Du2018 PDF

Wenjuan Du · Haifeng Wang · Siqi Bu
Small-Signal Stability
Analysis of Power
Systems Integrated
with Variable Speed
Wind Generators
Small-Signal Stability Analysis of Power Systems
Integrated with Variable Speed Wind Generators
Wenjuan Du • Haifeng Wang • Siqi Bu
Small-Signal Stability
Analysis of Power Systems
Integrated with Variable
Speed Wind Generators
Wenjuan Du Haifeng Wang
School of Electrical and Electronic School of Electrical and Electronic
Engineering Engineering
North China Electric Power University North China Electric Power University
Beijing, China Beijing, China
Siqi Bu
Department of Electrical Engineering
The Hong Kong Polytechnic University
Kowloon, Hong Kong
ISBN 978-3-319-94167-7 ISBN 978-3-319-94168-4 (eBook)

https://doi.org/10.1007/978-3-319-94168-4
Library of Congress Control Number: 2018947164
© Springer International Publishing AG, part of Springer Nature 2018

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or
information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the
editors give a warranty, express or implied, with respect to the material contained herein or for any errors
or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.
This Springer imprint is published by the registered company Springer Nature Switzerland AG.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Doubly fed induction generators and permanent magnet synchronous generators are
the two main types of variable speed wind generators. Changes brought about by the
grid connection of variable speed wind generators in an AC power system are
twofold. Firstly, unlike conventional synchronous generators, which are linked
directly to the power system, a variable speed wind generator is connected to the
grid via controlled voltage source converters. The dynamic behavior of the variable
speed wind generator is significantly different from that of a synchronous generator.
The impact of dynamic interactions introduced by the grid-connected variable speed
wind generator on power system stability needs to be examined carefully. In
addition, wind generation is variable due to the stochastic fluctuation of wind
speed. It is important to investigate how the random variation of wind speed may
affect power system stability. Secondly, VSC-based HVDC lines and multi-terminal
DC (MTDC) networks for large scale wind power generation, especially for the
offshore wind power, have many technical advantages. Stability of a VSC-based
DC/AC power system with wind power generation is an important aspect of the
impact brought about by grid-connected variable speed wind generator and needs to
be studied.
We started the investigation on the small-signal stability of the power system
affected by the grid-connected variable speed wind generators about a decade ago.
Our investigation has covered the two important aspects mentioned above, i.e., the
small-signal stability of the power system considering the impact of dynamic
interactions introduced by the VSWGs, the stochastic variations of wind speed,
and the integration of VSC-based HVDC lines and MTDC network. In this book, we
introduce the small-signal stability analysis of the power system integrated with the
variable speed wind generators by mainly presenting the results of our investigation
made in the last decade. Two main methods for the investigation are the damping
torque analysis and modal analysis. Their applications to examine the impact of grid-
connected variable speed wind generators on power system small-signal stability are
comprehensively introduced in the book. It is worth mentioning particularly one
important phenomenon we have found recently. That is the possibility of strong
v
vi Preface
dynamic interactions introduced by the grid-connected variable speed wind gener-

ators to degrade the small-signal stability of the power system under the condition of
open-loop modal resonance. Although we are still studying the phenomenon from
the aspects of theoretical analysis and practical applications, we introduce the initial
results of our investigation in the book. We hope the potential instability risk brought
about by the open-loop modal resonance should be aware of and understood fully by
more power system researchers and engineers.
We would like to thank the contributions from our research students to the book.
They are Mr. Xiao CHEN, Mr. Jintian BI, Mr. Qiang FU, Mr. Xubin WANG, and
Mr. Wenkai DONG.
Beijing, China Wenjuan Du

Beijing, China Haifeng Wang
Kowloon, Hong Kong Siqi Bu
April 2018
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Impact of Grid-Connected VSWGs on the Small-Signal
Angular Stability of Power System . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Impact of the VSWGs: A Brief Review . . . . . . . . . . . . . . . 2
1.1.2 Impact of the VSWGs: Discussions . . . . . . . . . . . . . . . . . . 5
1.2 Small-Signal Stability of the VSC-Based DC/AC
Power System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.1 Small-Signal Stability of a Single Grid-Connected
VSC System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.2 Small-Signal Stability of Integrated VSC-Based
DC/AC Power Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 16
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2 Linearized Model of a Power System with a Grid-Connected
Variable Speed Wind Generator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1 Linearized Model of a PMSG . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1.1 Linearized Model of the Permanent Magnet SG . . . . . . . . . 27
2.1.2 Linearized Model of Machine Side Converter
and Associated Control System . . . . . . . . . . . . . . . . . . . . . 30
2.1.3 Linearized Model of Grid Side Converter
and Associated Control System . . . . . . . . . . . . . . . . . . . . . 33
2.1.4 Linearized Model of the Entire PMSG System . . . . . . . . . . 39
2.2 Linearized Model of a DFIG . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.1 Linearized Model of the Induction Generator
and Two-Mass Shaft Rotational System . . . . . . . . . . . . . . 41
2.2.2 Linearized Model of the RSC and Associated
Control System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.2.3 Linearized Model of the GSC and Associated
Control System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.2.4 Linearized Model of the Entire DFIG . . . . . . . . . . . . . . . . 53
vii
viii Contents
2.3 Linearized Model of the Power System . . . . . . . . . . . . . . . . . . . . 55

2.4 Closed-Loop Interconnected Model of the Power System
with a Grid-Connected VSWG . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.4.1 Closed-Loop Interconnected Model of the Power
System with a PMSG . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
2.4.2 Closed-Loop Interconnected Model of the Power
System with a DFIG . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3 Damping Torque Analysis of Small-Signal Angular Stability
of a Power System Affected by Grid-Connected Wind Power
Induction Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.1 Impact of Grid-Connected Wind Power Induction
Generators on Small-Signal Angular Stability . . . . . . . . . . . . . . . . 63
3.2 System Modeling for Damping Torque Analysis
of a Power System Affected by Grid-Connected Wind Power
Induction Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2.1 Phillips-Heffron Model of a Power System
with Wind Power Induction Generators as a Generic
Open-Loop Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.2.2 Aggregate Model and Transfer Function of Wind
Power Induction Generators . . . . . . . . . . . . . . . . . . . . . . . 69
3.3 Implementation Framework for Damping Torque Analysis
of a Power System Affected by Grid-Connected Wind
Power Induction Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.4 Analytical Comparison on Damping Effectiveness
And Robustness of Type 1 and 3 Wind Power
Induction Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.4.1 Step 1: Detailed DFIG Model . . . . . . . . . . . . . . . . . . . . . . 73
3.4.2 Step 2: DFIG Model Without RSC Dynamics . . . . . . . . . . 75
3.4.3 Step 3: DFIG Model with Offset Rotor Voltage Only . . . . . 77
3.4.4 Step 4: DFIG Model with Constant Rotor Voltage . . . . . . . 77
3.4.5 Step 5: FSIG Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.4.6 Main Findings in Analytical Comparison on Damping
Mechanism of Type 1 and 3 Wind Power Induction
Generators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.5 Numerical Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.5.1 Example 3.1 (Base Case Comparison Study) . . . . . . . . . . . 80
3.5.2 Example 3.2 (Comparison Study Under Different
Wind Penetration Conditions) . . . . . . . . . . . . . . . . . . . . . . 84
3.6 Summary of Analytical and Numerical Comparison Analysis . . . . 86
Contents ix
Appendix 3.1: A Typical Example of a SMIB System

with Interface Equations of a WPIG . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Appendix 3.2: Data of Examples 3.1 and 3.2 . . . . . . . . . . . . . . . . . . . . 88
Example 16-Machine 68-Bus New York and New England
Power System [36] (Tables 3.4, 3.5 and 3.6) . . . . . . . . . . . . . . . . . 88
Data of DFIG and FSIG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4 Modal Analysis of Small-Signal Angular Stability of a Power System
Affected by Grid-Connected DFIG . . . . . . . . . . . . . . . . . . . . . . . . . . 97
4.1 Decomposed Modal Analysis of the Impact of a Grid-Connected
DFIG on Power System Small-Signal Angular Stability . . . . . . . . 97
4.1.1 Method of Decomposed Modal Analysis . . . . . . . . . . . . . . 97
4.1.2 Example 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.2 Impact of Dynamic Interactions Introduced by a Grid-Connected
DFIG on Power System Small-Signal Angular Stability . . . . . . . . 110
4.2.1 The Index of Dynamic Interactions (IDI) . . . . . . . . . . . . . . 111
4.2.2 The Impact of Dynamic Interactions Introduced
by the DFIG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.2.3 Example 4.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Appendix 4.1: Data of Example 4.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Power System [3] (Tables 4.10, 4.11, 4.12) . . . . . . . . . . . . . . . . . 138
Data of DFIG [4] (Tables 4.13, 4.14, 4.15) . . . . . . . . . . . . . . . . . . 143
Data of New England Power System [13] (Tables 4.16, 4.17
and 4.18) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
Data of DFIG [16] (Tables 4.19, 4.20 and 4.21) . . . . . . . . . . . . . . 146
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
5 Small-Signal Angular Stability of a Power System Affected
by Strong Dynamic Interactions Introduced from
a Grid-Connected VSWG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.1 Impact of Strong Dynamic Interactions Introduced
by a Grid-Connected PMSG . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
5.1.1 Strong Dynamic Interactions Introduced
by the Grid-Connected PMSG and the Impact . . . . . . . . . . 149
5.1.2 Example 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.2 Impact of Strong Dynamic Interactions Introduced
by a Grid-Connected DFIG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
by the Grid-Connected DFIG and the Impact . . . . . . . . . . . 173
5.2.2 Example 5.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
x Contents

Data of New England Power System . . . . . . . . . . . . . . . . . . . . . . 199
Data of DFIG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
Data of PMSG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6 Small-Signal Stability of a Power System with a VSWG
Affected by the PLL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
6.1 Impact of Open-Loop Modal Resonance Caused
by the PLL for a Grid-Connected PMSG . . . . . . . . . . . . . . . . . . . 203
6.1.1 Closed-Loop Interconnected Model of a Power
System with a PMSG . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6.1.2 Open-Loop Modal Resonance Caused by the PLL . . . . . . . 206
6.2 Small-Signal Stability of a Grid-Connected PMSG Wind
Farm Affected by the PLL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
6.2.1 Example 6.1: A PMSG Connected to an External
Power System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.2.2 Example 6.2: A PMSG Wind Farm Connected
to an External Power System . . . . . . . . . . . . . . . . . . . . . . 219
6.3 Example 6.3: Electromechanical Oscillation Modes
of a Power System Affected by the PLL . . . . . . . . . . . . . . . . . . . . 222
6.3.1 Test 1: Changes of Power System
Operating Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
6.3.2 Test 2: Variation of PLL Gains (SRF PLL for PMSG1) . . . 227
6.3.3 Test 3: Variation of SCR and Wind Power
Penetration (SRF PLL for PMSG1) . . . . . . . . . . . . . . . . . . 232
6.3.4 Test 4: DSOGI PLL for PMSG . . . . . . . . . . . . . . . . . . . . . 235
6.3.5 Test 5: The Case of DFIG . . . . . . . . . . . . . . . . . . . . . . . . 236
Appendix 6.1: Data of Example 6.1 (Tables 6.5, 6.6 and 6.7) . . . . . . . . . 240
Data of PMSG1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Line Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
Data of New England Power System . . . . . . . . . . . . . . . . . . . . . . 240
Data of PMSG (Tables 6.8, 6.9 and 6.10) . . . . . . . . . . . . . . . . . . . 240
Data of SRF PLL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
7 Small-Signal Stability of a Power System Integrated with an
MTDC Network for the Wind Power Transmission . . . . . . . . . . . . . 243
7.1 Small-Signal Angular Stability of an AC Power System
Affected by the Integration of an MTDC Network
for the Wind Power Transmission . . . . . . . . . . . . . . . . . . . . . . . . 244
7.1.1 The Open-Loop Modal Resonance (OLMR)
in the MTDC/AC Power System . . . . . . . . . . . . . . . . . . . . 244
7.1.2 Example MTDC/AC Power Systems . . . . . . . . . . . . . . . . . 256
Contents xi
7.2 Small-Signal Stability of an MTDC Network for the Wind

Power Transmission Affected by the Dynamic Interactions
Between the VSCs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
7.2.1 Small-Signal Stability Analysis . . . . . . . . . . . . . . . . . . . . . 268
7.2.2 Example MTDC Power Systems . . . . . . . . . . . . . . . . . . . . 277
7.3 Method of Open-loop Modal Analysis to Examine the Impact
of Dynamic Interactions Introduced by a Selected VSC
Control on the Small-Signal Stability of an MTDC/AC
Power System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
7.3.1 Method of Open-Loop Modal Analysis . . . . . . . . . . . . . . . 286
7.3.2 An Example MTDC/AC Power System
for Applying the Open-Loop Modal Analysis . . . . . . . . . . 295
Appendix 7.3: Data of Example 7.3 (Table 7.7) . . . . . . . . . . . . . . . . . . 300
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
8 Probabilistic Analysis of Small-Signal Stability of a Power System
Affected by Grid-Connected Wind Power Generation . . . . . . . . . . . . 303
8.1 Probabilistic Analysis of Small-Signal Stability
of a Power System: An Introduction . . . . . . . . . . . . . . . . . . . . . . . 304
8.1.1 Framework of Probabilistic Small-Signal
Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
8.1.2 Methodologies of Probabilistic Small-Signal
Stability Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
8.2 Probabilistic Analysis of Small-Signal Stability of a Power
System Affected by Directly Grid-Connected Onshore
Wind Power Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
8.2.1 Cumulant Theory Based Analytical Method
of Probabilistic Small-Signal Stability Analysis
of a Conventional AC Power System . . . . . . . . . . . . . . . . 308
8.2.2 Example 8.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
of a Power System Affected by MTDC–Connected
Offshore Wind Power Generation . . . . . . . . . . . . . . . . . . . . . . . . 320
of Probabilistic Small-Signal Stability Analysis
of a Hybrid AC/DC Power System . . . . . . . . . . . . . . . . . . 320
8.3.2 Example 8.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
8.3.3 Variance Indices Analysis of Probabilistic
Small-Signal Stability of a Hybrid AC/DC
Power System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
8.3.4 Example 8.3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
xii Contents
Appendix 8.1: Data of Examples 8.1, 8.2 and 8.3 . . . . . . . . . . . . . . . . . 341

Example 16-Machine 68-Bus New York and New England Power
System [34] (Tables 8.5, 8.6 and 8.7) . . . . . . . . . . . . . . . . . . . . . . 341
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Chapter 1
Introduction
This chapter introduces the small-signal stability analysis of a power system inte-
grated with variable speed wind generators (VSWGs) by reviewing the following
two subjects which are covered in this book: (1) The impact of grid connection of the
VSWGs on the small-signal angular stability of the power system, i.e., the
low-frequency electromechanical power oscillations (LEPOs); (2) Small-signal sta-
bility of the integrated VSC-based DC/AC power systems.
1.1 Impact of Grid-Connected VSWGs on the Small-Signal

Angular Stability of Power System
The small-signal angular stability of the power system is mainly about the problem
of low-frequency electromechanical power oscillations (LEPOs). In this section, the
review starts with four early published papers which proposed two main strategies to
examine the impact of the VSWGs on power system LEPOs: (1) The VSWGs
displacing the SGs; (2) The VSWGs being connected to the power system.
Following-on work to those four representative papers is reviewed separately
according to the two different strategies of examination. Then, detailed analysis on
each of strategies of examination is carried out. It revealed that in each of two
strategies of examination used, inherently there are two factors affecting the results
of examination. Mixture of two affecting factors has caused the diversification of
results of examination achieved so far about the impact of grid-connected VSWGs
on power system small-signal angular stability.
© Springer International Publishing AG, part of Springer Nature 2018 1

W. Du et al., Small-Signal Stability Analysis of Power Systems Integrated with
Variable Speed Wind Generators, https://doi.org/10.1007/978-3-319-94168-4_1
2 1 Introduction
1.1.1 Impact of the VSWGs: A Brief Review
1.1.1.1 Representatives of Earlier Study
Representatives of earlier published study about the impact of grid-connected

VSWGs on power system small-signal angular stability are [1–4]. Simple example
power systems were used to examine the change of system small-signal angular
stability when the VSWGs were connected. The examination was carried out by
computing system electromechanical oscillation modes (EOMs). In [1], power
output from the VSWGs displaced that of a SG gradually. The results indicated
that the effect of the VSWGs on the damping of system inter-area EOM was “limited
and varied” [1]. The approach adopted in [2] was different, where a SG was
displaced by a DFIG. Damping of inter-area EOMs of the example system with
the displaced SG and the displacing DFIG was compared. It was found that the
maximum difference of the damping was about 15% which was significant. In [3, 4],
the focus of study was the impact of change of system load flow caused by the grid-
connected VSWGs. The results show that the impact on the damping of system
oscillation modes was very small [3] and could be positive or negative [4].
Grid connection of a VSWG to a power system changes the system load flow. The
change brought about by the VSWG is balanced by the power output from the SGs at
the slack nodes in the power system. This is equivalent to having the power output
from the VSWG to replace that of the SGs. Under certain circumstance, some SGs
may withdraw from the power system to give way of generation to cleaner wind
power. Earlier work presented in [1–4] represented some of the features of the
consequence of grid connection of the VSWG. Slootweg and Kling [1] investigated
the case with the power output from the VSWGs to replace that of the SGs at same
locations in power systems. Thus, load flow of example power systems did not
change. Sanchez-Gasca et al. [2] examined the case of displacement of the SGs by
the VSWGs at the same locations. Load flow of example power systems remained
the same after the displacement. While the case considered in [3] and [4] was the
addition of the VSWGs in the example power systems. The change of load flow of
the power systems was balanced by the power output from the SGs at the slack bus.
Those three different operational scenarios of the grid connection of the VSWG can
be illustrated by Figs. 1.1, 1.2, and 1.3.
From the practical point of view of grid connection of wind power generation, the
case of Fig. 1.3 is more general. Figure 1.1 represents the case of wind-thermal
Fig. 1.1 Case considered in

[1]: Pw0 replacing Pg0
Pw0
VSWG A power
Pg0 system
SG
1.1 Impact of Grid-Connected VSWGs on the Small-Signal Angular Stability. . . 3
A power A power
system system
Node A Node A
SG VSWG
With the displaced SG With the displacing VSWG
Fig. 1.2 Case considered in [2]: the VSWG replacing the SG
Fig. 1.3 Case considered in VSWG

[3, 4]: Addition of a VSWG
in the power system A power
system
Pw0 + jQw0
bundled power generation and supply. The case shown by Fig. 1.2 is rare in practice
but useful for understanding the mechanism about how the VSWG affects power
system small-signal angular stability differently with the SG. In principle, the cases
considered in [1] (Fig. 1.1) and [3] (Fig. 1.3) are same that the wind power
generation in Fig. 1.3 replaces the power output from the SGs in the slack node in
the power system. Their difference is that the case considered in Fig. 1.1 may bring
about much less change of system load flow than that in Fig. 1.3. Hence Fig. 1.1 may
be considered as a special case of Fig. 1.3.
For years, the investigation on the power system small-signal angular stability as
affected by grid-connected VSWGs has been carried out in the similar ways to that
used in [1–4]. Mainly two types of grid connection of wind power generation have
been considered. The first one is the displacement of the SGs by the VSWGs at the
same locations as shown by Fig. 1.2. The other is the simple addition of the VSWGs
in the power system as shown by Fig. 1.3.
1.1.1.2 Representatives Study: The VSWG Displacing the SG
Strategy of investigation via a VSWG displacing a SG in a power system is shown

by Fig. 1.2. The objective of investigation is to compare the dynamic function
between the VSWG and the SG so as to understand the mechanism on how the
VSWG affects the small-signal angular stability of the power system differently to
the conventional SG.
In [5, 6], relatively more complicated example power systems were used to
examine the effect when a grid-connected DFIG displaced a SG. Computational
results of modal analysis in [5, 6] indicated that the effect could be small or large as
well as positive to enhance or negative to reduce the damping of power system
EOMs. The results were dependent of the locations and dynamic importance of the
SGs being displaced as well as the load conditions of the example power systems.
4 1 Introduction
Gautam et al. [7] is one of the most representative work adopting the approach of
the displacement. Authors of [7] thoughtfully pointed out that a DFIG affects power
system angular stability in four ways:
1. “displacing synchronous machines thereby affecting the modes;
2. impacting major path flows thereby affecting the synchronizing forces;
3. displacing synchronous machines that have power system stabilizers;
4. DFIG controls interacting with the damping torque on nearby large synchronous
generators” [7].
Approach proposed and used in [7] is “to convert the DFIG machines into
equivalent conventional round rotor synchronous machines and then evaluate the
sensitivity of the eigenvalues with respect to inertia. In this regard, modes that are
both detrimentally and beneficially affected by the change in inertia are identified”
[7]. Results presented in [7] indicated that the effect of displacement can be
detrimental or beneficial to the system small-signal angular stability. The results
were highly related with the locations of displacement and load conditions of power
systems.
Vittal et al. [8] and Vittal and Keane [9] further extended the strategy of
displacement to investigate the effect of supplementary reactive power/voltage
control of DFIGs on power system small-signal angular stability. Authors of [8, 9]
considered that “the reactive power control strategy of the wind farm can have a
significant impact on the rotor angle stability of conventional synchronous units in
the system. Controlling reactive power to achieve a targeted voltage can improve the
rotor angle stability of the existing synchronous units in the system”. It was
hypothesized in [9] “that when the synchronous wind farms with high participation
factors are replaced by a VSWT wind farm with terminal voltage control enabled, the
rotor angle stability of the system will be improved”. However, the hypothesis was
not generally proved to be true in [9].
1.1.1.3 Representatives Study: Adding the VSWG in the Power System
Strategy of investigation on the case that a VSWG is simply added in a power system
is illustrated by Fig. 1.3. As it is pointed out previously, Fig. 1.1 could be a special
case of Fig. 1.3. Tsourakis et al. [10] presents the study of a simple single-machine
infinite-bus power system connected with two DFIGs for wind power generation.
Power output from the DFIGs increased with the load and thus the power output
from the SG remained unchanged. Under this circumstance of system operation,
results of modal analysis indicated that the DFIGs with reactive power control or
voltage control affected the system oscillation modes differently with the change of
penetration level of the wind. Oliveira et al. [11] studied a Brazilian real power
distribution network. Results in [11] showed that different operational mode of the
DFIG, i.e., power factor control mode or voltage control mode, did not affect
significantly the electromechanical oscillations.
Quintero et al. [12] used a large-scale real power system as an example to study
the effect of converter control-based generators (CCBGs) on system small-signal
angular stability. The focus of study in [12] was the participation of CCBGs in power
system electromechanical oscillation modes so as to examine the scale of dynamic
interactions between the CCBGs and synchronous generators (SGs). Results in [12]
indicated that at higher level of wind penetration (13.2%), participation of CCBGs in
the electromechanical oscillation modes was observable. It was found that there were
oscillation modes caused by the CCBGs which were highly sensitive to the control
parameter variations of the CCBGs. Those modes were name as converter oscillation
modes. It was reported that when the converter oscillation modes were of the
frequency within the range of frequency of power system electromechanical oscil-
lation modes, the SGs participated considerably in some of those converter oscilla-
tion modes.
Ding et al. [13] studied the impact of adding a VSWG (DFIG and PMSG) into a
power system and presented the comparative results of modal analysis of an example
power system with the VSWG using models of different level of details. Main
conclusion made in [13] was that when the VSWG was modelled as a constant
power source, the results of modal analysis were close to that when dynamic models
of VSWG were used.
Knuppel1 et al. [14] and Kunjumuhammed et al. [15] examined both cases of
displacement by and addition of the VSWGs using example power systems. Results
in [14] showed that without the displacement, the participation of the SGs in the
example power system did not change. While with the displacement, it changed
significantly. Work in [15] used two-area test system and a representative Great
Britain test system. Various possible operating scenarios with different active power
combinations between synchronous generators and DFIGs, in some cases synchro-
nous generator disconnected, were examined. It indicated that the reduction in output
of synchronous generators with high participation in the power system electrome-
chanical oscillation modes resulted in higher improvement in modes’ damping, and
the change in inter-area power flow caused the change of frequency of inter-
area EOMs.
1.1.2 Impact of the VSWGs: Discussions
1.1.2.1 Strategy of Adding the VSWG in the Power System
This strategy reflects the real scenario that a VSWG is connected to a power system.
Investigation is to compare the system EOMs before and after the VSWG is added as
it did in [10–15]. The strategy can be illustrated by a simple power system shown by
Fig. 1.4 when no VSWG is connected.
After a VSWG is added to the power system of Fig. 1.4, the system configuration
is shown by Fig. 1.5. Obviously, the addition brings about two changes that may
affect system small-signal angular stability. Firstly, addition of the VSWG changes
6 1 Introduction
Fig. 1.4 A single-machine Infinite bus

infinite-bus power system SG1
Fig. 1.5 Addition of a Infinite bus

VSWG in power system of SG1
Fig. 1.4
Pw + jQw
VWSG
the load flow condition and configuration of the power system. Consequently,
system small-signal angular stability may change accordingly. This affecting factor
is well demonstrated by the results presented in [13]. When the VSWG was
modelled as the constant power source, effect of the VSWG dynamics was not
included [13]. Grid integration of the VSWG only changed the system load flow
condition to affect the damping of the power system EOMs. Results of case study in
[13] indicated the improvement of the damping of the EOMs when wind penetration
increased.
Secondly, addition of the VSWG introduces dynamic interactions of the VSWG
with the SG in the power system which could affect system small-signal angular
stability. In [13], results of modal computation with the VSWGs’ dynamics being
considered were compared with that with the VSWGs being modelled as constant
power sources (hence dynamic interactions were excluded). Comparison indicated
that both results were very close. This meant that the impact of the VSWG’s dynamic
interactions was small. Results of comparison presented in [13] fitted the usual
recognition on the dynamic effect of the VSWGs, because it is normally believed
that VSWGs are inertia-less due to the function of the VSC control and hence the
VSWGs’ dynamic interactions with SGs are weak. Results in [14] indicated that the
addition of the VSWGs did not change the participation of the SGs in the EOMs.
This implied that the dynamic impact of the SGs was not affected by the VSWGs’
dynamics. Hence results in [14] indirectly confirmed the recognition that normally
the VSWGs dynamically interact with the SGs weakly and affect little the small-
signal angular stability of the power system.
However, results presented in [12] challenged the normal recognition that the
VSWGs’ dynamic involvement in the EOMs is small. When the wind penetration
was high, [12] found the existence of considerable dynamic interactions between the
VSWGs and the SGs from the computational results of the participation factors (PFs)
of the EOMs [12]. The findings reported in [12] implied that the dynamic interac-
tions introduced by the VSWGs may become strong under certain special conditions
which may affect power system small-signal angular stability significantly.
In the fifth chapter of the book, the special conditions under which the dynamic
interactions between the VSWGs and the SGs may become strong are investigated.
The impact of the strong dynamic interactions introduced by the VSWGs is
examined.
1.1.2.2 Separate Examination of Affecting Factors
From the discussions in the above subsection, it can be concluded that when a
VSWG is added in a power system, the effect of addition on system small-signal
angular stability is the mixed result of two affecting factors: load flow change and
dynamic interactions brought about by the VSWG. Both load flow change and
dynamic interactions vary with the change of power output from the VSWG (wind
penetration level). Results of investigation on the effect of adding the VSWG in the
power system obtained so far have been varied. For example, [3, 4] indicated that
effect of change of power output from the VSWGs was small; but according to [14] it
was not. The reasons of such diversification are in two folds. Firstly, those results
were dependent of example power systems used and their operating conditions.
Secondly, they were mixed effect of two affecting factors. Strict theoretical analysis
can provide example-independent results and conclusions. At present if this is
difficult, separate examination of two affecting factors is a correct way forward
towards more detailed and elaborated investigation.
Quintero et al. [12] adopted the conventional tool of the PFs to examine the
dynamic engagement of the VSWGs in power system dynamic and vice versa. It was
an attempt to individually investigate the affecting factor of dynamic interactions
between the VSWGs and the SGs. However, the PFs can only estimate the scale of
dynamic involvement, not the direction to be positive to improve or negative to
reduce the power system small-signal angular stability. This inherent drawback of
the PFs may limit their applications in the investigation [13] used model of constant
power source to represent the VSWG. This model effectively enabled the separate
examination of the effect of power flow change introduced by the VSWG.
In [16], a method of separate examination of two affecting factors of grid
connection of the DFIGs on power system small-signal angular stability was pro-
posed. The method considered the dynamic variation of power exchange between a
grid-connected DFIG and power system to be ΔPw þ jΔQw. A closed-loop state
space model of the power system was established as shown by Fig. 1.6, where
the DFIG was treated as the feedback controller. Based on the model, impact of
the DFIG’s dynamics on an EOM of interests was estimated to be Δλ by use of the
damping torque analysis (DTA). In the third and fourth chapter of the book, how
the DTA is applied to examine the impact of the DFIG is introduced and demon-
strated in details.
8 1 Introduction
Fig. 1.6 Closed-loop state

Power d DVpcc
space model of power ΔXg = Ag ΔXg
system with DFIG [16] system dt
DPw
GP (s)
DFIG
DQw
GQ (s)
When there is no dynamic interaction between the DFIG and power system,
ΔPw þ jΔQw ¼ 0. The DFIG is degraded into a constant power source Pw0 þ jQw0
and the system is represented by its open-loop model. Hence it was concluded in [16]
that with the constant power model of the DFIG, effect of load flow change
introduced by the DFIG can be determined by calculating the EOM from system
open-loop state matrix Ag to be λ0. Thus, two affecting factors of the DFIG can be
examined separately. The total impact is Δλ þ λ0 ¼ λ.
It was derived in [16] that the dynamic variation of power exchange between the
DFIG and power system was ΔPw ¼ GP(s)ΔVpcc, ΔQw ¼ GQ(s)ΔVpcc where ΔVpcc
is the dynamic variation of magnitude of voltage at the point of common coupling
(PCC) of the DFIG. The derivation indicated that the dynamic variation of power
exchange should normally be small, because in power system ΔVpcc is usually small.
Hence normally the impact of the DFIG’s dynamics, Δλ, is small. Estimation of the
impact from the constant power model of the DFIG should approximately be equal
to the total impact, i.e., λ0 λ. This conclusion made in [16] confirmed the results
presented in [13] and is useful in practical applications. For example, by use of the
constant power model of the DFIG, the most dangerous scenarios of the grid
connection of the DFIG can be identified. This will be very helpful in planning the
grid connection of the wind farms when the dynamics of the wind farms are often
unknown but their impact needs to be predicted. However, feasibility of such
application still needs to be investigated. The obstacle is the finding reported in
[12] that the considerable dynamic interactions between the VSWG and the power
system may occur. Method proposed in [16] may be helpful in overcoming the
obstacle because the estimation of the impact of the dynamic interactions between
the VSWG and the power system by use of Δλ can tell not only the scale of impact as
the PFs do (which was used in [12]), but also the direction of impact. For example, if
the impact of dynamic interactions is predicted to be positive, in this case, the
constant power model of the DFIG can definitely identify the most dangerous
scenarios of grid connection of the DFIG. In the fourth chapter of the book, the
method proposed in [16] is introduced and demonstrated.
1.1.2.3 Strategy of the VSWG Displacing the SG
A VSWG displacing a SG at a same location rarely happens in practical power

systems. However, it is considered that this strategy of investigation is helpful in
Fig. 1.7 Power system with Infinite bus

SG2 to be displaced by a SG1
VSWG as shown by Fig. 1.5
DPg + jDQg
SG2
understanding the difference between the dynamic function of the VSWG and the
SG. The strategy can be explained by use of the simple power system of Fig. 1.5 with
the VSWG being displaced by SG2 as shown in Fig. 1.7. Modal analysis can be
carried out to compare the small-signal angular stability of system shown by
Figs. 1.5 and 1.7 to examine the impact of displacement.
Obviously, load flow condition of system shown by Figs. 1.5 and 1.7 can be kept
exactly same. The VSWG displacing the SG can happen without causing any change
of system load flow condition and configuration. Thus the affecting factor of the load
flow change caused by the grid connection of VSWG is excluded. The difference of
the impact of the displacement is from the different dynamics of the VSWG and the
SG. Results of investigation by the displacement so far were also diversified as
reviewed in the previous section [2, 4–7]. However, it has been found that the
dynamic importance of the SGs being displaced seemed playing more role in
determining the results of the displacement [5, 6, 15].
In fact, the displacement can be considered being implemented in two separate
steps. First, the dynamics of the displaced SG are withdrawn from the power system.
Second, the dynamics of the displacing VSWG are added. Hence, there are two
factors affecting the impact of the VSWG displacing SG on power system small-
signal angular stability. The first affecting factor is the dynamics of the SG and the
second is the dynamics of the VSWG. If the impact of the VSWG’s dynamics is
small as indicated by the results in [13, 16], the dynamics of the displaced SG are
more influential on the total impact of the displacement as reported in [5, 6, 15]. In
addition, if the impact of those two affecting factors is in the opposite direction, their
impact may cancel each other to result in little impact. Therefore, indeed the problem
is complicated and it is difficult to make a definite conclusion from the displacement.
In order to carry out in-depth investigation on the impact of the displacement of
the SG by the VSWG so as to understand the difference between the SG’s and the
VSWG’s dynamic function, it looks that methods to rank the dynamic importance of
the SGs need to be developed. The PFs can give indication of dynamic involvement
of the SGs in the EOMs of interests. However, their relationship with the dynamic
importance of the SGs has been unclear. In addition, they cannot give the indication
of direction of the impact.
10 1 Introduction
Method proposed in [7] to calculate the modal sensitivity to the constant of inertia
of the displaced SG was an attempt to develop an index to predict the impact of the
displacement without knowing the dynamics of the displacing VSWG. The index
can provide some aspects of dynamic importance of the SGs. However, some
important aspects of the SGs’ dynamics, such as the function of the automatic
voltage regulators (AVRs) and the power system stabilizers (PSSs), are excluded
from the index. In addition, application of the index is based on the assumption that
the relation between the EOMs of interests and the constant of inertia of the SGs is
linear which may not be true sometimes.
1.1.2.4 Impact of the VSWG Reactive Power/Terminal Voltage Control
A grid-connected VSWG can adopt the fixed reactive power output control or fixed
terminal voltage control. Impact of those two types of control on power system
small-signal angular stability may be different. This important issue was discussed in
[8–11] with diversified results. Conclusion made in [8, 9] was that the terminal
voltage control was more beneficial to system small-signal angular stability. Results
from [10] indicated that the impact was variable and dependent of system operating
conditions. It was concluded in [11] that the difference of impact between the
reactive power and terminal voltage control was small.
When the VSWG operates with the fixed reactive power output or the fixed
terminal voltage control, reactive power output from the VSWG at steady-state
operation is different. Thus, change of system load flow condition brought about
the VSWG is different when the reactive power or the terminal voltage control is
used. In addition, dynamic interactions introduced by the VSWG should also be
different. Hence, it is a problem mixed with two affecting factors (load flow
condition and dynamic interactions) and more careful comparison needs to be
carried out by examining each of the affecting factors individually. The comparison
is important in planning the grid connection of wind farms. Hence, it is more
desirable that the comparison is carried out without knowing the dynamics of
wind farms and their associated control systems. In the fourth chapter of the book,
an index to predict the difference of the impact between the reactive power and
terminal voltage control is introduced and evaluated.
1.1.2.5 Effect of the Phase Locked Loop (PLL)
Vector control is the most widely implemented scheme by the VSCs for the grid
connection of the VSWGs. Implementation of vector control relies on the phase
tracking of the PCC voltage at the VSWGs’ terminal. A phase locked loop (PLL) is a
control system being specially designed to fulfil the task of phase tracking of the
voltage. It plays a very important role for the grid connection of the VSWGs.
Functionally, the PLL links the VSWG with the power system. Hence, the impact
of the PLL should be in two folds. On one hand, dynamics of the PLL will affect the
1.2 Small-Signal Stability of the VSC-Based DC/AC Power System 11
performance of grid-connected VSWG. Recent study has shown that when the
VSWG is weakly connected with the power system, the PLL’s dynamics may
impose negative impact on the VSC-based system. Improper tuning of the PLL’s
parameters could even cause the small-signal instability of the VSWG itself. Hence,
careful tuning of the PLL’s parameters is needed to consider its effect on the
operational stability of the VSWG. In the study, the power system is represented
by a three-phase voltage source connected to the VSWG via transmission imped-
ance. Dynamics of power system are not considered. In Sect. 1.2.1, study on the
small-signal stability of grid-connected single VSC system under weak system
condition is reviewed.
On the other hand, dynamics of the PLL may also affect power system dynamic
performance, for example, the small-signal angular stability. So far, impact of the
VSWGs on power system small-signal angular stability has been examined by
assuming that the PLL dynamic performance is perfect, which can provide accurate
and instantaneous phase tracking of terminal voltage of the VSWGs. Thus, the
examination has been carried out under the known relative position of d-q coordinate
of the VSWGs to that of power system common coordinate without the effect of the
PLL’s dynamics being considered. Error of the PLL phase tracking has not been
included in the examination. Normally, ignorance of the phase tracking error of the
PLL may be acceptable if the PLL is designed properly. However, unusual cases
may occur if parameters of the PLL are not tuned properly.
In [17], impact of the PLL’s dynamics on power system small-signal angular
stability was examined. Results of modal analysis based on an example power
system were presented to indicate that a high-gain and fast-acting PLL affected little
the electromechanical dynamics of the power system. However, when the frequency
of the PLL oscillation mode was close to that of an EOM of the power system and at
the same time the damping of the PLL oscillation mode was light, the PLL partic-
ipated the power system electromechanical dynamics significantly. This implied
considerable dynamic interactions between the PLL and the power system. The
phenomenon reported in [17] is examined in the sixth chapter of the book.
1.2 Small-Signal Stability of the VSC-Based DC/AC Power

System
Voltage source converter (VSC) based HVDC lines and multi-terminal DC (MTDC)
networks are of many technical advantages over the conventional LCC HVDC lines.
They are the favorable solution to the large-scale transmission of renewable power
generation, such as the offshore wind power. In addition, the VSCs are the key
components for the grid connection of renewable power generation. Hence, small-
signal stability of an AC power system integrated with the VSCs has been a topic of
study attracted interests of many power system researchers and engineers. The
potential instability risk that might be brought about by the integration of the
12 1 Introduction
VSCs, the VSC-based HVDC line or MTDC network, has been investigated in
recent years from various standpoints.
There have been two main concerned problems about the small-signal stability of
an integrated VSC-based DC/AC power system. The first one is the small-signal
stability of the VSC-based system itself, such as a single grid-connected VSC system
or a MTDC network. The second problem is the small-signal stability of the
integrated VSC-based DC/AC power system, such as the AC power systems inte-
grated with the renewable power generation, the VSC-HVDC/AC or the MTDC/AC
power systems. The key issue is how the grid connection of the VSCs may affect the
small-signal stability of the integrated VSC-based DC/AC power system. Those two
problems are not covered by the conventional theory and analysis of small-signal
stability of normal AC power systems, where the dominant dynamic components are
the synchronous generators (SGs). In the VSC-based DC/AC power system, the
most important dynamic components are the VSCs, in addition to the SGs. Dynamic
interactions among the multiple VSCs and SGs determine the small-signal stability
of the VSC-based DC/AC power system. However, the conventional stability theory
and analysis of AC power system did not tell how the VSCs may interact among
themselves and with the SGs to affect the small-signal stability of the VSC-based
DC/AC power system.
This section reviews the research progress and main conclusions obtained so far
about the small-signal stability of three types of integrated VSC-based DC/AC
power systems. They are the grid-connected single VSC system, the AC power
system integrated with the VSC-based HVDC line and the AC power system
integrated with the MTDC network.
1.2.1 Small-Signal Stability of a Single Grid-Connected VSC

System
A single VSC system is the simplest VSC-based DC/AC power system. It could be a
grid-connected wind farm which is represented by a single VSWG being connected
by the VSC to the AC power system. It could also be a VSC of the HVDC line or the
MTDC network. Study on the small-signal stability of the single VSC system is
about the stability of the VSC system itself without considering the impact of its
dynamic interactions with the SGs in the external power system. The small-signal
stability of the single VSC system is found being mainly affected by the converter
control and the PLL.
1.2.1.1 Main Affecting Factors: Weak Grid Connection and the PLL
Figure 1.8 shows the configuration of a single grid-connected VSC system, where
the external power system is represented by an infinite busbar at node b. An
VSC
Renewable power
generation or the Exteral
HVDC line or the power system
MTDC b
PWM
Inner loops PLL
Outer loops Power

calculation
Fig. 1.8 Configuration of a single grid-connected VSC system
important factor affecting the small-signal stability of the VSC system is the short
circuit ratio (SCR). Denote Prated as the rated power of the VSC system. The SCR is
defined to be
Vpcc 2
SCR ¼ ð1:1Þ
XL Prated
When the SCR is smaller than 3, connection of the VSC with the AC power
system is considered being weak. It has been found that under the condition of weak
grid connection, the VSC system may suffer from the problem of small-signal
instability.
Early study in [18–20] based on the state-space model of the VSC system by
using the modal analysis indicated that the small-signal stability of the VSC system
was affected by the dynamics of the PLL. Maximum power transmission capability
decreased when the SCR was reduced. In [21, 22], Lennart Harnefors examined
comprehensively the small-signal stability of the VSC system from the perspective
of control system parameter conditions. The examination was conducted by using
the passivity theory on the basis of input impedance matrix model of the VSC
system. Results of examination demonstrated that increase of PI gains of current
control inner loops of the VSC may cause system instability. In addition, high
bandwidth of the DC voltage control outer loop and the PLL was unfavorable to
the small-signal stability of the VSC system.
Long-distance transmission of large-scale wind power was found to be prone to
instability problem [23–25]. Study cases by modal analysis in [26] showed that when
the SCR was reduced below a certain value, dominant poles of the PMSG system
moved into the right half of the complex plane. Computation of participation factors
indicated that unstable dominant poles were associated with the AC voltage control
outer loop of the grid side converter (GSC). Similar conclusions were obtained from
the study cases in [27]. However, main source of system instability was pointed to
14 1 Introduction
the PLL in [27]. Results presented in [18–27] consistently confirmed that the weak
grid connection was the key condition under which the VSC system may lose the
small-signal stability.
Investigation in [28–30] was focused on finding the source of instability and
confirmed that the PLL was the main dynamic component to cause the instability of
the VSC system when the grid connection was weak. Based on the quasi-steady state
model of the PLL, impact of transmission impedance and load impedance on the
low-frequency dynamics of the PLL was investigated comprehensively in
[28, 29]. Results of investigation showed that when the transmission impedance
was high, the VSC system may suffer from the problem of growing oscillations.
Results of root loci computation in [30] clearly indicated that when the SCR was
small, increase of PI gains of the PLL led to the instability of the VSC system. Work
in [28–30] confirmed the role played by the PLL to cause the loss of small-signal
stability of weakly grid-connected VSC system. In [31], investigation based on the
impedance model of the VSC system indicated that the range of negative resistance
of the VSC impedance increased when the bandwidth of the PLL increased. The
investigation attributed the impact of the PLL to the contribution of negative
resistance.
Work presented in [28–31] not only confirmed that the main factors affecting the
small-signal stability of the VSC system are the weak grid connection and PLL’s
dynamics, but also represented a series of effort to reveal the mechanism about how
the PLL affects the small-signal stability of the VSC system. In order to understand
why the PLL may affect the small-signal stability considerably, the impact of the
dynamic interactions introduced by the PLL has become a special topic of study. It
has been considered that the impact of dynamic interactions introduced by the PLL
may be the in-depth reason and can reveal the mechanism about why the VSC
system may lose the small-signal stability.
1.2.1.2 Impact of Dynamic Interactions Introduced by the PLL
Dynamic interactions are the mutual exciting and responding of two dynamic
components in a system. A good illustration is the closed-loop system consisted of
two open-loop subsystems shown by Fig. 1.9. Subsystem A responds to the excita-
tion from subsystem B, u. At the same time, its output, y, is the excitation to
subsystem B. If u ¼ 0 or y ¼ 0, dynamic interactions between subsystems are
Fig. 1.9 Closed-loop

y
system G(s)
+
Subsystem A
u Subsystem B
H(
H s)
H(s)
zero such that the closed-loop system is stable as long as the open-loop subsystems
are stable. If u 0 or/and y 0, dynamic interactions between subsystems are weak
and would normally affect little the closed-loop system stability. Only the existence
of strong dynamic interactions may cause the system instability. Hence, by examin-
ing the degree of dynamic interactions introduced by the PLL, it is possible to find
why the PLL may bring about the small-signal instability risk to the VSC system.
The bandwidths of power control outer loops and current control inner loops of
the VSC system in Fig. 1.1 are normally around 10 Hz and 100 Hz respectively.
Hence, usually the dynamic interactions between the outer and inner control loops
are weak and affect little the small-signal stability of the VSC system. The band-
width of the PLL can vary in a wide range to possibly interact with either the outer
loops or inner loops of VSC control. Hence, the dynamic interactions between the
PLL and either the power control outer loops or current control inner loops may
cause the small-signal instability of the VSC system.
Study in [32] demonstrated that the PLL and outer loops of VSC control
participated in the dominant oscillation modes of the VSC system which became
unstable when the grid connection was weak. This indicated that the dynamic
interactions between the PLL and outer loops caused system instability. In [33],
damping torque analysis was applied to examine the impact of dynamic interactions
between the DC and AC voltage control outer loop. Results of examination revealed
that when the lag caused by voltage measurement was considered, the AC voltage
control outer loop may contribute negative damping torque to the DC voltage control
outer loop. Contribution of negative damping torque increased when the SCR was
reduced. Results of modal analysis in [34] showed that under the condition of weak
grid connection, dynamic interactions between the active, reactive control outer loop
and the PLL became strong.
Small-signal stability of a grid-connected PMSG was examined in [35]. The
examination indicated that the dynamic interactions between the PLL, DC and AC
voltage control outer loop were harmful to the stability of the PMSG system as the
interactions were similar to the function of positive feedback. In addition, when the
bandwidth of the PLL and the DC voltage control outer loop was close, the degree of
dynamic interactions was maximum.
Dynamic interactions between the PLL and current control inner loops of the
VSC may also affect the small-signal stability of VSC system considerably. Results
of modal analysis and simulation in [36, 37] demonstrated that when the grid
connection was weak, the VSC system may lose the small-signal stability, which
was related to the dynamic interactions between the PLL and current control inner
loops. The small-signal stability of the PMSG with weak grid connection during the
low voltage ride through was examined in [38]. Results of examination showed that
the system stability was dependent of the dynamic interactions between the PLL and
current inner loops. When the bandwidth of the PLL and inner loops was close, the
degree of dynamic interactions increased.
16 1 Introduction
1.2.1.3 Summary: The Single VSC System
Study so far has confirmed the following two main conclusions about the small-
signal stability of the VSC system.
1. Under the condition of weak grid connection, dominant oscillation modes of the
VSC system may move into the right half of complex plane. The PLL is highly
related with the dominant modes. Hence, the PLL may very likely be responsible
for the small-signal instability of the VSC system.
2. Under the condition of weak grid connection, the dynamic interactions between
the PLL and either outer loops or inner loops of VSC control may become strong.
The strong dynamic interactions may be detrimental to the small-signal stability
of the VSC system. The strong dynamic interactions occur when the bandwidth of
the PLL and control loops is close.
However, the study has been carried out mainly by numerical computation,
simulation and experiment on the case-by-case basis. Damping torque analysis and
frequency-domain analysis based on the impedance model of VSC system attributed
the instability risk to the occurrence of negative damping torque or negative resis-
tance under the condition of weak grid connection. The analysis was very helpful for
understanding and exposing the insight about why the PLL may bring about the
instability risk under the condition of weak grid connection. However, the analysis
normally was also based on the numerical results on the case-by-case basis. There-
fore, the general applicability of the conclusions still needs the confirmation by
analytical examination, which may reveal the in-depth insight about why the VSC
system may lose the small-signal stability as affected by the PLL under the condition
of weak grid connection.
1.2.2 Small-Signal Stability of Integrated VSC-Based DC/AC

