0.043 20 0 1 0.043 20 0

2 0.270 21.6 0 2 0.25 20 0 2 0.25 20 0

3 0.011 43.2 0 3 0.01 40 0 3 0.01 40 0

4 0.011 43.2 0 4 0.01 40 0 4 0.01 40 0

5 0.011 43.2 0 5 0.01 40 0 5 0.01 40 0

tem loading conditions (using a standard π transmission line model, this corresponds to
setting R = G = B = 0 pu and X = 0.2 pu).

4.3 Study Results

The reported studies: 1) compare the dynamics of the fully-DC-segmented and partially-
DC-segmented configurations, 2) evaluate the performances of GSC1 and GSC2 in the
fully-DC-segmented configuration, and 3) investigate the sensitivity of GSC1 and GSC2 in
the fully-DC-segmented configuration with respect to variations in the system operating
condition and disturbance location. For each case study, prior to a disturbance, the
system steady-state operating point is obtained from an AC-DC power flow analysis. A
steady-state solution of the fully-DC-segmented configuration is obtained based on the
following conditions:

1. HVDC1, HVDC2, and HVDC3 transfer 85%, 53%, and 35.1% of their rated powers.

2. Bus number 1 is the slack bus within each AC segment.

3. The active power and voltage regulation setpoints of all the generators (aside from
the slack bus generators) are as specified in the IEEE 14-bus system description.

In all of the reported case studies, the system oscillatory dynamics are excited by a
5-cycle, self-cleared, L-L-L-G fault.
Chapter 4. Control of the Propagation of Oscillatory Modes 52

(a)
0.2 Fully−dc−segmented System
Partially−dc−segmented System

P C3−seg1 (pu)
0.1

−0.1

−0.2
0 1 2 3 4 5 6
(b) t(s)
P C3−seg2 (pu) 0.04

0.02

−0.02

−0.04
0 1 2 3 4 5 6
(c) t(s)
0.05
P C3−seg3 (pu)

−0.05
0 1 2 3 4 5 6
t(s)

Figure 4.3: Active power deviations of T-G unit C3 in each segment, due to an L-L-L-G
fault on bus 10 of segment 1, for the partially- and fully-DC-segmented systems (with no
supplementary controller).

4.3.1 Dynamics of the Fully-DC-Segmented and Partially-DC-

Segmented Configurations

Fig. 4.3 shows the active power deviations (from the steady-state pre-fault value) of T-G
unit C3 in each of the three segments due to a fault. The fault occurs at bus 10 of
segment 1 and no HVDC supplementary controller is in service. Figs. 4.3 (b)-(c) show
that the impact of the segment 1 disturbance on the other two segments is significantly
reduced when the segments are only connected via HVDC links. However, the presence of
oscillatory dynamics in segments 2 and 3, even for the fully-DC-segmented case, illustrates
that the operation of the HVDC links, based on the primary (CC and CV) controllers,
does not provide complete isolation of the segments and segments 2 and 3 experience the
impact of the fault in segment 1.
Chapter 4. Control of the Propagation of Oscillatory Modes 53

GSC1 Disabled
GSC1 Enabled− Current Order Modulation (case 1)
GSC1 Enabled− Voltage Reference Modulation (case 2)
GSC1 Enabled− Current Order and Voltage Reference Modulation (case 3)
(a)

0.71

P 2−seg1 (pu)
0.7
0.69
0.68
0.67

2 2.5 3 3.5 4 4.5 5 5.5 6

(b) time(s)
0.672
P 2−seg2 (pu)

0.67

0.668

0.666
2 2.5 3 3.5 4 4.5 5 5.5 6
(c) time(s)
0.556

0.555
P 2−seg3 (pu)

0.554

0.553

0.552

2 2.5 3 3.5 4 4.5 5 5.5 6

time(s)

Figure 4.4: Active power deviations of AC transmission line 2 in each segment, due to
an L-L-L-G fault on bus 10 of segment 1, with and without GSC1 in service.

4.3.2 GSC1 in the Fully-DC-Segmented System

An often-cited benefit of DC segmentation [11] is that it can partially impede the trans-
fer of disturbances between segments. However, preventing the transfer of oscillatory
dynamics of the perturbed segment to the unperturbed segments can have detrimental
effects on the system stability. One potential goal of an HVDC supplementary controller
is achieving the desired isolation between segments while, at the same time, improving
the overall system damping.
GSC1 is intended to (i) ensure system stability and (ii) reduce power oscillatory dy-
namics in the DC ties, and, as a result, minimize the transfer of oscillatory dynamics
from the perturbed segment to the other segments. This is achieved based on one im-
portant feature of the LQG-based supplementary controller which enables the control of
Chapter 4. Control of the Propagation of Oscillatory Modes 54

targeted states by selecting relatively large weights for those states in the Q matrix of
(3.3). Based on the selected states, different performance objectives can be achieved.
To enable GSC1 to reduce the oscillatory dynamics in the inter-segment HVDC lines,
the weights corresponding to the DC line states (rectifier and inverter currents, DC line
midpoint voltages, and inverter voltages) are set to unity in the Q matrix of (3.3) and
the weights corresponding to the rest of the states are set to zero.
Fig. 4.4 shows the active power deviation of AC transmission line 2 within each
segment due to the fault on bus 10 of segment 1, with and without GSC1 in service.
When GSC1 is in service, the following three HVDC power modulation strategies are
examined:
Case 1: Only the rectifier current orders of the three HVDC links are mod-
ulated.
Case 2: Only the inverter voltage references of the three HVDC links are
modulated.
Case 3: Both the current orders and voltage references of the three HVDC
links are modulated.
Fig. 4.4 (a) indicates that the choice of modulation strategy has no noticeable impact
on segment 1 (where the fault is initiated); however, Figs. 4.4 (b)-(c) show that the
modulation strategy can impact the amplitude of oscillations in segments 2 and 3 and
consequently the ability of GSC1 to isolate those segments.

4.3.2.1 Case 1: GSC1 with Current Order Modulation

Figs. 4.4 (b)-(c) show that modulating the rectifier current order, under GSC1, signif-
icantly reduces the propagation of the oscillatory dynamics to segment 2 but does not
reduce the oscillatory dynamics in segment 3. The reason is that rectifier current mod-
ulation can compensate for the AC-side oscillations if the rectifier station of the HVDC
link is located in the perturbed segment. Thus, the oscillations excited in segment 1 are
prevented from being transferred to the other segment connected to HVDC1 (segment 2),
but the oscillations are not prevented in HVDC3 since its rectifier station is in segment
3, not segment 1.

4.3.2.2 Case 2: GSC1 with Voltage Reference Modulation

Fig. 4.4 (c) shows that voltage modulation of the three inverters prevents the propagation
of the oscillations to segment 3. The reason is that voltage modulation can compensate
for the AC-side oscillations only if the inverter station of the DC link is located in
the perturbed segment. Thus, the oscillatory dynamics excited in segment 1 are not
Chapter 4. Control of the Propagation of Oscillatory Modes 55

transferred to segment 3. Fig. 4.4 (b) shows the transfer of oscillations is not prevented
in HVDC1 since its inverter station is in segment 2, not segment 1.

4.3.2.3 Case 3: GSC1 with Current Order and Voltage Reference Modula-
tion

Figs. 4.4 (b)-(c) show that modulating both the rectifier current orders and inverter volt-
age references prevents the transfer of oscillations to both adjacent segments, regardless
of the rectifier or inverter location. Since the best performance is achieved when modu-
lation is applied to both the current order and voltage reference signals, the remainder
of the chapter focuses on this mode of operation for the supplementary controllers.

4.3.3 GSC2 in the Fully-DC-Segmented System

GSC1 improves the isolation between the AC segments, but this may not always be the
desired behavior. For example, to take advantage of other segments’ damping capabilities,
it may be advantageous to deliberately transfer oscillations from one segment to another.
Thus, an alternative global supplementary controller (GSC2) is introduced which allows
for the transfer of oscillatory dynamics between segments if this can help to improve the
overall system response to a disturbance.
To design GSC2, first, the states corresponding to the T-G units with participation
factors [1] greater than 0.01 for the significant oscillatory modes (i.e., modes with fre-
quencies less than 2 Hz and damping ratios less than 0.5) are identified. Second, these
states are filtered based on how quickly they return to their pre-disturbance steady state
values (this is to ensure that states with slower dynamics do not have an undue influence
on the feedback gain). The weights corresponding to the states which have not been
filtered out are set to unity in the Q matrix of (3.3), and the weights corresponding to
the rest of the states are set to zero.
Fig. 4.5 shows the active power deviation of AC transmission line 2 within each
segment due to the fault on bus 10 of segment 1, with GSC2 either in service or disabled.
Based on the above analysis for GSC1, only simultaneous modulation of current order
and voltage reference is considered. Fig. 4.5 shows that GSC2 does result in the transfer
of oscillatory dynamics from segment 1 to segments 2 and 3. This effectively increases the
damping in the perturbed segment 1 and decreases the damping in the other segments.
GSC2 can be particularly beneficial if at least one of the segments is able to provide
high damping for the oscillations initiated within the faulted segment (e.g., if one of the
unperturbed segments is substantially larger than the others). In such a system, the
Chapter 4. Control of the Propagation of Oscillatory Modes 56

GSC2 Disabled
GSC2 Enabled
(a)