Power Systems
Explanation is presented for Fig. 1.9 above about why the small-signal stability of
the VSC system is determined by the dynamic interactions between the converter
control and the PLL. The explanation is general and applicable to more complicated
AC power system integrated with the VSC-HVDC line and the MTDC network.
In either the VSC-HVDC/AC power system or the MTDC/AC power system,
there are dynamic interactions between multiple VSCs or/and between the VSCs and
the SGs, which affect the system small-signal stability. When the VSC-HVDC/AC
power system is investigated, the main strategy applied to simplify the investigation
has been to examine a portion of the VSC-HVDC/AC power system by ignoring
some of dynamic interactions. When the MTDC/AC power system is examined,
strategy of simplification has been to mainly consider either the MTDC or the AC
network. With the simplification, the examination has become simpler and
A B VSC-1 C D VSC-2 E F
SG-1 SG-2
DC/AC DC/DC interactions between DC/AC

interactions VSC-1 and VSC-2 interactions
Fig. 1.10 Configuration of an AC power system integrated with a VSC-based HVDC line
manageable. However, the applicability of the conclusions obtained is constrained

by the assumptions of simplification made in the examination.
1.2.2.1 Small-Signal Stability of AC Power System Integrated

with a VSC-Based HVDC Line
Figure 1.10 shows the configuration of an AC power system integrated with a

VSC-based HVDC line. Without loss of generality, it can be assumed that VSC-1
uses the DC voltage control and VSC-2 adopts the active power control. Although
the point-to-point VSC-based HVDC line is the simplest DC power transmission
system, the DC/AC power system shown by Fig. 1.10 is much more complicated
than the single VSC system shown by Fig. 1.8. In addition to the various dynamic
interactions within each of two VSCs, following dynamic interactions may also
affect the small-signal stability of the DC/AC power system shown by Fig. 1.10:
(1) the DC/AC interactions between VSC-1 and SG-1; (2) the DC/AC interactions
between VSC-2 and SG-2; (3) the DC/DC interactions between VSC-1 and VSC-2.
Hence, examination of small-signal stability of the AC power system with the
VSC-based HVDC line is much more difficult. The difficulty is not in assessing
the system stability, as the modal analysis can always be applied. For example,
results of modal analysis in [39, 40] indicated that the integration of the VSC-HVDC
line may impose positive or negative impact on the system small-signal stability to
improve or reduce the damping of system sub-synchronous oscillation (SSO) modes.
The difficulty is to investigate why and how the VSC-HVDC line may affect the
small-signal stability of integrated power system, for example, the damping of the
SSOs. For the investigation, the frequency-domain analysis can be used with some
simplifications made to the DC/AC power system.
In [41, 42], the small-signal stability of the VSC-HVDC line system between
point B and E in Fig. 1.10 was studied. In the study, the VSC-based HVDC line
system was divided into two parts at point C, the part with VSC-1 and that with
VSC-2 separately. The output impedance of VSC-1 and input impedance of VSC-2
were derived. By using the Middlebrook stability criterion, the system small-signal
stability was assessed. Stability region with the variation of parameters of
VSC-based HVDC system was calculated and validated by simulation. Obviously,
study in [41, 42] considered only the DC/DC dynamic interactions between VSC-1
18 1 Introduction
and VSC-2. The impact of DC/AC interactions between the SGs and VSCs were not
examined.
In [43], the DC/AC power system between point D and F in Fig. 1.10 was
examined. The DC voltage across C2 was assumed being constant such that the
dynamics of HVDC line and AC system on the left of point D were ignored. Results
of investigation in [43] showed that the system was always stable even if the grid
connection of VSC-2 was weak. This implied that the DC/AC dynamic interactions
between SG-2 and VSC-2 with active power control affected little the small-signal
stability of the DC/AC power system shown by Fig. 1.10. The implication is useful,
because point E can be simply assumed to be an infinite busbar. The assumption
should not change the small-signal stability of the DC/AC power system shown by
Fig. 1.10.
The DC/AC power system shown by Fig. 1.10 between point A and E was
considered in [44]. VSC-2 was modelled as a constant power source and hence
dynamic interactions between VSC-2 and SG-2 was ignored. This simplification was
allowable as being confirmed by [43]. A closed-loop model was derived with two
open-loop subsystems. One open-loop subsystem was comprised of control systems
of VSC-1 and SG-1, which represented the DC/AC dynamic interactions. The other
open-loop subsystem consisted of DC line. Interface variables between two open-
loop subsystems were variations of active power output from VSC-1 and DC voltage
across C1. Complex torque analysis was applied to examine the small-signal stability
of the AC power system with the VSC-based HVDC line. Results of examination
indicated that when the load on the DC line, length of DC line or the bandwidth of
the VSC control increased, stability of the AC power system with the VSC-based
HVDC line may degrade to lead to the system instability.
Frequency-domain analysis attributes the degradation of small-signal stability of
the AC power system integrated with the VSC-based HVDC line to the contribution
of negative damping torque or negative resistance from the VSC-HVDC line. Hence,
the analysis provides the insight into why the VSC- HVDC line may affect power
system stability negatively. However, the frequency-domain analysis is applied on
the basis of the closed-loop model. In order to derive the closed-loop model, an
appropriate point of separation in the system needs to be selected to divide the
system into two interconnected open-loop subsystems. In [45, 46], point C in
Fig. 1.10 was chosen and the derived closed-loop model was single-input and
single-output (SISO) such that the frequency domain analysis can be applied.
Normally, assumptions of simplification were made in order to derive a SISO
closed-loop model if the point of division was not particularly selected. Typical
such assumptions were to examine a portion of the DC/AC power system shown by
Fig. 1.10 and ignored some of VSC control functions [41–46]. When all the dynamic
interactions introduced from the entire VSC-HVDC line are considered, the closed-
loop model of the power system is multiple-input and multiple-output (MIMO)
[47, 48], which would be challenging for the frequency-domain analysis to handle.
In [49], the impact of SSO dynamic interactions introduced by a control loop of a
VSC in the AC power system integrated with the VSC HVDC line was investigated.
A SISO closed-loop model was derived, where the control loop of the VSC was the
Interactions through the Interactions through the AC

MTDC network A power system
VSC-1 V1
SG-1
VSC-2 V2
DC AC
...
Network ... ... Network SG-2
...
...
VSC-M VM
SG-n
Fig. 1.11 Configuration of an AC power system integrated with a VSC-based MTDC network
subsystem in the feedback loop and rest of the power system was the subsystem in
the forward path. The investigation indicated that if an open-loop SSO mode of the
subsystem of the control loop of the VSC was close to an open-loop SSO mode of
the subsystem of the rest of the power system on the complex plane, the small-signal
stability of the closed-loop power system was degraded as caused by the control loop
of the VSC. The closeness of open-loop SSO modes was referred to as the open-loop
modal coupling. Study cases were presented in [49] to show that under the condition
of open-loop modal coupling, strong dynamic interactions occurred between the
VSC and the SG to cause the loss of system small-signal stability.
1.2.2.2 Small-Signal Stability of AC Power System Integrated

with a VSC-Based MTDC Network

VSC-based MTDC network. The DC/AC power networks are connected via multi-
ple VSCs. Small-signal stability of the MTDC/AC power system is affected by the
AC/DC dynamic interactions between the VSCs as well as between the VSCs and
SGs. The dynamic interactions are very complicated because they are via the MTDC
network and the AC network. In order to examine how those dynamic interactions
affect the small-signal stability of the MTDC/AC power system, following simpli-
fications have been made in the examination.
1. Dynamics of the AC power system were ignored by assuming the AC system to
be an infinite busbar such that only the dynamics of the MTDC network was
20 1 Introduction
a
Interactions Interactions between the
between VSCs VSCs and SGs
VSC-1 V1
... SG-1
... ...
AC
...
Network
...
VSC-M VM
SG-n
b
Interactions Interactions between the
between VSCs VSCs and SGs
VSC-1 V1
SG-1
...
... ...
DC
Network
...
...
VSC-M VM
SG-n
Fig. 1.12 Configuration of simplified MTDC/AC power system. (a) Multi-infeed VSCs/AC power
system. (b) MTDC power system with multi-independent AC systems
considered. The examination was about the small-signal stability of the MTDC
network, the system on the left-hand side of line A in Fig. 1.11. Obviously the
stability is affected only by the dynamic interactions between the VSCs through
the MTDC network.
2. The MTDC network was not considered and the configuration of the MTDC/AC
power system was simplified as shown Fig. 1.12a. This is the multi-infeed VSCs/
AC power system. The system small-signal stability was affected by the dynamic
interactions between the VSCs as well as between the VSCs and SGs through the
AC network.
3. The AC network was ignored and the simplified configuration of the MTDC/AC
power system is shown by Fig. 1.12b. The system small-signal stability was
affected by the dynamic interactions between the VSCs as well as between the
VSCs and SGs through the MTDC network.
Two factors affecting the small-signal stability of a MTDC network are the load
flow condition and the dynamic interactions between the multiple VSCs through the
MTDC network. Study cases presented in [48] showed that when the direction of
load flow in the MTDC network changed, the system may encounter the instability
problem. In [50], load flow optimization was studied in order to increase the stability
of the MTDC network. The dynamic interactions in the MTDC network are related
with the control schemes adopted by the VSCs. Hence, modal analysis in [51]
focused on the effect of droop coefficient on the small-signal stability of the
MTDC network, which implemented the DC voltage droop control. In [52], more
detailed modal analysis assisted by computational results of participation factors,
etc. was conducted to identify the dynamic components in the MTDC network
responsible for the instability risk.
Work presented in [47–52] demonstrated how the modal analysis can be applied
to examine the small-signal stability of the MTDC network. In [53], an impedance
model of the MTDC network was derived. The results of small-signal stability of the
MTDC network by using the frequency-domain analysis indicated that when the
voltage droop control was implemented, impropriate setting of control parameters
may lead to the loss of small-signal stability of the MTDC network.
Dynamic interactions between the VSCs in the MTDC network were examined in
[54] under the condition of open-loop modal coupling when an open-loop oscillation
mode of a selected VSC was close to an open-loop oscillation mode of the rest of
MTDC network. Analysis in [54] indicated that there existed no open-loop oscilla-
tion mode in the VSC which adopted the active power control. Hence, it was
concluded in [54] that when the MTDC network implemented the master-slave
control, there was no possibility of open-loop modal coupling in the MTDC network
to cause the loss of system small-signal stability. However, when the voltage droop
control was used, the open-loop modal coupling may possibly happen between the
VSCs adopting the DC voltage droop control. In [54], study cases were presented to
show that the MTDC network with the DC voltage droop control lost the stability
under the condition of open-loop modal coupling. In the seventh chapter of this
book, detailed analysis about the impact of the open-loop modal coupling is
introduced.
In [55], a simple two-infeed VSC/AC power system shown by Fig.1.12a was
considered. Dynamic interactions between two VSCs through the AC network were
studied and routes of dynamic interactions were identified. Study by frequency-
domain analysis indicated that when the bandwidth of VSC control was close to the
frequencies of high-frequency oscillation modes of AC power system, system
instability may occur. However, the mechanism of such instability risk was not
examined in [55].
In the simplified MTDC/AC power system shown by Fig. 1.12a, the impact of
dynamic interactions between the VSCs through the MTDC network is expected to
be similar to that in the MTDC network being connected to the infinite AC busbar.
Hence, the effect of DC/AC dynamic interactions through the MTDC network
should be the focus of examination. Results of simulation presented in [47, 56]
indicated that when the VSCs used the DC voltage droop control, DC/AC dynamic
22 1 Introduction
interactions between the VSC and SG propagated through the DC network. This
implied that the two independent AC systems (the SGs) may interact via the DC
network. However, no further analysis was carried out in [47, 56] to indicate why the
interactions may happen and what their impact on system stability was. In the
seventh chapter of the book, how the MTDC network may function as the media
of DC/AC dynamic interactions to affect system stability is studied.
In order to handle the complicated MTDC/AC power system shown by Fig. 1.11,
a systematic and simple way to establish the linearized model for the study of small-
signal stability is essentially important. Work presented in [57] is the representative
of pioneering effort along this particular direction of research. In [57], a closed-loop
interconnected model of the MTDC/AC power system was derived, where the
MTDC network and the AC power system were modelled as two separate open-
loop subsystems. The interconnected model derived in [57] made it possible to
examine the dynamic interactions between the MTDC and the AC power system.
In [50, 58], the open-loop subsystems of the VSCs, the SGs, DC and AC networks,
were separately derived and interconnected to establish the model of the MTDC/AC
power system. This not only reduced the complexity in modelling the MTDC/AC
power system considerably, but also linked the system stability with the dynamic
interactions between the VSCs and SGs. Whilst in [59], a generic state-space model
of the MTDC/AC power system was derived. Stability of an example MTDC/AC
system was examined using the derived model. In [60], analysis was carried out to
investigate the effectiveness of damping control implemented by a pair of VSCs in a
MTDC/AC power system. Results of analysis provided insight into the small-signal
stability of the MTDC/AC power system.
Based on the closed-loop model of the MTDC/AC power system, damping torque
analysis was applied in [61] to examine the small-signal stability of the MTDC/AC
power system. Results of analysis indicated that under the condition of open-loop
modal coupling, i.e., an oscillation mode of the open-loop MTDC network being
close to an electromechanical oscillation mode of the open-loop AC power system
on the complex plane, dynamic interactions between the MTDC and AC power
system may become strong to degrade the small-signal stability of the MTDC/AC
power system. Thus, the open-loop modal coupling between the open-loop oscilla-
tion mode of the MTDC network and the electromechanical oscillation mode of the
open-loop AC power system was a modal condition under which the MTDC/AC
power system may lose the small-signal stability. Details of this study are to be
introduced in Chap. 7 of the book.
In [62], dynamic interactions between a selected control loop of a VSC and rest of
a MTDC/AC power system was examined. Due to the selection of the control loop of
the VSC, a SISO closed-loop model of the MTDC/AC power system was
established. Based on the SISO closed-loop model, theoretical proof was presented
to indicate that under the condition of open-loop modal coupling, the MTDC/AC
power system may possibly loss the small-signal stability as caused by the dynamic
interactions between the VSCs and the SGs. This study is also to be introduced in
Chap. 7 of the book.
References 23
1.2.2.3 Summary: The HVDC/AC and MTDC/AC Power System
Key issues in the study of small-signal stability of the VSC-based HVDC/AC or

MTDC/AC power system are the dynamic interactions between the VSCs and
between the VSCs and SGs through both the DC and AC network. Results of
study obtained so far can be summarized as follows.
1. Integration of VSC-HVDC line or MTDC network into an AC power system may
bring about the small-signal instability risk. The instability risk may be caused by
the contribution of negative damping torque or negative resistance from the
VSC-HVDC line or the MTDC network. This is the conclusion obtained by
using the frequency-domain analysis. The instability risk may occur under the
condition of open-loop modal coupling when two open-loop oscillation modes of
subsystems are close to each other on the complex plane. When the open-loop
modal coupling happens, the dynamic interactions between the VSCs and the
SGs may become strong to lead to the loss of small-signal stability of the
integrated HVDC/AC or MTDC/AC power system. This is the explanation
about why the VSC-based HVDC/AC or MTDC/AC power system may become
unstable from the perspective of system open-loop modal condition.
2. In a MTDC network, dynamic interactions between the VSCs adopting the DC
voltage droop control may cause the small-signal instability if the control param-
eters of the VSCs are not set carefully. For the MTDC network connected to the
independent SGs (Fig. 1.5b), AC dynamic transient can propagate through the
MTDC network. Hence, it seems that the MTDC network implementing the DC
voltage droop control is more liable to the loss of small-signal stability because
the DC voltage droop control links the dynamics of the VSCs implementing the
DC voltage control.
3. The AC/DC dynamic interactions introduced by the VSC using the DC voltage
control may become strong to negatively affect the small-signal stability of the
HVDC/AC or the MTDC/AC power system. For the VSC adopting the active
power control, the AC/DC dynamic interactions are weak and affect little the
small-signal stability of the HVDC/AC or the MTDC/AC power system. Hence,
as far as the DC side is concerned, the VSC control systems seem to be the main
instability risk contributors. Design of the VSC control should consider the
impact on power system stability especially when the DC voltage control is used.
References
1. Slootweg JG, Kling WL (2003) The impact of large scale wind power generation on power
system oscillations. Elect Power Syst Res 67(1):9–20
2. Sanchez-Gasca JJ, Miller NW, Price WW (2004) A modal analysis of a two-area system with
significant wind power penetration. In: Proc power systems conf Expo 2, New York, USA, pp
1148–1152
3. Tang H, Wu J, Zhou S (2004) Modelling and simulation for small signal stability analysis of
power system containing wind farm. Power Syst Technol 28(1):38–41
24 1 Introduction
4. Mendonca A, Lopes JAP (2005) Impact of large scale wind power integration on small signal
stability. In: 2005 International Conference on Future Power Systems, Amsterdam, pp 1–5
5. Wang C, Shi LB, Yao LZ, Wang LM, Ni YX (2010) Small signal stability analysis of the large-
scale wind farm with DFIGs. Proc CSEE 30(4):63–70
6. Chen S, Chang X, Sun H, Zeng L (2013) Impact of grid-connected wind farm on damping
performance of power system. Power Syst Technol 37(6):1570–1577
7. Gautam D, Vittal V, Harbour T (2009) Impact of increased penetration of DFIG-based wind
turbine generators on transient and small signal stability of power systems. IEEE Trans Power
Syst 24(3):1426–1434
8. Vittal E, Malley MO, Keane A (2012) Rotor angle stability with high penetrations of wind
generation. IEEE Trans Power Syst 27(1):353–362
9. Vittal E, Keane A (2013) Identification of critical wind farm locations for improved stability and
system planning. IEEE Trans Power Syst 28(3):2950–2958
10. Tsourakis G, Nomikos BM, Vournas CD (2009) Contribution of doubly fed wind generators to
oscillation damping. IEEE Trans Energy Conver 24(3):783–791
11. Oliveira RV de, Zamadei JA, Hossi CH (2011) Impact of distributed synchronous and doubly-
fed induction generators on small-signal stability of a distribution network. In: Power and
energy society general meeting (PES), San Diego, pp 1–8
12. Quintero J, Vittal V, Heydt GT, Zhang H (2014) The impact of increased penetration of
converter control-based generators on power system modes of oscillation. IEEE Trans Power
Syst 29(5):2248–2256
13. Ding N, Lu Z, Qiao Y, Min Y (2013) Simplified models of large-scale wind and their
applications for small-signal stability. J Modern Power Syst Clean Energy 1(1):58–64
14. Knuppel T, Nielsen JN, Jensen KH, Dixon A, Ostergaard J (2012) Small-signal stability of wind
power system with full-load converter interfaced wind turbines. IET Renew Power Gener 6
(2):79–91
15. Kunjumuhammed LP, Pal BC, Anaparthi KK, Thornhill NF (2013) Effect of wind penetration
on power system stability. In: Power and energy society general meeting (PES), Vancouver, pp
1–5
16. Du WJ, Bi JT, Cao J, Wang HF (2016) A method to examine the impact of grid connection of
the DFIGs on power system electromechanical oscillation modes. IEEE Trans Power Syst 31
(5):3775–3785
17. Wang Z, Shen C, Liu F (2014) Impact of DFIG with phase lock loop dynamics on power
systems small signal stability. In: Power and Energy Soc General Meeting, National Harbor, pp
1–5
18. Konishi H, Takahashi C, Kishibe H, Sato H (2001) A consideration of stable operating power
limits in VSC-HVDC systems. In: Seventh international conference on AC-DC power trans-
mission, London, pp 102–106
19. Jovcic D, Lamont LA, Xu L (2003) VSC transmission model for analytical studies. In: IEEE
power engineering society general meeting, vol 3, Toronto, p 1742
20. Durrant M, Werner H, Abbott K (2003) Model of a VSC HVDC terminal attached to a weak AC
system. In: Proceedings of 2003 I.E. conference on control applications, vol 1, Istanbul, Turkey.
pp 178–182
21. Harnefors L, Bongiorno M, Lundberg S (2007) Input-Admittance Calculation and Shaping for
Controlled Voltage-Source Converters. IEEE Trans Ind Electron 54(6):3323–3334
22. Harnefors L, Wang X, Yepes AG, Blaabjerg F (2016) Passivity-based stability assessment of
grid-connected VSCs—an overview. IEEE J Emerg Sel Topics Power Electron 4(1):116–125
23. Rasmussen M, Jorgensen HK (2005) Current technology for integrating wind farms into weak
power grids. In: 2005 IEEE/PES transmission & distribution conference & exposition, Dalian,
China, pp 1–4
24. Yuan X, Wang F, Boroyevich D, Li Y, Burgos R (2009) DC-link voltage control of a full power
converter for wind generator operating in Weak-Grid systems. IEEE Trans Power Electron 24
(9):2178–2192
References 25
25. Liserre M, Teodorescu R, Blaabjerg F (2006) Stability of photovoltaic and wind turbine grid-
connected inverters for a large set of grid impedance values. IEEE Trans Power Electron 21
(1):263–272
26. Strachan NPW, Jovcic D (2010) Stability of a Variable-Speed Permanent Magnet Wind
Generator With Weak AC Grids. IEEE Trans Power Del 25(4):2779–2788
27. Zhou Y, Nguyen DD, Kjær PC, Saylors S (2013) Connecting wind power plant with weak
grid—challenges and solutions. In: 2013 I.E. power & energy society general meeting, Van-
couver, BC, pp 1–7
28. Dong D, Li J, Boroyevich D, Mattavelli P, Cvetkovic I, Xue Y (2012) Frequency behavior and
its stability of grid-interface converter in distributed generation systems. In: 2012 Twenty-
seventh annual IEEE applied power electronics conference and exposition (APEC) 1887–1893,
Orlando, FL
29. Dong D, Wen B, Boroyevich D, Mattavelli P, Xue Y (2015) Analysis of phase-locked loop
low-frequency stability in three-phase grid-connected power converters considering impedance
interactions. IEEE Trans Ind Electron 62(1):310–321
30. Zhou JZ, Ding H, Fan S, Zhang Y, Gole AM (2014) Impact of short-circuit ratio and phase-
locked-loop parameters on the small-signal behavior of a VSC-HVDC converter. IEEE Trans
Power Del 29(5):2287–2296
31. Wen B, Boroyevich D, Burgos R, Mattavelli P, Shen Z (2016) Analysis of D-Q small-signal
impedance of grid-tied inverters. IEEE Trans Power Electron 31(1):675–687
32. Zhou P, Yuan X, Hu J, Huang Y (2014) Stability of DC-link voltage as affected by phase locked
loop in VSC when attached to weak grid. In: 2014 I.E. PES general meeting|conference &
exposition, National Harbor, MD, pp 1–5
33. Huang Y, Yuan X, Hu J, Zhou P, Wang D (2016) DC-Bus voltage control stability affected by
AC-bus voltage control in VSCs connected to weak AC grids. IEEE J Emerg Sel Topics Power
Electron 4(2):445–458
34. Arani MFM, Mohamed YARI (2017) Analysis and performance enhancement of vector-
controlled VSC in HVDC links connected to very weak grids. IEEE Trans Power Syst 32
(1):684–693
35. Huang Y, Yuan X, Hu J (2015) Modeling of VSC connected to weak grid for stability analysis
of DC-link voltage control. IEEE J Emerg Sel Topics Power Electron 3(4):1193–1204
36. Midtsund T, Suul JA, Undeland T (2010) Evaluation of current controller performance and
stability for voltage source converters connected to a weak grid. In: The 2nd international
symposium on power electronics for distributed generation systems, Hefei, pp 382–388
37. Givaki K, Xu L (2015) Stability analysis of large wind farms connected to weak AC networks
incorporating PLL dynamics. In: International conference on renewable power generation (RPG
2015), Beijing, pp 1–6
38. Hu J, Hu Q, Wang B (2016) Small signal instability of PLL-synchronized type-4 wind turbines
connected to high-impedance AC grid during LVRT. IEEE Trans Energy Convers 31
(4):1676–1687
39. Prabhu N, Padiyar KR (2009) Investigation of subsynchronous resonance with VSC-based
HVDC transmission systems. IEEE Trans Power Del 24(1):433–440
40. Joseph T, Ugalde-Loo CE, Liang J (2015) Subsynchronous oscillatory stability analysis of an
AC/DC transmission system. In: 2015 I.E. Eindhoven PowerTech, Eindhoven, pp 1–6
41. Wildrick CM, Lee FC, Cho BH, Choi B (1995) A method of defining the load impedance
specification for a stable distributed power system. IEEE Trans Power Electron 10(3):280–285
42. Wu XG, Sun YF, Li GQ (2016) Analysis of impedance stability of VSC-HVDC systems.
Southern Power Syst Technol 10(5):75–79
43. Song Y, Breitholtz C (2016) Nyquist stability analysis of an AC-grid connected VSC-HVDC
system using a distributed parameter DC cable model. IEEE Trans Power Del 31(2):898–907
44. Stamatiou G, Bongiorno M (2016) Stability analysis of two-terminal VSC-HVDC systems
using the net-damping criterion. IEEE Trans Power Del 31(4):1748–1756
26 1 Introduction
45. Bidadfar A, Nee HP, Zhang L (2016) Power system stability analysis using feedback control
system modeling including HVDC transmission links. IEEE Trans Power Syst 31(1):116–124
46. Zhang L, Harnefors L, Nee HP (2011) Modeling and control of VSC-HVDC links connected to
Island systems. IEEE Trans Power Syst 26(2):783–793
47. Shen L, Wang W, Barnes M (2014) The influence of MTDC control on DC power flow and AC
system dynamic responses. In: 2014 I.E. PES general meeting, National Harbor, MD, pp 1–5
48. Pinares G, Bongiorno M (2016) Modeling and analysis of VSC-based HVDC systems for DC
network stability studies. IEEE Trans Power Del 31(2):848–856
49. Du WJ, Fu Q, Wang HF (2018) Sub-synchronous oscillations caused by open-loop modal
coupling between VSC-based HVDC line and power system. IEEE Trans Power Syst
33(4):3664–3677
50. Bucher MK, Wiget R, Andersson G (2014) Multiterminal HVDC networks—what is the
preferred topology? IEEE Trans Power Del 29(1):406–413
51. Chaudhuri NR, Chaudhuri B (2013) Adaptive droop control for effective power sharing in
multi-terminal DC (MTDC) grids. In: 2013 I.E. power & energy society general meeting,
Vancouver, BC, pp 1–1
52. Kalcon GO, Adam GP, Anaya-Lara O (2012) Small-signal stability analysis of multi-terminal
VSC based DC transmission systems. IEEE Trans Power Syst 27(4):1818–1830
53. Beerten J, D’Arco S, Suul JA (2016) Identification and small-signal analysis of interaction
modes in VSC MTDC systems. IEEE Trans Power Del 31(2):888–897
54. Du WJ, Fu Q, Wang HF (2018) Small-signal stability of an AC/MTDC power system as affected
by open-loop modal coupling between the VSCs. IEEE Trans Power Syst 33(3):3143–3152
55. Bayo-Salas A, Beerten J, Rimez J, Van Hertem D (2016) Analysis of control interactions in
multi-infeed VSC HVDC connections. IET Gener Transm Dis 10(6):1336–1344
56. Ndreko M, van der Meer AA, Gibescu M (2014) Impact of DC voltage control parameters on
AC/DC system dynamics under faulted conditions. In: 2014 I.E. PES general meeting|confer-
ence & exposition, National Harbor, MD, pp 1–5
57. Jovcic D, Pahalawaththa N, Zavahir M (1999) Analytical modelling of HVDC-HVAC systems.
IEEE Trans Power Syst 14(2):506–511
58. Pinto RT, Bauer P, Rodrigues SF, Wiggelinkhuizen EJ (2013) A novel distributed direct-
voltage control strategy for grid integration of offshore wind energy systems through MTDC
network. IEEE Trans Ind Electron 60(6):2429–2441
59. Chaudhuri N, Majumder R, Chaudhuri B, Pan J (2011) Stability analysis of VSC MTDC grids
connected to multimachine AC systems. IEEE Trans Power Del 26(4):2774–2784
60. Harnefors L, Johansson N, Berggren B (2014) Interarea oscillation damping using active-power
modulation of multiterminal HVDC transmissions. IEEE Trans Power Syst 29(5):2529–2538
61. Du WJ, Fu Q, Wang HF (2017) Strong dynamic interactions between multi-terminal DC
network and AC power systems caused by open-loop modal coupling. IET Generation Trans-
mission & Distribution 11(9):2362–2374
62. Du WJ, Fu Q, Wang HF (2018) Method of open-loop modal analysis for examining the
sub-synchronous interactions introduced by VSC control in an MTDC/AC system. IEEE
Trans Power Del 33(2):840–850
Chapter 2
Linearized Model of a Power System
with a Grid-Connected Variable Speed
Wind Generator
Linearized model of a power system with a grid-connected variable speed wind

generator (VSWG) is derived in this chapter in three steps. First, the linearized model
of the VSWG is established. That includes the establishment of linearized model of a
PMSG and a DFIG. Second, the linearized model of the power system is derived.
Finally, the linearized model of the power system with the VSWG is established by
combining the model of the VSWG and the power system.
2.1 Linearized Model of a PMSG
Figure 2.1 shows the configuration of a grid-connected PMSG, which is comprised

of three main parts: (1) The permanent magnet SG; (2) The machine side converter
(MSG) and associated control system; (3) The grid side converter (GSC) and
associated control system. Linearized models for each of three parts of the PMSG
are derived respectively in this section. Then, they are combined to form the
linearized model of the grid-connected PMSG.
2.1.1 Linearized Model of the Permanent Magnet SG
The voltage equations of stator windings of the permanent magnet SG in Fig. 2.1 are:
8
>
> d
< ψpsd ¼ ω0 Rps Ipsd ω0 Vpsd þ ω0 ωpr ψpsq
dt
ð2:1Þ
>
> d
: ψpsq ¼ ω0 Rps Ipsq ω0 Vpsq ω0 ωpr ψpsd
dt

28 2 Linearized Model of a Power System with a Grid-Connected Variable. . .
Vpsd +jVpsq Ipcd

MSC Pps Ppc GSC
+ jIpcq
N X pf
Cp Vpdc
S
Ipsd +jIpsq Pp +jQ p
Vpsdref Vpsqref Vpd Vpq
Vpcd Vpd + jVpq
Control Control
+ jVpcq (Vw q w )
Permanent SG system system
w prref Ipsdref Vpdcref Q pref
MSC and associated control system GSC and associated control system
Fig. 2.1 Configuration of a grid-connected PMSG
where Vpsd and Vpsq are the d and q component of voltage, ψpsd and ψpsq are the d
and q component of magnetic flux, Ipsd and Ipsq are the d and q component of current
of stator windings of the permanent magnet SG respectively; ω0 is the synchronous
speed; ωpr is the angular speed of the permanent magnet SG; Rpsis the resistance of
stator windings.
Motion equations of the rotor of the permanent magnet SG are:
dωpr
Jpr ¼ Tpm Tpe ð2:2Þ
dt
where Jpr is the constant of inertia of the rotor; Tpm andTpe are the mechanical torque
input to and the electromagnetic power output from the permanent magnet SG
respectively; and
Tpe ¼ ψpsq Ipsd ψpsd Ipsq ð2:3Þ
Flux linkage equations of the permanent magnet SG are

(
ψpsd ¼ Xpd Ipsd ψpm
ð2:4Þ
ψpsq ¼ Xpq Ipsq
where ψpm is the flux of permanent magnet, Xpd and Xpq are the d and q axis
reactance of stator windings respectively.
Linearization of Eq. (2.1) is
8
>
> d
< Δψpsd ¼ ω0 Rps ΔIpsd ω0 ΔVpsd þ ω0 ωpr0 Δψpsq þ ω0 ψpsq0 Δωpr
dt
ð2:5Þ
>
> d
: Δψpsq ¼ ω0 Rps ΔIpsq ω0 ΔVpsq ω0 ωpr0 Δψpsd ω0 ψpsd0 Δωpr
dt
2.1 Linearized Model of a PMSG 29
By neglecting the variation of the mechanical torque input to the permanent

magnet SG such that ΔTpm ¼ 0, linearization of Eq. (2.2) is obtained to be
dΔωpr
Jpr ¼ ΔTpm ΔTpe ¼ ΔTpe ð2:6Þ
dt
Substituting the linearization of Eq. (2.3) in Eq. (2.6), it can have
dΔωpr
Jpr ¼ ψpsq0 ΔIpsd þ ψpsd0 ΔIpsq Ipsd0 Δψpsq þ Ipsq0 Δψpsd ð2:7Þ
dt
ψpm in Eq. (2.4) is a constant. Hence, linearization of Eq. (2.4) is
(
Δψpsd ¼ Xpd ΔIpsd
ð2:8Þ
Δψpsq ¼ Xpq ΔIpsq
From Eqs. (2.7) and (2.8),
dΔωpr ψpsq0 ψpsd0

Jpr ¼ Δψpsd þ Δψpsq Ipsd0 Δψpsq þ Ipsq0 Δψpsd ð2:9Þ
dt Xpd Xpq
Writing Eqs. (2.5) and (2.9) together in the following form of state-space
representation,
d
ΔXp1 ¼ Ap1 ΔXp1 þ bp1 ΔVpsd þ bp2 Vpsq ð2:10Þ
dt
where
h i
Xp1 ¼ Δψpsd Δψpsq Δωpr ,
2 3
ω0 Rps
6 ω0 ωpr0 ω0 ψpsq0 7
6 Xpd 7 2 3 2 3
6 7 ω0 0
6 ω0 Rps 7 6 7 6 7
Ap1 ¼ 6
6 ω0 ωpr0 ω0 ψpsd0 7
7, bp1 ¼ 4 0 5, bq1 ¼ 4 ω0 5:
6 X pq 7
6 7 0 0
4 ψpsq0 Ipsq0 ψpsd0 Ipsd0 5
þ 0
Jpr Xpd Jpr Jpr Xpq Jpr
Active power output from the permanent magnet SG is
Pps ¼ Vpsq Ipsq þ Vpsd Ipsd ð2:11Þ
Linearization of (2.11) is
ΔPps ¼ Vpsq0 ΔIpsq þ Vpsd0 ΔIpsd þ Ipsq0 ΔVpsq þ Ipsd0 ΔVpsd ð2:12Þ
Substituting Eq. (2.8) in Eq. (2.12),
Vpsq0 Vpsd0
ΔPps ¼ Δψpsq þ Δψpsd þ Ipsq0 ΔVpsq þ Ipsd0 ΔVpsd ð2:13Þ
Xpq Xpd
Thus,
ΔPps ¼ cp1 T ΔXp1 þ dp1 ΔVpsd þ dp2 ΔVpsq ð2:14Þ
where

Vpsd0 Vpsq0
cp1 T ¼ 0 , dp1 ¼ Ipsd0 , dp2 ¼ Ipsq0 :
Xpd Xpq
By arranging Eqs. (2.10) and (2.14) together, the state-space model of the
permanent magnet SG is obtained to be
8
< d ΔXp1 ¼ Ap1 ΔXp1 þ bp1 ΔVpsd þ bp2 Vpsq
dt ð2:15Þ
:
ΔPps ¼ cp1 T ΔXp1 þ dp1 ΔVpsd þ dp2 ΔVpsq
2.1.2 Linearized Model of Machine Side Converter

and Associated Control System
The MSC implements the vector control. Configuration of its control system is
shown by Fig. 2.2. It can be seen that there are two current control inner loops in
Rotor speed control outer loop q-axis current control inner loop
w pr y pm
w prref – K pp1 I psqref + K pp 2 V
+ + – + psqref MSC
+ K pi1 + – K pi 2 + –
pr x p1 I psq xp2
s s w pr X pd I psd
I psdref K pp 3 Vpsdref
+ + –
MSC
– K pi 3 + +
I psd x p3
s w pr X pq I psq
d-axis current control inner loop
Fig. 2.2 Configuration of control system of the MSC of the PMSG
the control system to respectively control the d and q axis output current from the
stator windings of the permanent magnet SG, Ipsd and Ipsq. The speed control outer
loop controls the angular speed of the rotor of the permanent magnet SG, ωpr, to
generate the reference for the control of q axis current control inner loop. Normally,
the reference for the control of d axis current control inner loop is set to be zero, i.e.,
Ipsdref ¼ 0.
In Fig. 2.2, the outputs of three integral controllers are xp1, xp2 and xp3. It can have
8
> d
>
> xp1 ¼ Kpi1 ωpr ωprref
>
> dt
>
>
<
d
xp2 ¼ Kpi2 Ipsqref Ipsq ð2:16Þ
>
> dt
>
>
>
>
>
: d xp3 ¼ Kpi3 Ipsdref Ipsd
dt
where Kpi1, Kpi2 and Kpi3 are the gains of integral controllers, Ipsdref and Ipsqref are the
references of current control inner loops, ωprref is the reference of rotor speed control
outer loop. From Eq. (2.2),
8
> Ipsqref ¼ Kpp1 ωpr ωprref þ xp1
>
<
Vpsqref ¼ Kpp2 Ipsqref Ipsq xp2 ωpr ψpm þ ωpr Xpd Ipsd ð2:17Þ
>
>
:
Vpsdref ¼ Kpp3 Ipsdref Ipsd xp3 ωpr Xpq Ipsq
where Kpp1, Kpp2 and Kpp3 are the gains of proportional controller, Vpsdref and Vpsqref
are the output signals from the MSC control system, which are used as the d and q
axis references of the terminal voltage of the permanent magnet SG. Ignoring the
transient of the PWM of the MSC, it can have
(
Vpsdref Vpsd
ð2:18Þ
Vpsqref Vpsq
Linearization of Eq. (2.16) is (Δωprref ¼ 0,ΔIpsdref ¼ 0)

8
>
>
d
Δxp1 ¼ Kpi1 Δωpr
>
>
>
> dt
>
<
d
Δxp2 ¼ Kpi2 ΔIpsqref ΔIpsq ð2:19Þ
>
> dt
>
>
>
>
>
: d Δxp3 ¼ Kpi3 ΔIpsd
dt

8
> ΔIpsqref ¼ Kpp1 Δωpr þ Δxp1
>
>
>
>
< ΔVpsqref ¼ Kpp2 ΔIpsqref ΔIpsq Δxp2 ψpm Δωpr
ð2:20Þ
>
> þXpd Ipsd0 Δωpr þ ωpr0 Xpd ΔIpsd
>
>
>
:
ΔVpsdref ¼ Kpp3 ΔIpsd Δxp3 Xpq Ipsq0 Δωpr þ ωpr0 Xpd ΔIpsd
Substituting the first equation in Eqs. (2.20) and (2.8) in Eq. (2.19),
8
> d
>
> Δxp1 ¼ Kpi1 Δωpr
>
> dt
>
>
>
<d Kpi2 Δψpsq
Δxp2 ¼ Kpi2 Δxp1 þ Kpi2 Kpp1 Δωpr ð2:21Þ
>
> dt Xpq
>
>
>
> Kpi3 Δψpsd
>
> d
: Δxp3 ¼
dt Xpd
The above equation can be written as

d
ΔXp2 ¼ Ap2 ΔXp2 þ Bp1 ΔXp1 ð2:22Þ
dt
where
ΔXp2 ¼ ½ Δxp1 Δxp2 Δxp3 T ,

2 3
0 0 Kpi1
2 3
0 0 0 6 Kpi2 7
6 0 Kpi2 Kpp1 7
6 7 6 7
Ap2 ¼ 4 Kpi2 0 0 5, Bp1 ¼6 Xpq 7:
6 7
0 0 0 4 Kpi3 5
0 0
Xpd

(
ΔVpsdref ¼ ΔVpsd
ð2:23Þ
ΔVpsqref ¼ ΔVpsq
Substituting Eq. (2.22) and the first equation of Eq. (2.20) into the last two
equations of Eq. (2.20),
8
>
> ΔVpsd ¼ Kpp2 Kpp1 Δωpr Kpp2 Δxp1 þ Kpp2 ΔIpsq Δxp2 ψpm Δωpr
<
þXpd Ipsd0 Δωpr þ ωpr0 Xpd ΔIpsd ð2:24Þ
>
>
:
ΔVpsq ¼ Kpp3 ΔIpsd Δxp3 Xpq Ipsq0 Δωpr ωpr0 Xpq ΔIpsq

8
>
> ΔVpsd ¼ Kpp2 Δxp1 Δxp2 þ ωpr0 Δψpsd þ
Kpp2
Δψpsq
>
>
>
> Xpq
>
<
þ Xpd Ipsd0 Kpp2 Kpp1 ψpm Δωpr ð2:25Þ
>
>
>
>
>
> Kpp3
>
: ΔVpsq ¼ Δxp3 þ Δψpsd ωpr0 Δψpsq Xpq Ipsq0 Δωpr
Xpd
The above equation can be written as

(
ΔVpsd ¼ cp2 T ΔXp2 þ cp3 T ΔXp1
ð2:26Þ
ΔVpsq ¼ cp4 T ΔXp2 þ cp5 T ΔXp1
where
cp2 T ¼ ½ Kpp2 1 0 ,

Kpp2
cp3 T ¼ ωpr0 Xpd Ipsd0 Kpp2 Kpp1 ψpm ,
Xpq
cp4 T ¼ ½ 0 0 1 ,

Kpp3
cp5 T ¼ ωpr0 Xpq Ipsq0 :
Xpd
The state-space representation of the MSC and associated control system is

obtained by writing Eqs. (2.22) and (2.26) together as
8
>
> d
> ΔXp2 ¼ Ap2 ΔXp2 þ Bp1 ΔXp1
>
< dt
ΔVpsd ¼ cp2 T ΔXp2 þ cp3 T ΔXp1 ð2:27Þ
>
>
>
>
:
ΔVpsq ¼ cp4 T ΔXp2 þ cp5 T ΔXp1
2.1.3 Linearized Model of Grid Side Converter

and Associated Control System
From Fig. 2.1, the line voltage equations across the filter reactance, Xpf, can be
written as
8
>
> d ω0 Vpcd ω0 Vpd
< dtIpcd ¼ Xpf Xpf þ ω0 Ipcq
>
ð2:28Þ
>
> d ω0 Vpcq ω0 Vpq
>
: Ipcq ¼ ω0 Ipcd
dt Xpf Xpf
where Ipcd and Ipcq are the d and q axis component of output current from the GSC
respectively; Vpcd and Vpcq are the d and q axis component of terminal voltage of the
GSC; Vpd and Vpq are the d and q axis component of the point of common coupling
(PCC) of the PMSG.
Equation of the voltage across the capacitor is
dVpdc
Cp Vpdc ¼ Pps Ppc ð2:29Þ
dt
where Vpdc is the DC voltage across the capacitor; Cp is the capacitance; Ppc is the
active power output from the GSC; and
Ppc ¼ Vpcd Ipcd þ Vpcq Ipcq ð2:30Þ
The configuration of vector control system of the grid side converter (GSC) of the
PMSG is shown by Fig. 2.3. The current control inner loops control the d and q axis
output current from the GSC, Ipcd and Ipcq, respectively. The control outer loops
control the DC voltage across the capacitor, Vpdc, and the reactive power output from
the GSC, Qp, respectively to generate the current control references for the current
control inner loops.
In Fig. 2.3, the outputs of four integral controllers are xp4, xp5, xp6 and xp7. From
Fig. 2.3,
DC voltage control outer loop d-axis current control inner loop
K pp 4 K pp 5 V pd
V pdcref – I cdref + V
+ + + + pcdref
GSC
+ K pi 4 + – K pi 5 +
V pdc s xp 4 I pcd s x p5 X pf I pcq
K pp 6 K pp 7 V pq
Q pref – I cqref + Vpcqref
+ + + +
GSC
+ K pi 6 + – K pi 7 + +
Qp s xp 6 I pcq s xp7 X pf I pcd
Reactive power control outer loop q-axis current control inner loop
Fig. 2.3 Configuration of control system of the GSC of the PMSG

8
> d
>
> xp4 ¼ Kpi4 Vpdc Vpdcref
>
> dt
>
>
>
>
>
> d
>
< dtxp5 ¼ Kpi5 Ipcdref Ipcd
ð2:31Þ
>
> d
>
> xp6 ¼ Kpi6 Qp Qpref
>
>
>
> dt
>
>
>d
>
: xp7 ¼ Kpi7 Ipcqref Ipcq
dt
where Qpref is the reference of reactive power control outer loop; Ipcdref and Ipcqref are
the references of d and q axis current control inner loops respectively; Vpdcref is the
reference of DC voltage control outer loop; Qp is the reactive power output from the
GSC
Qp ¼ Vpq Ipcd Vpd Ipcq ð2:32Þ
From Fig. 2.3,

8
>
> Ipcdref ¼ Kpp4 Vpdc Vpdcref þ xp4
>
>
>
>
< Vpcdref ¼ Kpp5 Ipcdref Ipcd þ xp5 Xpf Ipcq þ Vpd
ð2:33Þ
>
> Ipcqref ¼ Kpp6 Qp Qpref þ xp6
>
>
>
>
:
Vpcqref ¼ Kpp7 Ipcqref Ipcq þ xp7 þ Xpf Ipcd þ Vpq
where Kpp4, Kpp5, Kpp6 and Kpp7 are the gains of proportional controllers.
Ignoring the transient of the PWM of the GSC,
(
Vpcdref ¼ Vpcd
ð2:34Þ
Vpcqref ¼ Vpcq
Linearization of Eqs. (2.28), (2.29) and (2.30) respectively is

8
> d ω0 ω0
>
< dtΔIpcd ¼ Xpf ΔVpcd Xpf ΔVpd þ ω0 ΔIpcq
>
ð2:35Þ
>
> d ω0 ω0
>
: ΔIpcq ¼ ΔVpcq ΔVpq ω0 ΔIpcd
dt Xpf Xpf
dΔVpdc
Cp Vpdc0 ¼ ΔPps ΔPpc ð2:36Þ
dt
ΔPpc ¼ Vpcd0 ΔIpcd þ Vpcq0 ΔIpcq þ Ipcd0 ΔVpcd0 þ Ipcq0 ΔVpcq0 ð2:37Þ

dΔVpdc
Cp Vpdc0 ¼ ΔPps Vpcd0 ΔIpcd Vpcq0 ΔIpcq Ipcd0 ΔVpcd Ipcq0 ΔVpcq
dt
ð2:38Þ
Qpref and Vpdcref are constants. Hence, linearization of Eq. (2.31) is

8
>
>
d
Δxp4 ¼ Kpi4 ΔVpdc
>
>
>
> dt
>
>
>
> d
>
>
< dtΔxp5 ¼ Kpi5 ΔIpcdref ΔIpcd
ð2:39Þ
>
> d
>
> Δxp6 ¼ Kpi6 ΔQp
>
>
>
> dt
>
>
>
>
: d Δxp7 ¼ Kpi7 ΔIpcqref ΔIpcq
dt
Linearization of Eqs. (2.32), (2.33) and (2.34) respectively is
ΔQp ¼ Vpq0 ΔIpcd Vpd0 ΔIpcq þ Ipcd0 ΔVpq Ipcq0 ΔVpd ð2:40Þ
8
> ΔIpcdref ¼ Kpp4 ΔVpdc þ Δxp4
>
>
>
>
< ΔVpcdref ¼ Kpp5 ΔIpcdref ΔIpcd þ Δxp5 Xpf ΔIpcq þ ΔVpd
ð2:41Þ
>
> ΔIpcqref ¼ Kpp6 ΔQp þ Δxp6
>
>
>
:
ΔVpcqref ¼ Kpp7 ΔIpcqref ΔIpcq þ Δxp7 þ Xpf ΔIpcd þ ΔVpq
(
Vpcdref ¼ Vpcd
ð2:42Þ
Vpcqref ¼ Vpcq
Substituting the first and the third equation of Eq. (2.41) into the second and the
fourth equation respectively and using Eq. (2.42),
(
ΔVpcd ¼ Kpp5 Kpp4 ΔVpdc þ Kpp5 Δxp4 Kpp5 ΔIpcd þ Δxp5 Xpf ΔIpcq þ ΔVpd
ΔVpcq ¼ Kpp7 Kpp6 ΔQp þ Kpp7 Δxp6 Kpp7 ΔIpcq þ Δxp7 þ Xpf ΔIpcd þ ΔVpq
ð2:43Þ
Substituting Eq. (2.40) in Eq. (2.43)

8
> ΔVpcd ¼ Kpp5 ΔIpcd Xpf ΔIpcq þ Kpp5 Kpp4 ΔVpdc þ Kpp5 Δxp4 þ Δxp5 þ ΔVpd
>
<
ΔVpcq ¼ Kpp7 Kpp6 Vpq0 þ Xpf ΔIpcd Kpp7 Kpp6 Vpd0 þ Kpp7 ΔIpcq
>
>
:
þKpp7 Δxp6 þ Δxp7 Kpp7 Kpp6 Ipcq0 ΔVpd þ Kpp7 Kpp6 Ipcd0 þ 1 ΔVpq
ð2:44Þ

8
> d ω0 Kpp5 ω0 Kpp5 Kpp4 ω0 Kpp5 ω0
>
> ΔIpcd ¼ ΔIpcd þ ΔVpdc þ Δxp4 þ Δxp5
>
> dt X X X X
>
> pf pf pf pf
>
>
< ω0 Kpp7 Kpp6 Vpd0 þ Kpp7
d ω0 Kpp7 Kpp6 Vpq0
ΔIpcq ¼ ΔIpcd ΔIpcq
>
> dt Xpf Xpf
>
>
>
>
> ωo Kpp7
> ω0 ω0 Kpp7 Kpp6 Ipcq0 ω0 Kpp7 Kpp6 Ipcd0
>
:þ Δxp6 þ Δxp7 ΔVpd þ ΔVpq
Xpf Xpf Xpf Xpf
ð2:45Þ
dΔVpdc
Cp Vpdc ¼ ΔPps Kpp5 Kpp4 Ipcd0 ΔVpdc Kpp5 Ipcd0 Δxp4 Ipcd0 Δxp5
dt

Ipcq0 Δxp7 þ Kpp5 Ipcd0 Vpcd0 Kpp7 Kpp6 Vpq0 Ipcq0 þ Xpf Ipcq0 ΔIpcd

þ Xpf Ipcd0 Vpcq0 þ Kpp7 Kpp6 Ipcq0 Vpd0 þ Kpp7 Ipcq0 ΔIpcq Kpp7 Ipcq0 Δxp6

þ Kpp7 Kpp6 I2pcq0 Ipcd0 ΔVpd Kpp7 Kpp6 Ipcd0 þ 1 Ipcq0 ΔVpq
ð2:46Þ
Substituting Eq. (2.40) into the third equation of Eq. (2.41),
ΔIpcqref ¼ Kpp6 Vpq0 ΔIpcd Kpp6 Vpd0 ΔIpcq þ Kpp6 Ipcd0 ΔVpq
Kpp6 Ipcq0 ΔVpd þ Δxp6 ð2:47Þ
Substituting Eq. (2.40), the first equation of Eqs. (2.41) and (2.47) in Eq. (2.39),
8d
>
> Δx ¼ Kpi4 ΔVpdc
>
> dt p4
>
>
>
>
>
> d
>
> Δxp5 ¼ Kpi5 Kpp4 ΔVpdc þ Kpi5 Δxp4 Kpi5 ΔIpcd
>
< dt
d
>
> Δxp6 ¼ Kpi6 Vpq0 ΔIpcd Kpi6 Vpd0 ΔIpcq þ Kpi6 Ipcd0 ΔVpq Kpi6 Ipcq0 ΔVpd
>
> dt
>
>
>
> d
>
> Δxp7 ¼ Kpi7 Kpp6 Vpq0 ΔIpcd Kpi7 Kpp6 Vpd0 þ Kpi7 ΔIpcq
>
>
>
: dt
þKpi7 Kpp6 Ipcd0 ΔVpq Kpi7 Kpp6 Ipcq0 ΔVpd þ Kpi7 Δxp6
ð2:48Þ
Arrange Eqs. (2.45), (2.46) and (2.48) in the following matrix form
d
ΔXp3 ¼ Ap3 ΔXp3 þ bp3 ΔPps þ bp4 ΔVpd þ bp5 ΔVpq ð2:49Þ
dt
where ΔXp3 ¼ ½ ΔIpcd ΔIpcq ΔVpdc Δxp4 Δxp5 Δxp6 Δxp7 ,
2.1.4 Linearized Model of the Entire PMSG System
Writing the first equation of Eq. (2.15), the first equation of Eqs. (2.27) and (2.49)
together,
8
> d
>
> ΔX ¼ Ap1 ΔXp1 þ bp1 ΔVpsd þ bp2 Vpsq
> dt p1
>
>
<
d
ΔXp2 ¼ Ap2 ΔXp2 þ Bp1 ΔXp1 ð2:50Þ
>
> dt
>
>
>
>
: d ΔXp3 ¼ Ap3 ΔXp3 þ bp3 ΔPps þ bp4 ΔVpd þ bp5 ΔVpq
dt
Substituting the last two equations of Eq. (2.27) in Eq. (2.50),
d
ΔXp ¼ Ap ΔXp þ bpd ΔVpd þ bpq ΔVpq ð2:51Þ
dt
where
2 3
Ap1 þ bp1 cp3 T þ bp2 cp5 T bp1 cp2 T þ bp2 cp4 T 0
6 7
Ap ¼ 4 Bp1 Ap2 0 5,
dp2 bp3 cp5 T þ bp3 cp1 T þ dp1 bp3 cp3 T dp1 bp3 cp2 T þ dp2 bp3 cp4 T Ap3
2 3 2 3
0 0
6 7 6 7
bpd ¼ 4 0 5, bpq ¼ 4 0 5:
bp4 bp5
Active power output from the PMSG is

Pp ¼ Vpd Ipcd þ Vpq Ipcq ¼ Ppc ¼ Vpcd Ipcd þ Vpcq Ipcq ð2:52Þ
From the linearization of Eqs. (2.52) and (2.40), variations of power output from
the PMSG can be obtained to be
" #
ΔPp
¼ Fpv ΔVpv þ Fpi ΔIpi ð2:53Þ
ΔQp
where

Ipcd0 Ipcq0
ΔVpv ¼ ½ ΔVpd ΔVpq T , ΔIpi ¼ ½ ΔIpcd ΔIpcq T , Fpv ¼ ,
Ipcq0 Ipcd0

Vpd0 Vpq0
Fpi ¼ :
Vpq0 Vpd0
ΔIpcd and ΔIpcq are the state variables (see Eq. (2.49)). Hence, Eq. (2.26) can be
written as
(
ΔPp ¼ cpp T ΔXp þ dpp1 ΔVpd þ dpp2 ΔVpq
ð2:54Þ
ΔQp ¼ cpq T ΔXp þ dpq1 ΔVpd þ dpq2 ΔVpq
Finally, the state-space model of the entire PMSG system is obtained from
Eqs. (2.51) and (2.54) to be
8
> d
>
> ΔX ¼ Ap ΔXp þ bpd ΔVpd þ bpq ΔVpq
< dt p
ΔPp ¼ cpp T ΔXp þ dpp1 ΔVpd þ dpp2 ΔVpq ð2:55Þ
>
>
>
:
ΔQp ¼ cpq T ΔXp þ dpq1 ΔVpd þ dpq2 ΔVpq
2.2 Linearized Model of a DFIG
Configuration of a DFIG for wind power generation is shown by Fig. 2.4. The DFIG
is of three main modules of electrical function: the induction generator and two-mass
shaft rotational system, the rotor side converter (RSC) and associated control system,
the gird side converter (GSC) and associated control system. Mechanical power is
converted to the electrical power by the induction generator via the shaft rotational
system. Function of the RSC and associated control system is to control the output
power from the DFIG. The GSC and associated control system functions to maintain
the DC voltage across the capacitor for the operation of the RSG and GSC.
Two-mass shaft Induction generator Vvsd +jVvsq

rotational system
I dsd +jI dsq (Vw q w )
Low- High-
speed speed
shaft shaft Pds +jQds
X df
I dcd +jI dcq
Pdc +jQdc
Pdr I drd +jI drq

Induction RSC Pdc1 GSC
generator and Vdrd + jVdrq
the shaft
system Cd Vddc
Vdcd + jVdcq
Vdrdref Vdrqref Vdcdref Vdcqref
RSC and Control Control GSC and

associated system system associated
control system Pdsref Qdsref Vddcref control system
Fig. 2.4 Configuration of a DFIG for wind power generation

2.2 Linearized Model of a DFIG 41
In this section, the linearized state-space representation of each functional module

of the DFIG is derived. Then, the linearized model of the DFIG in the d q
coordinate of the DFIG is established by combining the derived models of three
function modules. All the variables are expressed in per unit (p.u.) except that the
unit of constant of inertia of the rotor is second.
2.2.1 Linearized Model of the Induction Generator and Two-

Mass Shaft Rotational System
The voltage equations of stator windings of induction generator in Fig. 2.4 are:
8
>
> d
< ψdsd ¼ ω0 Rds Idsd þ ω0 Vdsd þ ω0 ψdsq
dt
ð2:56Þ
>
> d
: ψdsq ¼ ω0 Rds Idsq þ ω0 Vdsq ω0 ψdsd
dt
where Vdsd and Vdsq are the d and q component of voltage, ψdsd and ψdsq the d and q
component of flux, Idsd and Idsq the d and q component of current of stator windings
of the induction generator respectively; ω0 is the synchronous speed; Rdsis the
resistance of stator windings.
The voltage equations of rotor windings of induction generator are:
8
>
> d
< ψdrd ¼ ω0 Rdr Idrd þ ω0 Vdrd þ ω0 ð1 ωdr1 Þψdrq
dt
ð2:57Þ
>
> d
: ψdrq ¼ ω0 Rdr Idrq þ ω0 Vdrq ω0 ð1 ωdr1 Þψdrd
dt
where Vdrd and Vdrq are the d and q component of voltage, ψdrd and ψdrd the d and q
component of flux, Idrd and Idrq the d and q component of current of rotor windings of
the induction generator respectively; ωdr1 is the angular speed of high-speed shaft in
the two-mass shaft rotational system; Rdris the resistance of stator windings.
Flux equations of stator and rotor windings of the induction generator are
8
> ψdsd ¼ Xdss Idsd Xdm Idrd
>
>
>
< ψdsq ¼ Xdss Idsq Xdm Idrq
ð2:58Þ
>
> ψdrd ¼ Xdrr Idrd Xdm Idsd
>
>
:
ψdrq ¼ Xdrr Idrq Xdm Idsq
where Xdss and Xdrr are the self-inductance of stator and rotor windings respectively;
Xdm is the mutual-inductance between the stator and rotor windings.
Two-mass rotational system of induction generator consists of high-speed and
low-speed shaft. Two shafts are connected by a gear box. Motion equations of the
two-shaft rotational system are:
8
> d
>
> Jdr1 ωdr1 ¼ Kdm θdr Tde Ddm ðωdr1 ωdr2 Þ Ddr1 ωdr1
>
> dt
>
<
d
Jdr2 ωdr2 ¼ Tdm Kdm θdr Ddm ðωdr2 ωdr1 Þ Ddr2 ωdr2 ð2:59Þ
>
> dt
>
>
>
> d
: θdr ¼ ω0 ðωdr2 ωdr1 Þ
dt
where Jdr2 and Jdr1 are constants of inertia of high-speed and low-speed shaft
respectively; Tdm is the mechanical torque input to the low-speed shaft; ωdr2 is the
angular speed of low-speed shaft; Kdm and Ddm are the elastic coefficient and mutual
damping coefficient between the high-speed and low-speed shaft; Ddr2 and Ddr1 are
the self-damping coefficients of high-speed and low-speed shaft respectively;
θdr ¼ θdr2 θdr1 is the relative angular position of low-speed and high-speed shaft
(θdr1 is the angular position of high-speed shaft, is the angular position of low-speed
shaft); Tde is the electromagnetic output torque of the induction generator, which can
be expressed as:
Tde ¼ ψdsd Idsq ψdsq Idsd ð2:60Þ
Substitute Eq. (2.58) in Eq. (2.60),

Tde ¼ Xdm Idsd Idrq Idsq Idrd ð2:61Þ

8
>
> d
< Δψdsd ¼ ω0 Rds ΔIdsd þ ω0 ΔVdsd þ ω0 Δψdsq
dt
ð2:62Þ
>
> d
: Δψdsq ¼ ω0 Rds ΔIdsq þ ω0 ΔVdsq ω0 Δψdsd
dt
8
>
> d
< Δψdrd ¼ ω0 Rdr ΔIdrd þ ω0 ΔVdrd þ ω0 ð1 ωdr10 ÞΔψdrq ω0 ψdrq0 Δωdr1
dt
>
> d Δψ ¼ ω R ΔI þ ω ΔV ω ð1 ω ÞΔψ þ ω ψ Δω
: 0 dr drq 0 drq 0 dr10 0 drd0 dr1
dt drq drd
ð2:63Þ

8
> Δψdsd ¼ Xdss ΔIdsd Xdm ΔIdrd
>
>
>
< Δψdsq ¼ Xdss ΔIdsq Xdm ΔIdrq
ð2:64Þ
>
> Δψdrd ¼ Xdrr ΔIdrd Xdm ΔIdsd
>
>
:
Δψdrq ¼ Xdrr ΔIdrq Xdm ΔIdsq
Mechanical torque input to the induction generator is a constant, i.e., ΔTdm ¼ 0.

Thus, linearization of Eq. (2.59) is
8
> d
>
> Jdr1 dtΔωdr1 ¼ Kdm Δθdr ΔTde Ddm ðΔωdr1 Δωdr2 Þ Ddr1 Δωdr1
>
>
>
<
d
Jdr2 Δωdr2 ¼ Kdm Δθdr Ddm ðΔωdr2 Δωdr1 Þ Ddr2 Δωdr2 ð2:65Þ
>
> dt
>
>
>
>
: d Δθdr ¼ ω0 ðΔωdr2 Δωdr1 Þ
dt

ΔTde ¼ Xdm Idsd0 ΔIdrq Idsq0 ΔIdrd þ Idrq0 ΔIdsd Idrd0 ΔIdsq ð2:66Þ
Substituting Eq. (2.66) in (2.65),

8
>
>
d
> Jdr1 Δωdr1 ¼ Kdm Δθdr Xdm Idsd0 ΔIdrq Idsq0 ΔIdrd þ Idrq0 ΔIdsd Idrd0 ΔIdsq
>
>
> dt
>
>
>
< D dm ðΔωdr1 Δωdr2 Þ Ddr1 Δωdr1
> d
>
> Jdr2 Δωdr2 ¼ Kdm Δθdr Ddm ðΔωdr2 Δωdr1 Þ Ddr2 Δωdr2
>
> dt
>
>
>
> d
: Δθdr ¼ ω0 ðΔωdr2 Δωdr1 Þ
dt
ð2:67Þ
From Eq. (2.64)
ΔId ¼ M1 Δψd ð2:68Þ
where
2 3
Xdss 0 Xdm 0
6 7
6 0 Xdss 0 Xdm 7
M¼6
6 X
7,
4 dm 0 Xdrr 0 7 5
0 Xdm 0 Xdrr
T
ΔId ¼ ½ Idsd Idsq Idrd Idrq T , Δψd ¼ ψdsd ψdsq ψdrd ψdrq :
From Eqs. (2.62), (2.63), (2.67) and (2.68), the state equation of the induction
generator and two-mass shaft rotational system is obtained to be
d
ΔXd1 ¼ Ad1 ΔXd1 þ Bd1 Δzd1 þ bd1 ΔVdsq þ bd2 ΔVdsd ð2:69Þ
dt
where

ΔXd1 ¼ Δψdsd Δψdsq Δψdrd Δψdrq Δωdr1 Δωdr2 Δθdr
T
Δzd1 ¼ ½ ΔVdrd ΔVdrq ,
Active power output from the rotor side of the induction generator is
Pdr ¼ Vdrd Idrd þ Vdrq Idrq ð2:70Þ
Linearizing Eq. (2.70),
ΔPdr ¼ Vdrd0 ΔIdrd þ Vdrq0 ΔIdrq þ Idrd0 ΔVdrd þ Idrq0 ΔVdrq ð2:71Þ
From Eqs. (2.68) and (2.71),
ΔPdr ¼ ½ 0 0 Vdrd0 Vdrq0 M1 Δψd þ Idrd0 ΔVdrd þ Idrq0 ΔVdrq ð2:72Þ
According to Eq. (2.69), the above equation can be written as
ΔPdr ¼ cd1 T ΔXd1 þ cd2 T Δzd1 ð2:73Þ
where
cd1 T ¼ ½ K5 Vdrd0 K7 Vdrq0 K6 Vdrd0 K8 Vdrq0 0 0 0 ,

cd2 T ¼ ½ Idrd0 Idrq0 :
Writing Eqs. (2.69) and (2.73) together, the state-space model of the induction
generator and two-mass shaft rotational system is obtained to be
8
< d ΔX ¼ A ΔX þ B Δz þ b ΔV þ b ΔV
d1 d1 d1 d1 d1 d1 dsq d2 dsd
dt ð2:74Þ
:
ΔPdr ¼ cd1 ΔXd1 þ cd2 Δzd1
T T
2.2.2 Linearized Model of the RSC and Associated Control

System
The vector control is implemented by the RSC. Configuration of the control system
of the RSC is shown by Fig. 2.5, where current control inner loops control the d-axis
and q-axis stator current, Idsq and Idsd, respectively; the active and reactive power
control outer loops control the active and reactive power output from the stator side
2
Active power q-axis current X dm s X
control outer control inner loop
– sdr1 ( X drr – ) Idrd + dr1 dm Vdsq
X dss X dss
loop
K dp1 K dp 2
Pdsref Idsqref I drqref
+ + X + + – + Vdrqref
– dss RSC
– X dm
K di1 + – K di 2 +
Pds s xd 1 I drq s xd 2
K dp 3 I drdref K dp 4
Qdsref
+ + Idsdref X + + + – Vdrdref
– dss RSC
X dm +
– K di 3 + – – K di 4 +
Qds s xd 3 I drd s xd 4
Vdsq 2
X dm
Reactive power X dm sdr1 ( X drr – ) Idrq
control outer loop d-axis current control inner loop X dss
Fig. 2.5 Configuration of vector control system of the RSC of the DFIG
of the induction generator, Pds and Qds respectively. sdr1 ¼ 1 ωdr1 is slip of high-
speed shaft of the induction generator.
There are total four PI controllers in the RSC control system shown by Fig. 2.2.
Take the active power control outer loop as an example, the state equation related
with the integral control is
d
xd1 ¼ Kdi1 ðPdsref Pds Þ ð2:75Þ
dt
where xd1 is the state variable and output of integral control; Kdi1 is the integral gain;
Pdsref is the control reference of active power output from the stator side of the
induction generator.
Similarly, the output of integral control of q-axis rotor current control inner loop,
the reactive power control outer loop and d-axis rotor current control inner loop is
denoted as xd2, xd3 and xd4 respectively. Following state equations can be obtained.
8
> d
>
> xd2 ¼ Kdi2 Idrqref Idrq
>
> dt
>
<
d
xd3 ¼ Kdi3 ðQdsref Qds Þ ð2:76Þ
>
> dt
>
>
>
>
: d xd4 ¼ Kdi4 ðIdrdref Idrd Þ
dt
where Kdi2, Kdi3 and Kdi4 are the gains of corresponding integral controllers; Qdsref,
Idrqref and Idrdref are control reference of reactive power output from the stator side,
q-axis and d-axis of rotor current respectively.
From Fig. 2.5, following algebraic equations can be obtained.
8
> Idsqref ¼ Kdp1 ðPdsref Pds Þ þ xd1
>
>
>
>
>
> I ¼ Kdp3 ðQdsref Qds Þ þ xd3
> dsdref
>
>
>
>
> Idrqref ¼
Xdss
>
> Idsqref
>
> Xdm
>
>
>
> Xdss Vdsq
>
> I ¼ Idsdref
>
< drdref Xdm Xdm
ð2:77Þ
>
> Vdrqref ¼ Kdp2 Idrqref Idrq xd2
>
>

>
>
>
> X2dm sdr1 Xdm
>
> s X I þ Vdsq
> dr1 drr Xdss drd
> Xdss
>
>
>
>
>
> Vdrdref ¼ Kdp4 ðIdrdref Idrd Þ
>
>

>
>
>
> X2
: xd4 þ sdr1 Xdrr dm Idrq
Xdss
where Kdp1, Kdp2, Kdp3 and Kdp4 are the proportional gains; Idsqref and Idsdref are the
control reference of q and d component of current of stator windings; Vdrqref and
Vdrdref are the control reference output of the RSC for controlling the q and d
component of voltage of rotor windings respectively; Normally, transient of modu-

lation control can be ignored such that

Vdrdref ¼ Vdrd
ð2:78Þ
Vdrqref ¼ Vdrq
Active and reactive power outputs from the stator side of the induction generator
are

Pds ¼ Vdsq Idsq þ Vdsd Idsd
ð2:79Þ
Qds ¼ Vdsq Idsd Vdsd Idsq
Power control references are constant such that ΔPdsref ¼ 0 and ΔQdsref ¼ 0.
Thus, linearization of Eqs. (2.75) and (2.76) is
8
>
> d
Δxd1 ¼ Kdi1 ΔPds
>
>
>
> dt
>
>
>
> d
>
< Δxd2 ¼ Kdi2 ΔIdrqref ΔIdrq
dt
ð2:80Þ
>
> d
>
> Δxd3 ¼ Kdi3 ΔQds
>
> dt
>
>
>
>
>
: d Δxd4 ¼ Kdi4 ðΔIdrdref ΔIdrd Þ
dt
8

>
> X2dm
>
> ΔVdrqref ¼ Kdp2 ΔIdrqref ΔIdrq Δxd2 sdr10 Xdrr ΔIdrd
>
> Xdss
>
>
>
>
2
>
>
> þIdrd0 Xdrr Xdm Δωdr1 þ sdr10 Xdm ΔVdsq Vdsq0 Xdm Δωdr1
>
>
>
>
>
Xdss Xdss Xdss
>
>

>
> X2dm
>
> ΔVdrdref ¼ Kdp4 ðΔIdrdref ΔIdrd Þ Δxd4 þ sdr10 Xdrr ΔIdrq
>
>
>
> Xdss
>
<

X2dm ð2:81Þ
> Idrq0 Xdrr Δωdr1
>
> Xdss
>
>
>
>
> ΔIdsdref ¼ Kdp3 ΔQds þ Δxd3
>
>
>
>
> Xdss
>
> ΔIdrqref ¼ ΔIdsqref
>
>
>
> Xdm
>
>
>
> Xdss ΔVdsq
>
> ΔIdrdref ¼ ΔIdsdref
>
> X Xdm
>
> dm
:
ΔIdsqref ¼ Kdp1 ΔPds þ Δxd1
where Δωdr1 ¼ Δsdr1.


ΔVdrdref ¼ ΔVdrd
ð2:82Þ
ΔVdrqref ¼ ΔVdrq

ΔPds ¼ Vdsq0 ΔIdsq þ Vdsd0 ΔIdsd þ Idsq0 ΔVdsq þ Idsd0 ΔVdsd
ð2:83Þ
ΔQds ¼ Vdsq0 ΔIdsd Vdsd0 ΔIdsq þ Idsd0 ΔVdsq Idsq0 ΔVdsd
From Eqs. (2.68), (2.80), (2.81), (2.82) and (2.83), following state equation of the
RSC and associated control system is obtained
8
< d ΔX ¼ A ΔX þ B ΔX þ b ΔV þ b ΔV
d2 d2 d2 d2 d1 d3 dsq d4 dsd
dt ð2:84Þ
:
Δzd1 ¼ Cd1 ΔXd2 þ Cd2 ΔXd1 þ Cd3 ΔVdsq þ Cd4 ΔVdsd
where
ΔXd2 ¼ ½ Δxd1 Δxd2 Δxd3 Δxd4 ,

2 3
0 0 0 0
6 Kdi2 Xdss 7
6 07
6 X 0 0 7
6 dm 7
Ad2 ¼ 6 7,
6 0 0 0 0 7
6 7
4 Kdi4 Xdss 5
0 0 0
Xdm
2 Kdp4 Xdss 3
0 0 1
6 Xdm 7
Cd1 ¼ 4 K 5,
dp2 Xdss
1 0 0
Xdm
2 3
Kdp3 Kdp4 Xdss Xdrr Vdsq0 þ Kdp4 Xdm 2 Kdp2 Kdp1 Xdss Xdrr Vdsd0 sdr10 Xdm
6
6 Xdm Xdrr Xdss Xdm 2 Xdm Xdrr Xdss Xdm 2 Xdss 77
6 7
6 27
6 Kdp3 Kdp4 Xdss Xdrr Vdsd0 sdr10 Xdm Kdp2 Kdp1 Xdss Xdrr Vdsq0 þ Kdp2 Xdm 7
6 þ 7
6 Xdm Xdm 2 Xdss Xdrr Xdss Xdm Xdss Xdrr Xdm 2 7
6 7
6 7
6 Kdp3 Kdp4 Xdss Vdsq0 þ Kdp4 Xdss Kdp2 Kdp1 Xdss Vdsd0 7
6 þ s 7
6 Xdm 2 Xdrr Xdss Xdm 2 Xdrr Xdss
dr10
7
Cd2 T ¼6
6
7
6 Kdp3 Kdp4 Xdss Vdsd0 Kdp2 Kdp1 Xdss Vdsq0 þ Kdp2 Xdss 7 7
6 sdr10 7
6 Xdss Xdrr Xdm 2 Xdm 2 Xdss Xdrr 7
6

7
6 Xdm 2 2
Vdsq0 Xdm 7
6 Xdm 7
6 Idrq0 Xdrr Idrd0 Xdrr 7
6 Xdss Xdss Xdss 7
6 7
6 7
4 0 0 5
0 0
2 3 2 3
Kdp4 Kdp3 Kdp4 Xdss Idsd0 Kdp3 Kdp4 Xdss Idsq0
6 Xdm 7 6 Xdm 7
Cd3 ¼6
4s X
7, Cd4 ¼ 6
5 4
7:
dr10 dm Kdp2 Kdp1 Xdss Idsq0 Kdp2 Xdss Kdp1 Idsd0 5

Xdss Xdm Xdm
ΔXd1 and Δzd1 are defined by Eq. (2.69).
2.2.3 Linearized Model of the GSC and Associated Control

System
Configuration of the vector control system of the GSC is shown by Fig. 2.6, which is
comprised of q-axis and d-axis current control loop as well as the DC voltage control
outer loop. The q-axis and d-axis current control loop controls the output current of
the GSC, Idcq and Idcd, respectively. DC voltage control outer loop controls the DC
voltage across the capacitor, Vddc.
In the d-q coordinate, the voltage equations across the filter reactance, Xfd, on the
side of the GSC are
8
> d ω0 Vdcd ω0 Vdsd
>
< dtIdcd ¼ X þ ω0 Idcq
df Xdf
ð2:85Þ
>
> d ωV ω V
: Idcq ¼ 0 dcq 0 dsq ω0 Idcd
dt Xdf Xdf
DC voltage control outer loop q-axis current control inner loop
Kdp 5 Kdp 6 Vdsq

Vddcref – + I dcqref+ + + + Vdcqref
GSC
+ Kdi 5 + – Kdi 6 + +
Vddc I dcq X df I dcd
s xd 5 s xd 6
Vdsd
Kdp 7
I dcdref
+ + + +Vdcdref GSC
– Kdi 7 + –
I dcd X df I dcq
s xd 7
d-axis current control inner loop
Fig. 2.6 Configuration of vector control system of the GSC of the DFIG
where Xdf is the reactance of the output filter, Idcd and Idcq are the d and q component
of output current of the GSC, Vdcd and Vdcq are the d and q component of terminal
voltage of the GSC.
Power balance equation associated with the charging and discharging of the
capacitor is
dVddc
Cd Vddc ¼ Pdc1 Pdr ð2:86Þ
dt
where Cd is the capacitance, Vddc is the DC voltage across the capacitor, Pdr is the
active power output from the rotor side of the induction generator and expressed by
Eq. (2.70), Pdc1 is the active power injected from the DC side to the GSC and can be
expressed as
Pdc1 ¼ Vdcd Idcd þ Vdcq Idcq ð2:87Þ
The output variables of integrators in Fig. 2.3 are xd5, xd6 and xd7. It can have
8d
>
> x ¼ Kdi5 ðVddc Vddcref Þ
>
> dt d5
>
<
d
xd6 ¼ Kdi6 Idcqref Idcq ð2:88Þ
> dt
>
>
>
>
: d x ¼ K ðI
d7 di7 dcdref Idcd Þ
dt
where Kdi5, Kdi6 and Kdi7 are the integral gains of PI controllers in Fig. 2.6, Vddcref,
Idcqref and Idcdref are respectively the control reference of DC voltage control, q and d
component of output current o the GSC.
Following algebraic equations can be obtained from Fig. 2.6

8
> I ¼ Kdp5 ðVddcref Vddc Þ þ xd5
< cqref
Vdcqref ¼ Kdp6 Idcqref Idcq þ xd6 þ Vdsq þ Xdf Idcd ð2:89Þ
>
:
Vdcdref ¼ Kdp7 ðIdcdref Idcd Þ þ xd7 þ Vdsd Xdf Idcq
where Kdp5, Kdp6 and Kdp7 are the proportional gains of PI controllers in Fig. 2.6
Vdcdref and Vdcqref are the control reference output of the GSC for controlling the q
and d component of terminal voltage of the GSC respectively; Normally, transient of
modulation control can be ignored such that

Vdcdref ¼ Vdcd
ð2:90Þ
Vdcqref ¼ Vdcq

8
> d ω0 ΔVdcd ω0 ΔVdsd
>
< dtΔIdcd ¼ X þ ω0 ΔIdcq
df X df
ð2:91Þ
>
> d ω ΔV ω ΔV
: ΔIdcq ¼ 0 dcq 0 dsq ω0 ΔIdcd
dt Xdf Xdf

dΔVddc
Cd Vddc0 ¼ ΔPdr ΔPdc1 ð2:92Þ
dt
ΔPdc1 ¼ Vdcd0 ΔIdcd þ Vdcq0 ΔIdcq þ Idcd0 ΔVdcd þ Idcq0 ΔVdcq ð2:93Þ
Substituting Eq. (2.93) in Eq. (2.92), it can have

dΔVddc
Cd Vddc0 ¼ ΔPdr Vdcd0 ΔIdcd Vdcq0 ΔIdcq Idcd0 ΔVdcd Idcq0 ΔVdcq
dt
ð2:94Þ
Control references in Eqs. (2.88) and (2.89) are constant such that ΔVddcref ¼ 0
and ΔIdcdref ¼ 0. Thus, linearization of Eq. (2.88) is
8
> d
>
> Δx ¼ Kdi5 ΔVddc
> dt d5
>
>
<
d
Δxd6 ¼ Kdi6 ΔIdcqref ΔIdcq ð2:95Þ
>
> dt
>
>
>
>
: d Δxd7 ¼ Kdi7 ΔIdcd
dt

8
> ΔI ¼ Kdp5 ΔVddc þ Δxd5
< dcqref
ΔVdcqref ¼ Kdp6 ΔIdcqref ΔIdcq þ Δxd6 þ ΔVdsq þ Xdf ΔIdcd ð2:96Þ
>
:
ΔVdcdref ¼ Kdp7 ΔIdcd þ Δxd7 þ ΔVdsd Xdf ΔIdcq

ΔVdcdref ¼ ΔVdcd
ð2:97Þ
ΔVdcqref ¼ ΔVdcq
Substituting the first equation in Eq. (2.96) in Eq. (2.95),

8
> d
>
> Δxd5 ¼ Kdi5 ΔVddc
>
> dt
>
<
d
Δxd6 ¼ Kdi6 Kdp5 ΔVddc þ Kdi6 Δxd5 Kdi6 ΔIdcq ð2:98Þ
>
> dt
>
>
>
>
: d Δxd7 ¼ Kdi7 ΔIdcd
dt
Substituting Eq. (2.97) and the first equation of Eq. (2.96) in the second and third
equation of Eq. (2.96),
(
ΔVdcq ¼ Kdp6 Kdp5 ΔVddc þ Δxd5 ΔIdcq þ Δxd6 þ ΔVdsq þ Xdf ΔIdcd
ΔVdcd ¼ Kdp7 ΔIdcd þ Δxd7 þ ΔVdsd Xdf ΔIdcq
ð2:99Þ

8
>
> d ω0 Kdp7 ΔIdcd þ ω0 Δxd7
< dtΔIdcd ¼
>
Xdf
ð2:100Þ
>
> d ω0 Kdp6 Kdp5 ΔVddc þ Δxd5 ΔIdcq þ ω0 Δxd6
>
: ΔIdcq ¼
dt Xdf
The state equations of the GSC and associated control system is obtained from
Eqs. (2.94), (2.98) and (2.100) to be
d
ΔXd3 ¼ Ad3 ΔXd3 þ bd5 ΔVdsd þ bd6 ΔVdsq þ bd7 ΔPdr ð2:101Þ
dt
where
ΔXd3 ¼ ½ ΔIdcd ΔIdcq ΔVddc Δxd5 Δxd6 Δxd7 ,

2.2.4 Linearized Model of the Entire DFIG
By combining the linearized models of three main modules of the DFIG derived in
the above three subsections, the linearized model of the DFIG can be obtained as
follows.
Firstly, by writing the state equations of Eqs. (2.74), (2.84) and (2.101) together,
it can have
8d
>
> ΔX ¼ Ad1 ΔXd1 þ Bd1 Δzd1 þ bd1 ΔVdsq þ bd2 ΔVdsd
>
> dt d1
>
<
d
ΔXd2 ¼ Ad2 ΔXd2 þ Bd2 ΔXd1 þ bd3 ΔVdsq þ bd4 ΔVdsd ð2:102Þ
>
> dt
>
>
>
: d ΔX ¼ A ΔX þ b ΔV þ b ΔV þ b ΔP
d3 d3 d3 d5 dsd d6 dsq d7 dr
dt
Substituting the second equation of Eq. (2.84) in the second equation of
Eq. (2.74),
ΔPdr ¼ cd1 T ΔXd1 þ cd2 T Cd1 ΔXd2 þ cd2 T Cd2 ΔXd1

ð2:103Þ
þcd2 T Cd3 ΔVdsq þ cd2 T Cd4 ΔVdsd
From the second equation of Eqs. (2.84) and (2.103), Δzd1 and ΔPdr in
Eq. (2.102) can be replaced. With the replacement, Eq. (2.102) becomes
8d
>
> ΔXd1 ¼ Ad1 ΔXd1 þ Bd1 Δzd1 þ bd1 ΔVdsq þ bd2 ΔVdsd
>
>
>
<
dt
d
ΔXd2 ¼ Ad2 ΔXd2 þ Bd2 ΔXd1 þ bd3 ΔVdsq þ bd4 ΔVdsd ð2:104Þ
>
> dt
>
>
>
: d ΔX ¼ A ΔX þ b ΔV þ b ΔV þ b ΔP
d3 d3 d3 d5 dsd d6 dsq d7 dr
dt
where
T
ΔXd ¼ ΔXd1 T ΔXd2 T ΔXd3 T ,
2 3
Ad1 þ Bd1 Cd2 Bd1 Cd1 0
6 7
Ad ¼ 4 Bd2 Ad2 0 5,
bd7 cd2 T Cd2 þ bd7 cd1 T bd7 cd2 T Cd1 Ad3
2 3 2 3
Bd1 cd4 þ bd2 Bd1 cd3 þ bd1
6 7 6 7
bdd ¼ 4 bd4 5, bdq ¼ 4 bd3 5:
bd7 cd2 T cd4 þ bd5 bd6 þ bd7 cd2 T cd3
Active and reactive power injected from the DFIG to the power system are

Pd ¼ Pds þ Pcd ¼ Vdsd Idsd þ Vdsq Idsq þ Vdsd Idcd þ Vdsq Idcq
ð2:105Þ
Qd ¼ Qds þ Qcd ¼ Vdsq Idsd Vdsd Idsq þ Vdsq Idcd Vdsd Idcq
where Pcd and Qcd are the active and reactive power injected from the GSC to the
power system.
Linearization of Eq. (2.105) can be written as

ΔPd
¼ Fdsv ΔVdsv þ Fdsi ΔIdsi þ Fdci ΔIdci ð2:106Þ
ΔQd
where
T T T
ΔVdsv ¼ ½ ΔVdsq ΔVdsd , ΔIdsi ¼ ½ ΔIdsq ΔIdsd , ΔIdci ¼ ½ ΔIdcq ΔIdcd ,

Idsq0 þ Idcq0 Idsd0 þ Idcd0 Vdsq0 Vdsd0
Fdsv ¼ , Fdsi ¼ ,
Idsd0 þ Idcd0 Idsq0 Idcq0 Vdsd0 Vdsq0

Vdsq0 Vdsd0
Fdci ¼ :
Vdsd0 Vdsq0
Substituting Eq. (2.68) in Eq. (2.106) with ΔIdsi being replaced, Eq. (2.106)
becomes
(
ΔPd ¼ cdp T ΔXd þ ddp1 ΔVdsq þ ddp2 ΔVdsd
ð2:107Þ
ΔQd ¼ cdq T ΔXd þ ddq1 ΔVdsq þ ddq2 ΔVdsd
2.3 Linearized Model of the Power System 55
where
Writing Eqs. (2.104) and (2.107) together, the state-space model of the DFIG is
8
> d
>
< dtΔXd ¼ Ad ΔXd þ bdq ΔVdsq þ bdd ΔVdsd
>
ΔPd ¼ cdp T ΔXd þ ddp1 ΔVdsq þ ddp2 ΔVdsd ð2:108Þ
>
>
>
:
ΔQd ¼ cdq T ΔXd þ ddq1 ΔVdsq þ ddq2 ΔVdsd
2.3 Linearized Model of the Power System
Figure 2.7 shows the configuration of a multi-machine power system with N

synchronous generators, G1, G2. . .GN. The load is represented by impedance
model. A variable speed wind generator, either the PMSG or the DFIG, is connected
at node W. ΔPw and ΔQw denote the small variations of active and reactive power
injected from the variable speed wind generator. According to the notations used in
the previous sections, when the variable speed wind generator is the PMSG,
ΔPw ¼ ΔPp and ΔQw ¼ ΔQp. For the DFIG, ΔPw ¼ ΔPd and ΔQw ¼ ΔQd. In
Fig. 2.7 Configuration of a

power system integrated G1
with a PMSG or DFIG
G2
Transmission network
GN
W ΔVw Δqw
ΔPw ΔQw
this section, linearized state-space model of the power system is derived, where ΔPw
and ΔQw are treated as the input variables and small variations of the PCC voltage,
ΔVw ∠ Δθw, is treated as the output variables. Derivation of the model is carried out
in the common x-y coordinate of the power system.
(
ΔPw ¼ Ix0 ΔVx þ Iy0 ΔVy þ Vx0 ΔIx þ Vy0 ΔIy
ð2:109Þ
ΔQw ¼ Iy0 ΔVx þ Ix0 ΔVy þ Vy0 ΔIx Vx0 ΔIy
where ΔIx and ΔIy are the small variations of d and q component of the output
current from node W injected into the transmission network, ΔVx and ΔVy are the
small variations of d and q component of the terminal voltage at node W.
Express Eq. (2.109) in the following matrix form,

ΔPw
¼ Fw1 ΔI þ Fw2 ΔV ð2:110Þ
ΔQw
where

ΔIx ΔVx Vx0 Vy0 Ix0 Iy0
ΔI ¼ , ΔV ¼ , Fw1 ¼ , Fw2 ¼
ΔIy ΔVy Vy0 Vx0 Iy0 Ix0
Denote ΔIgxj and ΔIgyj as the small variations of d and q component of the output
current from Gj. Denote ΔVgxj and ΔVgyj as the small variations of d and q
component of the voltage at the terminal of Gj. Define the following output current
vector ΔIgj and terminal voltage vector ΔVgj of Gj.
( T
ΔVgj ¼ ½ ΔVgxj ΔVgyj
ð2:111Þ
T
ΔIgj ¼ ½ ΔIgxj ΔIgyj
Network equation is
" # " #" # " #
ΔIg Ygg Ygw ΔVg ΔVg
¼ ¼Y ð2:112Þ
ΔI Ywg Yww ΔV ΔV
T T
where ΔVg ¼ ΔVg1 T ΔVgN T , ΔIg ¼ ΔIg1 T ΔIgN T ,

Ygg Ygw
Y¼ is the admittance matrix of the transmission network with only
Ywg Yww
the generators’ nodes being kept.
The state-space model of Gj is
8
< d ΔX ¼ A ΔX þ B ΔV
gj gj gj gj gj
dt ð2:113Þ
:
ΔIgj ¼ Cgj ΔXgj þ Dgj ΔVgj
2.3 Linearized Model of the Power System 57
where ΔXgj is the vector of all the state variables of Gj. From Eq. (2.113), the state-
space model for N generators can be obtained by writing N state-space models of Gj
together as
8
< d ΔX ¼ A ΔX þ B ΔV
g g g g g
dt ð2:114Þ
:
ΔIg ¼ Cg ΔXg þ Dg ΔVg
where
T T
ΔXg ¼ ΔXg 1 T ΔXg N T , ΔIg ¼ ΔIg 1 T ΔIg N T ,
T
ΔVg ¼ ΔVg 1 T ΔVg N T ,
Ag ¼ diag½ Ag1 Ag2 . . . AgN , Bg ¼ diag½ Bg1 Bg2 . . . BgN ,
Cg ¼ diag½ Cg1 Cg2 . . . CgN , Dg ¼ diag½ Dg1 Dg2 . . . DgN ,
diag[] denotes either a diagonal matrix or a block diagonal matrix.