0.71

(pu)
0.7

2−seg1
0.69
0.68

P
0.67

2 2.5 3 3.5 4 4.5 5 5.5 6

time(s)
(b)

0.68
(pu)

0.67
2−seg2

0.66
P

0.65
2 2.5 3 3.5 4 4.5 5 5.5 6
(c) time(s)

0.56
(pu)
2−seg3

0.55
P

0.54
2 2.5 3 3.5 4 4.5 5 5.5 6
time(s)

Figure 4.5: Active power deviations of AC transmission line 2, due to an L-L-L-G fault
on bus 10 of segment 1, with GSC2 either enabled or disabled

oscillations can be deliberately transferred from the faulted segment to the segment with
higher damping, if the goal is to enhance the overall system stability.
An example of such a system can be obtained by modifying the original fully-DC-
segmented system, as described in section 4.2, by replacing each T-G unit in segment
2 with two identical T-G units in parallel, each of which delivers half the power of the
original T-G unit. In this system (referred to below as the “high inertia” case), the
inertia of each equivalent T-G unit in segment 2 is twice that of the corresponding T-G
units in segments 1 and 3.
For the high inertia case, the dynamic response to a fault on bus 10 of segment 3
demonstrates how GSC2 can improve the overall system stability. Fig. 4.6 shows the
active power deviation of T-G unit G1 in all three segments due to a fault on bus 10 of
segment 3 of the high inertia case, with GSC1, GSC2 and no GSC enabled. Fig. 4.6 (c)
shows that when GSC2 is not in service, segment 3 becomes unstable. By distributing
Chapter 4. Control of the Propagation of Oscillatory Modes 57

(a) GSC Disabled

2.44 GSC1 Enabled
GSC2 Enabled

PG1−seg1(pu)
2.42

2.4

2.38

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

time(s)
(b)
2
PG1−seg2(pu)

1.9

1.8

1.7

1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6

time(s)
(c)

2.3
PG1−seg3(pu)

2.2
2.1
2
1.9
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
time(s)

Figure 4.6: Active power deviation of T-G unit G1, due to an L-L-L-G fault on bus 10
of segment 3 of the high inertia case, with GSC1, GSC2, and no GSC enabled

the disturbance to segments 1 and 2 (as shown in Figs. 4.6 (a)-(b)) and exploiting the
damping of the entire system—particularly the high inertia of segment 2—GSC2 is able
to prevent the system from becoming unstable.

4.3.4 Performance Indices and Sensitivity Analyses of the GSC1

and GSC2

For a controller to be usable in practice, its performance should not substantively de-
grade as the system conditions change. To evaluate the sensitivity of GSC1 and GSC2
performance to system changes, the dynamic response of the fully-DC-segmented system
with respect to changes in the system operating point and fault location are investigated.
Chapter 4. Control of the Propagation of Oscillatory Modes 58

GSC Disabled
GSC1 Enabled
GSC2 Enabled
(a) (d) (g)
0.8 0.05 0.025

0.6 0.04 0.02

0.03 0.015
J seg1

J seg1

seg1
0.4
0.02

J
0.01
0.2
0.01 0.005
0 0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number (Segment 1 Faults) Fault Number (Segment 2 Faults) Fault Number (Segment 3 Faults)
(b) (e) (h)
0.1 0.4 0.04
0.08
0.3 0.03
0.06
seg2

J seg2

seg2
0.2 0.02
J

0.04

J
0.02 0.1 0.01

0 0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number (Segment 1 Faults) Fault Number (Segment 2 Faults) Fault Number (Segment 3 Faults)
(c) (f) (i)
0.08 0.04 0.4

0.06 0.03 0.3

seg3

J seg3

0.04 J seg3
0.02 0.2
J

0.02 0.01 0.1

0 0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number (Segment 1 Faults) Fault Number (Segment 2 Faults) Fault Number (Segment 3 Faults)
(j)
0.25

0.2

0.15
J total

0.1

0.05

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number

Figure 4.7: Performance of GSC1 and GSC2 as the fault location is varied

This evaluation is based on two performance indices:

kseg,i
 
 t2
(yl (t)−ylss )2 dt
Jseg,i 
l=1 t1
kseg,i
and (4.1)

nseg
kseg,s ×Jseg,s
Jtotal  s=1

nseg , (4.2)
kseg,s
s=1
Chapter 4. Control of the Propagation of Oscillatory Modes 59

where kseg,i is the number of AC lines in segment i; yl and ylss are, respectively, the
per-unit active instantaneous power flow and steady-state power flow of line l; t1 is one
second after the fault inception time instant; and t2 is the time when the least damped
oscillatory power signal reaches 1% of its post-disturbance steady-state value. In (4.1),
yl (t) − ylss is the deviation of each line power flow from its steady-state value and the
integral term is a measure of the energy associated with the oscillations within the line
for t ∈ [t1 , t2 ]. This energy is summed over all lines and averaged for each segment (Jseg,i )
or the entire system (Jtotal ). The net energy associated with the power oscillations of all
the lines (Jtotal ) or the lines within a particular segment (Jseg,i ) provide a single value
that captures how fast the oscillations are damped out in the entire system or within
a particular segment. Also, note that this definition is analogous to the state deviation
error function used in the LQR design (3.3).

4.3.4.1 Sensitivity to the Fault Location

To evaluate the sensitivity of the GSC to the fault location, the system is subjected to
faults 1 to 5, 6 to 10, and 11 to 15, which correspond to faults at buses 10 to 14 of segment
1, 2 and 3, respectively. Fig. 4.7 shows the changes in Jtotal and Jsegi corresponding to
various fault locations for both GSC schemes.
Figs. 4.7 (a)-(c) show that under GSC1 for the faults initiated in segment 1 (faults
numbered 1-5), Jseg1 increases while Jseg2 and Jseg3 decrease. This indicates that GSC1
effectively reduces the transfer of the oscillations to the neighboring segments. However,
in achieving the improved isolation, the overall system damping does not noticeably
improve as indicated by the increase in Jtotal with GSC1 enabled (Fig. 4.7 (j)). For the
same fault, GSC2 reduces Jtotal , as shown in Fig. 4.7 (j). The corresponding increases
in Jseg,2 and Jseg,3 with GSC2 enabled indicate that this improved damping is obtained
by distributing the disturbance to the other segments of the system. Faults initiated in
segments 2 and 3 follow the same trend as the faults initiated in segment 1 (Figs. 4.7
(d)-(f) and Figs. 4.7 (g)-(i), respectively).
Fig. 4.8 shows the variations of Jsegi and Jtotal as functions of the fault location for
the high inertia case. For the unstable case studies, t2 of (4.1)-(4.2) is assumed to be
the same as the one for the stable case studies. Figs. 4.8 (a), (e) and (i) show that the
damping ratio of segment 2 is more than those of segments 1 and 3, since the amplitude
of the oscillations caused by a fault in segment 2 is less than those of segments 1 and 3.
For faults in segments 1 and 2, Fig. 4.8 shows similar results as those of Fig. 4.7—
GSC1 reduces the oscillations transferred to the unperturbed segments, and GSC2 in-
creases the overall system damping by transferring oscillations to the unperturbed seg-
Chapter 4. Control of the Propagation of Oscillatory Modes 60

GSC Disabled
GSC1 Enabled
GSC2 Enabled
(a) (d) (g)
0.8 0.04 0.2

0.6 0.03 0.15

seg1

J seg1

J seg1
0.4 0.02 0.1
J

0.2 0.01 0.05

0 0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number (Segment 1 Faults) Fault Number (Segment 2 Faults) Fault Number (Segment 3 Faults)
(b) (e) (h)
0.2 0.4 0.2

0.15 0.3 0.15

J seg2

seg2

J seg2
0.1 0.2 0.1
J

0.05 0.1 0.05

0 0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number (Segment 1 Faults) Fault Number (Segment 2 Faults) Fault Number (Segment 3 Faults)

(c) (f) (i)

0.08 0.2 8

0.06 0.15 6
J seg3

J seg3

seg3

0.04 0.1 4
J

0.02 0.05 2

0 0 0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number (Segment 1 Faults) Fault Number (Segment 2 Faults) Fault Number (Segment 3 Faults)

(j)
2.5

1.5
J total

0.5

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fault Number

Figure 4.8: Performance of GSC1 and GSC2 as the fault location is varied (high inertia
case)
Chapter 4. Control of the Propagation of Oscillatory Modes 61

(a)
GSC1 Enabled
GSC2 Enabled

(pu)
0.86

HVDC1
0.84

P
0.82
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
time (s)
(b)
0.7

0.6
P HVDC2 (pu)

0.5

0.4

0.3
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
time (s)
(c)
0.5
P HVDC3 (pu)

0.4

0.3

0.2
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
time (s)