1 ΔPw
ΔV ¼ ðFw2 þ Fw1 Yww Þ ðFw2 þ Fw1 Yww Þ1 Fw1 Ywg ΔVg ð2:115Þ
ΔQw
From Eq. (2.112) and the output equation in Eq. (2.114)

Ygg Dg ΔVg þ Ygw ΔV ¼ Cg ΔXg ð2:116Þ

ΔPw
ΔVg ¼ F3 ΔXg þ F4 ð2:117Þ
ΔQw
where
h i1
F3 ¼ Ygg Dg Ygw ðFw2 þ Fw1 Yww Þ1 Fw1 Ywg Cg ,
h i1
F4 ¼ Ygg Dg Ygw ðFw2 þ Fw1 Yww Þ1 Fw1 Ywg Ywg ðFw2 þ Fw1 Yww Þ1 :
Substituting Eq. (2.117) in the first equation of Eq. (2.114),

d ΔPw
ΔXg ¼ A1 ΔXg þ B1 ð2:118Þ
dt ΔQw
where A1 ¼ Ag + BgF3 and B1 ¼ BgF4.

" #
ΔPw
ΔV ¼ F5 ΔXg þ F6 ð2:119Þ
ΔQw
where
F5 ¼ ðFw2 þ Fw1 Yww Þ1 Fw1 Ywg F3 ,

F6 ¼ ðFw2 þ Fw1 Yww Þ1 ðFw2 þ Fw1 Yww Þ1 Fw1 Ywg F4 :
Small variations of the magnitude and phase of the voltage at node W where the
variable speed wind generator is connected with the power system are ΔVw and
Δθw. It can have
8
>
> Vx0 Vy0
>
< ΔV w ¼ ΔV ¼ f w1 T ΔV
Vw0 Vw0
ð2:120Þ
>
> Vy0 Vx0
>
: Δθw ¼ ΔV ¼ f w2 ΔV
T
V2w0 V2w0

8 " #
>
> ΔPw
>
> ΔVw ¼ C1 ΔXg þ D1
>
< ΔQw
" # ð2:121Þ
>
> ΔPw
>
>
: Δθw ¼ C2 ΔXg þ D2
>
ΔQw
where C1 ¼ fw1TF5, D1 ¼ fw1TF6, C2 ¼ fw2TF5, D2 ¼ fw2TF6.

Writing Eqs. (2.118) and (2.121) together,
8 " #
>
> d ΔPw
>
> ΔXg ¼ A1 ΔXg þ B1
>
> dt ΔQw
>
>
>
> " #
>
< ΔPw
ΔVw ¼ C1 ΔXg þ D1 ð2:122Þ
>
> ΔQw
>
>
>
> " #
>
> ΔPw
>
>
> Δθw ¼ C2 ΔXg þ D2
:
ΔQw
Hence, the state-space model of the power system can be expressed as

8
> d
>
< dtΔXg ¼ Ag ΔXg þ bp ΔPw þ bq ΔQw
>
ΔVw ¼ cgv T ΔXg þ dvp ΔPw þ dvq ΔQw ð2:123Þ
>
>
>
:
Δθw ¼ cgθ T ΔXg þ dθp ΔPw þ dθq ΔQw
2.4 Closed-Loop Interconnected Model of the Power System. . . 59
2.4 Closed-Loop Interconnected Model of the Power

System with a Grid-Connected VSWG
2.4.1 Closed-Loop Interconnected Model of the Power System

with a PMSG
Consider that the VSWG connected at node W in Fig. 2.7 is a PMSG. Figure 2.8
shows the relation between the d-q coordinate of the PMSG and the common x-y
coordinate of the power system. It can be seen that the direction of the terminal
voltage of the PMSG, ΔVw ∠Δθw, is taken as the direction of d axis in the common
x-y coordinate. Hence, ΔVwd ¼ ΔVw and ΔVwq ¼ 0.
According to the notations to derive the state-space model of the PMSG in Sect.
2.1, ΔPw ¼ ΔPp, ΔQw ¼ ΔQp, ΔVpd ¼ ΔVwd and ΔVpq ¼ ΔVwq. Thus,
ΔVpd ¼ ΔVwd ¼ ΔVw and ΔVpq ¼ ΔVwq ¼ 0. The state-space model of the
PMSG shown by Eq. (2.55) becomes
8
> d
>
< dtΔXp ¼ Ap ΔXp þ bpd ΔVw
>
ΔPw ¼ cpp T ΔXp þ dpp1 ΔVw ð2:124Þ
>
>
>
:
ΔQw ¼ cpq T ΔXp þ dpq1 ΔVw
From Eq. (2.124), the transfer function model of the PMSG can be written as
(
ΔPw ¼ Hpp ðsÞΔVw
ð2:125Þ
ΔQw ¼ Hpq ðsÞΔVw
where Hpp(s) ¼ cppT(sI Ap)1bpd + dpp1, Hpq(s) ¼ cpqT(sI Ap)1bpd + dpq1.

The state-space model of the power system shown by Eq. (2.123) can be written
as
Fig. 2.8 Relation between y

the d-q coordinate of the
PMSG and the common x-y q
coordinate
Vw qw
qw
x
Fig. 2.9 Closed-loop Subsystem of rest of power system

interconnected model of the
power system with the ΔPw
PMSG Gvp (s)
ΔVw
+
Gvq (s)
ΔQw
Hpq (s)
Hpp (s)
PMSG subsystem
8
< d ΔX ¼ A ΔX þ b ΔP þ b ΔQ
g g g p w q w
dt ð2:126Þ
:
ΔVw ¼ cgv T ΔXg þ dvp ΔPw þ dvq ΔQw
From Eq. (2.126), the transfer function model of the power system can be written
as
ΔVw ¼ Gvp ðsÞΔPw þ Gvq ðsÞΔQw ð2:127Þ
where Gvp(s) ¼ cgvT(sI Ag)1bp + dvp, Gvq(s) ¼ cgvT(sI Ag)1bq + dvq.

From Eqs. (2.125) and (2.127), the closed-loop interconnected model of the
power system with the PMSG is obtained and shown by Fig. 2.9. The state-space
model of the closed-loop interconnected system can be derived from Eqs. (2.124)
and (2.126) to be
d
ΔXgp ¼ Agp ΔXgp ð2:128Þ
dt
T
where ΔXgp ¼ ΔXg T ΔXp T ,
2 3
dpq1 bq þ dpp1 bp cgv T dpq1 bq þ dpp1 bp dvp cpp T þ dvq cpq T
6 Ag þ T T
bp cpp þ bq cpq þ 7
6 1 dvp dpp1 dvq dpq1 1 dvp dpp1 dvq dpq1 7
Agp ¼ 6
6
7
7
4 bpd cgv T dvp bpd cpp T þ dvq bpd cpq T 5
Ap þ
1 dvp dpp1 dvq dpq1 1 dvp dpp1 dvq dpq1
2.4 Closed-Loop Interconnected Model of the Power System. . . 61
2.4.2 Closed-Loop Interconnected Model of the Power System

with a DFIG
When the VSWG connected at node W in Fig. 2.7 is a DFIG, a closed-loop

interconnected model of the power system with the DFIG can be derived similarly.
Normally, the direction of the terminal voltage of the DFIG, ΔVw ∠Δθw, is taken as
the direction of q axis in the common x-y coordinate as shown by Fig. 2.10. Hence,
ΔVwq ¼ ΔVw and ΔVwd ¼ 0.
According to the notations to derive the state-space model of the DFIG in Sect.
2.2, ΔPw ¼ ΔPd, ΔQw ¼ ΔQd, ΔVdsq ¼ ΔVwq and ΔVdsd ¼ ΔVdsd. Thus,
ΔVdsq ¼ ΔVwq ¼ ΔVw and ΔVdsd ¼ ΔVwd ¼ 0. The state-space model of the
DFIG shown by Eq. (2.108) becomes
8
> d
>
< dtΔXd ¼ Ad ΔXd þ bdq ΔVw
>
ΔPw ¼ cdp T ΔXd þ ddp1 ΔVw ð2:129Þ
>
>
>
:
ΔQw ¼ cdq T ΔXd þ ddq1 ΔVw
From Eq. (2.129), the transfer function model of the DFIG is obtained to be
(
ΔPw ¼ Hdp ðsÞΔVw
ð2:130Þ
ΔQw ¼ Hdq ðsÞΔVw
where Hdp(s) ¼ cdpT(sI Ad)1bdq + ddp1Hdq(s) ¼ cdqT(sI Ad)1bdq + ddq1.

The state-space and transfer function model of the power system are expressed by
Eqs. (2.126) and (2.127) respectively. From Eqs. (2.127) and (2.130), the closed-
loop interconnected model of the power system with the DFIG is obtained and
shown by Fig. 2.11. From Eqs. (2.126) and (2.129), the state-space model of the
closed-loop interconnected system shown by Fig. 2.11 is derived to be
d
ΔXgd ¼ Agd ΔXgd ð2:131Þ
dt
Fig. 2.10 Relation between y

the d-q coordinate of the
DFIG and the common x-y q
coordinate
Vw qw
qw
x

power system with the ΔPw
DFIG Gvp (s)
ΔVw
+
Gvq (s)
ΔQw
Hdq (s)
Hdp (s)
DFIG subsystem
where
T
ΔXgd ¼ ΔXg T ΔXd T ,
2 3
ddq1 bq þ ddp1 bp cgv T ddq1 bq þ ddp1 bp dvp cdp T þ dvq cdq T
6 Ag þ bp cdp T þ bq cdq T þ 7
6 1 dvp ddp1 dvq ddq1 1 dvp ddp1 dvq ddq1 7
Agd ¼ 6
6
7
7
4 bdq cgv T dvp bdq cdp T þ dvq bdq cdq T 5
Ad þ
1 dvp ddp1 dvq ddq1 1 dvp ddp1 dvq ddq1
Chapter 3
Damping Torque Analysis of Small-Signal
Angular Stability of a Power System
Affected by Grid-Connected Wind Power
Induction Generators
Induction generator based wind power generation has been dominating the wind
market since the rise of wind power industry at the end of last century and will be
continuously in a favorable position for large-scale grid connection given its lower
cost and more mature technology compared with other wind generation for the
foreseeable future [1]. Fixed-speed induction generator (FSIG-Type 1 Wind Gen
Model) and doubly-fed induction generator (DFIG-Type 3 Wind Gen Model) are
two main types of induction generator adopted for wind power generation especially
considering the fact that DFIG is the most frequently-used technology to date.
The increasing penetration of wind power generation has significantly affected
power system dynamics. Due to the difference in rotor structures and excitation
principles, FSIG and DFIG possess different dynamic behaviors during system
disturbances and hence impact the power system dynamics differently, which has
posed a critical challenge for the real-time system operation and therefore deserves a
careful investigation.
3.1 Impact of Grid-Connected Wind Power Induction

Generators on Small-Signal Angular Stability
The impact of the integration of FSIG and DFIG on power system small-signal
angular stability have been extensively examined from early this century. A com-
prehensive study regarding the influence of FSIG on power system oscillation is
presented in [2] by modal analysis, which considers multiple impact factors includ-
ing length of transmission interface, load condition, wind penetration level and wind
farm configuration etc. It is concluded that in most cases FSIG introduces a negative
damping to the system and additional reactive power compensation could mitigate
the negative impact of FSIG on small-signal angular stability. This conclusion is
supported by modal analysis in [3] but contradicted by [4]. Compared with FSIG,

64 3 Damping Torque Analysis of Small-Signal Angular Stability of a Power. . .
DFIG is comparatively new and has a more flexible control in active and reactive
power, and thus most of research efforts are devoted to the grid connection study of
DFIG in recent decade. Various case studies have been implemented to address
different aspects of DFIG in affecting the small-signal angular stability such as
integration method [4–9], inertia or other sensitivity based approach [10–12], reac-
tive power/voltage control [13–17], operating condition [18], virtual inertia control
[19–21], additional damping control [22–30] and external energy storage system
[31, 32].
It can be seen that: (1) Most of the existing research is actually case-by-case
observation by using the two common ‘computation’ methods (modal analysis and
time domain simulation), and thus the essential reason for inconsistent study results
with different preconditions cannot be effectively and convincingly investigated by
these two ‘black box’ methods. No proper theoretical method is seen so far to clearly
reveal the essential damping mechanism of power system small-signal angular
stability as affected by FSIG and DFIG; (2) Most of the published research tends
to study the grid impact of FSIG and DFIG separately and there is no systematic
analytical theory to compare the damping effectiveness and robustness of these two
wind power induction generators (WPIGs) and dig deeper information about their
essential difference and inner connection in affecting power system oscillations,
which certainly provides a better understanding of their individual damping
mechanisms.
Taking account of the points above, a generic methodology based on damping
torque analysis theory in order to analyze the damping mechanisms of different
WPIGs is proposed in this chapter, with the aim of giving a physical insight that how
the different rotor structures and excitation systems of FSIG and DFIG affect their
damping mechanisms. The rest of the chapter is organized as follows: In Sect. 3.2,
system modeling for damping torque analysis of a power system affected by grid-
connected wind power induction generators is carried out. Then a general imple-
mentation framework of explicit damping torque analysis of Phillips-Heffron model
based multi-machine power system with WPIGs is presented in Sect. 3.3. Hence, the
closed-form solution of damping torque contribution from the main internal dynamic
components of wind generators to each synchronous generator (SG) can be derived.
In Sect. 3.4, two typical linearized models and explicit transfer functions of different
WPIGs are proposed to facilitate the analytical comparison on the impact mecha-
nisms of DFIG and FSIG, where FSIG is treated as a special case of DFIG with rotor
side short-circuit (i.e., rotor voltage equal to zero). Unlike the existing numerical
comparison (case-by-case study), the purely analytical comparison of damping
mechanisms between different WPIGs can reveal their essential difference and
inner connection in damping mechanisms. In Sect. 3.5, the proposed methodology
is demonstrated in a 16-machine test system and then employed to perform a
numerical comparison under different wind penetration conditions, so that the
conclusions of analytical comparison analysis from Sect. 3.4 is testified. Time
domain simulation is employed to prove the accuracy of the proposed methodology
in frequency domain.
3.2 System Modeling for Damping Torque Analysis of a Power System. . . 65
3.2 System Modeling for Damping Torque Analysis of a

Power System Affected by Grid-Connected Wind Power
Induction Generators
Philips-Heffron model is widely used for damping torque analysis of a power system
[33], which is able to facilitate a deeper understanding of damping mechanism of
WPIGs. Therefore, different from the linearized model presented in Chap. 2 for
modal analysis, Philips-Heffron model of a power system with grid-connected wind
power induction generators is established in Sect. 3.2.
3.2.1 Phillips-Heffron Model of a Power System with Wind

Power Induction Generators as a Generic Open-Loop
Controller
The network equation of a multi-machine power system before the connection of a

WPIG can be expressed as
2 3
2 12 13 32 V
1 3
0 Y11 Y Y
6 7 6 22 23 7 6 7
4 0 5 ¼ 4 Y21 Y Y 54 V25 ð3:1Þ
Ig 31
Y 32
Y 33
Y g
V
where Vg and Ig is the vector of terminal voltage and current associated with SGs,
is the system
terminal 1 and 2 is the potential location for installing the WPIG, and Y
admittance matrix. After the WPIG is connected, according to Fig. 3.1, the network
can be written as
V1
Node 1 X1L
XwL
Multi- I1L
machine WPIG
Node L
Power VL Iw
Vw
System I2L
Node 2 X2L
V2
Fig. 3.1 Diagram of a multi-machine power system connected with a WPIG

8" # " #" # " # " #

> 0 0
Y 0 1
V I1L 13
Y
>
> 11 g
>
> ¼ þ þ V
< 0 0 Y0 2
V I2L 23
Y
22
" # ð3:2Þ
>
> V 1
>
> 32 g
33 V
: Ig ¼ Y31
> Y
2
þY
V
where Y 0 and Y 0 exclude X12 ¼ X1L þ X2L.

11 22
" # " #1 " # " # !
I1L jðX1L þ XwL Þ jXwL 1
V 1
Since ¼ w , it
V
I2L jXwL jðX2L þ XwL Þ 2
V 1
obtains
" # " # " # !
1
V 1 13
Y
¼Y 1 w Y
V L g
V ð3:3Þ
2
V
w
1 23
Y
where
" #" #
jðX1L þ XwL Þ jXwL 0
Y 0
w ¼ I þ
Y
11
,
jXwL jðX2L þ XwL Þ 0 0
Y22

L ¼ jðX1L þ XwL Þ
Y
jXwL
:
jXwL jðX2L þ XwL Þ
By substituting (3.3) into the second equation of (3.2), it should have

1 1 1 Y13
Ig ¼ Y31
Y
Y32 w
V þ Y33 Y31
Y
Y32 w L
Y g
V ð3:4Þ
1
w
Y23
By linearizing (3.3), it gives

ΔV1 Y13
¼ Y1 YI ΔVw YL ΔVg ð3:5Þ
ΔV2 w Y23
where ΔV1 ¼ ½ ΔV1x ΔV1y T and same form applies to other variables.
1 0 1 0
T
YI ¼ 0 1 0 1 , Y1 1 and Y
and YL are the expanded form of Y L
w w
respectively. Equation (3.4) can be also linearized to be

33 ½ Y31 Y13
ΔIg ¼ ½ Y31 Y32 Y1
w Y I ΔVw þ Y Y32 Y1
w YL ΔVg ð3:6Þ
Y23

As L ¼ V
Iw ¼ I1L þ I2L and V w jXwL Iw , it can have

Iw ¼ Y 2 V1
w þ Y
1 V ð3:7Þ
2
V
1 ¼ jðX1L þ X2L Þ
where Y and
X1L X2L þ X2L XwL þ XwL X1L
2 ¼ j½ X2L X1L
Y . (3.7) can be linearized to be
X1L X2L þ X2L XwL þ XwL X1L

ΔV1
ΔIw ¼ Y1 ΔVw þ Y2 ð3:8Þ
ΔV2

ΔV1
By substituting (3.5) into (3.8) and eliminating , it gives
ΔV2

Y13
ΔIw ¼ Y1 þ Y2 Y1
w Y I ΔVw Y2 Y1
w YL ΔVg ð3:9Þ
Y23
As the standard algebraic interface equations of a WPIG can be written as

T
ΔIw ¼ CwΔXw + DwΔVw, where ΔXw ¼ ½ Δs ΔEd ΔEq , by eliminating
ΔIw, (3.9) becomes

1 Y13
ΔVw ¼ Y1 Dw þ Y2 Y1
w Y I Cw ΔX w þ Y Y1
2 w YL ΔVg ð3:10Þ
Y23
By substituting (3.10) into (3.6), ΔVw is eliminated

1
ΔIg ¼ ½ Y31 Y32 Y1 YI Y1 Dw þ Y2 Y1 YI Cw ΔXw þ
0 w w 1
33 ½ Y31 Y32 Y1 YL Y13 þ ½ Y31 Y32 Y1 YI
Y
B w Y23 w C ð3:11Þ
B CΔVg
@ 1 Y13 A
Y1 Dw þ Y2 Y1
w Y I Y Y1
2 w YL
Y23
Then the reference frame transformation from x-y to d-q is applied to (3.11) by
introducing Δδ. After substituting the linearized form of the SG equation V g ¼ E0
q
jX0d Ig j Xq X0d Iq into (3.11) under d-q frame, ΔVg in (3.11) is eliminated and
(3.11) should have the form
ΔIg ¼ Rδ Δδ þ RE0q ΔE0q þ RIq ΔIq þ RXw ΔXw ð3:12Þ
Hence, ΔIg is converted to ΔId and ΔIq and (3.12) becomes
ΔId ¼ Rδd Δδ þ RE0q d ΔE0q þ RIq d ΔIq þ RXw d ΔXw

ð3:13Þ
ΔIq ¼ Rδq Δδ þ RE0q q ΔE0q þ RIq q ΔIq þ RXw q ΔXw
Phillips-Heffron model of a multi-machine power system [34] is
Δδ_ ¼ ω0 Δω
Δω_ ¼ M1 ðΔTE DΔωÞ
ð3:14Þ
ΔE_0q ¼ T1
d0 ðΔEQ þ ΔE fd Þ

ΔE_0 ¼ ΔE K ΔV T1
fd fd A g A
where

ΔTE ¼ ΔIq E0q0 þ Iq0 ΔE0q þ ΔIq Xq X0d Id0 þ Iq0 Xq X0d ΔId ,

ΔEQ ¼ ΔE0q Xd X0d ΔId and ΔVgd ¼ XqΔIq, ΔVgq ¼ ΔE0q X0d ΔId .
By substituting (3.13) into (3.14), Phillips-Heffron model of a multi-machine

power system with the interface equations of a WPIG can be derived. The above
derivations can be extended to accommodate the case of multiple WPIGs. To
facilitate the understanding, a typical example of a single machine infinite bus
(SMIB) system with a WPIG is presented in Appendix 3.1.
Therefore, on this basis, the general form of the explicit Phillips-Heffron linear-
ized model of a multi-machine power system considering the algebraic interface
equations of WPIGs can be established
2 _ 3 2 32 3
Δδ 0 ω0 I 0 0 Δδ
6 7 6 76 7
6 Δω_ 7 6 M1 K1 M1 D M1 K2 0 76 Δω 7
6 7¼6 76 7
6 ΔE_0 7 6 T1 K Td0-1 K T1 76 ΔE0 7þ
4 q 5 4 d0 4 0 3 d0 54 q 5
ΔE_0fd T1
A KA K5 0 T1
A KA K6 TA 1 ΔE0fd
2 3 2 3 2 3
0 0 0
6 M1 K 7 6 M1 K 7 6 M1 K 7
6 ω1 7 6 ω2 7 6 ω3 7
6 7 6 7 6 7
6 T-1 KE0 1 7Δs þ 6 T1 KE0 2 7ΔEd þ 6 T1 KE0 3 7ΔEq
4 d0 q 5 4 d0 q 5 4 d0 q 5
-
T K K1 1
T K K 1
T K K
A A E fd 1 A A E fd 2 A A E fd 3
ð3:15Þ
where state variables (Δδ, Δω, ΔE0q and ΔEfd) and matrix elements (ω0, M, D,
K1~K6, Td0, KA and TA) of SGs are defined in Chapter 3.1 of [34], Δs, ΔEd and
ΔEq is the vector of variation of slip and direct/quadrant-axis electromotive force of
WPIGs, and the rest elements (Kω1, Kω2, Kω3, KE0q 1 , KE0q 2 , KE0q 3 , KE fd 1 , KE fd 2 and
KE fd 3 ) can be determined by the above derivations.
According to (3.15), it can be noted that: (1) The linearized model presented in
(3.15) is an open-loop system with Δs, ΔEd and ΔEq as its control variables, since
the internal dynamics of WPIGs is not included. Hence, (3.15) is also named system-
side linearized model; (2) Only the state variables of induction generator (Δs, ΔEd
K1
⎡ Δs ⎤ ⎡ Kω 1⎤
T
⎢ ⎥ ⎢K ⎥ Δω ω 0I Δδ
⎢Δ Ed ⎥ ⎢ ω 2⎥
− ( pM + D)−1 p
⎢Δ Eq ⎥ ⎢⎣ Kω 3⎥⎦
⎣ ⎦
K2 K4 K5
− ⎡ K fd 1⎤
T
⎡ Δs ⎤
ΔE ⎢ ⎥ ⎢ ⎥
( pTd 0 + K3 ) ∑ ΔE −
'
q
−1
( pTA + I )
−1
KA ⎢K fd 2⎥ ⎢Δ Ed ⎥
− fd ⎢K fd 3⎥
⎣ ⎦ ⎢Δ Eq ⎥
⎣ ⎦
K6
[KE' 1, KE' 2, KE' 3 ]
q q q
[Δs, ΔE , ΔE ] d q
T
Fig. 3.2 System-side linearized model diagram of a power system integrated with WPIGs
Fig. 3.3 Representation of Δs

WPIG internal dynamics in
frequency domain (WPIG- ΔEd
ΔVw GWPIG (p)
side linearized model)
ΔEq
and ΔEq) have a direct impact on the system damping and other state variables (e.g.,
state variables of DFIG converter controllers) affect system via Δs, ΔEd and ΔEq.
For FSIG, Δs does not directly contribute to the system damping either since FSIG
rotor is a closed circuit and thus physically separate from the grid. However, to keep
a consistent form for the demonstration of WPIGs, Δs can be retained in (3.15) but
with Kω1 ¼ KE0q 1 ¼ KE fd 1 ¼ 0. The linearized model in (3.15) is illustrated by
Fig. 3.2 in frequency domain and p is the frequency domain operator.
3.2.2 Aggregate Model and Transfer Function of Wind

Power Induction Generators
The internal dynamics of WPIGs includes dynamics of induction generators and

converter controllers (if DFIG), which can be described by a set of first-order
differential equations. In frequency domain, these equations can be converted and
presented in the form of a SIMO controller as shown in Fig. 3.3, which is explained
in details in Sect. 3.4. The input of the controller (ΔVw) is terminal voltage
associated with WPIG bus, and the outputs of the controller (Δs, ΔEd and ΔEq) are
three state variables of WPIGs as mentioned in (3.15). Without losing generality, the
transfer function is written in a general format as GWPIG ðpÞ ¼
T
Gs ðpÞ GEd ðpÞ GEq ðpÞ , which can represent any type of WPIG. The full
representation of GWPIG(p) is derived in Sect. 3.4 based on the internal dynamics
of different types of WPIG.
Figures 3.2 and 3.3 form a closed-loop linearized model with a clear physical
insight for damping torque analysis of a power system affected by grid-connected
WPIGs. If any system disturbance happens (represented by ΔVw), there should be a
dynamic response from WPIG (reflected by (Δs, ΔEd and ΔEq). ThenΔs, ΔEd and
ΔEq will in turn impact SGs and hence the system according to Fig. 3.2. It can be
seen that the internal dynamics (i.e., GWPIG(p)) of WPIGs determines their dynamic
response and plays a critical role in the dynamic interaction. Therefore, different
internal dynamics is actually considered to be the major cause for different types of
WPIG to have different damping mechanisms, which is carefully compared and
investigated in this chapter.
3.3 Implementation Framework for Damping Torque

Analysis of a Power System Affected by Grid-
Connected Wind Power Induction Generators
Based on the models given in Figs. 3.2 and 3.3, a generic implementation framework
of damping torque analysis to evaluate the damping torque contributions from
different internal dynamic components of WPIGs and their impact on the system
critical oscillation mode is established as follows.
The forward path from Δs, ΔEd and ΔEq to the electromechanical oscillation
loop of SGs can be obtained from Fig. 3.2
8 h i
>
> Fws ðpÞ ¼ KF ðpÞ ðpTA þ IÞ1 KA K fd1 þ KE0q1 Kω1
>
>
>
< h i
FwEd ðpÞ ¼ KF ðpÞ ðpTA þ IÞ1 KA K fd2 þ KE0q2 Kω2 ð3:16Þ
>
>
>
> h i
>
: FwE ðpÞ ¼ KF ðpÞ ðpTA þ IÞ1 KA K fd3 þ KE0 Kω3
q q3
where KF(p) ¼ K2[(pTd0 + K3) þ (pTA + I)1 + KAK6]1, and Fws(p), FwEd ðpÞ and
FwEq ðpÞ are three m 1 matrices, assuming there are totally m SGs and l WPIGs in
the system. Hence, the electric torque provided by the main dynamic components of
WPIGs to electromechanical oscillation loop of SGs is
8
> ΔTws ¼ Fws ðpÞGs ðpÞΔVw
<
ΔTwEd ¼ FwEd ðpÞGEd ðpÞΔVw ð3:17Þ
>
:
ΔTwEq ¼ FwEq ðpÞGEq ðpÞΔVw
3.3 Implementation Framework for Damping Torque Analysis of a Power. . . 71
where ΔTws, ΔTwEd and ΔTwEq include the electric torque contribution of WPIGs to
all SGs and thus are m-dimention vectors, and ΔTw ¼ ΔTws þ ΔTwEd þ ΔTwEq .
It can be revealed that all three dynamic components have their independent system
channels to contribute to the system damping torque, which together forms the total
damping impact ΔTw of WPIGs.
Assume that the ith eigenvalue λi is the critical oscillation mode in the system.
According to the algebraic equations of the linearized model of a multi-machine
power system with WPIGs, it can obtain
ΔVw ¼ CVw Xg ΔXg ð3:18Þ
where ΔXg is the vector of state variables associated with SGs. If λi and vi is the ith
eigenvalue and associated right eigenvector of state matrix in (3.15), it can have
Xn
vig ai Xn
vik ai
ΔXg ¼ , Δωk ¼ ð3:19Þ
i¼1
p λ i i¼1
p λi
where vig is the vector inside vi corresponding to ΔXg, and vik is the element of vi
corresponding to Δωk (angular speed of the kth SG). Based on (3.18) and (3.19), the
relationship between ΔVw and Δωk can be derived.

X !
X n
vig ai n
vik ai
ΔVw ¼ CVw Xg Δωk ¼ γik Δωk ð3:20Þ
i¼1
p λi i¼1
p λi
Equation (3.17) can be further factorized to torque contribution of dynamic

components of each WPIG to each SG, and hence the electric torque provided by
different dynamics of the jth WPIG to the kth SG can be written as
8
> ΔT ¼ Fwskj ðλi ÞGsj ðλi Þγijk Δωk
< wskj
ΔTwEd kj ¼ FwEd kj ðλi ÞGEd j ðλi Þγijk Δωk ð3:21Þ
>
:
ΔTwEq kj ¼ FwEq kj ðλi ÞGEq j ðλi Þγijk Δωk
where subscript k and j denote the kth row and jth column element of corresponding
matrices for Fwskj(λi), FwEd kj ðλi Þ and FwEq kj ðλi Þ, and the jth row of corresponding
matrices for Gsj(λi), GEd j ðλi Þ, GEq j ðλi Þ and γijk. As the electric torque contribution
from the jth WPIG to the kth SG is the linear superposition of each main dynamic
component, ΔTwkj ¼ ΔTwskj þ ΔTwEd kj þ ΔTwEq kj . Similarly, considering the con-
X l
tributions from all the WPIGs, the electric torque of the kth SG ΔTwk ¼ ΔTwkj ,
j¼1
which is the kth element of ΔTw.
The sensitivity of λi with respect to the electric torque coefficient of the kth SG can
be computed to be
∂λi
Sik ¼ ¼ wik vik ð3:22Þ
∂TCwk
where wik is the element of λi associated left eigenvector wi corresponding to Δωk.
Thus, the variation of the ith eigenvalue λi caused by insertion of WPIG dynamics
can be assessed by employing Sik
X
m X
m X
l
Δλi ¼ Sik TCwk ¼ Sik TCwkj
k¼1 k¼1 j¼1
ð3:23Þ
X
m X
l
¼ Sik TCwskj þ TCwEd kj þ TCwEq kj
k¼1 j¼1
where TCwk and TCwkj are the electric torque coefficients of ΔTwk and ΔTwkj, and
TCwskj, TCwEd kj and TCwEq kj are the coefficients of each main dynamic component in
(3.21) respectively.
On the basis of the derivations above, the generic implementation framework for
damping torque analysis of multi-machine power system with different types of
WPIGs is established, which is based in frequency domain but capable of providing
deeper understandings about damping mechanisms than modal analysis. It can be
seen that if closed-form solution of the WPIG transfer function is given, by
substituting (3.16) into (3.21) and (3.23), both the damping torque contributions of
WPIG dynamics components and eigenvalue variation should have an explicit
expression, so that the damping mechanism of WPIGs can be easily examined and
revealed. Therefore, a comprehensive comparison analysis is carried out in Sect. 3.4
based on the proposed framework.
3.4 Analytical Comparison on Damping Effectiveness And

Robustness of Type 1 and 3 Wind Power Induction
Generators
In the presented implementation framework of damping torque analysis, the external

damping contribution channels for different types of WPIGs are quite similar as
shown in (3.16) and (3.17), and hence their transfer functions (internal contribution
channels) become a crucial part of comparison analysis, which are derived and
investigated in this section. To facilitate the comparison and understanding of their
damping mechanisms, a five-step model transformation from DFIG to FSIG are
employed, in which FSIG is treated as a special case of DFIG. In addition to DFIG
and FSIG, three transitional wound rotor generator models are proposed for com-
parison purposes and the corresponding linearized models are established.
3.4 Analytical Comparison on Damping Effectiveness And Robustness of Type. . . 73
3.4.1 Step 1: Detailed DFIG Model
Since the grid-side converter (GSC) controller of DFIG does not really affect the
damping of system oscillation [10, 34, 35], its dynamics is ignored in this study. The
linearized model of DFIGs considering the dynamics of both induction generator
and rotor-side converter (RSC) controller is established
2 3
Δs_
6 7
6 ΔE_ d 7
6 7
6 7
6 ΔE_ q 7
6 7
6 _ 7
6 ΔX ps1 7
6 7
6 ΔX_ 7
6 ps2 7
6 7
6 ΔX_ 7
4 qs1 5
ΔX_ qs2
2 3
M1 w Dw M1w K1w 0 0 0 0 0
6 7
6 0 K2w 0 K3w K4w 0 0 7
6 7
6 7
6 0 0 K5w 0 0 K6w K7w 7
6 7
6 7
¼6 0 KpsI1 K8w 0 0 0 0 0 7
6 7
6 0 7
6 0 KpsI2 K9w 0 KpsI2 K10w 0 0 7
6 7
6 0 7
4 0 0 KqsI1 K11w 0 0 0 5
0 0 KqsI2 K12w 0 0 KqsI2 K13w 0
2 3 2 3
Δs M1
w Ks
6 7 6 7
6 ΔEd 7 6 KEd 7
6 7 6 7
6 7 6 7
6 ΔEq 7 6 KEq 7
6 7 6 7
6 7 6 7
6 ΔXps1 7 þ 6 KpsI1 KXps1 7ΔVw
6 7 6 7
6 ΔX 7 6 K K 7
6 ps2 7 6 psI2 Xps2 7
6 7 6 7
6 ΔX 7 6 K K 7
4 qs1 5 4 qsI1 Xqs1 5
ΔXqs2 KqsI2 KXqs2
ð3:24Þ
where Mw is the inertia time constant, Dw is the damping coefficient, Xps1, Xps2, Xqs1
and Xqs2 are the state variables of RSC Integral Controllers illustrated in Fig. 3.4, and
KpsI1, KpsI2, KqsI1 and KqsI2 are relevant parameters of Integral Controllers. The
explicit form of matrix elements (K1w~K13w, Ks, KEd , KEq , KXps1 , KXps2 , KXqs1 and
KXqs2 ) in (3.24) is given below
⎛ X2 ⎞ X V
− s ⎜⎜ X rr − m ⎟⎟ I rd + s m w − rrIrq
Ps Irq ⎝ X ss ⎠ X ss
− Kpsp1 − Kpsp2 +
+ I sqref Xss I rqref +
P s
ref
∑ KpsI1 +
−
Xm + ∑ K psI2
∑ Vrq
p X ps1 p Xps2
Kqsp1 Kqsp2
+ I sdref Xss + I rdref +
Qsref ∑ KqsI1 + Xm
− ∑ + ∑ KqsI2
∑ Vrd
− p Xqs1 − − p Xqs2 +
Qs Vw Ird ⎛ X2 ⎞
Xm s ⎜⎜ Xrr − m ⎟⎟ I rq − rrIrd
⎝ X ss ⎠
Fig. 3.4 Configuration diagram of RSC control system model of DFIGs
Xrr Eq0 Kpsp2 Xss

K1w ¼ Isd0 þ , K2w ¼ I þ Kpsp1 jVw0 j ,
Xss Xrr
X2m Xss Xrr Xm2
Kpsp2 Xss Xm Kqsp2 Xss

K3w ¼ , K4w ¼ , K5w ¼ I þ Kqsp1 jVw0 j ,
Xrr Xrr Xss Xrr Xm2
Kqsp2 Xss Xm Xrr jVw0 j

K6w ¼ , K7w ¼ , K8w ¼ ,
Xrr Xrr Xss Xrr X2m
Xss Xrr Xss
K9w ¼ I þ Kpsp1 jVw0 j , K10w ¼ ,
X2m Xss Xrr Xm Xm
Xrr jVw0 j Xss Xrr
K11w ¼ , K12w ¼ I þ Kqsp1 jVw0 j ,
X2m Xss Xrr Xss Xrr Xm Xm
2
Xss Xss Ed0 ½ Vwx0 Vwy0

K13w ¼ , Ks ¼ 2 ,
Xm Xm Xss Xrr jVw0 j
Kpsp1 Kpsp2 Xss Isq0 ½ Vwx0 Vwy0
KEd ¼ ,
Xrr jVw0 j

Xrr jVw0 j Xrr jVw0 j
Kqsp2 Xss þ Kqsp1 Isd0 ½ Vwx0 Vwy0
Xss Xrr X2m Xss Xrr X2m
KEq ¼
Xrr jVw0 j
Isq0 ½ Vwx0 Vwy0 Kpsp1 Xss Isq0 ½ Vwx0 Vwy0
KXps1 ¼ , KXps2 ¼ ,
jVw0 j Xm jVw0 j

Isd0 Xrr
KXqs1 ¼ þ ½ Vwx0 Vwy0 ,
jVw0 j Xss Xrr X2m

Xss Xrr jVw0 j Xrr
KXqs2 ¼ Kqsp1 Isd0 ½ Vwx0 Vwy0 :
Xm jVw0 j Xss Xrr X2m Xss Xrr X2m
where the subscript 0 denotes the steady-state value of variables, d-q denotes the
WPIG reference frame and x-y denotes the system reference frame.
As the study focus is on the induction generator and RSC basic control, there is no
additional damping control imposed to the RSC controller. The linearized model of
RSC control can also be illustrated in frequency domain as shown in Fig. 3.4.
According to Fig. 3.4, the transfer functions mentioned in previous section can be
extended to their full explicit representation as presented in (3.25).
8 8 2 391
>
> > KpsI2 K10w KpsI1 K8w >
>
> >
< þ >
7=
> K K K
1 6
4w psI2 9w
>
> 6 K K K p 7
>
> Ed
G ð p Þ ¼ I ð pI K Þ 4
3w psI1 8w
þ 5>
>
> >
>
2w
p p >
>
> : ;
>
>
>
> 2 3
>
> K K K K
>
> K4w
psI2 10w psI1 Xps1
þ KpsI2 KXps2
>
> 6 K3w KpsI1 KXps1 7
>
> ð Þ 1 6 þ þ
p 7
>
> pI K 2w 4 K E 5
>
>
d
p p
<
8 2 391
>
> < KqsI2 K13w KqsI1 K11w
þ =
>
> K K K K 7w KqsI2 K 12w
>
> GEq ðpÞ ¼ I ðpI K5w Þ1 4
6w qsI1 11w
þ
p
5
>
> : ;
>
> p p
>
> 2 3
>
> KqsI2 K13w KqsI1 KXqs1
>
> þ KqsI2 KXqs2
>
> 6
K7w
7
>
> 1 6
K6w K qsI1 KX p 7
>
> ðpI K Þ K
4 Eq þ qs1
þ 5
>
>
5w
p p
>
>
>
>
>
:
Gs ðpÞ ¼ ðpMw þ Dw Þ1 ½Ks þ K1w GEd ðpÞ
ð3:25Þ
It can be noted from Fig. 3.5 that: (1) The damping contributions of Δs, ΔEd and
ΔEq to the system mainly consist of two parts, i.e., the dynamics of induction
generator and RSC Integral Controller, which can be easily differentiated and
split. In (3.25), the induction generator dynamics part of transfer functions includes
the items associated with KEd and KEq , and the RSC controller dynamics part of
transfer functions is associated with RSC controller parameter KpsI1, KpsI2, KqsI1 and
KqsI2. That is to say, by setting these RSC controller parameters to zero, the damping
effect of RSC dynamics is removed; (2) ΔEd and ΔEq have completely separate
internal damping contribution channels due to the adoption of decoupled power
control of DFIG. ΔEd is related to the P-Control and ΔEq is related to the Q-Control;
(3) The damping contribution of Δs is only affected by its own dynamics and ΔEd as
the rotor motion is mainly determined by the active power control.
3.4.2 Step 2: DFIG Model Without RSC Dynamics
Normally, the damping effect of RSC Integral Controllers (i.e., converter-side

dynamics) is limited compared with the RSC Proportional Controllers which con-
tribute to the induction generator-side dynamics. In Step 2, converter-side dynamics
Induction Generator-side Dynamics Converter-side Dynamics
ΔVw Ks ∑ ( pM w + Dw)−1 Δs
Rotor Spin Dynamics ΔX qs1
∑ ΔVw
I
K6w p
KqsI1KXqs1
K1w
KqsI2K13w KqsI1K11w
Δ Eq ΔXqs2
( pI − K 5 w)
−1
∑ K7w
I
p ∑ KqsI2K12w
ΔVw KEq KqsI2KXqs2 ΔVw Q-Control
ΔX ps1
∑ ΔVw
I
K3w KpsI1KX ps1
p
KpsI2K10w KpsI1K8w
ΔE d ΔXps2
(pI − K 2 w )−1 ∑ ∑
I
K4w p
K psI2K9w
K Ed KpsI2KX ps2
ΔVw ΔVw P-Control
Fig. 3.5 WPIG-side linearized model diagram (internal dynamics of DFIG)
is not taken into account for the damping analysis and thus the first transitional
generator model is obtained. Hence, (3.24) is reduced to three first-order equations
associated with Δs, ΔEd and ΔEq only.
2 3 2 32 3 2 1 3
Δs_ M1w Dw M1w K1w 0 Δs M w Ks
6 _ 7 6 76 7 6 7
4 ΔE d 5 ¼ 4 0 K2w 0 54 ΔEd 5 þ 4 KEd 7
6
5ΔVw ð3:26Þ
_
ΔE q 0 0 K5w ΔEq KE q
The diagram of the linearized model in (3.26) only includes the left part of Fig. 3.5
(induction generator-side dynamics). However, the explicit form and value of
elements in (3.26) remain the same as in (3.24). As discussed in Step 1, the transfer
functions of this transitional generator model should become
8
>
> G ðpÞ ¼ ðpI K2w Þ1 KEd
< Ed
GEq ðpÞ ¼ ðpI K5w Þ1 KEq ð3:27Þ
>
>
:
Gs ðpÞ ¼ ðpMw Dw Þ1 ½Ks þ K1w GEd ðpÞ
Equation (3.27) can be also simply obtained by setting KpsI1 ¼ KpsI2 ¼