Figure 4.9: Active power transmitted on the HVDC lines due to a fault on bus 10 of
segment 3 in the high inertia case, with GSC1 or GSC2 enabled

ments. However, for faults in segment 3, the lack of a supplementary controller results
in system instability; this is indicated by the extremely high “GSC Disabled” bars in
Figs. 4.8 (h)-(i).
On the other hand, Fig. 4.6 (c) shows that GSC1 and GSC2 both stabilize the system,
as designed. Under GSC1, the propagation of oscillatory dynamics to segments 1 and 2
are not fully prevented, since this would result in system instability; instead, deliberate
transfer of oscillations occurs to maintain system stability. Under GSC2, oscillations in
the system (specifically segment 3 oscillations) are mitigated by transferring the oscil-
lations to the other two segments; as a result, the smallest values of Jtotal are obtained
with this operating mode.
To better understand the performance of the controllers for a fault in segment 3,
Fig. 4.9 shows the active power transferred through the HVDC lines for a fault on bus 10
of segment 3. This figure shows that, because the system would otherwise be unstable, the
first objective of the controllers–to maintain system stability–results in some oscillations
being transferred to segments 1 and 2 from segment 3. However, these graphs also show
Chapter 4. Control of the Propagation of Oscillatory Modes 62

GSC Disabled
GSC1 Enabled
(a) GSC2 Enabled

0.5

Jseg1
0
0.7 0.75 0.8 0.85 0.9 0.95 1
PDC1(pu)

(b)

0.1
Jseg2

0.05

0
0.7 0.75 0.8 0.85 0.9 0.95 1
PDC1(pu)

(c)

0.05
Jseg3

0
0.7 0.75 0.8 0.85 0.9 0.95 1
PDC1(pu)
(d)

0.2
Jtotal

0.1

0
0.7 0.75 0.8 0.85 0.9 0.95 1
PDC1(pu)

Figure 4.10: Dynamic performance of GSC1 and GSC2 as the HVDC1 operating point
is varied from 70% to 100% of its rated value, in steps of 5%.

that, as compared to GSC2, GSC1 does favor control actions that minimize the transfer
of oscillations. This behavior is particularly evident in Fig. 4.9 (a), which shows much
lower oscillations in HVDC1 when GSC1 is used instead of GSC2.

4.3.4.2 Sensitivity to the Operating Point

To evaluate the sensitivity of GSC1 and GSC2 to changes in the system operating point,
the active power flow of HVDC1 is changed from 70% to 100% of the rated value in steps
of 5%. Fig. 4.10 shows the sensitivity indices Jseg,i and Jtotal for each operating point
with the oscillations initiated by an L-L-L-G fault on bus 10 of segment 1. Figs. 4.10
(a)-(c) reveal that under GSC1, for various operating points, Jseg,1 increases and Jseg2
Chapter 4. Control of the Propagation of Oscillatory Modes 63

and Jseg3 decrease, which indicates that isolation among segments improves. Fig. 4.10
(d) shows that, under GSC1, the transfer of oscillatory dynamics to other segments is
prevented; as a result, Jtotal does not improve. Fig. 4.10 (d) also shows that under GSC2,
for various operating points, Jtotal increases, which indicates that GSC2 is able to improve
the overall system damping by the controlled transfer of oscillatory dynamics. Therefore,
the desired performance of GSC1 and GSC2 is maintained under the specified range of
operating points.

4.4 Conclusions
This chapter presents two HVDC global supplementary control strategies for a DC-
segmented AC system—GSC1 and GSC2—designed based on the LQG control method.
The controllers are designed to (i) improve the system stability and either
(ii-a) minimize the transfer of oscillations between segments (GSC1) or
(ii-b) allow for the controlled transfer of oscillations to further reduce system-wide oscilla-
tions (GSC2).
Simulation results obtained by applying these controllers to a 3-segment AC system show
that: (i) DC segmentation can effectively minimize the inter-segment transfer of distur-
bances due to a fault, (ii) both GSC1 and GSC2 are able to initiate stabilizing action,
(iii) as compared with GSC1 or the lack of any supplementary controller, GSC2 improves
the overall system damping by spreading the disturbance to adjacent segments in a con-
trolled manner, (iv) as compared to GSC2 and the lack of any supplementary controller,
GSC1 effectively isolates the segments from one another during a disturbance, and (v)
the performances of GSC1 and GSC2 meet the desired objectives under various operating
conditions and fault locations.
Chapter 5

Dynamic Stability Enhancement of a

DC-Segmented AC Power System
via HVDC Operating-Point
Adjustment

5.1 Introduction
A number of reported studies discuss various operating point tuning strategies to increase
the stability margin and mitigate low frequency oscillations of the power system [62–70].
The studies are based on (i) voltage control of synchronous machines [62], [63], (ii) reactive
power control [63–66], (iii) active power control [63, 67, 68, 71], and (iv) imposition of
stability constraints associated with the Hopf bifurcation to the conventional optimal
power flow (OPF) problem [69, 70]. None of the reported strategies use the HVDC set-
points adjustment to achieve stability improvements.
Methods relying on reactive power control require the presence of controllable reactive
power loads or FACTS controllers, which may not be available. Methods based on active
power control (generation redispatch) impose challenging issues for real-time applications,
since most generators are not on automatic generation control and can take up to several
minutes to respond to redispatch instructions [72]. Furthermore, the reported methods
of [63, 67, 68, 71] require both positive and negative sensitivities, which are not always
available, to ensure active and reactive power balance is maintained.
The higher degree of controllability of the HVDC link power flow, as compared with
that of an AC line, can be used to adjust the operating point of a DC-segmented AC

64
Chapter 5. HVDC Operating-Point Adjustment 65

system to move the real part of the critical eigenvalues (corresponding to the oscillatory
modes with the least damping ratio) further away from the imaginary axis and thus
increase the stability margin and damping ratios of the oscillatory modes. The central
idea of this chapter is to determine the optimum set-point value of the HVDC links such
that a sufficiently large margin to the Hopf point and a high damping ratio is ensured.
Hopf bifurcation theory is used to derive a strategy to control the power flow of the
HVDC links to change the system operating point.
Changing the DC active power set-points indirectly results in the redispatch of the
generators’ active and reactive power outputs based on their governor and exciter con-
trollers, respectively. This is in contrast to prior methods [63,67–71], which rely on direct
control of the generators. The proposed approach allows for a more rapid response to
real-time events such as line outages, since the generators automatically redispatch based
on their active and reactive power controllers. The reported studies in this chapter show
that the optimal HVDC set-points can be used to (i) temporarily restabilize the system
subjected to a fault or (ii) preemptively increase the system robustness to faults and
changes in the loading factor.

5.2 Hopf Sensitivity Calculations [71, 73, 74]

The power system model for the envisioned studies is of the form
   
ẋ f(x, y, λ)
= = F(z, λ), (5.1)
0 g(x, y, λ)

where x ∈ nx , y ∈ ny , and λ ∈ m are the dynamic state variable, algebraic variable
and parameter vectors, respectively. ẋ = f(x, y, λ) and 0 = g(x, y, λ) (f : nx +ny +m →
nx and g : nx +ny +m → ny ) represent the differential and algebraic equations, re-
spectively. In (5.1), F = [f T , gT ]T , F : nz +m → nz , nz = nx + ny , z  [xT , yT ]T and
F(z0 , λ0 ) = 0 defines an equilibrium (z0 , λ0 ) of the system.
A bifurcation occurs when a small change in a parameter (the bifurcation parameter)
causes a sudden change in the behaviour of the system. Based on the system dynamic
manifolds and equilibrium points, bifurcations can be local or global. This chapter focuses
on the study of local bifurcations, in particular Hopf bifurcations, which can be identified
by the eigenvalues of the linearized system DAEs. In a Hopf bifurcation, a complex pair
of eigenvalues crosses the imaginary axis of the complex plane as the system parameters
gradually change.
Chapter 5. HVDC Operating-Point Adjustment 66

The system parameters vector λ = [μ, p] represents the non-controllable parameter μ,

e.g., the system loading, and the set of controllable (design) parameters p, e.g., HVDC
set-points. When a non-controllable parameter increases in a direction of stress from
μ0 with the controllable parameters fixed at p0 , the system reaches a Hopf bifurcation
point for the non-controllable parameter value μ∗1 when the real part of a single pair
of complex eigenvalues becomes zero (beyond this point, the system is no longer small-
signal stable). The stability margin, defined as M  μ∗ − μ0 , can often be increased by
controlling p using the first order sensitivity of M with respect to p. The sensitivity of
M also identifies those parameters that strongly affect the stability margin.

The Hopf bifurcation hypersurface, ΣHopf , consists of the set λ∗ for which (5.1) has a
Hopf bifurcation at (z∗0 , λ∗ ) (i.e., where F∗z , the Jacobian of F with respect to z evaluated
at the Hopf point, has a single pair of complex eigenvalues with no real component and
all other eigenvalues have strictly negative real parts). To evaluate Fz at the Hopf point,
first the power flow equations are solved. Second, based on the power flow solution, the
equilibrium point of the dynamic model at the Hopf point, z∗0 , is determined and then
Fz is evaluated at z∗0 .