KqsI1 ¼ KqsI2 ¼ 0 in (3.25).
It can be revealed from Fig. 3.5 that although dynamics model (Integral Control-
lers) of RSC controller is removed and only the RSC algebraic model (Proportional
Controllers) is retained in this step, the basic configuration of linearized model and
damping contribution channels is not changed. In other words, the decoupled
structure of damping contribution of ΔEd and ΔEq is not determined by the
controller dynamics of RSC but actually by offset items of rotor voltage (or named
offset rotor voltage) in RSC controller, which is intensively studied in the next step.
3.4.3 Step 3: DFIG Model with Offset Rotor Voltage Only
On the basis of Step 2, by setting all the parameters of PI controllers of RSC to zero,
the second transitional generator model is established. Due to the only existence of
offset rotor voltage in RSC, both the rotor dynamic equations associated with ΔEd
and ΔEq are eliminated ( KEd ¼ KEq ¼ K2w ¼ K5w ¼ 0 ), and only rotor spin
dynamics is retained in this linearized model. Equation (3.26) becomes
Δs_ ¼ M1 1
w Dw Δs þ Mw Ks ΔVw ð3:28Þ
The diagram of the linearized model in (3.28) only includes the left upper corner
of Fig. 3.5 (rotor spin dynamics). Based on the diagram, the transfer function of the
second transitional generator model is derived to be
Gs ðpÞ ¼ ðpMw þ Dw Þ1 Ks ð3:29Þ
It can be seen from (3.28) that the elimination effect of the offset rotor voltage has
removed the ‘original’ dynamics of induction generator associated with ΔEd and
ΔEq and the ‘new’ generator-side dynamics of DFIG presented in Steps 1 and 2 are
actually brought by the RSC algebraic model (Proportional Controllers) in a
decoupled manner. This finding can be also proved by the explicit elements of
(3.24) in Sect. 3.4.1. The key elements of generator-side dynamics K2w and KEd
are mainly affected by the RSC P-Control algebraic model and K5w and KEq are
mainly affected by the RSC Q-Control algebraic model, after the introduction of
offset rotor voltage. In contrast, Ks and K1w are determined by the induction
generator parameter as well as the steady-state operating status, which are not
changed by the introduction of either RSC controller or offset rotor voltage.
3.4.4 Step 4: DFIG Model with Constant Rotor Voltage
Similar to Step 3, the effect of RSC PI controllers is ignored and rotor circuit is still
wound in Step 4. The third transitional wound rotor generator model with constant
rotor voltage output (sometimes also called open-loop control of RSC converter) is
proposed in this step. As the rotor voltage output remains constant, the offset items
of rotor voltage are removed and hence the ‘original’ generator-side dynamics is
reflected. As a result, the configuration of the linearized model is changed to Fig. 3.6
and the linearized equations become
2 3 2 32 3 2 1 3
Δs_ M1 w Dw M1
w K1w 0 Δs Mw K s
6 _ 7 6 76 7 6 7
4 ΔE d 5 ¼ 4 K2w K3w K4w 54 ΔEd 5 þ 6 7
4 0 5ΔVw ð3:30Þ
ΔE_ q K5w K6w K7w ΔEq KEq
where the elements of state matrix have been renumbered due to the change of state
matrix configuration in this case and their explicit forms are given below
Xrr Eq0 Rr Xss
K1w ¼ Isd0 þ , K2w ¼ Eq0 , K3w ¼ K7w ¼ ,
X2m Xss Xrr X2m Xss Xrr
Xss Ed0 ½ Vwx0 Vwy0
K4w ¼ s0 , K5w ¼ Ed0 , K6w ¼ s0 , Ks ¼ 2 ,
Xm Xss Xrr jVw0 j
Rr X2 ½ Vwx0 Vwy0
KEq ¼ 2 m
Xm Xss Xrr jVw0 j
where the subscript 0 denotes the steady-state value of variables, d-q denotes the
WPIG reference frame and x-y denotes the system reference frame. Compared with
the explicit elements of (3.24) in Sect. 3.4.1, most elements in (3.30) are changed and
determined by steady-state value of variables (e.g., K2w, K4w, K5w and K6w). Ks and
K1w remains the same as indicated previously.
It can be demonstrated from Fig. 3.6 that the ‘original’ induction generator-side
dynamics has three significant features: (1) re-coupling of ΔEd and ΔEq represented
by K4w and K6w; (2) Feedback support from Δs to ΔEd and ΔEq represented by and
K5w; (3) The removal of KEd owing to DFIG d-q reference frame setting. Therefore,
ΔVw Ks ∑ ( pMw + Dw)−1 Δ s

K1w K 5w
Δ Eq
( pI − K 7 w)
−1
∑ KEq ΔVw
K 4w
K2w
K6w
Δ Ed
( pI − K 3 w)−1 ∑
Fig. 3.6 WPIG-side linearized model diagram (internal dynamics of DFIG with constant rotor
voltage or FSIG)
the damping mechanism of the third transitional generator model has been dramat-
ically changed, which is also reflected by the transfer functions in (3.31), where all
three transfer functions include Ks and KEq , and hence are coupled with each other.
8 8 91
>
> < ðpI K3w Þ1 K2w ðpMw þ Dw Þ1 K1w =
>
> h i
>
> GEd ðpÞ ¼
>
> : K4w ðpI K7w Þ K5w ðpMw þ Dw Þ K1w þ K6w ;
1 1
>
>
< n h io
> K2w ðpMw þ Dw Þ1 Ks þ K4w ðpI þ K7w Þ1 K5w ðpMw þ Dw Þ1 Ks þ KEq
>
>
>
>
> G ðpÞ ¼ ðpM þ D Þ1 ½K þ K G ðpÞ
>
>
> s w w s 1w Ed
>
: 1
GEq ðpÞ ¼ ðpI K7w Þ K5w Gs ðpÞ þ K6w GEd ðpÞ þ KEq
ð3:31Þ
3.4.5 Step 5: FSIG Model
In particular condition, if constant rotor voltage in Step 4 is equal to zero, the third
transitional wound rotor generator model becomes FSIG with the rotor circuit
closed. As a result, the same linearized model ((3.30) and Fig. 3.6), transfer functions
(3.31) and damping mechanisms can be applied to FSIG. However, it is worthy to
mention that Δs of FSIG does not have a direct impact on the system damping as
stated in Sect. 3.2.1 due to the closed physical structure of rotor, although the
damping contribution from Δs is normally very small.
3.4.6 Main Findings in Analytical Comparison on Damping

Mechanism of Type 1 and 3 Wind Power Induction
Generators
By implementing the step-by-step model transformation analysis demonstrated

above, essential difference and inner connection between DFIG and FSIG in their
damping mechanisms of oscillation stability have been clearly revealed. The main
points from the comparison analysis are summarized as follows:
1. The damping contributions of DFIG dynamics are essentially from the RSC
control, which can be divided into two parts, i.e., the ‘new’ induction generator
dynamics (related to RSC Proportional Controllers) and RSC controller dynamics
(related to RSC Integral Controllers). The former represented by the RSC alge-
braic model normally accounts for the major part of the total damping contribu-
tions of DFIG.
2. The existence of the offset rotor voltage in RSC not only enables the decoupled
PQ control of DFIG, but also establishes the decoupled structure of internal
damping contribution channels of DFIG.
3. FSIG featured by the ‘original’ induction generator dynamics can be treated as a
special case of DFIG model only when the open-loop control of RSC is applied.
In this case, the damping contributions of FSIG only comes from the ‘original’
generator dynamics and are mainly determined by generator parameters and
changeable steady-state value of variables as proved by the explicit elements of
(3.30) in Sect. 3.4.4. The damping contributions channels of internal dynamics
are coupled with each other.
Therefore, based on the main points from the comparison of damping mecha-
nisms of DFIG and FSIG, it can be rigorously concluded that the damping effect of
DFIG is more robust since it is mainly determined by generator parameters and RSC
controller parameters as proved by the explicit elements of (3.24) in Sect. 3.4.1. On
the contrary, the damping impact of FSIG is comparatively limited and less control-
lable especially in the changing conditions due to the characteristic of ‘original’
induction generator dynamics. These significant empirical conclusions regarding
damping effectiveness and robustness of the two typical WPIGs have been rigor-
ously proved in an analytical manner.
By substituting the detailed format of transfer functions of DFIG and FSIG
derived in (3.25, 3.27, 3.29, 3.31) into (3.21) and (3.23), the damping torque analysis
can be carried out, the numerical comparison of which is designed and presented in
the next section.
3.5 Numerical Comparison
To demonstrate the proposed implementation framework and validate the findings

from the analytical comparison, 16-machine and 68-node NYPS-NETS power
system is employed and shown in Fig. 3.7. The models and parameters of the test
system are given in Appendix 3.2. For demonstration purposes, a WPIG-based wind
farm is planned to be connected to node 15 of the system. The connecting location
could be any other nodes in the system, which does not affect the demonstration and
validation. A typical set of WPIG parameters is provided in Appendix 3.2.
3.5.1 Example 3.1 (Base Case Comparison Study)
There are totally four inter-area oscillation modes in this test system and the selection
of critical mode would not actually affect the validation of the proposed method and
general results from the analytical comparison. Hence, the 31st eigenvalue λ31 is
selected to be the system critical inter-area oscillation mode in the case study here for
3.5 Numerical Comparison 81
NYPS
14 8
A3
66 1
60
25 26
41 40 48 47 53 28 29
61
2
27
18 9
42 1
17
67 38 16
31 32 30
3 15 24
15 62 63
2
9
69
21
4 14
46 10 11 22
A4
33 5 6 12
58
19 23
49 34 36 7 54 11
13 20 6 59
35 64 2 10 56 57 7
8 55
51 45
12 4 5
3
A5 50 39 37
52
A1
44
68 43 65 A2
16 13 NETS
Fig. 3.7 Diagram of 16-machine 68-bus NYPS-NETS test system integrated with a WPIG
demonstration purposes. Before the internal dynamics of the WPIG is taken into
account, the initial value of this critical inter-area oscillation mode is
ð0Þ
λ31 ¼ 0:1558 þ j3:3940, which is calculated from state matrix of the open-loop
system presented in (3.15). Two WPIG models (i.e., detailed DFIG model and FSIG
model) are demonstrated here.
By adopting (3.16), the forward paths from the main dynamic components (Δs,
ΔEd and ΔEq) of the DFIG or FSIG to the electromechanical oscillation loop of SGs
are obtained and compared in Table 3.1. It can be seen from Table 3.1 that the DFIG
has one more external damping contribution channel than FSIG due to the introduc-
tion of different algebraic models in (3.16) brought by different rotor structures, but
the sum of the weightings of the external damping contribution channels for the
DFIG and FSIG are approximately equal as a whole.
The internal dynamics (transfer functions) of the DFIG and FSIG is computed by
using (3.25) and (3.31) respectively and compared in Table 3.2.
Table 3.2 clearly demonstrates that the main difference of damping mechanisms
between the DFIG and FSIG actually lies in their internal damping contribution
channels especially in GEd ðλ31 Þ and GEq ðλ31 Þ. Also, since the transfer functions of
RSC integral controllers are less than 1% of each corresponding transfer function of
DFIG, they are not listed in Table 3.2 separately.
Then the eigenvalue variation caused by the introduction of the DFIG or FSIG
main dynamics components are estimated by (3.23) and shown in Table 3.3.
Based on Table 3.3, the estimation of the critical eigenvalue after the insertion of
the DFIG or FSIG dynamics can be obtained
Table 3.1 Comparison on norm of forward paths (external damping contribution channels) of
DFIG and FSIG to electromechanical oscillation loop of SGs
DFIG model FSIG model

|Fws(λ31)| jFwEd ðλ31 Þj FwE ðλ31 Þ jFwEd ðλ31 Þj FwE ðλ31 Þ
q q
G1 0.0031 0.0104 0.0151 0.0117 0.0154

G2 0.0034 0.0116 0.0245 0.0130 0.0248
G3 0.0043 0.0147 0.0283 0.0165 0.0287
G4 0.0067 0.0228 0.0502 0.0256 0.0508
G5 0.0041 0.0142 0.0341 0.0160 0.0345
G6 0.0057 0.0193 0.0376 0.0217 0.0382
G7 0.0062 0.0209 0.0380 0.0234 0.0385
G8 0.0038 0.0131 0.0285 0.0147 0.0289
G9 0.0014 0.0052 0.0298 0.0058 0.0301
G10 0.0011 0.0036 0.0081 0.0041 0.0082
G11 0.0020 0.0070 0.0194 0.0079 0.0196
G12 0.0005 0.0017 0.0046 0.0020 0.0047
G13 0.0003 0.0010 0.0020 0.0012 0.0021
G14 0.0001 0.0002 0.0011 0.0003 0.0011
G15 0.0000 0.0000 0.0004 0.0000 0.0004
G16 0.0000 0.0001 0.0011 0.0001 0.0011
Table 3.2 Comparison on DFIG model ΔVWX ΔVWY

transfer functions (Internal
Gs(λ31) 0.0037 j0.1560 0.0009 – j0.0381
Damping Contribution
Channels) of DFIG and FSIG GEd ðλ31 Þ 0.0454 j0.0049 0.0111 j0.0012
GEq ðλ31 Þ 1.1412 j0.1217 0.2786 – j0.0297
FSIG model ΔVWX ΔVWY
Gs(λ31) 0.0074 j0.0072 0.0018 j0.0018
GEd ðλ31 Þ 0.2208 þ j0.0098 0.0539 þ j0.0024
GEq ðλ31 Þ 0.0907 j0.2901 0.0221 j0.0708
Table 3.3 Eigenvalue

ΔλDFIG
31 ΔλFSIG
31
variation brought by main
dynamic components of DFIG Δs 0.0008 þ j0.0005 0
and FSIG ΔEd 0.0005 j0.0007 0.0016 j0.0029
ΔEq 0.0248 j0.0443 0.0067 þ j0.0061
Total 0.0261 j0.0446 0.0082 þ j0.0032
ð0Þ
λDFIG
31est ¼ λ31 þ Δλ31
DFIG
¼ 0:1819 þ j3:3494 ð3:32aÞ
ð0Þ
λFSIG
31est ¼ λ31 þ Δλ31
FSIG
¼ 0:1640 þ j3:3972 ð3:32bÞ
To verify the results in (3.32a, 3.32b), the eigenvalue of the closed-loop system
including the DFIG or FSIG dynamics is calculated to be
λDFIG
31 ¼ 0:1822 þ j3:3496 ð3:33aÞ
λFSIG
31 ¼ 0:1640 þ j3:3975 ð3:33bÞ
Therefore, by comparing the results of (3.32a, 3.32b) and (3.33a, 3.33b), the
implementation framework of damping torque analysis is validated. It can be
summarized from the above comparison analysis that: (1) Both DFIG and FSIG
play a positive role in damping power system oscillation for this study case owing to
the positive damping impact from each dynamic component; (2) Given the same
system network and loading condition, the external damping contribution channels
of the DFIG and FSIG tend to be roughly equal. The difference in the total damping
effectiveness is essentially brought by their different internal damping contribution
channels.
Time domain simulation is also carried out for verification of the proposed
damping torque analysis. A three-phase short-circuit fault is applied to node 1 at
0.2s and cleared at 0.3s. G5 and G15 are the main generators related to the critical
oscillation mode. Hence, the power angle difference between G5 and G15 is plotted
in Fig. 3.8.
Fig. 3.8 Observation of G5-G15 power angle curve with different types of WPIG models
Fig. 3.8 shows that the non-linear simulation results align with the eigenvalue
estimation from frequency domain analysis in Table 3.3 and (3.32a, 3.32b). The
damping contribution from the FSIG is quite limited, which does not make a notable
change to the power oscillation curve compared with the case of open-loop system.
The damping effect of the DFIG is better although it has a slightly bigger initial
oscillating amplitude.
3.5.2 Example 3.2 (Comparison Study Under Different Wind

Penetration Conditions)
The damping torque analysis to compare damping effectiveness in a single wind

speed condition is demonstrated above. In this section, in order to provide useful
guidance to system operator for the real-time operation of WPIGs, the damping
robustness of DFIG and FSIG is further assessed and compared under different wind
penetration conditions. To simulate the intermittence of the wind power in the real
case, the power output of DFIG and FSIG is set in a range from the cut-in power0.2 p.
u to the rating power 2.0 p. u. As G5 is the main related synchronous generator for
the critical oscillation mode, the total damping torque provided by the dynamic
components of DFIG and FSIG to G5 is computed by (3.21) respectively under
different wind power levels in Fig. 3.9. Then the real part of critical eigenvalue of the
closed-loop system considering DFIG and FSIG dynamics and open-loop system
without any WPIG dynamics are calculated and displayed in Fig. 3.10.
It can be demonstrated by results in Fig. 3.9 and Fig. 3.10 that:
1. Although the damping of open-loop system (calculated from (3.15)) decreases
with the increasing wind power, both DFIG and FSIG dynamics contribute a
positive damping torque to the critical oscillation mode in this case. Generally,
the damping effectiveness of FSIG is limited when compared with DFIG in the
same loading condition.
2. The damping torque contribution of DFIG and FSIG in Fig. 3.9 is in exact
proportion to their impact on critical eigenvalue variation in Fig. 3.10 (i.e., the
difference between ‘open-loop system’ and two ‘closed-loop systems’).
3. DFIG shows more robustness in damping contribution than FSIG under different
wind penetration conditions as it is indicated in Sect. 3.4 that the damping effect
of DFIG is mainly determined by generator parameters and RSC control while
that of FSIG is affected by generator parameters and changing FSIG operating
point. Hence, once the generator is designed, FSIG shows less robustness of
damping effect.
Fig. 3.9 Comparison on total damping torque provided by DFIG and FSIG under different wind
power output levels
Fig. 3.10 Comparison on the real part of critical eigenvalue of DFIG and FSIG connected systems
under different wind power output levels
3.6 Summary of Analytical and Numerical Comparison

Analysis
The main conclusions drawn from the analytical comparison in Sect. 3.4 and
validated by the numerical comparison in Sect. 3.5 are summarized in the following:
1. The damping effect of DFIG mainly determined by induction generator param-
eters and RSC control (‘new’ induction generator dynamics plus RSC controller
dynamics) is more robust under different wind conditions and also more effective
if the RSC control parameters are properly tuned;
2. The damping effect of FSIG determined by induction generator parameters and
changeable FSIG operating point (‘original’ induction generator dynamics) is
comparatively limited and less robust in the changing wind conditions;
3. The inner connection between DFIG and FSIG is that DFIG will degenerate to
FSIG and lose its advantageous properties in damping mechanism if open control
of RSC is applied. The presented work can effectively facilitate the system
operator’s understanding of different dynamics of DFIG and FSIG in a complex
operational environment.
Appendix 3.1: A Typical Example of a SMIB System

with Interface Equations of a WPIG
From Fig. 3.11, it can obtain
XtL XLb
Vt VL Vb
ItL ILb
SG
Iw XwL
Vw
WPIG
Fig. 3.11 Diagram of a SMIB power system connected with a WPIG

Appendix 3.1: A Typical Example of a SMIB System with Interface. . . 87

ILb ¼ ItL þ Iw ¼ ItL þ Vw VL ¼ ItL þ Vw Vt þ jXtL ItL ð3:34Þ
jXwL jXwL
t ¼ jXtL ItL þ jXLb ILb þ V

V b ð3:35Þ
Substituting (3.34) into (3.35) gives

ItL ¼ jXwL XLb XLb b
1þ Vt Vw V
XtL XLb þ XLb XwL þ XwL XtL XwL XwL
ð3:36Þ
jXwL XLb XLb
¼ 1þ Vt Vw 1
XtL XLb þ XLb XwL þ XwL XtL XwL XwL

As L ¼ V
Iw ¼ ItL þ ILb and V w jXwL Iw , it can have
Iw ¼ Y w þ Y
1 V t þ
2 V jXtL
ð3:37Þ
XtL XLb þ XLb XwL þ XwL XtL
1 ¼ jðXtL þ XLb Þ 2 ¼ jXLb

where Y ,Y .
XtL XLb þ XLb XwL þ XwL XtL XtL XLb þ XLb XwL þ XwL XtL
(3.37) can be linearized to be
ΔIw ¼ Y1 ΔVw þ Y2 ΔVt ð3:38Þ
As the standard algebraic linearized model of a WPIG can be written as

ΔIw ¼ CwΔXw + DwΔVw, by eliminating ΔIw, (3.38) becomes
ΔVw ¼ ðY1 Dw Þ1 Cw ΔXw ðY1 Dw Þ1 Y2 ΔVt ð3:39Þ
Substituting (3.39) into the linearized equation of (3.36) to eliminate ΔVw gives
ΔItL ¼ R1IB ΔVt þ R2IB ΔXw ð3:40Þ
Transforming (3.40) from the Infinite Bus reference frame to d-q reference frame
by introducing Δδ, (3.40) becomes
ΔItL ¼ R1 ΔVt þ R2 ΔXw þ R3 Δδ ð3:41Þ
Substituting the linearized SG equation ΔVtd ¼ XqΔItLq and ΔVtq ¼ ΔE0q X0d
ΔItLq into (3.41) and decomposing ΔItL to ΔItLd and ΔItLq gives
ΔItLd ¼ Rδd Δδ þ RE0q d ΔE0q þ RIq d ΔItLq þ RXw d ΔXw

ð3:42Þ
ΔItLq ¼ Rδq Δδ þ RE0q q ΔE0q þ RIq q ΔItLq þ RXw q ΔXw
Then (3.42) is substituted into the SMIB system linearized model in (3.43) and
the Phillips-Heffron model of a SMIB system with interface equations of a WPIG is
derived.
Δδ_ ¼ ω0 Δω
Δω_ ¼ M1 ðΔTE DΔωÞ
ð3:43Þ
ΔE_0q ¼ T1
d0 ðΔEQ þ ΔE fd Þ
ΔE_ fd ¼ ðΔE fd KA ΔVt ÞΔT1
A
where

ΔTE ¼ ΔItLq E0q0 þ Iq0 ΔE0q þ ΔItLq Xq X0d ItLd0 þ ItLq0 Xq X0d ΔItLd ,

ΔEQ ¼ ΔE0q Xd X0d ΔItLd , ΔVtd ¼ XqΔItLq and ΔVtq ¼ ΔE0q X0d ΔItLd .
Appendix 3.2: Data of Examples 3.1 and 3.2

Power System [36] (Tables 3.4, 3.5 and 3.6)
All the synchronous generators employ sixth-order detailed model with damping
D ¼ 0.0. The structure of first-order excitation system model is shown by Fig. 3.12.
The parameters of excitation system model are KA ¼ 3.95, TA ¼ 0.1s,
efdmax ¼ 10.0, efdmin ¼ 10.0.
Data of DFIG and FSIG
Induction Generator Parameters
Mw ¼ 3.4s, Dw ¼ 0, Rr ¼ 0.0007, Xs ¼ 0.0878, Xr ¼ 0.0373, Xm ¼ 1.3246,

Xr3 ¼ 0.05, Xss ¼ Xs þ Xm, Xrr ¼ Xr þ Xm, Pw ¼ 2.0 p. u., Vw ¼ 1.015 p. u.
Control Parameters of RSC
Kpsp1 ¼ Kqsp1 ¼ 0.2, Kpsp2 ¼ Kqsp2 ¼ 1,

KpsI1 ¼ KqsI1 ¼ 12.56 s1, KpsI2 ¼ KqsI2 ¼ 62.5 s1.
Appendix 3.2: Data of Examples 3.1 and 3.2 89
Table 3.4 Bus data

Reactive
Bus Active power Reactive power Active power power
no. Voltage Angle consumption consumption generation generation
1 – – 2.5270 1.1856 – –
2 – – 0.0000 0.0000 – –
3 – – 3.2200 0.0200 – –
4 – – 5.0000 1.8400 – –
5 – – 0.0000 0.0000 – –
6 – – 0.0000 0.0000 – –
7 – – 2.3400 0.8400 – –
8 – – 5.2200 1.7700 – –
9 – – 1.0400 1.2500 – –
10 – – 0.0000 0.0000 – –
11 – – 0.0000 0.0000 – –
12 – – 0.0900 0.8800 – –
13 – – 0.0000 0.0000 – –
14 – – 0.0000 0.0000 – –
15 – – 3.2000 1.5300 – –
16 – – 3.2900 0.3200 – –
17 – – 0.0000 0.0000 – –
18 – – 1.5800 0.3000 – –
19 – – 0.0000 0.0000 – –
20 – – 6.8000 1.0300 – –
21 – – 2.7400 1.1500 – –
22 – – 0.0000 0.0000 – –
23 – – 2.4800 0.8500 – –
24 – – 3.0900 0.9200 – –
25 – – 2.2400 0.4700 – –
26 – – 1.3900 0.1700 – –
27 – – 2.8100 0.7600 – –
28 – – 2.0600 0.2800 – –
29 – – 2.8400 0.2700 – –
30 – – 0.0000 0.0000 – –
31 – – 0.0000 0.0000 – –
32 – – 0.0000 0.0000 – –
33 – – 1.1200 0.0000 – –
34 – – 0.0000 0.0000 – –
35 – – 0.0000 0.0000 – –
36 – – 1.0200 0.1946 – –
37 – – 60.0000 3.0000 – –
38 – – 0.0000 0.0000 – –
39 – – 2.6700 0.1260 – –
40 – – 0.6563 0.2353 – –
41 – – 10.0000 2.5000 – –
(continued)
Table 3.4 (continued)

Reactive
42 – – 11.5000 2.5000 – –
43 – – 0.0000 0.0000 – –
44 – – 2.6755 0.0484 – –
45 – – 2.0800 0.2100 – –
46 – – 1.5070 0.2850 – –
47 – – 2.0312 0.3259 – –
48 – – 2.4120 0.0220 – –
49 – – 1.6400 0.2900 – –
50 – – 1.0000 1.4700 – –
51 – – 3.3700 1.2200 – –
52 – – 24.7000 1.2300 – –
53 1.0450 – – – 2.5000 –
54 0.9800 – – – 5.4500 –
55 0.9830 – – – 6.5000 –
56 0.9970 – – – 6.3200 –
57 1.0110 – – – 5.0520 –
58 1.0500 – – – 7.0000 –
59 1.0630 – – – 5.6000 –
60 1.0300 – – – 5.4000 –
61 1.0250 – – – 8.0000 –
62 1.0100 – – – 5.0000 –
63 1.0000 – – – 10.0000 –
64 1.0156 – – – 13.5000 –
65 1.0110 0.0000 – – – –
66 1.0000 – – – 17.8500 –
67 1.0000 – – – 10.0000 –
68 1.0000 – – – 40.0000 –
Table 3.5 Line data

Branch No. From bus To bus R X B Transformer ratio
1 1 2 0.0035 0.0411 0.6987 1
2 1 30 0.0008 0.0074 0.4800 1
3 2 3 0.0013 0.0151 0.2572 1
4 2 25 0.0070 0.0086 0.1460 1
5 2 53 0.0000 0.0181 0.0000 1.025
6 3 4 0.0013 0.0213 0.2214 1
7 3 18 0.0011 0.0133 0.2138 1
8 4 5 0.0008 0.0128 0.1342 1
9 4 14 0.0008 0.0129 0.1382 1
10 5 6 0.0002 0.0026 0.0434 1
11 5 8 0.0008 0.0112 0.1476 1
12 6 7 0.0006 0.0092 0.1130 1
13 6 11 0.0007 0.0082 0.1389 1
14 6 54 0.0000 0.0250 0.0000 1.07
15 7 8 0.0004 0.0046 0.0780 1
16 8 9 0.0023 0.0363 0.3804 1
17 9 30 0.0019 0.0183 0.2900 1
18 10 11 0.0004 0.0043 0.0729 1
19 10 13 0.0004 0.0043 0.0729 1
20 10 55 0.0000 0.0200 0.0000 1.07
21 12 11 0.0016 0.0435 0.0000 1.06
22 12 13 0.0016 0.0435 0.0000 1.06
23 13 14 0.0009 0.0101 0.1723 1
24 14 15 0.0018 0.0217 0.3660 1
25 15 16 0.0009 0.0094 0.1710 1
26 16 17 0.0007 0.0089 0.1342 1
27 16 19 0.0016 0.0195 0.3040 1
28 16 21 0.0008 0.0135 0.2548 1
29 16 24 0.0003 0.0059 0.0680 1
30 17 18 0.0007 0.0082 0.1319 1
31 17 27 0.0013 0.0173 0.3216 1
32 19 20 0.0007 0.0138 0.0000 1.06
33 19 56 0.0007 0.0142 0.0000 1.07
34 20 57 0.0009 0.0180 0.0000 1.009
35 21 22 0.0008 0.0140 0.2565 1
36 22 23 0.0006 0.0096 0.1846 1
37 22 58 0.0000 0.0143 0.0000 1.025
38 23 24 0.0022 0.0350 0.3610 1
39 23 59 0.0005 0.0272 0.0000 1
40 25 26 0.0032 0.0323 0.5310 1
41 25 60 0.0006 0.0232 0.0000 1.025
42 26 27 0.0014 0.0147 0.2396 1
(continued)

43 26 28 0.0043 0.0474 0.7802 1
44 26 29 0.0057 0.0625 1.0290 1
45 28 29 0.0014 0.0151 0.2490 1
46 29 61 0.0008 0.0156 0.0000 1.025
47 9 30 0.0019 0.0183 0.2900 1
48 9 36 0.0022 0.0196 0.3400 1
49 9 36 0.0022 0.0196 0.3400 1
50 36 37 0.0005 0.0045 0.3200 1
51 34 36 0.0033 0.0111 1.4500 1
52 35 34 0.0001 0.0074 0.0000 0.946
53 33 34 0.0011 0.0157 0.2020 1
54 32 33 0.0008 0.0099 0.1680 1
55 30 31 0.0013 0.0187 0.3330 1
56 30 32 0.0024 0.0288 0.4880 1
57 1 31 0.0016 0.0163 0.2500 1
58 31 38 0.0011 0.0147 0.2470 1
59 33 38 0.0036 0.0444 0.6930 1
60 38 46 0.0022 0.0284 0.4300 1
61 46 49 0.0018 0.0274 0.2700 1
62 1 47 0.0013 0.0188 1.3100 1
63 47 48 0.0025 0.0268 0.4000 1
64 47 48 0.0025 0.0268 0.4000 1
65 48 40 0.0020 0.0220 1.2800 1
66 35 45 0.0007 0.0175 1.3900 1
67 37 43 0.0005 0.0276 0.0000 1
68 43 44 0.0001 0.0011 0.0000 1
69 44 45 0.0025 0.0730 0.0000 1
70 39 44 0.0000 0.0411 0.0000 1
71 39 45 0.0000 0.0839 0.0000 1
72 45 51 0.0004 0.0105 0.7200 1
73 50 52 0.0012 0.0288 2.0600 1
74 50 51 0.0009 0.0221 1.6200 1
75 49 52 0.0076 0.1141 1.1600 1
76 52 42 0.0040 0.0600 2.2500 1
77 42 41 0.0040 0.0600 2.2500 1
78 41 40 0.0060 0.0840 3.1500 1
79 31 62 0.0000 0.0260 0.0000 1.04
80 32 63 0.0000 0.0130 0.0000 1.04
81 36 64 0.0000 0.0075 0.0000 1.04
82 37 65 0.0000 0.0033 0.0000 1.04
83 41 66 0.0000 0.0015 0.0000 1
84 42 67 0.0000 0.0015 0.0000 1
(continued)

85 52 68 0.0000 0.0030 0.0000 1
86 1 27 0.0320 0.3200 0.4100 1
Table 3.6 Machine data

Generator Bus Generator base
0 00 0
no. name (MVA) xa xd xd xd Td0 ðsÞ
G1 B53 100 0.0125 0.1000 0.0310 0.0250 10.20
G2 B54 100 0.0350 0.2950 0.0697 0.0500 6.56
G3 B55 100 0.0304 0.2495 0.0531 0.0450 5.70
G4 B56 100 0.0295 0.2620 0.0436 0.0350 5.69
G5 B57 100 0.0270 0.3300 0.0660 0.0500 5.40
G6 B58 100 0.0224 0.2540 0.0500 0.0400 7.30
G7 B59 100 0.0322 0.2950 0.0490 0.0400 5.66
G8 B60 100 0.0280 0.2900 0.0570 0.0450 6.70
G9 B61 100 0.0298 0.2106 0.0570 0.0450 4.79
G10 B62 100 0.0199 0.1690 0.0457 0.0400 9.37
G11 B63 100 0.0103 0.1280 0.0180 0.0120 4.10
G12 B64 100 0.0220 0.1010 0.0310 0.0250 7.40
G13 B65 200 0.0030 0.0296 0.0055 0.0040 5.90
G14 B66 100 0.0017 0.0180 0.00285 0.0023 4.10
G15 B67 100 0.0017 0.0180 0.00285 0.0023 4.10
G16 B68 200 0.0041 0.0356 0.0071 0.0055 7.80
G1 0.05 0.0690 0.0280 0.0250 1.50 0.035 42.0
G2 0.05 0.2820 0.0600 0.0500 1.50 0.035 30.2
G3 0.05 0.2370 0.0500 0.0450 1.50 0.035 35.8
G4 0.05 0.2580 0.0400 0.0350 1.50 0.035 28.6
G5 0.05 0.3100 0.0600 0.0500 0.44 0.035 26.0
G6 0.05 0.2410 0.0450 0.0400 0.40 0.035 34.8
G7 0.05 0.2920 0.0450 0.0400 1.50 0.035 26.4
G8 0.05 0.2800 0.0500 0.0450 0.41 0.035 24.3
G9 0.05 0.2050 0.0500 0.0450 1.96 0.035 34.5
G10 0.05 0.1150 0.0450 0.0400 1.50 0.035 31.0
G11 0.05 0.1230 0.0150 0.0120 1.50 0.035 28.2
G12 0.05 0.0950 0.0280 0.0250 1.50 0.035 92.3
G13 0.05 0.0286 0.0050 0.0040 1.50 0.035 248.0
G14 0.05 0.0173 0.0025 0.0023 1.50 0.035 300.0
G15 0.05 0.0173 0.0025 0.0023 1.50 0.035 300.0
G16 0.05 0.0334 0.0060 0.0055 1.50 0.035 225.0
Vtref efd0 efd max
+ +
Vt KA Δefd efd
Σ Σ
− 1+TAs +
efd min
Fig. 3.12 The first-order excitation system model of synchronous generator
References
1. Global Wind Energy Council (2016) Global Wind Report, Annual Market Update 2016. http://
www.gwec.net/
2. Thakur D, Mithulananthan N (2009) Influence of constant speed wind turbine generator on
power system oscillation. Electr Power Compo Syst 37:478–494
3. Fayek HM, Elamvazuthi I, Perumal N, Benkatesh B (2014) The impact of DFIG and FSIG
Wind Farms on the Small Signal Stability of a Power System. In: 5th International Conference
on Intelligent and Advanced Systems, Kuala Lumpur, pp 1–6.
4. Slootweg JG, Kling WL (2003) The impact of large scale wind power generation on power
system oscillations. Electr Power Syst Res 67:9–20
5. Mei F, Pal BC (2007) Modal analysis of a grid-connected doubly fed induction generator. IEEE
Trans Energy Convers 22(3):728–736
6. Wu F, Zhang XP, Godfrey K, Ju P (2007) Small signal stability analysis and optimal control of a
wind turbine with doubly fed Induction generator. IET Gener Transm Distrib 1(5):751–760
7. Sanchez-Gasca JJ, Miller NW, Price WW (2004) A modal analysis of a two-area system with
significant wind Power penetration. In: Power systems conference and exposition 2, New York,
pp 1148–1152
8. Mendonca A, Pecas Lopes JA (2005) Impact of large scale wind power integration on small
signal stability. Future Power Syst:1–5
Syst 29(5):2248–2256
10. Gautam D, Vittal V, Harbour T (2009) Impact of increased penetration of DFIG-based wind
turbine generators on transient and small signal stability of power systems. IEEE Trans Power
Syst 24(3):1426–1434
11. Jafarian M, Ranjbar AM (2012) Interaction of the dynamics of doubly fed wind generators with
power system electromechanical oscillations. IET Renew Power Gen 7(2):89–97
12. Garmroodi M, Hill DJ, Verbic G, Ma J (2016) Impact of tie-line power on inter-area modes with
increased penetration of wind power. IEEE Trans Power Syst 31(4):3051–3060
13. Vittal E, O'Malley M, Keane A (2012) Rotor angle stability with high penetrations of wind
15. Tsourakis G, Nomikos BM, Vournas CD (2009) Effect of wind parks with doubly fed
asynchronous generators on small-signal stability. Elect Power Syst Res 79(1):190–200
16. Fan L, Miao Z, Osborn D (2008) Impact of doubly fed wind turbine generation on inter-area
oscillation damping. In: IEEE Power Eng Soc General Meeting, Pittsburgh, PA, pp 1–8
References 95
18. Kunjumuhammed LP, Pal BC, Anaparthi KK, Thornhill NF (2013) Effect of wind penetration
on power system stability. In: IEEE power and energy Soc general meeting (PES), Vancouver,
BC, pp 1–5.
19. Gautam D, Goel L, Ayyanar R, Vittal V, Harbour T (2011) Control strategy to mitigate the
impact of reduced inertia due to doubly fed induction generators on large power systems. IEEE
Trans Power Syst 26(1):214–224
20. Arani MFM, El-Saadany EF (2013) Implementing virtual inertia in DFIG-based wind power
21. Ma J, Qiu Y, Zhang WB, Song ZX, Thorp JS (2016) Research on the impact of DFIG virtual
inertia control on power system small-signal stability considering the phase-locked loop. IEEE
22. Dominguez-Garcia J, Bianchi FD, Gomis-Bellmunt O (2013) Control signal selection for
damping oscillations with wind power plants based on fundamental limitations. IEEE Trans
Power Syst 28(4):4274–4281
23. Arani MFM, Mohamed Y (2015) Analysis and impacts of implementing droop control in
DFIG-based wind turbines on microgrid/weak-grid stability. IEEE Trans Power Syst 30
(1):385–396
24. Surinkaew T, Ngamroo I (2014) Coordinated robust control of DFIG wind turbine and PSS for
stabilization of power oscillations considering system uncertainties. IEEE Trans Sustain Energ
5(3):823–833
25. Surinkaew T, Ngamroo I (2016) Hierarchical co-ordinated wide area and local controls of DFIG
wind turbine and PSS for robust power oscillation damping. IEEE Trans Sustainable Energ 7
(3):943–955
26. Liu Y, Gracia JR, King TJ, Liu YL (2015) Frequency regulation and oscillation damping
contributions of variable-speed wind generators in the U.S. Eastern Interconnection (EI). IEEE
Trans Sustain Energ 6(3):951–958
27. Zeni L, Eriksson R, Goumalatsos S, Altin M, Sorensen P, Hansen A, Kjar P, Hesselbak B
(2016) Power oscillation damping from VSC-HVDC connected offshore wind power plants.
IEEE Trans Power Deliver 31(2):829–838
28. Singh M, Allen AJ, Muljadi E, Gevorgian V, Zhang YC, Santoso S (2015) Interarea oscillation
damping controls for wind power plants. IEEE Trans Sustain Energ 6(3):967–975
29. Mishra Y, Mishra S, Li FX, Dong ZY, Bansal RC (2009) Small-signal stability analysis of a
DFIG-based wind power system under different modes of operation. IEEE Trans Energy
Conver 24(4):972–982
30. Mokhtari M, Aminifar F (2014) Toward wide-area oscillation control through doubly-fed
induction generator wind farms. IEEE Trans Power Syst 29(6):2985–2992
31. Chandra S, Gayme DF, Chakrabortty A (2014) Coordinating wind farms and battery manage-
ment systems for inter-area oscillation damping: a frequency-domain approach. IEEE Trans
Power Syst 29(3):1454–1462
32. Jamehbozorg A, Radman G (2015) Small-signal analysis of power systems with wind and
energy storage units. IEEE Trans Power Syst 30(1):298–305
33. Yu YN (1983) Electric power system dynamics. Academic Press, New York
34. CIGRE Technical Brochure on Modeling and Dynamic Behavior of Wind Generation as it
Relates to Power System Control and Dynamic Performance, Working Group 01 (2006)
Advisory Group 6, Study Committee C4 Draft Rep
35. Rodriguez JM, Fernandez JL, Beato D, Iturbe R, Usaola J, Ledesma P, Wilhelmi JR (2002)
Incidence on power system dynamics of high penetration of fixed speed and doubly fed wind
energy systems: study of the Spanish case. IEEE Trans Power Syst 17(4):1089–1095
36. Rogers G (2000) Power system oscillations. Kluwer, Norwell, MA
Chapter 4
Modal Analysis of Small-Signal Angular
Stability of a Power System Affected by
Grid-Connected DFIG
This chapter introduces modal analysis of small-signal angular stability of a power

system affected by a grid-connected DFIG. First, a method of decomposed modal
analysis is proposed. The method can separately examine the impact of load flow
change and dynamic interactions brought about by the grid-connected DFIG. Sec-
ond, the impact of dynamic interactions introduced by the grid-connected DFIG on
power system small-signal angular stability is examined comprehensively. Two
indices to estimate the impact are introduced.
4.1 Decomposed Modal Analysis of the Impact of a Grid-

Connected DFIG on Power System Small-Signal
Angular Stability
Grid connection of a DFIG for the wind power generation brings about the change of
system load flow and introduces the dynamic interactions. Those are two main
factors that the grid-connected DFIG affects power system small-signal angular
stability. In this section, a method of decomposed modal analysis is presented
which can separately examine the impact of those two affecting factors introduced
by the DFIG. Separate examination provides a potential way to gain better under-
standing on and deeper insight into the mechanism about how the DFIG affects
power system small-signal angular stability.
4.1.1 Method of Decomposed Modal Analysis
For a multi-machine power system connected with a DFIG shown by Fig. 4.1, the
following linearized state-space model of the power system can be established

98 4 Modal Analysis of Small-Signal Angular Stability of a Power System. . .
DFIG
A multi- Pw + jQw Ps+ jQs
machine
power Pr + jQr
system Vw qw
GSC RSC
GSC control RSC control

system system
Fig. 4.1 A DFIG connected to a multi-machine power system

d ΔXg Agg Agd ΔXg ΔXg
¼ ¼A ð4:1Þ
dt ΔXd Adg Awd ΔXd ΔXd
where ΔXd is the vector of all the state variables associated with the DFIG, ΔXg is
the vector of all the state variables of the synchronous generators. By computing the
electromechanical oscillation modes (EOMs) of the power system from the state
matrix A, the effect of the DFIG on system small-signal angular stability can be
examined. Obviously, the computation gives the overall effect of the DFIG, which
includes that of change of system load flow brought about by the DFIG and that of
the dynamic interaction between the DFIG and the synchronous generators. Nor-
mally, separate examination of two affecting factors of the DFIG, i.e., the change of
system load flow and dynamic interaction, cannot be obtained from the conventional
modal analysis based on the state-space model of Eq. (4.1).
In order to examine the two affecting factors separately so as to understand how
the DFIG affects the EOMs of the power system, the closed-loop interconnected
model derived in 2.4 is used, where the DFIG is the open-loop subsystem in the
feedback loop and the rest of the power system (ROPS) is the open-loop subsystem
in the forward path. The closed-loop interconnected model is shown by Fig. 2.11 and
presented as Fig. 4.2. The effect of the dynamic interactions introduced by the DFIG
is individually estimated by applying the damping torque analysis (DTA). Thus, two
affecting factors of the DFIG can be examined separately.
4.1.1.1 Electric Torque Contributed by the Grid-Connected DFIG
In the power system of Fig. 4.1, the power output from the DFIG, Pw þ jQw, is the
physical exhibition of dynamic interactions between the DFIG and the synchronous
4.1 Decomposed Modal Analysis of the Impact of a Grid-Connected DFIG on. . . 99

power system with ΔPw
the DFIG Gvp (s)
ΔVw
+
Gvq (s)
ΔQw
Hdq (s)
Hdp (s)
DFIG subsystem
generators. Denote Pw0 þ jQw0 as the power output from the DFIG at the steady-state
operation of the power system. It can have
Pw þ jQw ¼ Pw0 þ jQw0 þ ΔPw þ jΔQw ð4:2Þ
Obviously ΔPw þ jΔQw characterizes the dynamic interactions between the

DFIG subsystem and the ROPS subsystem, i.e., the synchronous generators. The
state-space model of the ROPS subsystem is derived in Sect. 2.3 to be (2.126) and
given below
8
g g g p w q w
dt ð4:3Þ
:
ΔVw ¼ cgv ΔXg þ dvp ΔPw þ dvq ΔQw
T
Hence, the state-space model of the ROPS subsystem can be rewritten as

2 3 2 32 3 2 3 2 3
Δδ 0 ω0 I 0 Δδ 0 0
d6 7 6 76 7 6 7 6 7
4 Δω 5 ¼ 4 A21 A22 A23 54 Δω 5 þ 4 bP2 5ΔPw þ 4 bQ2 5ΔQw
dt ð4:4Þ
Δz A31 A32 A33 Δz bP3 bQ3

ΔVw ¼ Cg1 Cg2 Cg3 ΔXg þ dvp ΔPw þ dvq ΔQw
where Δδ and Δω is the vector of deviation of angular positions and speed variables
of the synchronous generators respectively, Δz is the vector of all the other state
variables of the synchronous generators in the power system, ΔVw is the deviation of
the magnitude of voltage at the terminal of the DFIG.
The transfer function model of the DFIG subsystem is derived in Sect. 2.4.2 to be
(see (2.130)),
(
ΔPw ¼ Hdp ðsÞΔVw
ð4:5Þ
ΔQw ¼ Hdq ðsÞΔVw
It is interesting to note that the DFIG only responds to the variation of the
magnitude of the terminal voltage, ΔVw, of the DFIG. In the power system, normally
dynamic variation of ΔVw is limited and so the dynamic response of the DFIG is
very small. This explains why normally the dynamic interactions introduced by the
DFIG are weak and the DFIG is considered to be “inertia-less”.
The closed-loop interconnected model shown by Fig. 4.4 can be redrawn in more
details from (4.4) and (4.5). The redrawn model is shown by Fig. 4.3. Figure 4.3
clearly depicts the two-ways dynamic interactions between the DFIG and the
synchronous generators. One of the ways is the response of the DFIG to ΔVw to
inject a variable power ΔPw þ jΔQw into the power system as described by (4.5).
The other way is the response of the power system to ΔPw þ jΔQw to generate ΔVw
as described by (4.4). From Fig. 4.3, the effect of the dynamic interactions intro-
duced by the DFIG can be estimated by applying the damping torque analysis (DTA)
as follows.
Transfer function matrix from the DFIG output power, ΔPw þ jΔQw, to its electric
torque contribution to the electromechanical oscillation loops of the synchronous
generators can be obtained from (4.4) or Fig. 4.3 to be
ROPS subsystem
A21
Δω w0 I Δδ
+ ( sI - A22 )-1
ΔTQ s
A23 A32
ΔTP
Δz
bP2 ( sI - A33 )-1 + A31
bP3 bQ3
bQ2
C g3 C g2 d vp C g1
H dp ( s )
D Pw DVw
+
DQw
H dq ( s )
d vq
DFIG subsystem
Fig. 4.3 Closed-loop interconnected model of the multi-machine power system with the grid-
connected DFIG
ΔTP
¼ GP ðsÞ ¼ bP2 þ A23 ðsI A33 Þ1 bP3
ΔPw
ð4:6Þ
ΔTQ
¼ GQ ðsÞ ¼ bQ2 þ A23 ðsI A33 Þ1 bQ3
ΔQw
Hence from (4.5) and (4.6), the electric torque contribution from the DFIG is
obtained to be

ΔTw ¼ ΔTP þ ΔTQ ¼ GP ðsÞHdp ðsÞ þ GQ ðsÞHdq ðsÞ ΔVw ð4:7Þ
Thus the electric torque provided by the DFIG to the kth synchronous generator in
the power system is

ΔTwk ¼ gPk ðsÞHdp ðsÞ þ gQk ðsÞHdq ðsÞ ΔVw ð4:8Þ
where gPk(s) and gQk(s)is the kth element of GP(s) and GQ(s) respectively.
4.1.1.2 Method of Decomposed Modal Analysis
Consider an assumed case that ΔPw þ jΔQw ¼ 0. This is when the DFIG is
completely decoupled dynamically with the ROPS. Thus, the only factor that the
DFIG affects the system small-signal angular stability is the change of system load
flow it introduces. Obviously, when ΔPw þ jΔQw ¼ 0, the DFIG is degraded into a
constant power source, Pw0 þ jQw0, as shown by (4.2), and the state-space model of
the ROPS of (4.4) becomes
2 3 2 32 3 2 3
Δδ 0 ω0 I 0 Δδ Δδ
d4
Δω 5 ¼ 4 A21 A22 A23 54 Δω 5 ¼ Ag 4 Δω 5 ð4:9Þ
dt
Δz A31 A32 A33 Δz Δz
Therefore, by modelling the DFIG as the constant power source and computing
the EOMs of the power system from state matrix Ag, the impact of the change of the
load flow brought about by the DFIG on system small-signal angular stability can be
determined. Denote λi ¼ – ξi þ jωi as the ith EOM of concern computed from the
open-loop state matrix Ag in (4.9). Denote ^λ i ¼ ^ξ i þ j^
ω i as the EOM of concern of
the closed-loop interconnected system of Fig. 4.3 corresponding to λi ¼ – ξi þ jωi.
The difference between the closed-loop and open-loop EOM of concern is Δλi ¼ ^λ i
λi which is caused by ΔPW þ jΔQw 6¼ 0 Δλi ¼ ^λ i λi can be estimated by the
DTA as follows.
Denote λi ¼ ξi þ jωi and vj, j ¼ 1, 2, M as the eigenvalues and associated
right eigenvectors of open-loop state matrix Ag in (4.9). Solution of (4.9) can be
expressed as
XM
v ja j XM
v jk a j
ΔXg ¼ , Δωk ¼ ð4:10Þ
j¼1
s λ j j¼1
s λj
where Δωk is the deviation of rotor speed of the kth synchronous generator and a
state variable, vjk is the element of vj corresponding to Δωk. From the second
equation of (4.4) and (4.5), it can have
Cg ΔXg
ΔVw ¼ ð4:11Þ
1 dvp Hdp ðsÞ dvq Hdq ðsÞ
From (4.10) and (4.11),
1 X
M
v ja j
Cg
ΔVw 1 dvp Hdp ðsÞ dvq Hdq ðsÞ j¼1 s λ j
¼
Δωk XM
v jk a j
j¼1
s λj
ð4:12Þ
1 XM
v ja j
ðs λi Þ Cg
1 dvp Hdp ðsÞ dvq Hdq ðsÞ j¼1
s λj
¼
XM
v jk a j
ð s λi Þ
j¼1
s λj
Let s ¼ λi in the above equation,
1 Cg vig
ΔVw ¼ Δωk ¼ γik ðλi ÞΔωk ð4:13Þ
1 dvp Hdp ðλi Þ dvq Hdq ðλi Þ vik
Hence, at the complex oscillation frequency λi ¼ ξi þ jωi, the damping torque

provided by the DFIG to the kth synchronous generator can be obtained from (4.8)
and (4.13) to be

ΔTwDk ¼ Re gPk ðλi ÞHdp ðλi Þ þ gQk ðλi Þ Hdq ðλi Þ γik ðλi Þ Δωk ¼ Dwk Δωk ð4:14Þ
where Re{} denotes the real part of a complex number. Let the sensitivity of the
EOM of concern to the damping torque coefficient of the kth synchronous generator
be
∂λi
Sik ¼ ð4:15Þ
∂Dwk
From (4.14) and (4.15),
X
N
Δλi ¼ Sik Dwk ð4:16Þ
k¼1
Thus, Δλi can be quantitatively calculated to estimate the impact of dynamic

interactions introduced by the DFIG on power system small-signal angular stability.
In addition, Eq. (4.16) clearly indicates that the DFIG contribute the damping torque
to the synchronous generators to affect the EOMs.
Based on the discussions above, the method to separately examine the impact of
change of system load flow and dynamic interactions brought about by the DFIG on
the EOMs can be summarized as follows.
1. Model the DFIG as a constant power source and derive the state matrix Ag.
Compute the EOM of concern from Ag, λi, to examine the effect of the change of
load flow brought about by the DFIG on the system small-signal angular stability.
2. Compute the impact of the dynamic interactions of the DFIG with the synchro-
nous generators on the oscillation mode to be Δλi by use of (4.16) with the full
dynamics of the DFIG being included.
3. The total impact of the DFIG is
^λ i λi þ Δλi ð4:17Þ
Integration of large number of wind farms, each of which may consist of tens
even hundreds of the DFIGs, will increase the dimension of state matrix, A, in (4.1)
dramatically. Numerical complexity and difficulty of modal analysis for A may grow
considerably in the examination of the impact of the integration of wind farms. The
proposed method of decomposed modal analysis provides a way to avoid the
problem. λi is computed from the state matrix Ag, dimension of which does not
change with the integration of the DFIGs because the DFIG’s dynamics are not
included. Thus the estimation of the impact of the dynamics of the DFIG, Δλi,
involves no modal analysis of state matrix with increased dimension. If Δλi is found
to be small, the DFIG can be modelled as a constant power source in the examina-
tion. This will reduce the computational burden significantly if the integration of
large number of wind farms needs to be examined.
Examination of dynamic interactions of the DFIG with the synchronous genera-
tors is very important to understand the involvement and behavior of the DFIG in
power system dynamic transient. The estimation can be carried out by calculating the
participation factors (PFs) of the state variables of the DFIG for the EOMs of
concern. Calculation of Δλi as proposed above provides an alternative way to assess
the dynamic interactions. The assessment does not need to carry out the modal
analysis of the closed-loop state matrix A with increased dimension. While nor-
mally, computation of the PFs is based on the modal analysis of A. In addition,
computation of Δλi can provide the estimation of not only the scale but also the
direction (positive to improve or negative to reduce the damping of the EOM of
concern) of the dynamic interactions.
4.1.2 Example 4.1
4.1.2.1 The Example Power System
Configuration of an example 16-machine 68-bus power system is shown by Fig. 4.4.

In this example, the simplified third-order model of the synchronous generator and
the first-order model of the AVR recommended in [1–3] were adopted. The bus data,
line data, machine data are given in Tables 4.10, 4.11, and 4.12 in Appendix 4.1
respectively. No PSS was installed on any synchronous generator. There are four
inter-area EOMs in the power system, which are crucial to the small-signal angular
stability of the example system. They are
ð0Þ ð0Þ
λ1 ¼ 0:1000 þ j5:3419, λ2 ¼ 0:1353 þ j4:6374,
ð0Þ ð0Þ
λ3 ¼ 0:0960 þ j3:9617, λ4 ¼ 0:2639 þ j3:0428
where superscript (0) indicates the value of the modes when no DFIG is connected to
the example power system.
In order to demonstrate the application of the method introduced in the previous
section, it is assumed that a DFIG is to be connected to the example power system.
The candidate locations of connection are node 8 and 16. Though the proposed
method is applicable for any type of the model of the synchronous generators and the
DFIG, in the example power system, a third-order model of the synchronous
generator and a simple first-order model of the automatic voltage regulator (AVR)
were used [1]. All the synchronous generators were not equipped with the power
NYPS
14 8
A3
66 1 60
53 25 26
41 40 48 47
2 28 29 61
18 27
1 9
42
67 38 17 16
30
31 32 15 24
3
15 62 63 9
14 21
46 10 11 4 22
A4
33 5 6 12 58
34 19 23
49 36 7 11
54 13
20 59
10 6
35 64 56 57
2 55
45 8 7
51 4
12 5
A5 37 3
50 39
A1
52
44 43
A2
68 65
NETS
16 13
Fig. 4.4 Configuration of example 16-machine 68-bus power system

system stabilizers (PSS). The loads were modelled as constant impedance. Nor-
mally, for the study of power system small-signal angular stability, such models of
the synchronous generator, the AVR and the loads are sufficient if the focus of the
study is not the impact of the synchronous generator, the AVR or the loads [1]. Load
flow change introduced by the DFIG was balanced by G13 at slack bus 65.
A detailed fourteenth-order model of the DFIG was used [4]. The DFIG adopted
the reactive power control with a fixed power factor (0.95). Parameters of the DFIG,
PI gains of RSC and GSC of DFIG are listed in Tables 4.13, 4.14, and 4.15 of
Appendix 4.1 respectively.
4.1.2.2 Decomposed Modal Analysis
Firstly, the DFIG was modelled as a constant power injection, Pw0, to examine the
effect of change of load flow introduced by the DFIG at the candidate connecting
locations. Pw0 increased gradually from Pw0 ¼ 0 to Pw0 ¼ 2 p. u. Qw0 changed
accordingly with a fixed power factor of 0.95. Changes of four inter-area oscillation
modes, λi, i ¼ 1, 2, 3, 4, were calculated from the open-loop system state matrix Ag
and are displayed in Figs. 4.5 and 4.6 as solid curves respectively, where the arrows
show the direction of change of λi, i ¼ 1, 2, 3, 4 when Pw0 increased.
Secondly, based on the closed-loop interconnected model of the example power
system shown by Fig. 4.2, impact of dynamic interactions introduced by the DFIG
was estimated by use of (4.16) to give Δλi, i ¼ 1, 2, 3, 4 when Pw0 increased. Results
of λi þ Δλi, i ¼ 1, 2, 3, 4 are presented in Figs. 4.5 and 4.6 as dashed curves
respectively.
In order to confirm the results obtained above by use of the method of
decomposed modal analysis, normal linearized model of the example power system
with the DFIG’s dynamics being included shown by (4.1) was established. The inter-
area EOMs were calculated from state matrix A to be ^λ i , i ¼ 1, 2, 3, 4 and displayed
as dashed curves with crosses in Figs. 4.5 and 4.6, which overlap the dashed curves
(λi þ Δλi, i ¼ 1, 2, 3, 4). Results of Figs. 4.5 and 4.6 confirm the correctness of
(4.16), i.e., the total effect of the DFIG, ^λ i , i ¼ 1, 2, 3, 4, is separated successfully by
the method of decomposed modal analysis into two parts, the effect of the change of
load flow brought about by the DFIG, λi, i ¼ 1, 2, 3, 4, and that of dynamic
interactions introduced by the DFIG, Δλi, i ¼ 1, 2, 3, 4.
4.1.2.3 The Impact of Dynamic Interactions
From Figs. 4.5 and 4.6, it can be seen that the effect of dynamic interactions
introduced by the DFIG on system small-signal angular stability can be negative
or positive. For example, Fig. 4.5(1) shows that the dash curve is on the right hand
side of the solid one for the first EOM from Pw0 ¼ 0 to Pw0 ¼ 2 p. u. This means
that the effect of the dynamic interaction on the mode (Δλ1) is negative, being
Imaginary axis Imaginary axis

5.342 4.65
(0)
l1 4.64 (0)
l2
4.63
5.3415 4.62
Pw0 increased
4.61
lˆ 1 4.6
Pw0 increased
l1
l1 + Δl1 l2
5.341 4.59
lˆ 2
4.58
Δl1 4.57 Δl2
l2 + Δl2
-0.1001 -0.1 -0.0999 -0.0998 -0.145 -0.135 -0.125 -0.115 -0.105
Real axis Real axis

3.9625 3.05
(0)
(0) 3.045 l4
3.962 l3
3.04
3.9615 Pw0 increased 3.035
3.961 lˆ 3 3.03
l3 + Δl3 3.025
3.9605
3.02
Pw0 increased
3.96 l3
3.015 l4 + Δl4 l4
3.9595 lˆ 4
Δl3 3.01
Δl4
-0.0995 -0.0985 -0.0975 -0.0965 -0.0955 -0.265 -0.26 -0.255 -0.25
Real axis Real axis
Fig. 4.5 Trajectories of inter-area EOMs when the level of wind penetration increases with the
DFIG being connected at node 8
detrimental to the small-signal angular stability of the example power system. Whilst
for other three EOMs, the effect of dynamic interactions is beneficial because adding
Δλi, i ¼ 2, 3, 4 on λi, i ¼ 2, 3, 4 makes the modes moving towards the left on the
complex plane.
Table 4.1 presents the computational results of the damping torque contribution
from the DFIG to the synchronous generators when it is connected at node 8 and
Pw0 ¼ 2 p. u.From Table 4.1, it can be seen that the DFIG contributed negative
damping torque to the majority of the synchronous generators for the first EOM. It
provided the positive damping torque to the majority of the synchronous generators
for other three EOMs. This explains why the effect of dynamic interactions intro-
duced by the DFIG was negative for the first EOM (Δλ1) and positive for other three
EOMs (Δλi, i ¼ 2, 3, 4).
From Figs. 4.5 and 4.6, it can be observed that the shift of Δλi, i ¼ 1, 2, 3, 4 in the
horizontal direction is much smaller than that of λi þ Δλi, i ¼ 1, 2, 3, 4 or
^λ i , i ¼ 1, 2, 3, 4. Hence, when the DFIG was connected at the candidate locations,
its impact on system small-signal angular stability was mainly from the change of

5.342 4.64 (0)
l1
(0) l2
5.3418 4.62
5.3416
Pw0 increased 4.6 Pw0 increased
5.3414
4.58
5.3412
lˆ 1 4.56 l2
5.341
l1 + Δl1 4.54 l2 + Δl2
5.3408 l1
4.52
lˆ 2
5.3406
Δl1 Δl2
-0.1001 -0.1 -0.0999 -0.0998 -0.14 -0.13 -0.12 -0.11 -0.1 -0.09 -0.08
Real axis Real axis
3.963 3.05
(0) (0)
3.962 l3 3.04 l4
3.03
3.961 Pw0 increased
3.02
3.96
lˆ 3
3.01
3.959 l3 + Δl3 Pw0 increased
3
3.958 l4
2.99 lˆ 4
3.957 2.98
l3 l4 + Δl4
3.956
Δl3 2.97 Δl4
-0.103 -0.101 -0.099 -0.097 -0.095 -0.27 -0.26 -0.25 -0.24 -0.23 -0.22
Real axis Real axis
Fig. 4.6 Trajectory of inter-area oscillation modes when the level of wind penetration increases
with the DFIG being connected at node 16
Table 4.1 Computational Mode 1 Mode 2 Mode 3 Mode 4

results of the damping torque
G1 1.0971 4.5036 2.2922 0.1205
contribution from the DFIG to
the synchronous generators G2 2.8677 11.2763 5.2071 0.2861
G3 2.8583 9.8262 4.8684 0.2965
G4 1.0519 2.4555 1.5013 0.1212
G5 0.6180 1.4108 0.8927 0.0740
G6 0.9118 2.1746 1.3770 0.1095
G7 0.8674 1.9086 1.1927 0.1002
G8 1.2503 3.3788 1.7510 0.1238
G9 0.7538 1.6448 1.0196 0.0816
G10 2.8498 2.3268 4.7636 0.1153
G11 16.1222 3.8338 4.9318 0.2951
G12 1.5291 1.9379 2.8577 0.2051
G13 1.7722 3.3472 5.4747 0.4192
G14 0.0729 6.1747 0.2198 0.2728
G15 0.0160 3.2344 0.3234 0.0578
G16 0.2304 8.0689 0.3815 0.3747
Table 4.2 Computational results of summation of the participation factors of all the state variables
of the DFIG
Mode 1 2 3 4
PFi107 0.39594 1.5378 9.6013 105.10
load flow it introduced. Small Δλi, i ¼ 1, 2, 3, 4 indicated weak dynamic coupling

between the DFIG and the synchronous generators. To confirm this, Table 4.2 gives
the computational results of summation of the participation factors of all the state
variables of the DFIG, PFi, i ¼ 1, 2, 3, 4, when it was connected at node 16 and
Pw0 ¼ 2.0 p. u. Obviously, very small numerical results in Table 4.2 confirm the
nearly decoupling of the DFIG with the synchronous generators.
Though the calculation of PFi, i ¼ 1, 2, 3, 4 can give an estimation of the dynamic
interactions between the DFIG and the synchronous generators, estimation by use of
Δλi, i ¼ 1, 2, 3, 4 is better in two aspects. (1) Δλi, i ¼ 1, 2, 3, 4 does not need to
perform the modal analysis of state matrix A with increased dimension to include the
DFIG’s dynamics as PFi, i ¼ 1, 2, 3, 4 does; (2) Δλi, i ¼ 1, 2, 3, 4 can not only
provide the estimation of the scale of dynamic interactions, but also indicate whether
the interactions are positive to improve or negative to reduce the small-signal angular
stability of the example power system.
4.1.2.4 The Impact of Load Flow Change
From Figs. 4.5 and 4.6, it can be observed that the effect of change of load flow due
to the DFIG (λi, i ¼ 1, 2, 3, 4, solid curves) on system small-signal angular stability
can be positive or negative. For example, from Fig. 4.5 it can be seen that when the
DFIG was connected at node 8, with the increase of the penetration level of the wind,
λ1 and λ3 moved towards left on the complex plane hence benefiting the small-signal
angular stability of the example power system. While λ2 and λ4 move towards right
which was detrimental to the small-signal angular stability of the example power
system. With the increase of penetration level of wind power, shift of the EOMs
towards right indicated the potential threat to system stability. Hence it is important
to identify the biggest movement towards right when the wind power is dispatched in
power system operation or when a wind farm is planned to be connected at various
candidate locations. The method of decomposed modal analysis suggests a simpler
way to identify the most dangerous scenarios of wind power dispatching and
connections as to be explained as follows.
By comparing the solid curves (λi) and the dash ones with crossed (^λ i) in Figs. 4.5
and 4.6, it can be seen that when the range of horizontal shift of ^λ i is relatively big,
the scale and direction of change of λi are very close to those of ^λ i . Thus, the biggest
negative impact of the DFIG on the system small-signal angular stability can be
estimated approximately from λi when the DFIG is modelled simply as a constant
power injection. For example, from Figs. 4.5 and 4.6 it can be seen that the range of
horizontal variation of λ2 and λ4 was bigger. The biggest shift of the modes to the
right can be identified from the solid curves (λ2 and λ4) when the DFIG was
modelled as a constant power injection.
Therefore, it can be proposed that in dispatching wind power or planning the
connection of wind farms, the wind generations are modelled firstly as constant
power injections. Thus, modal computation from state matrix Ag (λi) can provide a
primary and fast scanning to find the most dangerous cases of wind power
dispatching or connection. Secondly, the effect of dynamic interactions due to the
DFIGs’ dynamics is estimated for those most dangerous cases identified. By doing
so, the computational cost involved in the examination would be significantly
reduced as the dimension of state matrix Ag does not increase with the wind
generations modelled as constant power sources.
However, prescreening the most dangerous cases by modelling the wind power as
constant power source may miss the case when the dynamic interactions introduced
by the DFIG are strong. This special case is possible and will be discusses in Chap. 5
of the book.
4.1.2.5 The Impact of Reactive Power Control/Terminal Voltage

Control
Study has shown that the impact of reactive power/voltage control adopted by the
DFIG on the damping of power system EOMs was different [5–9]. A more detailed
examination on the difference can also be carried out by use of the proposed method
as to be demonstrated as follows.
Table 4.3 present the computational results when the DFIG was connected at
node 16 (Pw0¼3.0 p.u.) in the example power system. Two types of control adopted
by the DFIG are compared in Table 4.3: (1) the reactive power control with a fixed
power factor of 0.95; (2) the terminal voltage control with a reference value of
voltage to be 1.0 p.u. When the DFIG adopts reactive power or terminal voltage
control, not only the dynamic interactions introduced by the DFIG but also the
Table 4.3 Representative Modes Reactive power control Voltage control

computational results of the
λf1 0.0997 þ j5.3396 0.0998 þ 5.3396i
DFIG with the reactive power
or voltage control λ10 0.0998 þ j5.3396 0.0998 þ 5.3396i
Δλ1 0.0001 j0.0000 0.0000 þ 0.0000i
λf2 0.1332 þ j4.4231 0.0936 þ 4.4426i
λ20 0.0313 þ j4.3761 0.0795 þ 4.4504i
Δλ2 0.1091 þ j0.0506 0.0141 0.0078i
λf3 0.1098 þ j3.9578 0.1071 þ 3.9545i
λ30 0.1114 þ j3.9472 0.1063 þ 3.9537i
Δλ3 0.0037 þ j0.0099 0.0008 þ 0.0009i
λf4 0.2437 þ j2.8888 0.2131 þ 2.9231i
λ40 0.1684 þ j2.8761 0.2077 þ 2.9260i
Δλ4 0.0771 þ j0.0206 0.0054 0.0029i
Table 4.4 Representative Modes Reactive power control Voltage control

computational results of the
λf1 0.0997 þ j5.3396 0.0997 þ j5.3396
DFIG with the parameters of
controllers being changed λ10 0.0998 þ j5.3396 0.0998 þ j5.3396
Δλ1 0.0001 j0.0000 0.0000 þ j0.0000
λf2 0.1424 þ j4.4194 0.1414 þ j4.4164
λ20 0.0313 þ j4.3761 0.0795 þ j4.4504
Δλ2 0.1199 þ j0.0464 0.0623 j0.0374
λf3 0.1103 þ j3.9587 0.1106 þ j3.9587
λ30 0.1114 þ j3.9472 0.1063 þ j3.9537
Δλ3 0.0036 þ j0.0110 0.0043 þ j0.0059
λf4 0.2537 þ j2.8779 0.2529 þ j2.8735
λ40 0.1684 þ j2.8761 0.2077 þ j2.9260
Δλ4 0.0905 þ j0.0102 0.0490 j0.0584
change of load flow brought about by the DFIG is different. By using the method of
decomposed modal analysis, a mor.3-3, ^λ i , i ¼ 1, 2, 3, 4 were calculated from the
closed-loop state matrix A in (4.1) in order to confirm the correctness of the
computation provided by use of the method of decomposed modal analysis.
It is interesting to note that the impact of dynamic interactions introduced by the
DFIG with the terminal voltage control was always smaller than that with the
reactive power control (Δλi, i ¼ 1, 2, 3, 4). The same results have been reported in
[8, 9]. The reason may be that the terminal voltage control applied by the DFIG
constrained the variation of “feedback signal”, ΔVw, in Fig. 4.3, resulting in less
dynamic interactions between the DFIG and the synchronous generators and hence
less damping torque contribution from the DFIG.
To further validate the results presented in Table 4.3, parameters of both PI
reactive power controller and PI terminal voltage controller of the DFIG were
reduced by 100 times. Computational results are given in Table 4.4. From Tables 4.3
and 4.4, it can be seen that with the smaller PI gains, the control ability of controllers
were weakened, causing bigger dynamic interactions between the DFIG and syn-
chronous generators, which was beneficial to the system small-signal angular
stability.
4.2 Impact of Dynamic Interactions Introduced

by a Grid-Connected DFIG on Power System
Small-Signal Angular Stability
When impact of grid-connected wind farms on power system small-signal angular

stability is examined, a commonly-adopted strategy has been to represent a grid-
connected wind farm by a single variable speed wind generator (VSWG), such as a
DFIG. This, in fact, has assumed that an equivalent dynamic model of single VSWG
4.2 Impact of Dynamic Interactions Introduced by a Grid-Connected DFIG on. . . 111
can be used to examine the impact of wind farm on power system small-signal
angular stability. In this section, the conditions of equivalence under this assumption
are examined.
Firstly, the sensitivity of a power system EOM of concern to the dynamic
interactions introduced by a grid-connected DFIG for wind power generation is
derived. The derived modal sensitivity is proposed as an index to indicate the impact
of dynamic interactions (index of dynamic interactions, IDI) introduced by the DFIG
on the EOM of concern. The analytical expression of the IDI explicitly connects the
impact with the frequency response of the DFIG at the frequency of the EOM of
concern. From the analytical expression of the IDI, it can be clearly seen that when
the impact on the EOM of concern is examined, an equivalent dynamic model of
single DFIG can be used to represent a wind farm, as long as the frequency response
of this equivalent DFIG model around the frequency of the EOM of concern
approximately matches the frequency response of the wind farm. Thus, the condition
to represent the dynamics of the wind farm by the dynamic model of a single DFIG is
established.
Based on the proposed index, analysis is carried out in the section with following
generally-applicable conclusion obtained: Impact of dynamic interactions intro-
duced by the DFIG on power system EOM of concern increases when the level of
wind power penetration increases. Furthermore, an indicator of the DFIG control
(IDC) is derived for comparing the impact of the terminal voltage control and
reactive power control implemented by the DFIG. The derivation provides some
insight into the difference between the impact of terminal voltage and reactive power
control on power system small-signal angular stability. Main merit of the IDC is the
computational simplicity to determine that either the terminal voltage or reactive
power control is more beneficial to the power system small-signal angular stability,
as no knowledge of DFIG dynamic model is needed for computing the IDC.
4.2.1 The Index of Dynamic Interactions (IDI)
4.2.1.1 The Impact of The Grid-Connected DFIG
For a multi-machine power system connected with a DFIG shown by Fig. 4.1, the
closed-loop interconnected model was derived in Sect. 2.4.2 of Chap. 2 which
consists of the open-loop DFIG subsystem in the feedback loop and the ROPS
subsystem in the forward path. The closed-loop interconnected model is shown by
Fig. 4.2 and was used in the previous section to introduce the method of decomposed
modal analysis. The state-space model of ROPS subsystem is given by (4.3). The
transfer function model of the ROPS subsystem can be obtained from (4.3) to be
The transfer function model of the DFIG subsystem in Fig. 4.2 is (4.5). The state-
space representation of the DFIG subsystem is (2.124) and rewritten below
8
> d
>
> ΔXd ¼ Ad ΔXd þ bdq ΔVw
< dt
ΔPw ¼ cdp T ΔXd þ ddp1 ΔVw ð4:19Þ
>
>
>
:
ΔQw ¼ cdq T ΔXd þ ddq1 ΔVw
From (4.3) and (4.19), the closed-loop state-space model of the power system
with the DFIG can be obtained to be
d
ΔX ¼ AΔX ð4:20Þ
dt
where the vector of state variables, ΔX, and the closed-loop state matrix, A, are
given in (2.131).
Denote ^λ i as an EOM of concern in the power system with the DFIG. ^λ i is an
eigenvalue of closed-loop state matrix, A, in Eq. (4.20). Denote λni as the
corresponding EOM of concern when the DFIG is not connected to the power
system. Impact of grid connection of the DFIG on the EOM of concern is
^λ i λni . Examination in the previous section shows that ^λ i λni is caused by
change of load flow and dynamic interactions brought about by the DFIG. The
method of decomposed modal analysis introduced in the previous section can
examine those two affecting factors separately in the following two steps.
First, consider the case that there are no dynamic interactions between the DFIG
and the ROPS such that ΔPw þ jΔQw ¼ 0. In the closed-loop interconnected system
shown by Fig. 4.2, the feedback loop is open. In this case, the DFIG becomes a
constant power source, Pw0 þ jQw0, and power system model degrades from (4.3) to
d
ΔXg ¼ Ag ΔXg ð4:21Þ
dt
Denote λi as the EOM of concern corresponding to ^λ i when the DFIG is modelled
as the constant power source, Pw0 þ jQw0. λi is an eigenvalue of open-loop state
matrix Ag and hence is the open-loop EOM. Injection of constant power, Pw0 þ jQw0,
into the power system changes the system load flow to affect the open-loop EOM of
concern. Hence, the impact of load flow change introduced by the DFIG on the EOM
of concern is λi λni.
Second, consider the case that ΔPw þ jΔQw 6¼ 0 and the feedback loop in Fig. 4.2
is closed. The EOM of concern is ^λ i and hence is the closed-loop EOM. The
difference between the closed-loop and open-loop EOM of concern, Δλi ¼ ^λ i λi ,
is caused by the dynamic interactions brought about by the DFIG. Hence, it
measures the impact of dynamic interactions introduced by DFIG on the EOM of
concern, λi. Thus, the total impact of the DFIG is separated as
^λ i λni ¼ ^λ i λi þ λi λni ð4:22Þ

|fflfflfflffl{zfflfflfflffl} |fflfflffl{zfflfflffl} |fflfflfflffl{zfflfflfflffl}
Total impact Impact of dynamic interactions Impact of load flow change
In the previous section, the DTA is applied to estimate the impact of dynamic
interactions on the EOM of concern, Δλi ¼ ^λ i λi .
In fact, Δλi ¼ ^λ i λi can be calculated directly from eigensolution of open-loop
and closed-loop state matrix, Ag and A, to determine the exact impact of dynamic
interactions introduced by the DFIG. However, direct calculation can only examine
the impact on a case-by-case basis and would be difficult to gain deeper insight into
and better understanding on the impact. In order to carry out analytical examination
on the impact of dynamic interactions brought about by the DFIG to draw generally-
applied conclusions, an index of dynamic interactions (IDI) is derived in the next
subsection.
4.2.1.2 Derivation of the IDI
Denote λk, k ¼ 1, 2, n as the eigenvalues of open-loop state matrix Ag; pk and rk

are the corresponding left and right eigenvectors. According to the theory of modal
control [1], the transfer functions of the open-loop ROPS subsystem, i.e., the power
system, can be written as
8
> Xn
Rvpk
>
> G ð sÞ ¼ þ dvp
>
< vp ð s λk Þ
k¼1
ð4:23Þ
>
> Xn
Rvqk
> G ðsÞ ¼
> þ dvq
: vq ð s λk Þ
k¼1
where Rvpk ¼ pkT bP cgT rk and Rvqk ¼ pkT bQ cgT rk are the residues of input-output pair
ΔPw ΔVw and ΔQw ΔVw of the open-loop state-space model of the power
system respectively, which is given by (4.3).
From Fig. 4.2, the characteristic equation of closed-loop power system with the
DFIG can be obtained to be
1 ¼ Hdp ðλÞGvp ðλÞ þ Hdq ðλÞGvq ðλÞ ð4:24Þ
For the closed-loop EOM of concern, ^λ i , from Eqs. (4.23) and Eq. (4.24), it can
have
P
n
P n
Rvpk
Hdp ^λ i þ Rvqk
Hdq ^λ i
^λ i λk ^
k¼1 k¼1 λ i λk

¼1 ð4:25Þ
1 dvp Hdp λ i dvq Hdq ^λ i
^
Multiplying both sides of (4.25) by ^λ i λi and taking the limit ^λ i ! λi , it can

have [1].

Rvpi Hdp ðλi Þ þ Rvqi Hdq ðλi Þ
Δλi lim ^λ i λi ¼ ð4:26Þ
^λ i !λi 1 dvp Hdp ðλi Þ dvq Hdq ðλi Þ
where Rvpi and Rvqi are the residues corresponding to λi for the input-output pair
ΔPw ΔVw and ΔQw ΔVw in the open-loop state-space model of the power
system respectively.
For the open-loop EOM of concern, λi ¼ ξi þ jωi, normally |ωi| >> |ξi|. Thus,
Hdp(λi) Hdp(jωi) and Hdq(λi) Hdq(jωi). Hence from Eq. (4.26), following IDI is
proposed to indicate the impact of dynamic interactions introduced by the DFIG on
the EOM of concern
Rvpi Hdp ðjωi Þ þ Rvqi Hdq ðjωi Þ

Δλi IDIðjωi Þ ¼ ð4:27Þ
1 dvp Hdp ðjωi Þ dvq Hdq ðjωi Þ
The IDI derived in (4.27), in fact, is the modal sensitivity to the variations of
dynamic interactions between the DFIG and power system because of the derivation
of Eq. (4.26). Magnitude of IDI(jωi) indicates the level of impact of dynamic
interactions on the EOM of concern. Phase of IDI(jωi) specifies the direction of
impact of dynamic interactions, i.e., positive to improve or negative to reduce the
damping of EOM of concern. For example, if |phase of IDI(jωi)| < 900, the IDI
indicates that ^λ i is on the right hand side of λi on the complex plane. The dynamic
interactions introduced by the DFIG degrade the damping of the EOM of concern.
As the modal sensitivity, the IDI can be used as a pre-screening indicator when
large numbers of scenarios of integration and operation of the DFIGs are examined.
After potentially detrimental cases to power system stability are indicated by the IDI,
more detailed and in-depth examination can be carried out to determine the impact of
the DFIG more accurately.
Like any index of modal sensitivity, the IDI can be an approximation of the actual
impact, Δλi ¼ ^λ i λi , as long as the dynamic interactions between the DFIG and
power system, i.e., ΔPw þ jΔQw, are small. Normally in a power system, dynamic
variations of ΔVw is very small. Hence, from Eq. (4.5), it can be seen that ΔPw and
ΔQw should usually be small. This explains why the DFIG usually exhibits very
limited dynamic response to the small disturbances in the power system due to the
fast converter control implemented by the DFIG. Small dynamic interactions
between the DFIG and power system ensures that usually, the IDI is a good
approximation of the actual impact of dynamic interactions, i.e., Δλi ¼ ^λ i λi .
As a modal sensitivity, estimation error by the IDI may possibly become consid-
erable when the dynamic interactions introduced by the DFIG are not small and
when the IDI is not a constant. It is important to identify such case that the estimation
error of the IDI on Δλi ¼ ^λ i λi may be considerable when the IDI is applied. The
identification establishes the boundary about the reliability of the IDI. Beyond the
boundary, other means of more detailed examination can be conducted in addition to

the application of the IDI. A way to identify the case that the IDI might be unreliable
is introduced as follows.
Mathematical derivation of Eq. (4.26) implicitly means that

Hdp ^λ i ! Hdp ðλi Þ, Hdq ^λ i ! Hdq ðλi Þ, when ^λ i ! λi ð4:28Þ
Hence, when the IDI is used as an approximation of the impact, i.e.,

IDIðjωi Þ Δλi ¼ ^
λ i λi , it should have

Hdp ^λ i Hdp ðλi Þ, Hdq ^λ i Hdq ðλi Þ ð4:29Þ
Thus, if Hdp(jω) or/and Hdq(jω) changes dramatically around ωi, the approxima-
tion of (4.29) may yield considerable error such that IDIðjωi Þ Δλi ¼ ^λ i λi may
be of considerable error of estimation. Such unusual case of radical changes of
Hdp(jω) or/and Hdq(jω) around ωi is possible and can be elaborated as follows by
using Hdp(jω) as an example.
Denote λd ¼ ξd þ jωd as a complex pole of Hdp(s). It is a complex eigenvalue of
open-loop state matrix of the DFIG, Ad in Eq. (4.19). Since |Hdp(λd)| ¼ |
Hdp(ξd þ jωd)| ! 1, it is possible that |Hdp(jωi)| may increase significantly
when ωi is close to ωd. In this case, from Eq. (4.5) it can be seen that ΔPw may
become considerable around the oscillation frequency ωi, indicating substantial
dynamic interactions introduced by the DFIG. Thus, jΔλi j ¼ λ i λi may possibly
^
increase radically, indicating considerable impact of dynamic interactions intro-
duced by the DFIG on the EOM of concern. This is the case that the IDI may be
of considerable estimation error and even become unreliable. The case can be
identified from the magnitude of frequency response of the DFIG, |Hdp(jω)|, around
the oscillation frequency ωi, as being illustrated by Fig. 4.7.
In Fig. 4.7, |Hdp(jω)| changes radically within the interval [ω1, ω2] where ωd
locates. According to the analysis made above, if |Hdp(jω)| is at point A, the IDI
should be a good approximation of the impact. If |Hdp(jω)| is at point B, estimation
error by the IDI may be considerable. Hence, [ω1, ω2] defines the boundary. Inside
Fig. 4.7 Identification of

estimation error of the IDI Hdp ( jw )
w
w1 wd w2
[ω1, ω2], the IDI may become unreliable with considerable estimation error and more
detailed examination by other means is required.
Therefore, the IDI can be used jointly with |Hdp(jω)| in the pre-screening to find
out potential cases that the DFIG may threaten the system stability.
4.2.1.3 Further Discussions About the IDI
The IDI introduced in the previous subsection can provide an indication of the
impact of dynamic interactions introduced by the DFIG, both the scale and direction,
on the EOM of concern. Dynamic interaction of the DFIG with power system EOMs
can be estimated by calculating the participation factors of state variables of the
DFIG for the EOM of concern. However, the participation factors are a type of
dimensionless index and hence can only provide the estimation of scale of dynamic
interactions. Impact (both level and direction) of dynamic interactions on the EOM
of concern cannot be assessed directly from computational results of the participa-
tion factors.
Modal sensitivity to a certain parameter of the DFIG can indicate how variation of
the parameter affects the EOM of concern in both level and direction. Thus, an index
of modal-to-parameter sensitivity can show how a factor in the DFIG affects the
EOM of concern in both level and direction. However, like the participation factors,
index of modal-to-parameters sensitivity relies on the computational results and is
applicable for case-by-case study. Normally, it would be difficult to draw generally-
applicable conclusions by applying the index of modal-to-parameters sensitivity and
the participation factors.
In the analytical expression of the proposed IDI of Eq. (4.27), the frequency
response of the DFIG at the frequency of the EOM of concern, i.e., Hdp(jωi) and
Hdq(jωi), is explicitly linked with the impact of dynamic interactions introduced by
the DFIG. This is the main merit of the IDI as compared with the participation factors
and modal-to-parameters sensitivity. The other main merits of the IDI are further
elaborated as follows.
Firstly, the analytical expression makes theoretical analysis possible to draw
generally-applicable conclusion. This will be demonstrated in the next subsection.
Secondly, when the impact of a wind farm on power system small-signal angular
stability is examined, so far the equivalent dynamic model of a single DFIG has been
used to represent dynamics of the wind farm. Adequacy of representing dynamics of
the wind farm by dynamic model of the single DFIG has not been carefully
investigated. The analytical expression of the IDI establishes the condition under
which the equivalence of dynamic model of single DFIG and the wind farm for
studying the impact of the wind farm. According to (4.27), it can be concluded that
an equivalent dynamic model of single DFIG can be used to examine the impact of
the wind farm on the EOMs of concern, as long as the frequency response of this
equivalent DFIG model around the frequency of EOMs of concern approximately
matches the frequency response of wind farm. Study cases are presented in Sect.
4.2.3 to demonstrate this aspect of the merit of the IDI. It is shown that when a single
DFIG with increased capacity is simply used to represent a wind farm, wrong result
of assessment on the impact of the wind farm occurs. After the dynamic model of the
single DFIG is adjusted to meet the condition of equivalence established above,
correct assessment is obtained.
Thirdly, for a practical large-scale power system, dimension of state matrix could
be very high. Errors of system parameters caused by various factors may result in a
state-space model of the power system with errors. Those can cause considerable
problem in applying the modal analysis to assess the impact of the DFIG on the
EOM of concern. In the IDI shown by (4.27), line parameters, dvp and dvq, and the
residues, Rvpi and Rvqi, can be obtained from measuring data in field. Hdp(jw) and
Hdq(jω) can also be obtained from frequency-response test of the DFIG. Hence, the
IDI can become model-independent and be calculated from measuring data. Thus,
the problems of model errors and computational complexity due to high-dimensional
state matrix in modal analysis can be avoided.
The IDI is of two computational advantages as follows. First, from (4.27) it can be
seen that when the impact of changes of some key parameters of the DFIG is
examined, proposed index, IDI(jωi), can be calculated from changes of Hdp(jωi)
and Hdq(jωi) without need to perform eigensolution of closed-loop state matrix. This
is very useful when the control parameters of the DFIG are tuned and the effect of
tuning on the EOMs of concern need to be checked. In practical application, it is
usually much easier to obtain the frequency response of the DFIG at given oscillation
frequencies of the EOMs than to derive parametric dynamic model, which is
normally needed for the calculation of the modal-to-parameter sensitivity and the
participation factors. Second, compared with the method of direct calculation to find
Δλi ¼ ^λ i λi from eigensolution of closed-loop and open-loop state matrix, com-
putational burden of the IDI is significantly reduced when the scale of power system
is large and there are large numbers of the DFIGs to be examined. For each of the
DFIGs being examined, direct calculation needs to perform eigensolution of closed-
loop state matrix and to select correct ^λ i corresponding to λi from the results of
eigensolution. Hence, for N DFIGs to be examined, eigensolution of closed-loop
state matrix has to be carried out for N times. Especially, selection of correct ^λ i from
the results of eigensolution of closed-loop state matrix often is not a straightforward
work when the scale of power system is large. Whilst computation of the IDI needs
only to perform eigensolution of open-loop state matrix for once, no matter how
many DFIGs needs to be examined.
4.2.2 The Impact of Dynamic Interactions Introduced by

the DFIG
When the wind power penetration varies, impact of both the load flow change and
dynamic interactions introduced by the DFIG may vary accordingly. In (4.22), λni is
the EOM of concern when the DFIG is not connected; thus the active power output
from the DFIG, Pw0, is zero. λi is the EOM of concern when the DFIG is represented
by a constant power source, Pw0. Change of Pw0 is balanced by the change of the
active power output from the synchronous generators at the slack nodes in a power
system if the total system load keeps unchanged. Hence, the impact of load flow
change brought about by the DFIG on the EOM of concern, λi λni in Eq. (4.22), is
equivalent to that of change of active power output from the synchronous generators
at slack nodes in the power system from Pg0 to Pg0 Pw0. Normally, the more the
active power output from the synchronous generators varies, the more the EOMs are
affected. Hence, it is easy to understand that when the level of wind power penetra-
tion is higher (bigger Pw0), the impact of load flow change introduced by the DFIG
on the EOM of concern, i.e., λi λni in Eq. (4.22), is bigger. Of course, if the modal
sensitivity to the variation of active power output from the synchronous generators at
slack nodes in the power system is always very small, |λi λni| will be small,
implying the limited impact of load flow change introduced by the DFIG on the
EOM of concern, even if the wind power penetration is high.
In this section, impact of dynamic interactions introduced by the DFIG, i.e.,
Δλi ¼ ^λ i λi in Eq. (4.22), is examined. In order to draw generally-applicable
conclusions, a simplified transfer function model of the DFIG is derived from the
detailed 14th-order dynamic model of the DFIG introduced in Chap. 2. The exami-
nation then is carried out by using the IDI and the simplified transfer function model of
the DFIG.
4.2.2.1 Simplified Transfer Function Model of the DFIG
Direction of terminal voltage of the DFIG, Vw, is taken as that of q axis of d q

coordinate of the DFIG as shown by Fig. 2.10. Voltage and flux equations of the
rotor windings of the DFIG are given as (2.56) and (2.57) in Chap. 2. By omitting the
first subscript d in (2.56) and (2.57), those equations are rewritten below.
dψrd
¼ ω0 Vrd þ sw ψrq þ Rr Ird
dt
dψrq
ð4:30Þ
¼ ω0 Vrq sw ψrd þ Rr Irq
dt
ψrd ¼ Xm Isd Xrr Ird , ψrq ¼ Xm Isq Xrr Irq
where sw is the rotor slip, Rr and Xrr are the resistance and self-inductance of the
rotor windings, Xm is the magnetizing inductance, ψrd, ψrq, Vrd, Vrq, Ird and Irq are
the flux, voltage and current of the rotor windings expressed in the d q coordinate
of the DFIG respectively, ω0 is the synchronous speed.
Ignoring the transient and resistance of stator windings and from (2.55) and (2.57)
(note: the first subscript d in (2.55) and (2.57) is omitted),
Vwd ¼ 0 ¼ ψsq ¼ Xss Isq þ Xm Irq
ð4:31Þ
Vwq ¼ Vw ¼ ψsd ¼ Xss Isd Xm Ird
where Xss is the self-inductance of the stator windings, Irq Isd, Isq, Vwd and Vwq are
the d and q component of stator current and terminal voltage of the DFIG
respectively.
From (4.30) and (4.31)