Since F∗z is invertible [73], the implicit function theorem implies that there is a function
u : m → nz in the neighborhood of λ∗ which maps system parameters to the position
of the equilibrium and its nz × m Jacobian uλ is obtained from the solution of

Fz uλ = −Fλ , (5.2)

where Fλ is the Jacobian of F with respect to λ. The sensitivity of M with respect to

p is a scaled projection of a normal vector to ΣHopf . The normal vector to ΣHopf can
be determined by the sensitivities of the real parts of the eigenvalues with respect to the
parameters by
N = Re{d (F∗zz u∗λ + F∗zλ ) e}, (5.3)

where Fzz , an nz × nz × nz tensor, is the Hessian of F with respect to z and Fzλ ,

an nz × nz × m tensor, is the Hessian of F with respect to z and λ. d and e are
the normalized extended right and left eigenvectors of Fz corresponding to jω ∗ such that
de = 1 and |d| = 1. d and e are determined by solving the generalized eigenproblem [74].
Furthermore, if
N = [nμ , np1 , np2 , ..., np(m−1) ], (5.4)

1
All the parameters and variables at a Hopf point are identified by the superscript ∗.
Chapter 5. HVDC Operating-Point Adjustment 67

then
Mp|p0 = −(nμ )−1 [np1 , np2 , ..., np(m−1) ], (5.5)

where Mp|p0 is the sensitivity of the stability margin with respect to the controllable
parameters. An example illustrating the procedure for calculating the sensitivity of the
stability margin with respect to the design parameters is given in Appendix C.

5.3 Optimization Formulation

The objective of this study is to control the active power flows of the HVDC links of the
study systems of Fig. 5.2 and Fig. 5.3 to increase the stability margin subject to system
constraints. Thus, the active power set-points of the HVDC links are considered as the
controllable parameters p. Since Mp|p0 is a scaled projection of N, movement away from
the Hopf surface implies an increase in the stability margin and damping ratio of the
oscillatory modes. In the optimization process (see below), the damping ratio is easier
to calculate compared to the stability margin and increasing the damping ratio will tend
to increase the stability margin. Thus, to improve the stability margin, the objective is
to maximize the damping ratio of the least damped mode.
The optimum operating point of the HVDC system, which results in the highest
damping ratio of the oscillatory dynamics, is obtained from the following optimization
problem:
max ζmin (x, y, λ), (5.6)
pDC

s.t. g(x, y, λ) = 0, (5.7)

f(x, y, λ) = 0, (5.8)
hmin ≤ h(x, y, λ) ≤ hmax , (5.9)

where
ζmin = min ζj , (5.10)
j∈{1,...nmode}

λ = [μ, pDC ], (5.11)

and pDC and nmode are the vector of DC link active powers and the number of oscillatory
modes, respectively. The objective is to determine the optimum active power set-points
of the DC links to obtain the maximum damping ratio of the least damped oscillatory
modes. The equality constraints (5.7)-(5.8) ensure the system is in steady-state. The
inequality constraints (5.9) consist of the limits on the DC link power injections.
Chapter 5. HVDC Operating-Point Adjustment 68

A flowchart of the optimization process is shown in Fig. 5.1. In this process, the
initial operating point is assumed to be the solution of the conventional, cost-minimizing
OPF problem. At this operating point, the selected non-controllable parameter (e.g., a
load scaling factor) is varied until the system encounters a Hopf bifurcation, where the
0
damping ratio of the least damped mode, ζmin , is zero. Then, at the Hopf point, the
sensitivities of the real parts of the critical eigenvalues with respect to the active power
flow set-points of the HVDC links are determined. Finally, a line search is carried out to
identify the optimum controllable parameters:

(0)
P̃DCi (αopt ) = PDCi + αopt mPDCi
⎧
⎪
⎪ min min
⎪
⎪PDCi if P̃DCi < PDCi
⎨
opt
PDCi min
= P̃DCi (αopt ) if PDCi ≤ P̃DCi (αopt ) ≤ PDCi
max (5.12)
⎪
⎪
⎪
⎪
⎩P max max
if P̃DCi (αopt ) > PDCi
DCi

opt (0)
where mPDCi , αopt , PDCi and PDCi are the sensitivity of the stability margin with respect
to PDCi , optimum step size, and the optimum and initial flows of the ith HVDC link,
respectively. For the line search, first, the maximum and minimum values of the step
size, αmax and αmin , are determined based on mPDCi and the limits of the active power
set-points of the DC links, i.e., pDC max and pDC min . Next, K equally-spaced points in
the [αmin αmax ] interval are tested. During the line search process the governor-based
power flow (GBPF) [75], in which the generators’ output levels depend on the droop
characteristics, is solved after each update to the DC set-points. In the GBPF problem,
(i) the net DC load of each segment, defined as the difference between the DC powers
injected to and absorbed from the neighboring segments, is determined, (ii) the change
in the net DC load from the initial solution is calculated, and (iii) in each segment,
the change in the output of each T-G unit is determined by its droop characteristic.
The system model is linearized at the operating point obtained from the GBPF solution
and then used to calculate the damping ratios of the oscillatory modes and evaluate the
objective function (5.6).

5.4 Study Systems

To study the HVDC control operating point tuning scheme the IEEE 14-bus 3-segment
system, Fig. 5.3, and 2-segment system, Fig. 5.2, are used. The 3-segment system is
a more general case study compared to the 2-segment system; however, for illustrative
Chapter 5. HVDC Operating-Point Adjustment 69

Initialize
(0)
pDC = pbase
DC
k=1

? 
Begin
Optimization
 
?
Find the Hopf point
0
ζmin =0

?
Calculate mPDCi
i ∈ {1, ...ndc }
, αmin and αmax

-
?
(k−1)αmax +(K−k)αmin
α(k) = K−1
(k) (0)
pDC = pDC + α(k) MPDC

?
Solve the GBPF problem
Calculate the new equilibrium point
Linearize around the new operating point

?
(k)
Calculate ζmin

?
 X
XXX
 XX
X Is ζ
X (k) (k−1) X

XXX min < ζmin ?
X   
X
N
Y
?
αopt = α(k)
(opt) (k)
pDC = pDC

 ?XXX
 XXX
k =k+1  X
XXX Is k ≤ K? 
XXX 
X
6 Y
N
? 
End
 

Figure 5.1: Optimization flowchart

purposes, the 2-segment case study is also considered.

The 2-segment system is composed of two identical segments where each one is a
Chapter 5. HVDC Operating-Point Adjustment 70

Table 5.1: Cost function coefficients of the T-G units of the 2-segment system

Segment 1 Segment 2

i ci2 ci1 ci0 i ci2 ci1 ci0

1 0.0645 30 0 1 0.043 20 0

2 0.375 30 0 2 0.25 20 0

3 0.015 60 0 3 0.01 40 0

4 0.015 60 0 4 0.01 40 0

5 0.015 60 0 5 0.01 40 0

DC2
19 20 19 20

17 17
13 13

12 11 18 16 12 11 18 16
9 15 9 15
2

10 8 14 10 8 14
1 1
7 7
2
5 4 6 5 4 6

3 3
DC1

Figure 5.2: Schematic one-line diagram of the 2-segment system

copy of the IEEE 14-bus system. The segments are connected by two LCC HVDC links,
i.e., HVDC1 and HVDC2. The rectifier (R) and inverter (I) stations of the HVDC links
are also identified on Fig. 5.2. The cost coefficients for the T-G units of the 2-segment
system are given in Table 5.1.

5.5 Study Results

In this section, two case studies in the MATLAB/Simulink environment are reported.
The stability margin and eigenvalue plot of each test system operating at the optimum
point are compared with those of the system at the operating point obtained from the
conventional, cost-minimizing OPF solution. Hereafter, the conventional OPF solution,
Chapter 5. HVDC Operating-Point Adjustment 71

DC2
19 20 19 20 19 20
17 17 17
13 13 13

12 11 18 16 12 11 18 16 16
12 11 18
9 15 9 15 15
2 9
2 2

10 8 14 10 8 14 8 14
10
1 1 1
7 7 7

5 4 6 5 4 6 5 4 6

3 3 3
DC1

DC3

Figure 5.3: Schematic one-line diagram of the fully-DC-segmented test system

without any stability constraints, is referred to as the “base OPF” solution.

5.5.1 Case Study on the 2-Segment System

It is assumed that the HVDC lines of Fig. 5.2 are lossless, i.e., Prec = −Pinv ; therefore,
there are only two controllable parameters, PDC1 and PDC2 (within the range [-1,1] pu).
All loads in the system are varied with respect to the base values pL0 and qL0 as

pL = pL0 × (1 + β), (5.13)

qL = qL0 × (1 + β), (5.14)

where β is the loading factor. The initial active power set-points of the HVDC links are
PDC1 = −0.90 pu and PDC2 = −0.74 pu. At the initial operating point all the system
eigenvalues have negative real parts and the system is stable. When β is increased from 0
to 1.0 for the loads in segment 2 (the system operating point is obtained from the GBPF
solution), the system encounters a Hopf bifurcation at β ∗ = 1.0.
To stabilize the system, first, the sensitivity vector of the real parts of the critical
eigenvalues to PDC1 and PDC2 at the Hopf point is calculated. The sensitivity vector,
[mPDC1 , mPDC2 ] = [11.11, 11.25], shows that to increase the damping ratio of the os-
cillatory modes, the flows of HVDC1 and HVDC2 should be increased in the positive
direction. Second, knowing the sensitivity vector, a line search is used to calculate the
optimum changes to PDC1 and PDC2 . In this case study, PoptDC = [1, 1] pu.
To demonstrate the effectiveness of the proposed method, the real part of the closest
oscillatory eigenvalue to the imaginary axis, as a function of PDC1 and PDC2 , is shown
in Fig. 5.4. Fig. 5.4 shows that the proposed method is able to determine the HVDC
Chapter 5. HVDC Operating-Point Adjustment 72

0.2 0

−0.1
0
−0.2
−0.2
−0.3
−0.4 −0.4

σ(1/s)
−0.6 −0.5

−0.6
−0.8
−0.7
−1
−0.8
−1.2 X: 1
−1 −0.9
Y: 1
Z: −1.098
−1
0 −1
P DC1 (pu) −0.5
0
1 0.5
1 P DC2 (pu)

Figure 5.4: Real part of the closest eigenvalue to the imaginary axis versus the flows of
HVDC links

set-points that achieve the highest damping of the system. It should be noted that
since the frequency of the critical mode does not noticeably change with the operating
point, the real part of the eigenvalue associated with the critical mode is an appropriate
representative of the damping ratio.