Xrr X2m d X2m
Vrd ¼ Rr þ Ird þ sw Xrr Irq
ω0 ω0 Xss dt Xss
ð4:32Þ
Xrr X2m d X2m Xm
Vrq ¼ Rr þ Irq sw Xrr Ird þ sw Vw
ω0 ω0 Xss dt Xss Xss
Configuration of vector control system of the RSC of the DFIG is shown by

Fig. 2.5, which is displayed as Fig. 4.8. From (4.32) and Fig. 4.8, an equivalent
system to that shown by Fig. 4.8 can be obtained and shown by Fig. 4.9. From
Fig. 4.9, it can have
Xss
ΔIrq ¼ Gq ðsÞΔIsq
ref
Xm
Xss ref Vw ð4:33Þ
ΔIrd ¼ Gd ðsÞ ΔIsd
Xm Xm
where Gq(s) and Gd(s) are the transfer function from ΔIrqref and ΔIrdref to ΔIrq and ΔIrd
respectively in Fig. 4.9. Linearization of (4.31) is
Xm 1
ΔIsd ¼ ΔIrd ΔjVw j
Xss Xss
ð4:34Þ
Xm
ΔIsq ¼ ΔIrq
Xss
active power control q-axis current control X m2 X

outer loop inner loop
– sw ( X rr – ) I rd + sw m Vw
X ss X ss
Psref I sqref I rqref

+ X + – +
K pq ( s) – ss K iq ( s ) Vrq
– Xm –
Ps I rq
Qsref I sdref X I rdref –

+ + +
K pd ( s) – ss K id ( s) Vrd
– Xm – – +
Qs I rd
Vw X m2
reactive power d-axis current sw ( X rr – ) I rq
control outer loop Xm control inner loop X ss
Fig. 4.8 Configuration of the RSC vector control system of the DFIG
Gq ( s)
Psref I sqref I rqref 1 I rq

+ X + –
K pq ( s) – ss K iq ( s) X rrX2
– Xm Rr + ( – m )s
w0 w0 X ss
Pe
Qsref I sdref I rdref 1 I rd

+ X + +
K pd ( s) – ss Kid ( s) X rrX m2
– Xm – – Rr + ( – )s
w0 w0 X ss
Qs I rd
Vw
Xm Gd ( s)
Fig. 4.9 Equivalent system of RSC vector control system of the DFIG
From Fig. 4.9, (4.33) and (4.34), it can have
ΔIsq ¼ Gq ðsÞΔIsq
ref
¼ Gq ðsÞKpq ðsÞðΔPs Þ
1
ΔIsd ¼ Gd ðsÞΔIsd
ref
þ ½Gd ðsÞ 1ΔVw ð4:35Þ
Xss
1
¼ Gd ðsÞKpd ðsÞðΔQs Þ þ ½Gd ðsÞ 1ΔVw
Xss
where Ps and Qs are the active and reactive power output from the stator side of the
DFIG respectively (see Fig. 4.1) and they are
Ps ¼ Vwd Isd þ Vwq Isq ¼ Vw Isq
ð4:36Þ
Qs ¼ Vwq Isd Vwd Isq ¼ Vw Isd
Linearization of above equations is

ΔPs ¼ Vw0 ΔIsq þ Isq0 ΔVw
ð4:37Þ
ΔQs ¼ Vw0 ΔIsd þ Isd0 ΔVw
Active and reactive power output from the GSC of the DFIG is Pr (see Fig. 4.1)
and Pr swPs, Qr 0 [10–12], i.e.,
ΔPr sw0 ΔPs Ps0 Δsw , ΔQr 0 ð4:38Þ
From (4.35), (4.37) and (4.38), it can have

Vw0 Vw0
ΔQw ¼ ΔQs ¼ Hdq ðsÞΔVw ¼ Isd0 Vw0 Gd ðsÞKpd ðsÞ þ G d ðsÞ ΔVw
Xss Xss
ð4:39Þ
Rotor motion equations of the DFIG are (2.58) when the rotor is represented by a
two-mass shaft rotational system. The rotor can be simply modelled as a one-mass
shaft. The torque can be considered to be same to the power. Thus (2.58) is
simplified to obtain the following linearized rotor motion equation of the DFIG
[10–12].
1
Δsw ¼ ðΔPs þ ΔPr Þ ð4:40Þ
Js
where J is the constant of rotor inertia. From (4.38) and (4.40),

Ps0 Ps0
ΔPr ¼ sw0 þ 1þ ΔPs ¼ JðsÞΔPs ð4:41Þ
Js Js
From (4.35), (4.37), (4.38) and (4.41), it can have

Isq0 ½1 þ JðsÞ
ΔPw ¼ ΔPs þ ΔPr ¼ ΔVw ¼ Hdp ðsÞΔVw ð4:42Þ
1 þ Vw0 Gq ðsÞKpq ðsÞ
Thus, the simplified transfer function model of the DFIG, Hdq(s) and Hdp(s), are
derived as shown by Eqs. (4.39) and (4.42) respectively.
4.2.2.2 Impact of Dynamic Interactions Affected by the Level of Wind

Power Penetration
Both the q-axis and d-axis current control inner loop in the RSC control system of
the DFIG is derived from the electromagnetic transient of the rotor windings of the
DFIG. The transient is electromagnetic and much faster than the electromechanical
dynamic interactions between the DFIG and power system, which is the focus of
discussions in this section. If this fast electromagnetic transient is ignored, Ird ¼ Irdref
and Irq ¼ Irqref , which is equivalent to Gd(s) ¼ 1 and Gq(s) ¼ 1 as it can be seen from
Fig. 4.9. Thus, by ignoring the electromagnetic transient of the rotor windings of the
DFIG with Gd(s) ¼ 1 and Gq(s) ¼ 1, transfer function model of the DFIG shown by
(4.39) and Eq. (4.42) are simplified to be
Isq0 ½1 þ JðsÞ
Hdp ðsÞ ð4:43Þ
1 þ Vw0 Kpq ðsÞ
Isd0
Hdq ðsÞ ð4:44Þ
1 þ Vw0 Kpd ðsÞ
In the output equation of (4.3), dvp should be very small as the magnitude of
busbar voltage is usually affected little by variation of line active power. Hence, the
IDI given by (4.27) can be simplified to be
Rvpi Hdp ðjωi Þ þ Rvqi Hdq ðjωi Þ

Δλi IDIðjωi Þ ¼ ð4:45Þ
1 dvq Hdq ðjωi Þ
At steady state, reactive power output from the DFIG is Qw0 ¼ Vw0Isd0. Nor-
mally, the DFIG operates with high power factor such that Qw0 0. Hence, it can
have Isd0 0. From (4.44), it can be seen that Hdq(jωi) 0. Hence, the IDI given in
(4.45) can be further simplified) to be
IDIðjωi Þ Rvpi Hdp ðjωi Þ ð4:46Þ
From Pr swPs, Qr 0 [10–12], it can have

Pw0 ¼ Ps0 þ Pr0 ð1 sw0 ÞPs0 ¼ ð1 sw0 ÞVw0 Isq0 ð4:47Þ
Thus, when the amount of active power output from the DFIG at steady state
increases, Isq0 increases. From (4.43), it can be seen that |Hdp(jωi)| increases accord-
ingly with the increase of Pw0. Hence, it can be concluded that increase of wind
power penetration will lead to the increase of scale of the impact of dynamic
interactions introduced by the DFIG, which is ^λ i λi .
4.2.2.3 Impact of DFIG’s Terminal Voltage Control Vs. Reactive Power

Control
A DFIG can be equipped with the terminal voltage control instead of reactive power
control. In this case, the input signal in the reactive power control outer loop of
Fig. 4.8, Qsref Qs , is replaced by Vwref Vw . In this case, from the reactive power
control outer loop in Fig. 4.8, it can have
ΔIsd
ref
¼ Kpd ðsÞΔVw ð4:48Þ
From (4.33), (4.34) and (4.48), it can have

1
ΔIsd ¼ Gd ðsÞKpd ðsÞðΔVw Þ þ ½Gd ðsÞ 1ΔVw ð4:49Þ
Xss
From (4.37), (4.38) and (4.49)
ΔQw ¼ ΔQs ¼ Hdq ðsÞΔVw
Vw0 Vw0 ð4:50Þ
¼ Isd0 Vw0 Gd ðsÞKpd ðsÞ þ Gd ðsÞ ΔVw
Xss Xss
By ignoring the electromagnetic transient of the rotor windings of the DFIG with
Gd(s) ¼ 1, from (4.50) it can have
Hdq ðsÞ ¼ Isd0 Vw0 Kpd ðsÞ ð4:51Þ
Impact of the DFIG’s terminal voltage control as compared with the reactive
power control can be estimated by examining the changes of Kpd(s) as follows.
Firstly, consider the extreme case that the gain value of outer loop PI controller is
zero to have Kpd(s) ¼ 0. In this case, terminal voltage control results in Isd ref
¼ 0 as
shown by Fig. 4.8. A well-designed reactive power control with the unity power
factor (Qsref ¼ 0) also provides Isd
ref
¼ 0 if the controller’s transient is ignored. Hence,
it is expected that the impact of the DFIG with a normal reactive power control (with
high power factor and well-designed PI controller) on the EOM of concern should be
approximately equal to that of the DFIG with the terminal voltage control with gain
value of PI controller being equal to zero. Denote λR as the EOM of concern in
this case.
Secondly, consider the case of a well-designed outer loop PI controller for the
DFIG’s terminal voltage control with the EOM of concern being λV. The difference
between the impact of dynamic interactions as caused by the reactive power and
terminal voltage control can be determined by comparing λR and λV. The compar-
ison can be carried out by increasing the gain value of outer loop PI controller of the
DFIG’s terminal voltage control from zero and to estimate the change of the impact
by use of (4.45). From (4.51), it can be seen that |Hdq(jωi)| will increase when the
gain value of outer loop PI controller increases from zero to result in the EOM of
concern to move from λR towards λV on the complex plane. The trend of motion
from λR towards λV thus can be estimated from (4.45) to be
H ðjω Þ
Rvpi Hdpdq ðjωi Þ þ Rvqi Rvqi
lim λV λ R lim Δλi ¼ lim i
1dvp Hdp ðjωi Þ
¼
jHdq ðjωi Þj!1 jHdq ðjωi Þj!1 jHdq ðjωi Þj!1 Hdq ðjωi Þ dvq dvq
ð4:52Þ
From (4.52), it can be seen that if (Real part of Rvqi/dvq) > 0, λV is on the left hand
side of λR on the complex plane. Thus, damping of the EOM of concern is improved
if the DFIG is equipped with terminal voltage control instead of reactive power
control, and vice versa. Therefore, Rvqi/dvq can be used as an indicator of DFIG
control (IDC) to approximately determine whether the terminal voltage control
implemented by the DFIG is more beneficial to power system small-signal angular
stability than the reactive power control or not. Calculation of the IDC is simple as
no knowledge of DFIG’s dynamic model is required. Hence, it is useful at the stage
of planning the integration of a wind farm when the details of the wind farm are
unknown and the guidance on the selection of types of DFIG’s control is needed.
The IDC is derived from the IDI and hence also a modal sensitivity. The IDC is
for the comparison of the effect of |Hdq(jωi)| on the IDI between the reactive power
and terminal voltage control implemented by the DFIG. It is derived on the basis that
|Hdq(jωi)| of the terminal voltage control is much greater than |Hdq(jωi)| of the
reactive power control in an extreme case that |Hdq(jωi)| ! 1. Hence, the IDC is
an indication of the trend of the modal sensitivity. It is not an approximation of the

impact of either reactive power control or terminal voltage control implemented by
the DFIG. The greater |Hdq(jωi)| of the terminal voltage control is than |Hdq(jωi)| of
the reactive power control, the more accurate the IDC is.
4.2.3 Example 4.2
4.2.3.1 Impact of Dynamic Interactions as Affected by Level of Wind

Power Penetration
Figure 4.10 shows the configuration of the example New England power system.
Parameters and models of the system and synchronous generators given in [13] were
used. In all the study cases presented in this section, detailed 14th-order dynamic model
of the DFIG including its control systems was used [10–12]. The DFIGs operated with
unity power factor. Parameters of the example are given in Appendix 4.2.
The EOM of concern was an inter-area EOM between the tenth synchronous
generator and the rest of the synchronous generators in the example New England
power system as the inter-area EOM of concern was most lightly damped.
8
1 37
30
26 28 29
25
2
27 24
18 38
1 17 9
16
10 6
3 35
39 15
14 21
4 22
DFIG
5 19
6 12 23
9 DFIG
20
11 13 33 36
31
4 7
7 2 34
10
5
8
32
3
Fig. 4.10 Configuration of New England power system

Two DFIGs were connected to the example New England power system at node
22 and 39 respectively. Table 4.5 presents the computational results of the proposed
index, IDI(jωi), when the active power output from the DFIGs Pw0 ¼ 5 p. u. and
Pw0 ¼ 10 p. u. respectively for both DFIGs. Firstly, it can be seen from Table 4.5
that with the increase of active power output from both DFIGs, level of the impact of
dynamic interactions on the inter-area EOM of concern increased as indicated by IDI
(jωi). Correctness of IDI(jωi) was confirmed by direct calculation of ^λ i λi from
eigensolution of closed-loop and open-loop state matrix.
Secondly, it can be seen from Table 4.5 that |Hdp(jωi)| increased with the increase
of wind power penetration. |Hdq(jωi)| was approximately equal to zero. Hence,
computational results of |Hdp(jωi)| and |Hdq(jωi)| demonstrate the correctness of
analysis carried out in Sect. 4.2.2.2 to draw the generally-applicable conclusion
from the proposed index that level of impact of dynamic interactions increases with
increase of wind power penetration.
Thirdly, IDI(jωi) presented in Table 4.5 indicated that the impact of the DFIG
at node 22 was positive as the damping of inter-area EOM of concern was improved
( ^λ i λi þ IDIðjωi Þ). Whilst the impact of the DFIG at node 39 was negative to
reduce the damping of inter-area EOM of concern. In Table 4.5, PFdfig is the sum of
the participation factors of all the state variables of the DFIGs. Computational results
of PFdfig confirmed the increase of dynamic interactions between the DFIGs and
power system when the wind power penetration increased, though PFdfig cannot
indicate the direction of the impact of dynamic interactions.

Fourthly, the modal-to-Pw0 sensitivity, ∂ ^λ i λi =∂Pw0 , can also be used as an

index to estimate the impact of dynamic interactions introduced by the DFIGs on the
inter-area EOM of concern when the level of wind power penetration varied.
Computational results of this modal-to-Pw0 sensitivity are given in Table 4.5.
From Table 4.5, it can be seen that both level and direction of impact of DFIGs
indicated by the modal-to-Pw0 sensitivity were as same as those by IDI(jωi),
confirming the correctness of proposed index, IDI(jωi). However,

it is very difficult
to draw the generally-applicable conclusion by use of ∂ ^λ i λi =∂Pw0 that impact
of dynamic interactions increases with increase of wind power penetration which
was obtained from proposed index IDI(jωi).
Finally, IDI(jωi) was evaluated for the DFIG at node 22 and 39 when the level of
wind power penetration changed from no wind (Pw0 ¼ 0 p. u.) to the maximum
wind capacity allowed by the system operation (Pw0 ¼ 15 p. u., beyond which the
load flow calculation of the example New England power system did not converge).
Figures 4.11 and 4.12 show the results of evaluation with confirmation from the
direct calculation of ^λ i λi . From Figs. 4.11 and 4.12, it can be seen that the
indication by IDI(jωi) on the impact of dynamic interactions introduced by the
DFIGs at node 22 and 39 is correct. Estimation error of IDI(jωi) on the actual impact,
^λ i λi , is small.
Figure 4.13 presents the magnitude of frequency-response, |Hdp(jω)|, for the
DFIG at node 22 and 39 respectively. From Fig. 4.13, it can be seen that |Hdp(jω)|
is outside the boundary[ω1, ω2], within which the IDI may become unreliable with
126
Table 4.5 Impact of dynamic interactions when levels of wind power penetration varied
DFIG location Node 22 Node 39
Pw0 5 10 5 10
IDI(jωi) 0.02437 þ 0.018577j 0.0448 þ 0.0462j 0.001652 þ 0.000526j 0.004894 þ 0.00223j
b
λ i λi 0.0286 þ 0.0133j 0.06521 þ 0.0324j 0.00152 þ 0.0009j 0.00395 þ 0.004j
PFdfig 0.010632 0.029117 0.00048045 0.0016741
|Hdp(jωi)| 2.8245 6.0514 3.056 6.8517
|Hdq(jωi)| 2.1552e05 2.202e-05 2.7174e05 3.0923e-05

∂ b λ i λi 0.00576 þ 0.0022j 0.01312 þ 0.0097j 0.00014 þ 0.0001j 0.0012 þ 0.0013j

∂ Kpp þ Kpi
4 Modal Analysis of Small-Signal Angular Stability of a Power System. . .
0.3 160
Magnitude IDI ( jωi ) Phase angle
155
0.25
( lˆ – l )
i i 150
0.2 Pw0 = 5
145
0.15 Pw0 =10 140
Pw0 =10
Pw0 = 5 135
0.1
130
IDI ( jωi )
0.05
125 (lˆ – l )
i i
Active power output from the DFIG Active power output from the DFIG
0 120
1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15
Fig. 4.11 Evaluation of the IDI for the DFIG at node 22 when wind power penetration varied
0.018 70
Magnitude IDI ( jωi ) Phase angle
0.015 60
(lˆ – l )
i i 50 Pw0 =10
0.012
Pw0 =10 40
0.009 30
Pw0 = 5
0.006 20
10
IDI ( jωi )
Pw0 =5
0.003
0 (lˆ – l )
i i
Active power output from the DFIG Active power output from the DFIG
0 -10
1 3 5 7 9 11 13 15 1 3 5 7 9 11 13 15
Fig. 4.12 Evaluation of the IDI for the DFIG at node 39 when wind power penetration varied
70 70
ω1 ωd ω2 ω1 ωd ω2 Pw0 =10
60 Pw0 =10 60
Pw0 = 5 Pw0 = 5
50 50
Magnitude
Magnitude
40 40
30 30
A A
20 20
10 10
0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Frequency (Hz) Frequency (Hz)
DFIG at node 22 DFIG at node 39
Fig. 4.13 Frequency response of |Hdp(jω)|
considerable error. Around ωi, |Hdp(jω)| does not change radically. Results presented
in Figs. 4.11, 4.12 and 4.13 confirm the analysis made in Sect. 4.2.1.2 about the
estimation error of the IDI.
4.2.3.2 Dynamic Interactions as Affected by Parameters of RSC Active

Power Controller
Table 4.6 presents the computational results of proposed index, IDI(jωi), when
setting of PI parameters, Kpp and Kip, of the active power control outer loop of the
RSC shown by Fig. 4.8 changed. From Table 4.6, it can be seen that with increased
values of PI parameters, |Hdp(jωi)| decreased. Impact of dynamic interactions intro-
duced by both DFIGs at node 22 and 39 on the inter-area EOM of concern decreased
accordingly as indicated by IDI(jωi), though directions of impact did not change.
Indication by IDI(jωi) in Table 4.6 was confirmed being correct by the results of
direct calculation of ^λ i λi and computation of the participation factors, PFdfig.
Computational results of the index of modal-to-parameters sensitivity are also
presented in Table 4.6 for confirmation and comparison. Impact of changes of Kpp
and Kip on the inter-area EOM of concern presented in Table 4.6 can be elaborated as
follows.
In Fig. 4.8 and (4.43), Kpq(jωi) ¼ Kpp þ Kpi/jω. Hence, |Kpq(jωi)| increased when
Kpp and Kip increased. Accordingly, |Hdp(jωi)| decreased as it can be seen from
(4.43). Subsequently, |IDI(jωi)| decreased as indicated by (4.46). Hence, increase of
the gain values of active power PI controller of the RSC led to decreased impact of
dynamic interactions introduced by the DFIG on the inter-area EOM of concern.
Elaboration above demonstrates the advantage of proposed index in the estab-
lishment of explicit connection between impact and frequency response of the DFIG,
which makes analysis possible. Hence, computational results obtained in Table 4.6
can be explained and understood. Obviously, it would be difficult to carry out the
same analysis as demonstrated above by use of index of participation factors and the
modal-to-parameters sensitivity. In addition, the proposed index is of certain com-
putational advantage: When Kpp and Kip varied, values of Hdp(jωi) and Hdq(jωi) were
calculated from the DFIG’s dynamic model such that IDI(jωi) was calculated directly
by use of (4.27). Hence, there was no need to carry out the eigensolution of power
system state matrix every time when Kpp and Kip varied as the open-loop state space
model of power system shown by (4.3) did not change with the variations of Kpp and
Kip. Whilst computation of the participation factors and the modal-to-parameters
sensitivity needs to perform eigensolution of power system state matrix every time
when Kpp and Kip varied.
4.2.3.3 Considerable Dynamic Interactions Introduced by the DFIG
In order to demonstrate and evaluate the proposed index, IDI(jωi), when the dynamic
interactions introduced by the DFIG are unusually considerable, a DFIG was
connected at node 19 in the example New England power system. The active
power output from the DFIG at steady state was 10 p.u. The DFIG operated with
unity power factor. When the DFIG is modelled as a constant power injection,
Pw0 ¼ 10 p. u., the open-loop inter-area EOM of the example New England
Table 4.6 Impact of dynamic interactions as affected by control parameters of the RSC
DFIG at Node 22 Node 39
Kpp, Kip 1.15 0.1.5 1.15 0.1.5
IDI(jωi) 0.019826 þ 0.013702j 0.0448 þ 0.0462j 0.0016709 þ 0.0012639j 0.004894 þ 0.00223j
b
λ i λi 0.02897 þ 0.0029j 0.06521 þ 0.0324j 0.00106 þ 0.0021j 0.00395 þ 0.004j
PFdfig 1.6469 0.029117 0.091813 0.0016741
|Hdp(jωi)| 2.2021 6.0514 2.5821 6.8517

∂ λbi λi 0.0172–0.00433j 0.0676–0.096j 0.00143 0.0006668j 0.008 0.002j

∂ Kpp þ Kpi
4.2 Impact of Dynamic Interactions Introduced by a Grid-Connected DFIG on. . .
129
Fig. 4.14 Frequency 40

response of |Hdp(jω)| to
assess the reliability of 35
the IDI 30
Magnietud
25
ω1
20
15 H dp ( jωi )
10
5
ωd
ω2
0
0.2 1.2 2.2 3.2 4.2 5.2 6.2 7.2 8
Frequency (Hz)
power system was λi ¼ 0.33834 þ j3.5342; thus ωi ¼ 0.56. Computational results

of the IDI was IDI(jωi) ¼ 0.046267 j0.37745, indicating that the dynamic
interactions introduced by the DFIG may degrade the damping of the inter-area
EOM of concern.
For examining the estimation error by the IDI, magnitude of frequency response
of the DFIG was obtained and is shown by Fig. 4.14. From Fig. 4.14, |Hdp(jωi)| was
identified to be at point B within the boundary, [ω1, ω2]. Hence, it is possible that the
IDI may have considerable estimation error on the impact of dynamic interactions
introduced by the DFIG at node 19.
To confirm the assessment above using the IDI, the closed-loop inter-area EOM
was calculate to be ^λ i ¼ 0:22662 þ j3:2592. Hence, ^λ i λi ¼ 0:11172 j0:275.
The actual damping degradation estimation error by the IDI was indeed consider-
able, though the IDI indicated correctly that the dynamic interactions introduced by
the DFIG at node 19 degraded the damping of the inter-area EOM of concern.
To eliminate the detrimental effect of dynamic interactions introduced by the
DFIG, control parameters of the DFIG were tuned such that the complex pole of
Hdp(s), λd ¼ ξd þ jωd, moved upwards on the complex plane and ωd increased. After
the control parameters were tuned, magnitude of frequency response of the DFIG
was obtained and is shown by Fig. 4.15. From Fig. 4.15, it can be seen that |Hdp(jωi)|
decreased and is at point A outside the boundary, [ω1, ω2]. This implies that firstly,
the scale of dynamic interactions introduced by the DFIG decreased. Secondly, the
IDI should be more accurate.
Computational results of the IDI for the case of point A in Fig. 4.15 was IDI
(jωi) ¼ 0.013282 0.036996j, indicating reduced impact of dynamic interactions.
Confirmation from the computation of closed-loop inter-area EOM of concern was
^λ i ¼ 0:33834 þ 3:5333j. Hence, ^λ i λi ¼ 0:0162 0:0347j. Indeed, the dynamic
interactions introduced by the DFIG decreased and the effect of damping degrada-
tion was reduced. In addition, the IDI was a good approximation of the actual
impact ^λ i λi .
Fig. 4.15 Frequency 70

response of |Hdp(jω)| with
increased ωd 60
50
Magnitude
40
30 A: H dp ( jωi )
20 ω1
10 ωd
ω2
0
0 1 2 3 4 5 6 7 8
Frequecy (Hz)
Fig. 4.16 Participation 0.6

factors for the closed-loop Case A
inter-area EOM of concern 0.5
Case B
Participation factors
0.4
0.3
0.2
0.1
0
sg1 sg2 sg3 sg4 sg5 sg6 sg7 sg8 sg9 sg10 dfig
Generators
To further confirm the examination carried out above, the participation factors for
the closed-loop inter-area EOM were calculated and computational results are
displayed in Fig. 4.16. In Fig. 4.16, sgi, i ¼ 1, 2, . . .10 indicates the participation
factors of the ith synchronous generator and dfig is the sum of the participation
factors of all the state variables of the DFIG.
Figure 4.17 presents the results of non-linear simulation. At 1 s of simulation,
mechanical torque of the DFIG dropped by 10% and the drop was recovered in
100 ms. In Figs. 4.16 and 4.17, case B is the first case discussed above when the IDI
was of considerable error, indicating strong dynamic interactions introduced by the
DFIG. Case A is the second case when the control parameters of the DFIG were
tuned such that the dynamic interactions introduced by the DFIG decreased.
From Fig. 4.16, it can be observed that in case B, the inter-area EOM of concern
was participated considerably by the DFIG, indicating strong dynamic interactions
introduced by the DFIG. From Fig. 4.17, it can be seen that damping degradation
caused by the strong dynamic interactions is noticeable in case B when the active
power output from the DFIG varies visibly.
Fig. 4.17 Confirmation by -1

non-linear simulation Case A
-2
Case B
-3
d10 – d1 (degree)
-4
-5
-6
-7
-8
0 1 2 3 4 5 6 7 8 9 10
Time(s)
10.4
Active power output of DFIG
10.2
10
9.8
9.6
9.4 Case A
Case B
9.2
0 1 2 3 4 5 6 7 8 9 10
Time(s)
4.2.3.4 Representation of Wind Farm by a Single DFIG
In Sect. 4.2.1.3, following analytical conclusion is made from the proposed index,
IDI(jωi): An equivalent dynamic model of single DFIG can be used to examine the
impact of wind farm on EOMs of concern, as long as frequency response of this
equivalent DFIG model around frequency of electromechanical oscillations of
concern approximately matches the frequency response of wind farm. In this
sub-section, tests were carried out to demonstrate and validate this analytical
conclusion.
The tests considered that node 22 and 39 in the example New England power
system were connected with a wind farm respectively instead of a DFIG. Each of
wind farms consisted of 20 DFIGs, among which 13 DFIGs used the reactive power
control and 7 DFIGs adopted terminal voltage control. Parameters and operating
conditions of those 20 DFIGs were different and grouped into three clusters.
Parameters and operating conditions of DFIGs in each cluster were same. Thus
transfer function models of DFIGs in each cluster were same and denoted as
Hdpk(s) and Hdqk(s), k ¼ 1, 2, 3.
Table 4.7 Impact of dynamic interactions introduced by the wind farms

Locations of wind farms Node 22 Node 39
b
λ i λi (first test: model of wind farms 0.0303 þ 0.037j 0.0028 þ 0.0035j
included all 20 DFIGs)
b
λ i λi (second test: model of single 0.0130 þ 0.0023i 0.0017 þ 0.0024i
DFIG was used)
b
λ i λi (third test: model of single DFIG 0.043635 þ 0.045579j 0.0037496 þ 0.0028666j
with tuned parameters was used)
ID( jωi) 0.04288 þ 0.046001j 0.0032252 þ 0.0029898j
In the first test, dynamic model of each of wind farms was established with
detailed 14th-order dynamic models of all 20 DFIGs being used. Impact of wind
farms on the inter-area EOM of concern was assessed by direct calculation of Δλi
¼ ^λ i λi from eigensolution of open-loop and closed-loop state matrix. Results of
assessment are presented in the second row of Table 4.7. It can be seen that impact of
the wind farm at node 22 was negative as real part of Δλi > 0 such that the dynamic
interactions introduced by the wind farm caused the decrease of damping of inter-
area EOM of concern. Whilst impact of the wind farm at 39 was positive. In this test,
frequency response of two wind farms at node 22 and 39, FP k(jω) and FQ k(jω),
k ¼ 22, 39was obtained by simulation and are shown by solid curves in Figs. 4.18
and 4.19.
In the second test, transfer function model of a single DFIG in the first cluster of
wind farms, Hdp1(s) and Hdq1(s), with increased power output equal to that of wind
farms was simply used to represent wind farms. Impact of wind farms was assessed
by direct calculation of Δλi ¼ ^λ i λ from eigensolutions of open-loop and closed-
loop state matrix. Results of assessment are presented in the third row of Table 4.7.
Comparing results in the second and third row of Table 4.7, it can be seen that
representation of the wind farms by dynamic model of single DFIG produced
incorrect assessment of the impact of the wind farms.
Frequency response of the single DFIG with increased capacity to represent the
wind farms at node 22 and 39 is CkHdp1(jω) and CkHdq1(jω), k ¼ 22, 39, where Ck,
k ¼ 22, 39 are constants to represent the increased capacity of the single DFIG. In
Figs. 4.18 and 4.19, CkHdp1(jω), k ¼ 22, 39 are shown by dash-dotted curves. From
Figs. 4.18 and 4.19, it can be seen that at ω ¼ ωi, both the magnitude and phase
ofFP k(jω) and FQ k(jω), k ¼ 22, 39 are different with those of CkHdp1(jω) and
CkHdq1(jω), k ¼ 22, 39. That is why wrong assessment on the impact of the DFIG at
node 22 and 39 was made by using the model of single DFIG.
In the third test, PI gains of the RSC active and reactive power control outer loop
of the DFIGs, Kpp, Kip, Kpq and Kiq, were tuned. Transfer function model of the
DFIGs, CkHdp1(jω) and CkHdq1(jω), k ¼ 22, 39, changed with the tuning of the
gains. Objective of tuning PI gains is to make
Ck Hdp1 ðjωÞ FPk ðjωi Þ
ð4:53Þ
Ck Hdq1 ðjωÞ FQk ðjωi Þ, k ¼ 22, 39
8 65
Magnitude Phase angle FP–22 ( j ω)
7 60
C22G P0 –22 ( jω)
FP–22 ( j ω) 55
6 C22 H dp1 ( j ω)
C22GP 0–22 ( jω) 50
5 C22 H dp1 ( j ω) 45
4 ωi 40 ωi
35
3 30
Frequency(Hz) Frequency(Hz)
2 25
0.4 0.6 0.8 1 1.2 1.4 1.6 0.4 0.6 0.8 1 1.2 1.4 1.6
25 110
Magnitude FQ –22 ( jω) 100 Phase angle
20 C22GQ 0–22 ( jω) 90
C22 H dq1 ( jω) 80 FQ –22 ( j ω)
15 70 C22GQ 0–22 ( jω)
10 60 C22 H dq1 ( j ω)
ωi 50 ωi
5 40
Frequency(Hz) Frequency(Hz)
30
0 20
0.4 0.6 0.8 1 1.2 1.4 1.6 0.4 0.6 0.8 1 1.2 1.4 1.6
Fig. 4.18 Frequency response of the wind farm at node 22
With (4.53), the frequency response of the equivalent DFIGs’ model around
frequency of electromechanical oscillations of concern approximately matches the
frequency response of the wind farms. Hence, following objective function was set
up

¼ min CkHdp1 ðjωi Þ jF
Pk ðjωi Þj þ ∠Ck Hdp1 ðjωi Þ ∠FPk ðjωi Þ
Obj

þ Ck Hdq1 ðjωi Þ FQk ðjωi Þ þ ∠Ck Hdq1 ðjωi Þ ∠FQk ðjωi Þ , k ¼ 22, 39
ð4:54Þ
The Hooke-Jeeves direct searching algorithm [14] was used to find the solution of
the above objective function. Initial searching step was 0.01. Solutions of objective
functions by the direct searching were Kpp ¼ 0.12, Kip ¼ 5.2, Kpq ¼ 3.2 and Kiq ¼ 58
for the DFIGs. Then, PI gains of the DFIGs were changed to the searching solutions.
Transfer function models of the equivalent single DFIG for the wind farms at node
22 and 39 were changed to CkGP0 k(s), CkGQ0 k(s), k ¼ 22, 39. Using this
equivalent model of single DFIG, computational results of Δλi ¼ ^λ i λi are given
in the fourth row of Table 4.7. In addition, computational results of IDI(jωi) by using
CkGP0 k(jωi), k ¼ 22, 39 are presented in the fifth row of Table 4.7. Comparing
results in the second, fourth and fifth row of Table 4.7, it can be seen that estimation
of impact of wind farms by using the dynamic model of single DFIG with parameters
being tuned according to (4.53) is very close to the correct results.
8 65
Magnitude Phase angle FP −39 ( j ω)
7 60
FP −39 ( j ω) 55 C39G P0 −39 (j ω)
6 C39G P0 −39 (j ω) C39 H dp1 ( j ω)
50
C39 H dp1 ( j ω)
5 45
4 ωi 40 ωi
35
3
Frequency(Hz) 30
Frequency(Hz)
2 25
0.4 0.6 0.8 1 1.2 1.4 1.6 0.4 0.6 0.8 1 1.2 1.4 1.6
25 110
Magnitude Phase angle
FQ −39 ( jω) 100
20 C39GQ0 −39 (j ω) 90
80 FQ −39 ( jω)
C39 H dq1 ( jω)
15 70 C39GQ 0−39 (jω)
60 C39 H dq1 ( jω)
10 ωi ωi
50
5 40 Frequency(Hz)
Frequency(Hz)
30
0 20
0.4 0.6 0.8 1 1.2 1.4 1.6 0.4 0.6 0.8 1 1.2 1.4 1.6
Fig. 4.19 Frequency response of the wind farm at node 39
Therefore, above tests confirmed the analytical conclusion drawn in Sect. 4.2.1.3
about how correctly the dynamics of a wind farm can be represented by the dynamic
model of single DFIG.
4.2.3.5 The IDI Obtained by Using the Measuring Data
The IDI can be applied to assess the impact of wind farms from measuring data
instead of using parametric model of power system. This merit of the IDI is
demonstrated in this subsection. Measuring data were obtained from the simulation
tests which were polluted with white noises. In the simulation, the wind farms at
node 22 and 39 were constructed by 20 DFIGs respectively.
Firstly, frequency-response tests at the terminals of the wind farms at node 22 and
39 were conducted. Thus, the values of frequency response of the wind farms at the
frequency, ωi, were obtained and are given in the first and second row of Table 4.8.
Secondly, a small step increment of active power and reactive power output from
the wind farms at node 22 and node 39, ΔPw and ΔQw, was applied. Subsequent
increments of the magnitude of terminal voltage at node 22 and 39, ΔVw, were
measured. Thus, dvp and dvq were obtained from the ratio of ΔPw and ΔQw to ΔVw
respectively. The results are presented in the third and fourth row of Table 4.8. To
confirm the results obtained from the measuring data and given in Table 4.8,
136
Table 4.8 Impact of dynamic interactions assessed by the measuring data

Locations of wind farms Node 22 Node 39
Results from measuring data Results from parametric model Results from measuring data Results from parametric model
FP k(jωi) 2.9238 þ 4.0322j 2.9921 þ 4.0539j 2.9825 þ 4.0809j 3.6246 þ 4.2322j
FQ k(jωi) 1.1229 þ 17.838j 1.1122 þ 17.545j 1.1018 þ 17.7j 0.99652 þ 17.284j
dvp 0.0210 0.0160 0.0084 0.0082
dvq 0.0422 0.0378 0.0301 0.0287
Rvpi 0.0013 þ 0.0079j 0.0018 þ 0.0112j 0.0005–0.0031j 0.0006–0.00029j
Rvqi 0.0009 ‑ 0.0044j 0.0014 ‑ 0.0064j 0.0009 þ 0.0011j 0.0002 þ 0.0003j
IDI(jωi) 0.0274 þ 0.0318j 0.0429 þ 0.0460j 0.0076þ 0.0035j 0.0032 þ 0.0030j
4 Modal Analysis of Small-Signal Angular Stability of a Power System. . .
Table 4.9 Computational results of index to estimate impact of DFIG’s terminal voltage control as
compared with reactive power control
Node 22 Node 15 Node 1
Rvqi/dvq 0.045899 0.23573j 0.01666 0.082519j 0.012428 þ 0.018207j
computational results from parametric model of the example New England power
system with the wind farms are listed Table 4.8.
Finally, residues were obtained from the measuring data by using the method
proposed in [15] instead of parametric model of power system. Results are presented
in the fifth and sixth row of Table 4.8. Hence from the measuring data, the IDI was
calculated and given in the last row of Table 4.8. Comparing the results of the IDI
obtained from the measuring data and parametric model given in Table 4.8, it can be
seen that the IDI obtained from the measuring data provides correct assessment on
the impact of the wind farms at node 22 and 39.
4.2.3.6 Terminal Voltage Control Vs. Reactive Power Control
In order to demonstrate and validate the IDC, Rvqi/dvq, to predict the impact of
DFIG’s terminal voltage control as compared with the reactive power control, two
more optional connecting locations for the DFIG at node 22 were examined. Those
two optional locations were node 15 and node 1 in the example New England power
system. Computational results of Rvqi/dvq are given in Table 4.9. From Table 4.9, it
can be seen that (Real part of Rvqi/dvq) < 0 if the DFIG was connected at node 22 and
15, which indicated λR to be on the left hand side of λV on the complex plane. Hence,
the impact of DFIG’s reactive power control was more beneficial to the damping of
the inter-area EOM of concern as compared with the reactive power control. Whilst
(Real part of Rvqi/dvq) > 0 for node 1. Hence, damping of inter-area EOM of concern
was benefited if the terminal voltage control was adopted by the DFIG connected at
node 1 instead of the reactive power control.
To confirm the correctness of prediction from the IDC, as given in Table 4.9,
firstly the DFIG was equipped with a well-designed reactive power controller. Gain
value of PI reactive power controller in the outer loop (see Fig. 4.8) was set to be
Kpq ¼ 5, Kiq ¼ 50. The DFIG was connected at node 1, 15 and 22 respectively. The
inter-area EOM of concern was calculated and indicated as λR on the complex plane
in Fig. 4.20. Secondly, the reactive power controller was replaced by the terminal
voltage controller. Gain values of terminal voltage PI controller were set to be zero,
i.e., Kpq ¼ 0, Kiq ¼ 0, such that Kpd(s) ¼ 0. The inter-area EOM of concern was
calculated and indicated as λV(0) in Fig. 4.20. Finally, gain values of terminal
voltage PI controller were increased from Kpq ¼ 0.0, Kiq ¼ 0 to Kpq ¼ 1000,
Kiq ¼ 10000. Trajectories of inter-area EOM of concern, λV, starting from
λV(0) are presented in Fig. 4.20.
From Fig. 4.20, it can be seen that λV(0) almost overlaps λR and the direction and
trend of the variation of λR towards λV agrees well with the prediction by the IDC
4.1
Bus 22 Bus 15
4.0 Bus 1
Rvqi
3.9 λR – dvq
Imagiary axis
3.8
3.7 direction of λR moving

towards λV
λR
3.6
λV (0)
trajectory of λV
3.5
-0.36 -0.34 -0.32 -0.3 -0.28 -0.26
Real axis
Fig. 4.20 Trajectories of the EOM when PI parameters of voltage controller increased
presented in Table 4.9. Hence, Fig. 4.20 confirms the analysis carried out to derive
the IDC in Sect. 4.2.2.3 and the correctness of relative position of λV in respect to λR
as being predicted by the computational results given in Table 4.9.
Appendix 4.1: Data of Example 4.1

Power System [3] (Tables 4.10, 4.11, 4.12)
Appendix 4.1: Data of Example 4.1 139
Table 4.10 Bus data

Bus Active power Reactive power Active power Reactive power
no. consumption consumption generation generation
1 2.527 1.1856 – –
3 3.22 0.02 – –
4 5 1.84 – –
7 2.34 0.84 – –
8 5.22 1.77 – –
9 1.04 1.25 – –
12 0.09 0.88 – –
15 3.2 1.53 – –
16 3.29 0.32 – –
18 1.58 0.3 – –
20 6.8 1.03 – –
21 1.74 1.15 – –
23 1.48 0.85 – –
24 3.09 0.92 – –
25 2.24 0.47 – –
26 1.39 0.17 – –
27 2.81 0.76 – –
28 2.06 0.28 – –
29 2.84 0.27 – –
33 1.12 0 – –
36 1.02 0.1946 – –
37 60 3 – –
39 2.67 0.126 – –
40 0.6563 0.2353 – –
41 10 2.5 – –
42 11.5 2.5 – –
44 2.6755 0.0484 – –
45 2.08 0.21 – –
46 1.507 0.285 – –
47 2.0312 0.3259 – –
48 2.412 0.022 – –
49 1.64 0.29 – –
50 2 1.47 – –
51 4.37 1.22 – –
52 24.7 1.23 – –
53 – – 2.5 0.4789
54 – – 5.45 2.5134
55 – – 6.5 2.7087
56 – – 6.32 2.0328
57 – – 5.052 1.6098
58 – – 7 1.8488
59 – – 5.6 0.4723
(continued)

60 – – 5.4 0.0897
61 – – 8 0.2378
62 – – 5 0.0543
63 – – 10 0.2462
64 – – 13.5 1.9941
65 – – 35.91 8.9526
66 – – 17.85 0.5714
67 – – 10 0.70456
68 – – 40 4.7864
Table 4.11 Line data

Branch no. From bus To bus R X B Transformer ratio
1 1 2 0.007 0.0822 0.3493 1
2 1 30 0.0008 0.0074 0.48 1
3 2 3 0.0013 0.0151 0.2572 1
4 2 25 0.007 0.0086 0.146 1
5 2 53 0 0.0181 0 1.025
6 3 4 0.0013 0.0213 0.2214 1
7 3 18 0.0011 0.0133 0.2138 1
8 4 5 0.0008 0.0128 0.1342 1
9 4 14 0.0008 0.0129 0.1382 1
10 5 6 0.0002 0.0026 0.0434 1
11 5 8 0.0008 0.0112 0.1476 1
12 6 7 0.0006 0.0092 0.113 1
13 6 11 0.0007 0.0082 0.1389 1
14 6 54 0 0.025 0 1.07
15 7 8 0.0004 0.0046 0.078 1
16 8 9 0.0023 0.0363 0.3804 1
17 9 30 0.0019 0.0183 0.29 1
18 10 11 0.0004 0.0043 0.0729 1
19 10 13 0.0004 0.0043 0.0729 1
20 10 55 0 0.02 0 1.07
21 12 11 0.0016 0.0435 0 1.06
22 12 13 0.0016 0.0435 0 1.06
23 13 14 0.0009 0.0101 0.1723 1
24 14 15 0.0018 0.0217 0.366 1
25 15 16 0.0009 0.0094 0.171 1
26 16 17 0.0007 0.0089 0.1342 1
27 16 19 0.0016 0.0195 0.304 1
28 16 21 0.0008 0.0135 0.2548 1
(continued)