Fig. 5.5 shows the system eigenvalues associated with the oscillatory modes with
frequencies less than 2 Hz and damping ratios less than 0.5 when β ∗ = 1.0. When
the system operates at the base OPF solution, the system is at the Hopf bifurcation.
However, after the HVDC set-points are adjusted, the real part of the closest oscillatory
eigenvalue to the imaginary axis is −1.1.

The trade-off between the increased generation cost and increased damping ratio is
illustrated in Fig. 5.6. Fig. 5.6 (a) shows the real part of the closest oscillatory eigenvalue
to the imaginary axis with respect to α. Fig. 5.6 (a) illustrates that the solution obtained
from the optimization problem (moving in the direction of the sensitivity vector) results
in an increase in the damping ratio of the oscillatory modes (αopt = 0.17). However,
this increase in the damping is accompanied by an increase in the generation cost due to
changes in the generators’ output levels, as shown in Fig. 5.6 (b) (note: the generation
costs in this figure are normalized with respect to the cost obtained from the base OPF
solution).
Chapter 5. HVDC Operating-Point Adjustment 73

15

10

ω(rad/s)
Base OPF After Load Variation
0 After HVDC Set−Point Variation

−5

−10

−15
−6 −5 −4 −3 −2 −1 0 1
σ(1/s)

Figure 5.5: Eigenvalues plot of the system corresponding to the two operating points
obtained from the base OPF and optimization problem
(a)
0.5

0
σ(1/s)

−0.5

−1

−1.5
0 0.05 0.1 0.15 0.2
α
(b)
1.6

1.4
cost(pu)

1.2

0.8
0 0.05 0.1 0.15 0.2
α

Figure 5.6: Generation cost and σ versus the optimization step size

5.5.2 Case Studies on the 3-Segment System

The controllable parameters in this case study are PDC1 , PDC2 and PDC3 . The system
is forced to Hopf bifurcation due to either load variations (case 1) or line impedance
changes (case 2). At the Hopf bifurcation point, the sensitivity of the real parts of the
critical eigenvalues with respect to the controllable parameters is calculated and, using
Chapter 5. HVDC Operating-Point Adjustment 74

−0.1

−0.2

−0.3

σ(1/s)
−0.4

−0.5

−0.6

−0.7
0 0.005 0.01 0.015 0.02 0.025 0.03
α

Figure 5.7: σ versus the optimization step size

the sensitivity information, the HVDC flows are adjusted to move the eigenvalues away
from the imaginary axis.

5.5.2.1 Case 1: Load Variations

Using generator and HVDC set-points determined from the standard OPF, increas-
ing β for the loads in segment 3 from 0 to 1.9 leads the system to a Hopf bifurca-
(0)
tion. At the Hopf point, PDC = [−0.72, −1, −0.80] pu and the sensitivities of the
real parts of the critical eigenvalues with respect to the controllable parameters are
[mPDC1 , mPDC2 , mPDC3 ] = [0, 72.31, −70.89]. Based on the optimum direction, the op-
DC = [−0.72, 0.45, −1] pu. If the
timum step size is calculated (αopt = 0.02) and Popt
HVDC links transfer the optimum flows, the real part of the closest oscillatory eigen-
value to the imaginary axis is less than −0.68. Fig. 5.7 shows the real part of the closest
oscillatory eigenvalues to the imaginary axis with respect to α, and illustrates that the
solution obtained from the optimization problem results in an increase in the damping
ratios of the oscillatory modes.
Fig. 5.8 shows the system eigenvalues associated with the oscillatory modes with
frequencies less than 2 Hz and damping ratios less than 0.3. When the system operates
at the base OPF solution, all the eigenvalues are on the left side of the imaginary axis.
When the load is increased, the eigenvalues move toward the imaginary axis until the
system encounters a Hopf bifurcation. Based on the HVDC set-point adjustment, the
oscillatory modes move away from the imaginary axis and the damping ratios of the
Chapter 5. HVDC Operating-Point Adjustment 75

15

10

After HVDC Set−Point Variation (β=β*=1.9)

ω(rad/s) 0 Base OPF Before Load Variation (β=0)
*
Base OPF After Load Variation (β=β =1.9)

−5

−10

−15
−3 −2.5 −2 −1.5 −1 −0.5 0
σ(1/s)

Figure 5.8: Eigenvalues plot of the system corresponding to three operating points ob-
tained from the optimization problem, and base OPF solution after and before load
variation

oscillatory modes increase.

5.5.2.2 Case 2: Line Impedance Change

The effects of changes in transmission system impedance on system stability have been
reported in the technical literature [76]. In this section, the effect of such changes (rather
than load scaling, as in case 1) on the system stability is studied and the application of
the proposed operating point tuning method to stabilize the system, subsequent to a line
outage, is demonstrated. In this case study β for loads in segments 1, 2 and 3 are 0, 0,
and 0.6, respectively. To represent the line outage impact, the impedance of line 3 within
each segment is increased to 100 times the nominal value. To stabilize the system due to
the line outage, the sensitivity of the stability margin with respect to the set-points of
the HVDCs is calculated and then, based on a line search, the optimum operating point
is determined.
Fig. 5.9 shows the loci of the two least damped eigenvalues (mode 1 and 2 have the
least damping as the line opens) for various line impedances at the base OPF solution
and demonstrates that as the line impedance increases the system stability decreases (at
the Hopf point, mode 1 crosses the imaginary axis). The eigenvalues at the Hopf point
and the line outage are identified on Fig. 5.9. Fig. 5.9 also shows that the system stability
improves as the DC flows are changed in the direction of the sensitivity. The loci of the
Chapter 5. HVDC Operating-Point Adjustment 76

Figure 5.9: Loci of the eigenvalues associated with the least damped modes due to
changes in line impedances. The dashed line indicates the change in operating conditions
determined via the method detailed in Fig. 5.1

two eigenvalues associated with the least damped modes due to an increase in the line 3
impedance is depicted in Fig. 5.9. Fig. 5.9 shows that at the optimum operating point,
mode 1 moves further away from the imaginary axis (increased damping) and mode 2
moves toward the imaginary axis (decreased damping). However, since the damping of
the critical mode (mode 1) is increased, the system becomes more stable. Fig. 5.9 also
shows that at the optimum operating point, for various line impedance values and even
when the line three of each segment is disconnected (Z3 → ∞), the system remains stable
at the new operating point.

5.6 Conclusions
This chapter proposes a new method, based on tuning the HVDC control set-points
(reference values) of a DC-segmented AC system, to increase the stability margin. The
system optimum operating point is obtained from an optimization process in which the
set-points of the active power, transferred through the HVDC links, are the optimization
variables. To solve the optimization problem, the sensitivity of the stability margin
with respect to the controllable parameters at the Hopf point is used to determine the
Chapter 5. HVDC Operating-Point Adjustment 77

direction in which the HVDC set-points should be changed and a line search is carried
out to determine the optimum step size. At the optimum operating point, the eigenvalues
associated with the least damped oscillatory modes are shifted away from the imaginary
axis and the stability margin is improved.
The use of the optimum solution as a temporary operating point can help to restabilize
the system subsequent to a change, e.g., a line outage. In the absence of the optimum
solution, the system will definitely become unstable; however, the optimum solution
directs the system toward a stable equilibrium. It can also be used as a preemptive
operating point to increase the robustness of the system to changes, e.g., a line outage.
Simulation results show that the proposed method can improve the stability margin and
damping ratio caused by various events such as changes in the loading factor and line
outages.
Chapter 6

Conclusions

6.1 Thesis Summary

This thesis proposes a control strategy for small-signal dynamic stability enhancement
of a DC-segmented AC power system, in which the AC grid is decomposed into smaller
AC segments connected through DC links, whereas the main power corridors among
the segments are the DC links. Although there have been studies on the conceptual
segmentation of the AC grid with HVDC links, to the best of our knowledge, this thesis
is the first study which investigates the small-signal stability enhancement in such a
configuration.
The proposed control strategy includes four control schemes based on HVDC sup-
plementary control and modification of the operating condition of the HVDC system.
Depending on the system characteristics, one or more of the proposed control schemes
may be effective for the mitigation of the system oscillations.
Control scheme 1 provides a supplementary current control of an LCC HVDC link,
based on MPC method to damp the oscillatory dynamics within a segment of a DC-
segmented AC system. Control schemes 2 and 3 stabilize the system and provide co-
ordinated LQG-based supplementary control to minimize the propagation of dynamics
among the segments and selectively permit the propagation of the oscillations from the
perturbed segment to the other segments, respectively. In control scheme 4, the Hopf
bifurcation theory is utilized to steer the system away from instability, increase the sta-
bility margin and improve the damping of oscillatory modes by changing the HVDC
set-points in the optimum direction determined based on the sensitivities of the Hopf
stability margin to the set-point values of the HVDC links.
Since power system software tools exhibit limitations for advanced control design due
to (i) inadequacy of a full model for the AC-DC system, (ii) lack of transparency of

78
Chapter 6. Conclusions 79

model structures, (iii) inability to automatically generate a linearized state-space model,

and (iv) inflexibility to accommodate custom-made models; this thesis also addresses the
above issues and presents a methodology based on MATLAB/Simulink software to (i)
systematically construct the nonlinear differential-algebraic model of an AC-DC system,
and (ii) automatically extract a linearized state space model of the system. The developed
framework is used for the design of the proposed control schemes of this thesis and a
platform for the time-domain simulation of power system dynamics.

6.2 Thesis Conclusions

This thesis concludes that

the developed MATLAB/Simulink-based framework (i) provides accurate time-

domain simulations, (ii) enables for building an accurate nonlinear AC-DC power
system model from the complete DAEs of the system (iii) automatically derives the
linear model of the AC-DC system which can be used for linear control design, (iv)
provides modeling flexibility, and (v) enables for easy inclusion of additional com-
ponent models, e.g., those of FACTS controllers and custom-specified apparatus.

the developed intra-segment MPC-based supplementary controller mitigates the

dominant inter-area oscillatory modes of the AC-DC system while addressing actu-
ator limitations and enabling operation closer to the system constraints on inputs,
outputs and states. Study results show that the MPC-based supplementary con-
troller has a time-varying gain which provides faster recovery times after faults
compared to the traditional LQG-based supplementary controller. Furthermore,
the MPC-based supplementary controller has satisfactory performance even if the
communication delay is accounted for.

the developed inter-segment LQG-based supplementary controllers improve the sys-

tem small-signal stability by (i) minimizing the propagation of the oscillatory dy-
namics among the segments or (ii) selectively distributing the oscillatory dynamics
among the segments. Furthermore, the sound performance of the inter-segment
supplementary control schemes is maintained for various system operating points
and fault locations.

the developed HVDC control set-points tuning scheme increases the system stability
margin and improves the damping ratio for Hopf bifurcations caused by various
events such as changes in the loading factor and line impedances, e.g., line outages.
Chapter 6. Conclusions 80

The optimum operating point, obtained from an optimization process, can be used
as a temporary or permanent operating point to restabilize the system subsequent
to a change, e.g., a line outage (which leads the system to instability) or to increase
the robustness of the system to changes, e.g., a line outage, respectively.

6.3 Thesis Contributions

The contributions of this thesis are:

developing a control strategy to damp oscillatory dynamics in a DC-segmented AC

system including

1. an intra-segment supplementary classical HVDC controller based on the MPC

methodology.
2. an inter-segment coordinated supplementary classical HVDC controller based
on the LQG methodology to stabilize the system and minimize the transfer of
oscillatory dynamics among the segments.
3. an inter-segment coordinated supplementary classical HVDC controller based
on the LQG methodology to distribute the oscillatory dynamics among the
segments.
4. an HVDC control system operating-point tuning scheme to shift the eigenval-
ues associated with the least damped oscillatory modes away from the imagi-
nary axis and therefore enhance the stability margin.

presenting a systematic approach to design each of the above control schemes.

developing a MATLAB/Simulink-based framework for control design and time-

domain simulation of the AC-DC power system transients.

6.4 Future Works

Further research in continuation of this work includes the following:

inclusion of various component models to the MATLAB/Simulink-based frame-

work, e.g., wind farms, power system stabilizers (PSS), and FACTS controllers.

investigating the performance of other types of intra-segment supplementary con-

trollers, e.g., adaptive [77] and robust controllers, to damp oscillatory dynamics.
Chapter 6. Conclusions 81

investigating the performance of other types of inter-segment supplementary con-

trollers, e.g., adaptive and robust controllers, to improve system stability.

developing a hierarchical control structure consisting of the proposed control schemes

of this thesis, which coordinates all the proposed control schemes, to further im-
prove the system stability. The hierarchical control structure should identify when
each control scheme starts and stops coping with the oscillatory dynamics during
a fault or disturbance scenario.
Appendix A

MPC Optimization Procedure

Consider a system of the form

x(k + 1) = Ax(k) + Bu(k) + (k), (A.1)

y(k) = Cx(k) + ν(k). (A.2)

In the optimization phase of the MPC control design, first the future outputs of the
system are predicted and then the objective function is minimized. For prediction, the
state space model of the system is used and the future outputs are calculated by
⎡ ⎤ ⎡ ⎤
y(k + 1) Δu(k|k)
⎢ ⎥ ⎢ ⎥
⎢ .. ⎥
⎥ = Sx x(k) + Su1 u(k − 1) + Su ⎢ ⎥
..
⎢ . ⎣ . ⎦ (A.3)
⎣ ⎦
y(k + p) Δu(k + p − 1|k)

where ⎡ ⎤
⎡ ⎤ CB
CA ⎢ ⎥
⎢ ⎥ ⎢CB + CAB⎥
⎢CA2 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
Sx = ⎢ . ⎥ , Su1 = ⎢ .
. ⎥, (A.4)
⎢ .. ⎥ ⎢ . ⎥
⎣ ⎦ ⎢ ⎥
⎣ p−1
% h
⎦
p
CA CA B
h=0
⎡ ⎤
CB 0 ... 0
⎢ ⎥
⎢CB + CAB CB ... 0 ⎥
⎢ ⎥
⎢ ⎥
Su = ⎢ .
. .
. .. ⎥. (A.5)
⎢ . . . 0 ⎥
⎢ ⎥
⎣ p−1
% %
p−2 ⎦
h h
CA B CA B . . . CB
h=0 h=0

82
Appendix A. MPC Optimization Procedure 83

Substituting y(k + 1), . . . , y(k + p) from (A.3) in (3.10)-(3.12), the MPC optimization
problem is converted to a quadratic programming (QP) form given by

min(zT Hz + 2f T z), (A.6)

z

subject to
bmin ≤ Λz ≤ bmax , (A.7)

where
H = WΔu + STu Wy Su , (A.8)

⎤T ⎡
r(k + 1)
 ⎢ ⎥ 
⎢ .. ⎥ Wy Su ,
f = u(k − 1) Su1 + x(k) Sx − ⎣
T T T T T
. ⎦ (A.9)
r(k + p)

⎡ ⎤
Δu(k|k)
⎢ .. ⎥
z=⎢
⎣ . ⎥,
⎦ (A.10)
Δu(k + p − 1|k)
Δu Δu Δu Δu
WΔu = diag(w0,1 , w0,2 , . . . , w0,n u
, . . . , wp−1,n u
), (A.11)
y y y y
Wy = diag(w0,1 , w0,2 , . . . , w0,n y
, . . . , wp−1,n y
), (A.12)
⎡ ⎤
u max (0) − u(k − 1)
⎡ ⎤
1 0 ... 0 ⎢ min
⎥
⎢ .. ⎥
⎢ ⎥ ⎢ . ⎥
⎢1 1 . . . 0⎥ ⎢ ⎥
⎢ ⎥ ⎢ max ⎥
⎢. . . . ⎥ ⎢ min
u (p − 1) − u(k − 1) ⎥
Λ=⎢ . .
⎢. .
. . .. ⎥ , b max = ⎢
⎥ min ⎢
⎥.
⎥ (A.13)
⎢ ⎥ ⎢ Δu max (0) ⎥
⎢1 1 . . . 1⎥ ⎢ min
⎥
⎣ ⎦ ⎢ .. ⎥
⎢ . ⎥
Ip×p ⎣ ⎦
Δu min (p − 1)
max

In the above equations Ip×p is the identity matrix of size p.

Appendix B

System Differential-Algebraic
Equations (DAEs)

This section describes how the HVDC system affects the AC-DC system algebraic equa-
tions in the optimization formulation of chapter 5. The algebraic and differential variables
of the AC-DC system of (5.1) are also presented.
In (5.1), g(x, y, λ) = 0 represents the algebraic equations of the stator circuitry of
the SMs (gSMac ), (2.1)-(2.2), network nodal equations (gN E ), (B.1), coordinate transfor-
mation equations (gCOac ) for all the generator buses, (2.30)-(2.33), and the DC algebraic
equations. The network nodal algebraic equations represent the AC transmission network
as
Īinj = Yred V̄inj , (B.1)

where Īinj and V̄inj are the vectors of current and voltage phasors at the injection buses
and Yred is the reduced network admittance matrix [36]. Injection buses include the
generator and HVDC buses. The DC algebraic equations include (B.2)-(B.3) for all
HVDC buses.
Pconv − VDconv IDconv − VQconv IQconv = 0, (B.2)

Qconv − VQconv IDconv + VDconv IQconv = 0. (B.3)

In this study, the vectors associated with the overall DAEs of a multi-machine AC-DC
system represented in (5.1) are
 T
x = xT1 xT2 . . . xTnT G , (B.4)

 
xi = [Eqi Edi Ψ1di Ψ2qi δi ωi Ef di Rf i VRi Tsvi Tmi ]T , (B.5)

84
Appendix B. System Differential-Algebraic Equations (DAEs) 85

 
yac
y= , (B.6)
ydc
 T
yac = yT1ac yT2ac . . . yTnT G ac , (B.7)
 T
yiac = VDi VQi Vdi Vqi IDi IQi Idi Iqi , (B.8)
 T
ydc = yT1dc yT2dc . . . yTndcdc , (B.9)

yidc = [VDirec VDiinv VQirec VQiinv IDirec IDiinv IQirec IQiinv ]T . (B.10)
Appendix C

Calculating the Sensitivity of

Stability Margin with Respect to
Parameters

To demonstrate the procedure of calculating the sensitivity of the stability margin with
respect to the parameters, a simple example where the damping ratio of the oscillatory
modes has a closed form expression is provided. For a system of the form
    
ẋ1 α1 − α22 β x1
= , α2 ≥ 0 (C.1)
ẋ2 −β α1 − α22 x2
 
α1 − α22 β
with λ = [μ p]T , μ = α1 , p = α2 , z = [x1 x2 ]T , and Fz = ,
−β α1 − α22
the eigenvalues are κ = α1 − α22 ± iβ. When α1 increases from α1 to α1∗ = α22 , the
(0)

system encounters a Hopf bifurcation and κ∗ = ±iβ ∗ . Therefore, the Hopf bifurcation
hypersurface can be written as:
&     2
'
ΣHopf = [α1∗ α2∗ ]T : α1∗ = α2∗ .

2
The stability margin is then defined as M = α1∗ − α1 = α2∗ − α1 . The left and
(0) (0)

−i
right eigenvectors corresponding to the Hopf point are d = [ √12 √ 2
] and e = [ √12 √i2 ]T ,
respectively. In this example, Fzz = 0 and Fzλ is a 2 × 2 × 2 tensor where
   
1 0 −2α2 0
Fzλ1 = , Fzλ2 = ,
0 1 0 −2α2

86
Appendix C. Sensitivity of Stability Margin with Respect to Parameters87

  
1 −i 1 0 √1
nα1 = Re([ √ √ ] 2
) = 1,
2 2 0 1 √i
2
  
1 −i −2α2 0 √1
nα2 = Re([ √ √ ] 2
) = −2α2 ,
2 2 0 −2α2 √i
2
 
1
N= ,
−2α2
(0)
Mα2 |α(0) = 2α2 .
2

Mα2 |α(0) indicates that in order to increase the stability margin, α2 should be increased
2
as expected, given that α22 defines the limit on α1 which results in a Hopf bifurcation.
Increasing α2 moves the eigenvalues further to the left of the imaginary axis and improves
α2 −α
the damping ratio (ζ = √ 2 2 2 1 2 ).
β +(α2 −α1 )
Bibliography

[1] P. Kundur. Power System Stability and Control. McGraw-Hill, 1994.

[2] Final report on the August 14, 2003 blackout in the United States
and Canada: causes and recommendations. [Online]: Available
https://reports.energy.gov/BlackoutFinal-Web.pdf, 2004.

[3] C. Kim, V. K. Sood, G. Jang, S. Lim, and S. Lee. HVDC Transmission: Power
Conversion Applications in Power Systems. IEEE Press-John Wiley and Sons (Asia)
Pte Ltd, 2009.

[4] ABB HVDC reference projects. [Online]: Available

http://www.abb.com/industries/us/9AAF400191.aspx?country=00, 2013.

[5] HVDC grid feasibility study. Technical report, Cigre Working Group B4-52, 2012.

[6] J. Pan, R. Nuqui, L. Tang, and P. Holmberg. VSC-HVDC control and application in
meshed AC networks. In IEEE Power Engineering Society General Meeting, 2008.

[7] J. Pan, R. Nuqui, K. Srivastava, T. Jonsson, P. Holmberg, and Y. Hafner. AC grid

with embedded VSC-HVDC for secure and efficient power delivery. In IEEE Energy
2030 Conference, pages 1–6, 2008.

[8] S. Zhou, J. Liang, J. Ekanayake, and N. Jenkins. Control of multi-terminal VSC-

HVDC transmission system for offshore wind power generation. In Proceedings of
the 44th International Universities Power Engineering Conference (UPEC), pages
1–5, 2009.

[9] T. Ding, C. Zhang, Z. Hu, and Z. Duan. Coordinated control strategy for multi-
terminal VSC-HVDC based wind farm interconnection. In International Conference
on Sustainable Power Generation and Supply (SUPERGEN), pages 1–6, 2009.

88
Bibliography 89

[10] D. Van Hertem and M. Ghandhari. Multi-terminal VSC-HVDC for the European
supergrid: Obstacles. Renewable and Sustainable Energy Reviews, 14:3156–3163,
2010.

[11] H. Clark, A. Edris, M. El-Gasseir, K. Epp, A. Isaacs, and D. Woodford. Softening

the blow of disturbances. IEEE Power and Energy Magazine, 6:30–41, 2008.

[12] F. Xinghao and J. H. Chow. Upgrading AC transmission to DC for maximum

power transfer capacity. In Power System Conference, 2008. MEPCON 2008. 12th
International Middle-East, pages 1–7, 2009.

[13] L. O. Barthold, H. Clark, and D. Woodford. Principles and applications of current-

modulated HVDC transmission systems. In Transmission and Distribution Confer-
ence and Exhibition, 2005/2006 IEEE PES, pages 1429–1435, 2006.

[14] H. Clark, M. M. El-Gasseir, H. D. K. Epp, and A. Edris. The application of segmen-

tation and grid shock absorber concept for reliable power grids. In 12th International
Middle-East Power System Conference (MEPCON), pages 34–38, 2008.

[15] G. C. Loehr. Take my grid, please! A daring proposal for electric transmission.
Technical report, IEEE HVDC-FACTS SUBCOMMITTEE MEETING, 2006 IEEE-
PES Technical Committee Meeting, 2006.

[16] R. Eriksson and V. Knazkins. Nonlinear coordinated control of multiple HVDC

links. In IEEE 2nd International Power and Energy Conference, pages 497–501,
2008.

[17] M. Noroozian, L. Angquist, M. Ghandhari, and G. Andersson. Improving power

system dynamics by series-connected FACTS devices. Power Delivery, IEEE Trans-
actions on, 12:1635 – 1641, 1997.

[18] Z. Huang, Y. Ni, C. M. Shen, F. F. Wu, S. Chen, and B. Zhang. Application of

unified power flow controller in interconnected power systems-modeling, interface,
control strategy, and case study. Power Systems, IEEE Transactions on, 15:817 –
824, 2000.

[19] S. Jiang, U. D. Annakkage, and A. M. Gole. A platform for validation of FACTS

models. Power Delivery, IEEE Transactions on, 21:484 – 491, 2006.

[20] R. Rouco and F. L. Pagola. An eigenvalue sensitivity approach to location and con-
troller design of controllable series capacitors for damping power system oscillations.
Power Systems, IEEE Transactions on, 12:1660 – 1666, 1997.
Bibliography 90

[21] N. Martins, H. J. C. P. Pinto, and J. J. Paserba. Using a TCSC for line power
scheduling and system oscillation damping-small signal and transient stability stud-
ies. In IEEE Power Engineering Society Winter Meeting, pages 1455–1461, 2000.

[22] S. Arabi, G. J. Rogers, D. Y. Wong, P. Kundur, and M. G. Lauby. Small signal

stability program analysis of SVC and HVDC in AC power systems. Power Systems,
IEEE Transactions on, 6:1147 – 1153, 1991.

[23] I. Kamwa, J. Beland, G. Trudel, R. Grondin, C. Lafond, and D. McNabb. Wide-area

monitoring and control at Hydro-Quebec: past, present and future. In IEEE Power
Engineering Society General Meeting, 2006.

[24] M. R. Iravani, P. L. Dandeno, K. H. Nguyen, D. Zhu, and D. Maratukulam. Applica-

tions of static phase shifters in power systems. Power Delivery, IEEE Transactions
on, 9:1600 – 1608, 1994.

[25] K. V. Patil, J. Senthil, J. Jiang, and R. M. Mathur. Application of STATCOM for

damping torsional oscillations in series compensated AC systems. Energy Conver-
sion, IEEE Transactions on, 13:237 – 243, 1998.

[26] P. Kundur, M. Klein, G. J. Rogers, and M. S. Zywno. Application of power sys-

tem stabilizers for enhancement of overall system stability. Power Systems, IEEE
Transactions on, 4:614 – 626, 1989.

[27] I. Kamwa, R. Grondin, and Y. Hebert. Wide-area measurement based stabilizing

control of large power systems-a decentralized/hierarchical approach. Power Sys-
tems, IEEE Transactions on, 16:136 – 153, 2001.

[28] D. Dotta, A. S. e Silva, and I. C. Decker. Wide-area measurements-based two-

level control design considering signal transmission delay. Power Systems, IEEE
Transactions on, 24(1):208–216, 2009.

[29] M. Xiao-ming, Z. Yao, G. Lin, and W. Xiao-chen. Coordinated control of interarea

oscillation in the China Southern power grid. Power Systems, IEEE Transactions
on, 21:845 – 852, 2006.

[30] I. Kamwa, G. Trudel, and L. Gerin-Lajoie. Robust design and coordination of mul-
tiple damping controllers using nonlinear constrained optimization. Power Systems,
IEEE Transactions on, 15:1084 – 1092, 2000.
Bibliography 91

[31] P. Zhang and A. H. Coonick. Coordinated synthesis of PSS parameters in multi-

machine power systems using the method of inequalities applied to genetic algo-
rithms. Power Systems, IEEE Transactions on, 15:811 – 816, 2000.

[32] R. L. Cresap and W. A. Mittelstadt. Small-signal modulation of the Pacific HVDC

intertie. Power Apparatus and Systems, IEEE Transactions on, 95:536 – 541, 1976.

[33] T. Smed and G. Anderson. Utilising HVDC to damp power oscillations. Power
Delivery, IEEE Transactions on, 8:620–627, 1993.

[34] J. H. Chow and K. W. Cheung. A toolbox for power system dynamics and control
engineering education and research. Power Systems, IEEE Transactions on, 7:1559–
1564, 1992.

[35] MATLAB. The MathWorks Simulink User’s Guide R2012a. The MathWorks Inc.,
2010.

[36] P. W. Sauer and M. A. Pai. Power System Dynamics and Stability. Prentice-Hall,
Inc, 1998.

[37] R. Preece and J. V. Milanovic. Comparison of dynamic performance of meshed

networks with different types of HVDC lines. In 9th IET International Conference
on AC and DC Power Transmission, 2010. ACDC., 2010.

[38] F. Milano. Power System Modeling and Scripting. Springer, 2010.

[39] R. M. Brandt, U. D. Annakkage, D. P. Brandt, and N. Kshatriya. Validation of a

two-time step HVDC transient stability simulation model including detailed HVDC
controls and DC line L/R dynamics. In IEEE Power Engineering Society General
Meeting, 2006.

[40] Y. Y. Hong and W. C. Wu. New approach using optimization for tuning parameters
of power system stabilizers. Energy Conversion, IEEE Transactions on, 14:780–786,
1999.

[41] X. Lei, E. N. Lerch, and D. Povh. Optimization and coordination of damping controls
for improving system dynamic performance. Power Systems, IEEE Transactions on,
16:473–480, 2001.

[42] A. L. B. Do Bomfim, G. N. Taranto, and D. M. Falcao. Simultaneous tuning of

power system damping controllers using genetic algorithms. Power Systems, IEEE
Transactions on, 15:163–169, 2000.
Bibliography 92

[43] Y. N. Yu, K. Vongsuriya, and L. N. Wedman. Application of an optimal control

theory to a power system. Power Apparatus and Systems, IEEE Transactions on,
PAS-89:55–62, 1970.

[44] Y. Fan, C. Chen, and W. Xitian. Observer-based decentralized control of inter-

area oscillation in multi-infeed HVDC system. In IEEE/PES Transmission and
Distribution Conference and Exposition, pages 1–4, 2008.

[45] W. D. Yang, Z. Xu, and Z. X. Han. A co-ordinated large-signal modulation strategy

for multi-infeed HVDC systems. In Seventh International Conference on AC-DC
Power Transmission, pages 338–343, 2002.

[46] D. Q. Mayne, J. B. Rawlings, C. V. Rao, and P. O. M. Scokaert. Constrained model

predictive control: Stability and optimality. Automatica, 36:789–814, 2000.

[47] S. Joe Qin and T. A. Badgwell. A survey of industrial model predictive control
technology. Control Engineering Practice, 11:733–764, 2003.

[48] G. Welch and G. Bishop. An introduction to the Kalman filter. Technical Report
TR 95-041, University of North Carolina, Department of Computer Science, 1995.

[49] E. F. Camacho and C. Bordons. Model Predictive Control. Springer, 2004.

[50] C. Schmid and L. T. Biegler. Quadratic programming methods for reduced hessian
SQP. Computers and Chemical Engineering, 18:817–832, 1994.

[51] J. H. Chow, J. J. Sanchez-Gasca, H. Ren, and S. Wang. Power system damping

controller design-using multiple input signals. IEEE Control Systems, 20(4):82–90,
2000.

[52] K. Morrison, L. Wang, and P. Kundur. Power system security assessment. IEEE
Power and Energy Magazine, 2(5):30–30, 2004.

[53] R. Schainker, P. Miller, W. Dubbelday, P. Hirsch, and G. Zhang. Realtime dynamic

security assessment. Power Systems, IEEE Transactions on, 4(2):51–58, 2006.

[54] K. Morrison, L. Wang, and P. Kundur. Implementation of online security assessment.

IEEE Power and Energy Magazine, 4(5):46–59, 2006.

[55] G. De Nicolao, L. Magni, and R. Scattolini. On the robustness of receding hori-

zon control with terminal constraints. Automatic Control, IEEE Transactions on,
41(3):451–453, 1996.
Bibliography 93

[56] H. Michalska and D. Q. Mayne. Robust receding horizon control of constrained

nonlinear systems. Automatic Control, IEEE Transactions on, 38(11):1623–1633,
1993.

[57] G. De Nicolao, L. Magni, and R. Scattolini. Stability and robustness of nonlinear

receding horizon control. Journal of Nonlinear model predictive control, 26:3–22,
2000.

[58] P. M. Anderson and A. A. Fouad. Power System Control and Stability. IEEE Press,
1994.

[59] R. E. Kalman. Mathematical description of linear systems. SIAM Journal on Control

and Optimization, 1:152–192, 1963.

[60] S. Skogestad and I. Postlethwaite. Multivariable Feedback Control: Analysis and

Design. John Wiley and Sons, 1996.

[61] Y. Halevi. Stable LQG controllers. IEEE Transactions on Automatic Control,

39:2104–2106, 1994.

[62] C. D. Vournas, M. A. Pai, and P. W. Sauer. The effect of automatic voltage regu-
lation on the bifurcation evolution in power systems. Power Systems, IEEE Trans-
actions on, 11:1683–1688, 1996.

[63] A. A. P. Lerm and A. Silva. Avoiding Hopf bifurcations in power systems via set-
points tuning. Power Systems, IEEE Transactions on, 19:1076–1084, 2004.

[64] A. A. P. Lerm. Control of Hopf bifurcation in multi-area power systems via a sec-
ondary voltage regulation scheme. In Power Engineering Society Summer Meeting,
2002 IEEE, pages 1615–1620, 2002.

[65] S. Wang, Y. Yi, Q. Jiang, X. Chen, and Y. Cao. On-line control of Hopf bifurcations
in power systems. In IEEE Power and Energy Society General Meeting - Conversion
and Delivery of Electrical Energy in the 21st Century, pages 1–7, 2008.

[66] N. Mithulananthan, C. A. Canizares, J. Reeve, and G. J. Rogers. Comparison of

PSS, SVC, and STATCOM controllers for damping power system oscillations. Power
Systems, IEEE Transactions on, 18:786–792, 2003.

[67] A. I. Zecevic and D. M. Miljkovic. The effects of generation redispatch on Hopf

bifurcations in electric power systems. Circuits and Systems I: Fundamental Theory
and Applications, IEEE Transactions on, 49:1180–1186, 2002.
Bibliography 94

[68] H. Zhenyu, N. Zhou, F. Tuffner, C. Yousu, D. Trudnowski, W. Mittelstadt, J. Hauer,

and J. Dagle. Improving small signal stability through operating point adjustment.
In IEEE Power and Energy Society General Meeting, pages 1–8, 2010.

[69] S. K. M. Kodsi and C. A. Canizares. Application of a stability-constrained optimal

power flow to tuning of oscillation controls in competitive electricity markets. Power
Systems, IEEE Transactions on, 22:1944–1954, 2007.

[70] G. G. Lage, G. R. M. da Costa, and C. A. Canizares. Limitations of assigning general

critical values to voltage stability indices in voltage-stability-constrained optimal
power flows. In 2012 IEEE International Conference on Power System Technology
(POWERCON), pages 1–6, 2012.

[71] A. A. P. Lerm. Control of Hopf bifurcation in power systems via a generation

redispatch. In IEEE Porto Power Tech Proceedings, pages 1–8, 2001.

[72] Independent Electricity System Operator. Review of the dispatch

algorithm’s compliance with market rules. [Online]: Available
http://www.ieso.ca/imoweb/pubs/dso-2012.pdf, June 2012.

[73] I. Dobson, F. Alvarado, and C. L. DeMarco. Sensitivity of Hopf bifurcations to

power system parameters. In Proceedings of the 31st Conference on Decision and
Control, 1992.

[74] T. Smed. Feasible eigenvalue sensitivity for large power systems. Power Systems,
IEEE Transactions on, 8:555–563, 1993.

[75] M. Lotfalian, R. Schlueter, D. Idizior, P. Rusche, S. Tedeschi, L. Shu, and A. Yaz-

dankhah. Inertial, governor, and AGC/economic dispatch load flow simulations of
loss of generation contingencies. Power Apparatus and Systems, IEEE Transactions
on, PAS-104:3020–3028, 1985.

[76] D. J. N. Limebeer, R. G. Harley, and M. A. Lahoud. Suppressing subsynchronous

resonance with static filters. Generation, Transmission and Distribution, IEE Pro-
ceedings C, 128:33–44, 1981.

[77] S. Lefebvre, M. Saad, and R. Hurteau. Adaptive control for HVDC power transmis-
sion systems. Power Apparatus and Systems, IEEE Transactions on, PAS-104:2329–
2335, 1985.