29 16 24 0.0003 0.0059 0.068 1
30 17 18 0.0007 0.0082 0.1319 1
31 17 27 0.0013 0.0173 0.3216 1
32 19 20 0.0007 0.0138 0 1.06
33 19 56 0.0007 0.0142 0 1.07
34 20 57 0.0009 0.018 0 1.009
35 21 22 0.0008 0.014 0.2565 1
36 22 23 0.0006 0.0096 0.1846 1
37 22 58 0 0.0143 0 1.025
38 23 24 0.0022 0.035 0.361 1
39 23 59 0.0005 0.0272 0 1
40 25 26 0.0032 0.0323 0.531 1
41 25 60 0.0006 0.0232 0 1.025
42 26 27 0.0014 0.0147 0.2396 1
43 26 28 0.0043 0.0474 0.7802 1
44 26 29 0.0057 0.0625 1.029 1
45 28 29 0.0014 0.0151 0.249 1
46 29 61 0.0008 0.0156 0 1.025
47 9 30 0.0019 0.0183 0.29 1
48 9 36 0.0022 0.0196 0.34 1
49 9 36 0.0022 0.0196 0.34 1
50 36 37 0.0005 0.0045 0.32 1
51 34 36 0.0033 0.0111 1.45 1
52 35 34 0.0001 0.0074 0 0.946
53 33 34 0.0011 0.0157 0.202 1
54 32 33 0.0008 0.0099 0.168 1
55 30 31 0.0013 0.0187 0.333 1
56 30 32 0.0024 0.0288 0.488 1
57 1 31 0.0016 0.0163 0.25 1
58 31 38 0.0011 0.0147 0.247 1
59 33 38 0.0036 0.0444 0.693 1
60 38 46 0.0022 0.0284 0.43 1
61 46 49 0.0018 0.0274 0.27 1
62 1 47 0.0013 0.0188 1.31 1
63 47 48 0.0025 0.0268 0.4 1
64 47 48 0.0025 0.0268 0.4 1
65 48 40 0.002 0.022 1.28 1
66 35 45 0.0007 0.0175 1.39 1
67 37 43 0.0005 0.0276 0 1
68 43 44 0.0001 0.0011 0 1
69 44 45 0.0025 0.073 0 1
70 39 44 0 0.0411 0 1
(continued)

71 39 45 0 0.0839 0 1
72 45 51 0.0004 0.0105 0.72 1
73 50 52 0.0012 0.0288 2.06 1
74 50 51 0.0009 0.0221 1.62 1
75 49 52 0.0076 0.1141 1.16 1
76 52 42 0.004 0.06 2.25 1
77 42 41 0.004 0.06 2.25 1
78 41 40 0.006 0.084 3.15 1
79 31 62 0 0.026 0 1.04
80 32 63 0 0.013 0 1.04
81 36 64 0 0.0075 0 1.04
82 37 65 0 0.0033 0 1.04
83 41 66 0 0.0015 0 1
84 42 67 0 0.0015 0 1
85 52 68 0 0.003 0 1
86 1 27 0.032 0.32 0.41 1

Generator No. Xd X0d Td0 Xq M D KA TA
G1 1.8 0.3582 10.2 1.242 2.3333 10 10 0.01
G2 1.8 0.3252 6.56 1.7207 3.9494 10 10 0.01
G3 1.8 0.283 5.7 1.7098 3.9623 10 10 0.01
G4 1.8 0.2994 5.69 1.7725 6.5629 13 10 0.01
G5 1.8 0.36 5.4 1.6909 3.7667 8 10 0.01
G6 1.8 0.3543 7.3 1.7079 2.9107 14 10 0.01
G7 1.8 0.2989 5.66 1.7817 3.3267 2 10 0.01
G8 1.8 0.3537 6.7 1.7379 2.915 10 10 0.01
G9 1.8 0.2871 4.79 1.7521 2.0365 1.8 0 0.01
G10 1.8 0.5867 9.37 1.2249 2.9106 30 0 0.01
G11 1.8 0.6531 4.1 1.7297 2.0053 30 0 0.01
G12 1.8 0.5524 7.4 1.6931 5.1791 70 0 0.01
G13 1.8 0.4344 5.9 1.7392 9.0782 100 0 0.01
G14 1.8 0.485 4.1 1.73 3 90 0 0.01
G15 1.8 0.585 4.1 1.73 3 70 0 0.01
G16 1.8 0.6589 7.8 1.6888 4.45 60 0 0.01
Data of DFIG [4] (Tables 4.13, 4.14, 4.15)
Table 4.13 Basic data of Jdr1 Jdr2 Ddm Ddr1 Ddr2 Rds
DFIG
8 8 0 0 0 0
Rdr Xdm Xdss Xdrr Cd Xfd
0.0145 2.4012 0.1784 0.1225 13.29 5
Table 4.14 PI gains of RSC Kdp1 Kdi1 Kdp2 Kdi2 Kdp3 Kdi3 Kdp4 Kdi4
of DFIG
0.2 12.56 1 62.5 0.2 12.56 1 62.5
Table 4.15 PI gains of GSC Kdp5 Kdi5 Kdp6 Kdi6 Kdp7 Kdi7
of DFIG
0.2 12.56 1 62.5 1 62.5
Data of New England Power System [13] (Tables 4.16, 4.17

and 4.18)
Table 4.16 Line data

1 1 2 0.0035 0.0411 0.6987 1
2 1 39 0.001 0.025 0.75 1
3 2 3 0.0013 0.0151 0.2572 1
4 2 25 0.007 0.0086 0.146 1
5 3 4 0.0013 0.0213 0.2214 1
6 3 18 0.0011 0.0133 0.2138 1
7 4 5 0.0008 0.0128 0.1342 1
8 4 14 0.0008 0.0129 0.1382 1
9 5 6 0.0002 0.0026 0.0434 1
10 5 8 0.0008 0.0112 0.1476 1
11 6 7 0.0006 0.0092 0.113 1
12 6 11 0.0007 0.0082 0.1389 1
13 7 8 0.0004 0.0046 0.078 1
14 8 9 0.0023 0.0363 0.3804 1
15 9 39 0.001 0.025 1.2 1
16 10 11 0.0004 0.0043 0.0729 1
17 10 13 0.0004 0.0043 0.0729 1
18 13 14 0.0009 0.0101 0.1723 1
19 14 15 0.0018 0.0217 0.366 1
20 15 16 0.0009 0.0094 0.171 1
21 16 17 0.0007 0.0089 0.1342 1
22 16 19 0.0016 0.0195 0.304 1
23 16 21 0.0008 0.0135 0.2548 1
24 16 24 0.0003 0.0059 0.068 1
25 17 18 0.0007 0.0082 0.1319 1
26 17 27 0.0013 0.0173 0.3216 1
27 21 22 0.0008 0.014 0.2565 1
28 22 23 0.0006 0.0096 0.1846 1
29 23 24 0.0022 0.035 0.361 1
30 25 26 0.0032 0.0323 0.513 1
31 26 27 0.0014 0.0147 0.2396 1
32 26 28 0.0043 0.0474 0.7802 1
33 26 29 0.0057 0.0625 1.029 1
34 28 29 0.0014 0.0151 0.249 1
35 12 11 0.0016 0.0435 0 1.006
36 12 13 0.0016 0.0435 0 1.006
37 2 30 0 0.0181 0 1.025
38 6 31 0 0.025 0 1.07
39 10 32 0 0.02 0 1.07

Generator no. Xd X0d Xq J D Tf Kf KA
G1 0.1 0.031 0.069 84 40 10.2 0.06 10
G2 0.295 0.0647 0.282 60.6 40 6.56 0.05 10
G3 0.2495 0.0531 0.237 71.6 40 5.7 0.06 10
G4 0.262 0.0436 0.258 57.2 40 5.69 0.06 10
G5 0.67 0.132 0.62 52 40 5.4 0.02 10
G6 0.254 0.05 0.241 52.8 40 7.3 0.02 10
G7 0.295 0.049 0.292 69.6 40 5.66 0.02 10
G8 0.29 0.057 0.28 48.6 40 6.7 0.02 10
G9 0.2106 0.057 0.205 69 40 4.79 0.02 10
G10 0.2 0.006 0.019 1000 40 7 0.01 10
Table 4.18 Bus data

3 3.22 0.024 – –
4 5 1.84 – –
7 2.338 0.84 – –
8 5.22 1.76 – –
13 0.085 0.88 – –
15 3.2 1.53 – –
16 3.294 0.323 – –
18 1.58 0.3 – –
20 6.8 1.03 – –
21 2.74 1.15 – –
23 2.475 0.846 – –
24 3.086 0.922 – –
25 2.24 0.472 – –
26 1.39 0.17 – –
27 2.81 0.755 – –
28 2.06 0.276 – –
29 2.835 0.269 – –
30 – – 2.5 1.0475
31 0.092 0.046 – –
32 – – 5.5 0.9831
33 – – 3.32 0.9972
34 – – 5.08 1.0123
35 – – 3.5 1.0493
36 – – 5.6 1.0635
37 – – 5.4 1.0278
38 – – 5.3 1.0265
39 11.04 2.5 10 1.03
Data of DFIG [16] (Tables 4.19, 4.20 and 4.21)
Table 4.19 Basic data of Xdss Xdm Xdrr Jdr1 Jdr2 Rdr Rds Xdf Cd
DFIG
2.58 2.40 2.52 3s 3s 0.015 0 0.05 30
Table 4.20 PI gains of RSC Kdp1 Kdi1 Kdp2 Kdi2 Kdp3 Kdi3 Kdp4 Kdi4
of DFIG
0.2 10 0.4 15 0.2 10 0.4 15
Table 4.21 PI gains of GSC Kdp5 Kdi5 Kdp6 Kdi6 Kdp7 Kdi7
of DFIG
0.1 5 10 62.5 0.2 10
References

2. Yu YN (1979) Power system dynamics. Academic Press, New York
3. Kundur P (1994) Power system stability and control. McGraw-Hill, New York
4. Abad G, Lopez J, Rodriguez M, Marroyo L, Iwanski G (2011) Doubly fed induction machine:
modeling and control for wind energy generation. John Wiley & Sons, Hoboken
5. Tsourakis G, Nomikos BM, Vournas CD (2009) Effect of wind parks with doubly fed
asynchronous generators on small-signal stability. Electr Power Syst Res 79(1):190–200
6. Fan L, Miao Z, Osborn D (2008) Impact of doubly fed wind turbine generation on inter-area
oscillation damping. In: Proc IEEE Power Eng Soc Gen Meeting, Pittsburgh, PA, pp 1–8
8. Vittal E, O’Malley M, Keane A (2012) Rotor angle stability with high penetrations of wind
10. Ekanayake JB, Holdsworth L, Jenkins N (2003) Comparison of 5th order and 3rd order machine
models for doubly fed induction generator (DFIG) wind turbines. Elect Power Syst Res 67
(3):207–215
11. Feijóo A, Cidrás J, Carrillo C (2000) A third order model for the doubly-fed induction machine.
Elect Power Syst Res 56(2):121
12. Ko HS, Yoon GG, Kyung NH, Hong WP (2008) Modeling and control of DFIG-based variable-
speed wind-turbine. Elect Power Syst Res 78(11):1841–1849
13. Padiyar KR (1996) Power system dynamics stability and control. Wiley, New York
14. Avriel M (1976) Nonlinear programming analysis and method. Prentice Hall, Englewood Cliffs
15. Nudelland TR, Chakrabortt A (2014) A graph-theoretic algorithm for localization of forced
harmonic oscillation inputs in power system networks. In: 2014 American Control Conference
(ACC), Portland, OR, pp 1334–1340
16. MATLAB Simulink wind farm-DFIG detailed model. Wind Farm (DFIG Phasor Model)
Chapter 5
Small-Signal Angular Stability of a Power
System Affected by Strong Dynamic
Interactions Introduced
from a Grid-Connected VSWG
Small-signal angular stability of a power system may be affected considerably by a

grid-connected VSWG under the condition of open-loop modal resonance. This
chapter introduces the concept of open-loop modal resonance. Why strong dynamic
interactions may be induced by the grid-connected VSWG when the open-loop
modal resonance happens is examined. Analysis is presented to indicate that the
open-loop modal resonance may very likely degrade the small-signal angular sta-
bility of the power system. Both the grid-connected PMSG and DFIG are studied.
5.1 Impact of Strong Dynamic Interactions Introduced by

a Grid-Connected PMSG
A PMSG for the wind power generation is connected to a power system through
fully controlled voltage source converter (VSC). Dynamic interactions between such
VSC-controlled generation system and AC grid have been an actively pursued
research topic in recent years. Impact of dynamic interactions between
VSC-controlled system and external power system is twofold. One is on the dynamic
performance of VSC-controlled system itself, which is reviewed in Chap. 1. Another
is on the dynamic performance of external power system. Electromechanical
low-frequency oscillation is one of typical power system dynamic problems and
characterized by damping of power system EOMs.
In [1], the Western Electricity Coordinating Council (WECC) power system with
PMSGs in the United States for years 2010, 2020 and 2022 was examined. Com-
putational results of the participation factors in [1] indicated that state variables
associated with reactive power control outer loops of some PMSGs took part in the
power system EOMs considerably. Converter oscillation modes of those PMSGs
were also participated by some synchronous generators noticeably. Phenomenon of
such considerable dual participations of the PMSGs and synchronous generators in
the converter oscillation modes and the power system EOMs were found to be

148 5 Small-Signal Angular Stability of a Power System Affected by Strong. . .
related with the nearness of frequencies of involved converter oscillation modes and
the power system and adoption of type 4 reactive power control by the PMSGs [1].
Grid connection of a PMSG for the wind power generation brings about changes
of system load flow and introduces dynamic interactions as a grid-connected DFIG
does, which has been discussed in the previous chapter. Those are two main factors
that the PMSG affects power system small-signal angular stability. It has been
recognized that the dynamic interactions between the PMSG and the power system
are normally weak. Power system EOMs usually are not expected to be affected
considerably by the dynamic interactions introduced by the grid-connected PMSG.
This has been examined in the previous chapter for the grid-connected DFIG.
However, the observable dual participations of the PMSGs and synchronous gener-
ators reported in [1] in fact meant possible strong dynamic interactions between the
PMSG and power system. Such strong dynamic interactions introduced by the
PMSG may significantly affect power system small-signal angular stability.
In this section, the strong dynamic interactions between a grid-connected PMSG
and power system and their impact on power system small-signal angular stability
are examined. The examination is based on the closed-loop interconnected dynamic
model of the power system with the PMSG, which is derived in Sect. 2.3 of Chap. 2.
Then, the damping torque analysis is applied to examine the damping torque
contributions from the PMSG to the synchronous generators in power system,
which has been introduced and applied for the grid-connected DFIG in Chaps. 3
and 4. Analytical results indicate that under a special condition that an open-loop
converter oscillation mode of the PMSG is close to an open-loop EOM of the power
system on the complex plane, strong dynamic interactions between the PMSG and
power system may occur to considerably affect power system small-signal angular
stability. This special condition is referred to as the open-loop modal resonance
(OLMR). The consequence of the OLMR is examined to indicate that the modal
resonance may possibly lead to the closed-loop converter oscillation mode and
power system EOM moving towards the opposite directions in respect to the
positions of resonant open-loop converter oscillation mode and the EOM on the
complex plane. Thus, damping of either closed-loop converter oscillation mode or
the EOM may decrease such that power system small-signal angular stability is
affected negatively by the strong dynamic interactions introduced from the PMSG
under the condition of the OLMR. Further analysis is conducted in the section to
indicate that the responsible open-loop converter oscillation mode to cause the
OLMR is related to the grid-side converter (GSC) control systems of the PMSG.
Both the reactive power control loop with type 4 model and the active power control
loop of the GSC may provide responsible converter oscillation mode to cause the
OLMR. Finally, an example multi-machine power system with the PMSGs is
presented to demonstrate and validate the analysis and conclusions made in the
section.
5.1 Impact of Strong Dynamic Interactions Introduced by a Grid-Connected PMSG 149
5.1.1 Strong Dynamic Interactions Introduced by the Grid-

Connected PMSG and the Impact
5.1.1.1 Strong Dynamic Interactions Introduced by the Grid-

Connected PMSG
In Sect. 2.4.1 of Chap. 2, a closed-loop interconnected model of a power system with

a grid-connected PMSG is derived. The derived closed-loop interconnected model is
shown by Fig. 2.9 and redrawn as Fig. 5.1.
The closed-loop interconnected model is comprised of the PMSG subsystem and
the subsystem of the rest of power system (ROPS). State-space and transfer function
models of the PMSG subsystem are (2.124) and (2.125), which are re-presented as
follows.
8
> d
>
< dtΔXp ¼ Ap ΔXp þ bpd ΔVw
>
ΔPw ¼ cpp T ΔXp þ dpp1 ΔVw ð5:1Þ
>
>
>
:
ΔQw ¼ cpq T ΔXp þ dpq1 ΔVw
(
ΔPw ¼ Hpp ðsÞΔVw
ð5:2Þ
ΔQw ¼ Hpq ðsÞΔVw
where
1
Hpp ðsÞ ¼ cpp T sI Ap bpd þ dpp1
1 :
Hpq ðsÞ ¼ cpq T sI Ap bpd þ dpq1
Fig. 5.1 Closed-loop ROPS subsystem

power system with ΔPw
the PMSG Gvp (s)
ΔVw
+
Gvq (s)
ΔQw
Hpq (s)
Hpp (s)
PMSG subsystem
The state-space and transfer function models are (2.126) and (2.127), which are
re-listed as follows.
8
g g g p w q w
dt ð5:3Þ
:
ΔVw ¼ cgv ΔXg þ dvp ΔPw þ dvq ΔQw
T
where
1
Gvp ðsÞ ¼ cgv T sI Ag bp þ dvp
1 :
Gvq ðsÞ ¼ cgv T sI Ag bq þ dvq
The state-space model of the closed-loop interconnected system is (2.128), i.e.,

d
ΔXgp ¼ Agp ΔXgp ð5:5Þ
dt
where ΔXgp ¼ [ΔXgT ΔXpT]T.
Denote ^λ i a power system EOM of concerns with dynamics of PMSG being
included. Hence, ^λ i is an eigenvalue of closed-loop state matrix, Agp, in (5.5). If
ΔPw þ jΔQw ¼ 0, there are no dynamic interactions between the PMSG and the
power system. In this case, the PMSG degrades into a constant power source,
Pw0 þ jQw0, and the linearized model of the power system of (5.3) becomes
d
dt
Denote λi as the EOM corresponding to ^λ i when the PMSG is modelled as
constant power source, Pw0 þ jQw0, when ΔPw þ jΔQw ¼ 0. Obviously, λi is an
eigenvalue of open-loop state matrix of system, Ag. Difference of closed-loop and
open-loop EOM, Δλi ¼ ^λ i λi , is caused by the dynamic interactions introduced by
the PMSG. Hence, Δλi ¼ ^λ i λi measures the impact of dynamic interactions on the
EOM of concerns. The closed-loop interconnected model shown by Fig. 5.1 estab-
lishes the link between the impact of the PMSG on EOM of concerns and the
dynamic interactions introduced by the PMSG which exhibit as ΔPw þ jΔQw.
The damping torque analysis can be applied to estimate Δλi ¼ ^λ i λi , as it is
introduced in Chaps. 3 and 4 for the grid-connected DFIG. The estimation on the
impact of dynamic interactions introduced by the PMSG by use of the damping
torque analysis is
XN
Δλi ¼ Sik Tk ,
nh
k¼1 i o ð5:7Þ
Tk ¼ Re gpk ðλi ÞHpp ðλi Þ þ gqk ðλi ÞHpq ðλi Þ γk ðλi Þ
where gpk(s) and gqk(s) is the transfer function of the forward path from ΔPw and
ΔQw to the electromechanical oscillation loop of the kth synchronous generator in
the power system respectively, γk(s) is the reconstructing transfer function of ΔVw
by the rotor speed of the kth synchronous generator, Δωk, i.e., ΔVw ¼ γk(s)Δωk, Sik
is the sensitivity of the EOM of concern to the addition of damping torque on the kth
synchronous generator, Re{} denotes the real part of a complex number. In Eq. (5.7),
Tk is the coefficient of the amount of damping torque provided by the PMSG to the
kth synchronous generator, TkΔωk. Derivation of (5.7) is similar to that of (4.16) for
the grid-connected DFIG from (4.6) to (4.16) presented in Chap. 4.
The derived (5.7) indicates that under a special condition, ΔPw þ jΔQw can be
significant such that PMSG may introduce significant dynamic interactions with
power system, as to be elaborated as follows.
Denote λvsc as an open-loop converter oscillation mode of the PMSG. It is a
complex pole of transfer function Hpp(s) or/and Hpq(s) such that |Hpp(λvsc)| ¼ 1
or/and |Hpq(λvsc)| ¼ 1. Thus, if λvsc is close to the EOM of concerns, λi, on the
complex plane, i.e., λvsc λi, |Hpp(λi)| or/and |Hpq(λi)| can be very big to result in
significant ΔPw þ jΔQw at the complex frequency λi as it can be seen from Fig. 5.1
and (5.2). Significant dynamic interactions between the PMSG and power system
may occur under this special condition.
According to (5.7), under the special
condition that λvsc λi, jΔλi j ¼ ^λ i λi is no longer small due to the significant
increase of amount of the damping torque contributions from the PMSG, TkΔωk.
Hence, the EOM of concerns may be affected and participated by the PMSG
considerably.
5.1.1.2 The Open-Loop Modal Resonance (OLMR)
For the convenience of discussion, the special condition that the open-loop converter
oscillation mode of the PMSG, λvsc, is close to the open-loop EOM of concern, λi, on
the complex plane, i.e. λvsc λi, is referred to as the open-loop modal resonance
(OLMR). More generally, for the closed-loop interconnected system shown by
Fig. 5.1, closeness of two open-loop oscillation modes separately from two sub-
systems on the complex plane is named as the condition of the OLMR.
Normally, when the dynamic interactions between the PMSG and power system
are weak, closed-loop EOM, ^λ i , is close to λi on the complex plane. However, under
the condition of the OLMR, ^λ i could move away from the position where λvsc λi
on the complex plane because jΔλi j ¼ ^λ i λi may not be small any more.
It might be very difficult to analytically identify the exact position of ^λ i on the
complex plane under the condition that λvsc λi to strictly determine the impact of
strong dynamic interactions caused by the OLMR. However, how ^λ i moves away in
the neighborhood of λvsc λi (^λ i ! λi) on the complex plane when λvsc λi can be
estimated as follows.
From Fig. 5.1, the characteristic equation of closed-loop interconnected system

can be obtained to be
Gvp ðλÞHpp ðλÞ þ Gvq ðλÞHpq ðλÞ ¼ 1 ð5:8Þ
Without loss of generality, assume that λi and λvsc is the pole of Gvp(s) and
Hpp(s) respectively. Express Gvp(s) and Hpp(s) as
gp ð s Þ hp ð s Þ
Gvp ðsÞ ¼ , Hpp ðsÞ ¼ ð5:9Þ
ð s λi Þ ðs λvsc Þ
^λ i should satisfy (5.8). Thus from (5.8) and (5.9), it can have

^λ i λi ^λ i λvsc
ð5:10Þ
¼ gp ^λ i hp ^λ i þ ^λ i λi ^λ i λvsc Gvq ^λ i Hpq ^λ i
Under the condition of the OLMR, i.e., λvsc λi, from (5.10) it can have
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
lim Δλi ¼ lim ^λ i λi gp ðλi Þhp ðλi Þ ð5:11Þ
^λ i !λi ^λ i !i
Denote ^λ vsc as the closed-loop converter oscillation mode corresponding to λvsc.

^λ vsc should also satisfy (5.8). Taking the derivation similar to that from (5.8) to
(5.11), it can have

lim Δλvsc ¼ lim ^λ vsc λvsc
^λ vsc !λvsc ^λ vsc !λvsc
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð5:12Þ
gp ðλvsc Þhp ðλvsc Þ gp ðλi Þhp ðλi Þ
From (5.11) and (5.12), it can be seen that when λvsc λi, Δλi, and Δλvsc are
approximately of the same value but opposite signs. This means that on the complex
plane, the phase difference between vector Δλi and Δλvsc is about 180 . Closed-loop
oscillation modes, ^λ i and ^λ vsc , intend to move away from each other in the opposite
directions from λvsc λi on the complex plane. Hence, it is very likely that either ^λ i
or ^λ vsc will move towards the right on the complex plane from λvsc λi. The OLMR
may cause power system small-signal angular stability to decrease.
Mathematically, derivations of (5.11) and (5.12) only stand correct for ^λ i and ^λ vsc
in the neighborhood of λvsc λi. However, they indicate how ^λ i and ^λ vsc intend to
move away from λvsc λi on the complex plane when the OLMR occurs. In (5.10),
denote

f ^λ i ¼ gp ^λ i hp ^λ i þ ^λ i λi ^λ i λvsc Gvq ^λ i Hpq ^λ i ð5:13Þ

Taylor series expansion of f ^λ i at point λi is

f ^λ i ¼ f ðλi þ Δλi Þ ¼
0 00
ð5:14Þ
f ðλi Þ þ f ðλi ÞΔλi þ f ðλi ÞΔλi 2 þ
From (5.13), it can have f(λi) ¼ gp(λi)hp(λi). Hence, (5.11) and (5.12),
in fact, are
derived by only taking the first item in Taylor series expansion of f ^λ i . If ^λ i and
^λ vsc are far away from the neighborhood of λvsc λi, (5.11) and (5.12) may not stand
strictly, i.e., the phase difference between vector Δλi and Δλvsc may not exactly be
180 on the complex plane. Nevertheless, when the OLMR occurs, decreases of
damping of ^λ i or ^λ vsc is quite possible. This will be demonstrated by examples in
Sect. 5.1.2.
In addition, if |gp(λi)hp(λi)| is small, the OLMR may not result in considerable
movement of ^λ i and ^λ vsc on the complex plane as it can be seen from (5.11) and
(5.12).
5.1.1.3 Existence of Converter Oscillation Modes
To examine whether Hpp(s) or Hpq(s) is of any complex pole to be a converter

oscillation mode, a simplified version of transfer functions, Hpp(s) or Hpq(s), is
derived as follows.
The configuration of a PMSG being connected to an external power system is
shown by Fig. 5.2, where Xf is the reactance of the filter.
From Fig. 5.2, the linearized voltage equation across Xf is obtained to be
dΔId
Xf ¼ ω0 ðΔVcd ΔVd Þ þ ω0 ΔIq
dt
ð5:15Þ
dΔIq
Xf ¼ ω0 ΔVcq ΔVq þ ω0 ΔId
dt
Direction of terminal voltage of the PMSG, Vw ∠θw, is taken as that of d axis of
!
d q coordinate of the GSC (see Fig. 2.8). Thus Vq ¼0, Vd ¼ Vw . Variation of
power injected from the PMSG to the external power system is
ΔPw ¼ Id0 ΔVw þ Vd0 ΔId , ΔQw ¼ Iq0 ΔVw Vd0 ΔIq ð5:16Þ
Pin Pw
Vcd + jVcq Vw qw = Vd + jVq
External
A PMSG Cd Vdc power
Xf
system
I d + jI q Pw +jQw
GSC
Fig. 5.2 A PMSG connected to an external power system

1 2
K p_vdc K p_id Vd
Vdcref – + I dref+ + V ref
+ + cd
+ – –
Ki_vdc + K i_id +
Vdc Id X f Iq
s xvdc s xid
PWM
3 4
Vq
Qwref – I qref+ Vcqref
K i_q K i_iq + +
K p_q + K p_iq +
+ s – s +
Qw = -Vd I q Iq X f Id
1 active power control outer loop 2 d-axis current control inner loop
3 reactive power control outer loop 4 q-axis current control inner loop
Fig. 5.3 Configuration of the GSC control system of the PMSG
Linearized dynamic equation of DC voltage across the capacitor in Fig. 5.2 is (see
Sect. 5.2.1)
dΔVdc
Cd Vdc0 ¼ ΔPin ΔPw ð5:17Þ
dt
The configuration of the converter control system of the GSC of the PMSG is
shown by Fig. 2.3 and is represented as Fig. 5.3 below.
Ignore the fast transient of the pulse width modulation (PWM) such that (see
2.33))
Vcd ¼ Vcd
ref
, Vcq ¼ Vcq
ref
ð5:18Þ
From Fig. 5.3, (5.15) and (5.18), following equations can be obtained
ω0 Kp id s þ Ki id ω0
ΔId ¼ ΔI ref ¼ Gid ðsÞΔIdref
X f s2 þ ω0 Kp id s þ Ki id ω0 d
ð5:19Þ
Kp vdc s þ Ki vdc
ΔIdref ¼ ΔVdc
s
ω0 Kp iq s þ Ki iq ω0
ΔIq ¼ ΔI ref ¼ Giq ðsÞΔIqref
X f þ ω0 Kp iq s þ Ki iq ω0 q
s2
ð5:20Þ
Kp q s þ Ki q
ΔIqref ¼ Iq0 ΔVpcc þ Vd0 ΔIq
s
ΔPin in (5.17) is determined completely by the state variables of the permanent
magnet synchronous generator which are affected by the wind speed only. Thus,
ΔPin is irrelevant with power system electromechanical dynamics, neither taking
DV w + D Pw
I d0
+
K i_vdc -1
Vd0 Gid (s ) K p_vdc +
DI d DI ref
d s DVdc CdVdc0 s
Fig. 5.4 Active power control path of the GSC
Fig. 5.5 Reactive control DV w D Qw

path of the GSC –
I q0
–
Ki_q
Vd0 Giq ( s) ref K p_q +
DI q DI q
s
part in nor being affected by the dynamic interactions between the PMSG and power
system. Hence, in Eq. (5.17) ΔPin ¼ 0 and (5.17) becomes
dΔVdc
Cd Vdc0 ¼ ΔPw ð5:21Þ
dt
From the first equation of (5.16), (5.19) and (5.21), the active power control
implemented by the GSC can be shown by Fig. 5.4.
From Fig. 5.4
ΔPw ¼ Hpp ðsÞΔVw ð5:22Þ
Id0 Cd Vdc0 s2
where Hpp ðsÞ ¼ C .
d0 Gid ðsÞðKp vdc sþKi vdc Þ
2 þV
d Vdc0 s
Similarly, from the second equation of (5.16) and (5.20), the reactive power
control implemented by the GSC can be shown by Fig. 5.5.
From Fig. 5.5,
ΔQw ¼ Hpq ðsÞΔVw ð5:23Þ
Gqv ðsÞ¼Iq0
where Kiq
.
1þVd0 Giq ðsÞ Kp q þ s
A simplification can be made for the transfer function model of the PMSG by
ignoring the electromagnetic transient of the output current from the PMSG. This
simplification is appropriate for the study of electromechanical dynamics of power
system. With the simplification, the current control inner loops of the GSC control
system of the PMSG can be removed to have ΔId ¼ ΔIdref and ΔIq ¼ ΔIqref , that is,
Gid(s) ¼ 1 and Giq(s) ¼ 1 in (5.19) and (5.20). Thus, from (5.22) and (5.23), the
simplified transfer function model of the PMSG is obtained to be
Id0 Cd Vdc0 s2
Hpp ðsÞ ¼ ð5:24Þ
Cd Vdc0 s2 þ Vd0 Kp vdc s þ Vd0 Ki vdc
Iq0 s
Hpq ðsÞ ¼ ð5:25Þ
1 þ Vd0 Kp q s þ Vd0 Ki q
From (5.24), it can be seen that Hpp(s) may have a pair of complex poles to be a
converter oscillation mode. Setting of PI parameters of the active power control outer
loop, Kp_vdc and Ki_vdc, affects the position of the converter oscillation mode on the
complex plane. Hence, if strong dynamic interactions occur with considerable
variations of active power output from PMSG, ΔPw, due to the OLMR, retuning
of PI parameters of the active power control outer loop of the GSC can change the
position of the converter oscillation mode on the complex plane to dismiss the strong
dynamic interactions.
Since Vw ∠ θw is on the d axis of d q coordinate of the GSC, Pw0 ¼ Vd0Id0.
With the increase of the PMSG active power output, Id0 will increase. From (5.2) and
(5.24), it can be seen that |Hpp(λi)| and thus ΔPw will increase accordingly with the
increase of active power output from the PMSG to lead to increased dynamic
interactions between the PMSG and power system.
From (5.25), it can be seen that Hpq(s) is first-order and hence will not provide a
converter oscillation mode. In fact, since Qw0 ¼ Vd0Iq0 and normally the PMSG
operates with high power factor, Qw0 0 such that Iq0 0. From (5.25), it can be
seen that Hpq(s) 0. Hence, ΔQw should usually be very small.
However, if type 4 model of reactive power control [1] rather than that shown in
Fig. 5.3 is used, Hpq(s) may provide a converter oscillation mode. An equivalent
reactive power control path of type 4 model can be derived as shown by Fig. 5.6.
A similar simplified transfer function to (5.25) can be derived from Fig. 5.6 for type
4 model to be
Iq0 s2 þ Vd0 Kvi s

Hpq ðsÞ ¼ ð5:26Þ
s2 þ Kqi Kvi Vd0
Obviously in this case, Hpq(s) is of a pair of complex poles. Hence when type
4 mode is used, setting of PI parameters of reactive power control outer loop of the
PMSG may cause the OLMR to lead to strong dynamic interactions between the
PMSG and power system. In this case, strong dynamic interactions occur with
considerable variations of reactive power output from the PMSG, ΔQw. Retuning
of PI parameters of reactive power control outer loop of the PMSG can terminate the
strong dynamic interactions.
DVw D Qw
I q0 –
–
D xq
D xv K vi + K qi
Vd0 Giq (s )
DIq D I qref s – s
Fig. 5.6 Reactive control path of the GSC with the generic type 4 model
5.1.2 Example 5.1
5.1.2.1 Example Power System
Figure 5.7 shows the configuration of example New England power system with the
PMSGs. Parameters of network and synchronous generators given in [2] were used.
Detailed seventh-order model of synchronous generator and a third-order model of
the AVR were adopted [2, 3]. Loads were modelled as constant impedance. Two
PMSGs for the wind power generation were already connected at node 6 and 25.
Three more PMSGs (PMSG1, PMSG2 and PMSG3) were to be connected at node
22 and 10. There were total nine EOMs in the example New England power system.
The inter-area EOM of concern was of the lowest oscillation frequency and the
related inter-area power oscillation was between the tenth synchronous generator
and the rest of synchronous generators. Impact of dynamic interactions introduced
by the grid connections of PMSG1, PMSG2 and PMSG3 on power system EOMs,
especially the inter-area EOM, was examined. Parameters of the example power
system are given in Appendix 5.1.
8
1 37 PM
SG
30
26 28 29
25
2
27 24
18 38
1 17 9
16
10 6
3 35
39 15
14 21
4 PM 22
SG 19
5 PM 23
6 12
SG1 PM
9
20 SG2
11 13 33 36
31
4 7
7 2 34
10
5
8
PM 32
SG3
3
Fig. 5.7 Configuration of example New England power system with the PMSGs
700
600
40
Imaginary axis
20
Kp_vdc increased
0
-20
-40
Detailed model
-600
(5.30) used
-700
-800 -700 -600 -500 -30 -25 -20 -15 -10 -5 0
Real axis
Fig. 5.8 Trajectories of converter oscillation modes of PMSG1 with variation of parameters of the
GSC active power control outer loop
5.1.2.2 Converter Oscillation Modes
All the PMSGs were modelled by the detailed dynamic model presented in Chap. 2.
Typical parameters of the PMSG given in [4] were used. Configuration of the GSC
control system shown by Fig. 5.3 was adopted by PMSG1. PMSG2 used the type
4 model of reactive power control of the GSC and typical parameters of type 4 model
given in [5] were used.
State-space model described by (5.1) for both PMSG1 and PMSG2 was
established. Converter oscillation modes were calculated from the open-loop state
matrix Ap in (5.1). When the outer loop PI parameters of the GSC control systems of
PMSG1, Kp_vdc, Ki_vdc, Kp_q, and Ki_q shown in Fig. 5.3, varied with Ki_vdc ¼
cKp_vdc, Ki_vdc ¼ cKp_vdc, Ki_q ¼ cKp_q and c ¼ 40, computational results of the
converter oscillation modes are displayed in Figs. 5.8 and 5.9 as the solid curves.
Dotted areas in Figs. 5.8 and 5.9 are where power system EOMs usually locate with
the oscillation frequency between 0.1 and 2.5 Hz.
In Figs. 5.8 and 5.9, trajectories of the converter oscillation modes of PMSG1,
being estimated by using (5.24) and (5.25), are also presented as the dashed curves.
Computational results presented in Figs. 5.8 and 5.9 confirm the following analysis
and conclusions made in the previous Sect. 5.1.1.3 about the existence of converter
oscillation modes.
700
650
600
Imaginary axis
0.3 Kp_q increased
-0.3
-600
Detailed model
-650 (5.31) used
-700
-1000 -900 -700 -500 -20 -15 -10 -5 0
Real axis
GSC reactive power control outer loop
1. Simplification made in the analysis is appropriate since the estimated converter

oscillation modes (dashed curves) were very close to those calculated from the
detailed state-space model of PMSG1 (solid curves).
2. A pair of converter oscillation modes were from Hpp(s). Hence, they were related
with GSC active power control outer loop. The converter oscillation modes can
get close to power system EOMs with the variations of PI parameters of active
power control outer loop (Fig. 5.8), which may cause the OLMR.
3. Hpq(s) did not have any converter oscillation modes (Fig. 5.9). Hence, the
variations of PI parameters of reactive power control outer loop will not cause
the OLMR when the control configuration of reactive power control shown by
Fig. 5.3 is used.
When parameters in type 4 model of reactive power control outer loop of the GSC
of PMSG2 varied with Kiv ¼ cKiq and c ¼ 13.5, trajectories of converter oscillation
modes of PMSG2 were calculated from (5.26) and are displayed in Fig. 5.10 (the
dashed line on the imaginary axis). Trajectories of complex poles of PMSG2 by
using the detailed state-space model of PMSG2 are shown as the solid curve. It can
be clearly seen that a pair of converter oscillation modes entered the area where
power system EOMs locate. Hence, it is confirmed that when type 4 model is used,
the GSC reactive power control system of the PMSG may cause the OLMR. This
was reported in [1] and is explained analytically in Sect. 5.1.1.3.
30
Kiv increased
20
10
Imaginary axis
-10
-20 Detailed model

(5.32) used
-30
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
Real axis
GSC reactive power control outer loop (type 4 model used)
5.1.2.3 Dynamic Interactions of PMSG1 with Power System
In this sub-section, impact of grid connection of PMSG1 was examined. Table 5.1
presents the results of modal computation when the power output from PMSG1 was
Pvsc0 ¼ 0 p. u. , Qvsc0 ¼ 0 p. u. The inter-area EOM of concern was
λi ¼ 0.2841 þ 3.3356j. In case A, PI parameters of active power control outer
loop of the GSC given in [4] were used (Kp_vdc ¼ 0.7147, Ki_vdc ¼ 7.1515). In
case B, settings of PI parameters were changed toKp_vdc ¼ 0.22, Ki_vdc ¼ 3.
From Table 5.1, it can be seen that in case A when λvsc was away from
λi ¼ 0.2841 þ 3.3356j on the complex plane, dynamic interactions between
PMSG1 and power system were small as measured by jΔλi j ¼ ^λ i λi . Impact of
PMSG1 on the inter-area EOM of concerns was small because ^λ i only changed
slightly from λi with the dynamics of PMSG1 being included. However, in case B
when λvsc was close to λi ¼ 0.2841 þ 3.3356j on the complex plane, the inter-area
EOM of concern was affected significantly and its damping decreased considerably.
In this case, dynamic interactions introduced by PMSG1 imposed a serious threat to
power system small-signal angular stability.
Participation factors of state variables for the EOMs of concerns are the indication
of dynamic interactions between the PMSG and power system, which were used in
[1] extensively to examine the involvement of the PMSG in power system EOMs.
Last column in Table 5.1 gives the computational results of sum of participation
factors of all state variables of PMSG1, PFPMSG, for the inter-area EOM of concern.
It clearly shows that when λvsc was close to λi (case B), PFPMSG increased
Table 5.1 Impact of dynamic interactions introduced by PMSG1
Case λvsc b
λi |Δλi| PFPMSG(%)
A 4.2882 þ 8.2115j 0.29245 þ j3.2749j 0.061272 6.142
B 0.37006 þ j3.5007j 0.18242 þ j2.481j 0.86063 74.648
a b
0.4 0.35
0.35 0.3
0.3
0.25
0.25
0.2
0.2
0.15
0.15
0.1
0.1
0.05 0.05
0 0
DVdc Dxvdc Dxid 1 2 3 4 5 6 7 8 9 10
State variables Generator number
Fig. 5.11 Computational results of participation factors in case B. (a) Participation of PMSG1 in
inter-area oscillation mode; (b) participation of SGs in the converter oscillation mode of PMSG1
significantly, indicating considerable participation of PMSG1 in the inter-area EOM

of concerns.
A further split of total participation of PMSG1, PFPMSG, to individual state
variables can help identifying the cause of participation [1]. Figure 5.11a presents
the participation of three state variables of PMSG1 which contributed most to
PFPMSG in case B. From Fig. 5.11a, it can be seen that indeed those state variables
were associated with the active power control system of the GSC shown by Fig. 5.3
where Δxvdc and Δxid are indicated. Figure 5.11b is the participation of synchronous
generators in the converter oscillation mode in case B. It shows considerable
participations of synchronous generators in the converter oscillation mode.
Figures 5.12 and 5.13 present further confirmation from the results of non-linear
simulation. In the simulation, a 5% drop of PMSG1 output active power took place
which was recovered in 0.1 s. From Fig. 5.12, it can be seen that in case B, active
power exchange between PMSG1 and power system was significant. This confirms
the occurrence of strong dynamic interactions and analysis made in Sect. 5.1.1.3
about strong dynamic interactions caused by the OLMR when the converter oscil-
lation mode is the complex pole of Hpp(s). Consequently, damping of inter-area
power oscillation decreased as shown by Fig. 5.13.
In order to more clearly demonstrate how the OLMR caused the strong dynamic
interactions between PMSG1 and power system, setting of PI parameters of active
power control outer loop of PMSG1 in case A were changed gradually to that in case
B. Trajectories of open-loop and closed-loop COMs and inter-area EOM with
changes of parameters setting were calculated and are displayed in Fig. 5.14.
From Fig. 5.14, it can be seen that when the open-loop converter oscillation
mode, λvsc, got closer to the open-loop EOM of concern,λi, on the complex plane, the
11
Case A
10.8 Case B
10.6
Active power output of PMSG1
10.4
10.2
10
9.8
9.6
9.4
9.2
9
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.12 Exchange of active power between PMSG1 and power system
0
Case A
Case B
-2
-4
d10–d1
-6
-8
-10
-12
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.13 Relative rotor positions between the tenth and first SG
closed-loop EOM, ^λ i , moved away from λi towards the right and the damping of
inter-area EOM decreased. In Fig. 5.14, estimated positions of ^λ i and closed-loop
converter oscillation mode, ^λ vsc , by using (5.11) and (5.12) are indicated by crosses,
showing their opposite positions in respect to λi λvsc on the complex plane. Hence,
results presented in Fig. 5.14 confirm the analytical conclusion made in Sect. 5.1.1.3
that the OLMR may cause the damping of the EOM of concern to decrease.
Case A
4.5
λi – g p ( λi )hp ( λi )
4 λˆvsc
Imaginary axis
λvsc
Case B
3.5
λi
3
2.5 λî
λi + g p ( λi )h p ( λi )
2
-0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1
Real axis
Fig. 5.14 The OLMR between PMSG1 and power system
Table 5.2 Impact of dynamic interactions at different levels of active power output from PMSG1
Pw0 λi b
1 0.27396 þ 3.2786j 0.28222 þ 3.2194j 0.059773 13.87
5 0.28184 þ 3.3858j 0.29196 þ 3.0798j 0.30617 47.686
10 0.37006 þ j3.5007j 0.18242 þ 2.481j 0.86065 74.648
5.1.2.4 Dynamic Interactions Affected by the Level of Wind Power

Penetration
Analysis in Sect. 5.1.1.3 has concluded that when the active power output from the
PMSG increases, dynamic interactions introduced by the PMSG are expected to
become stronger. This conclusion is confirmed by the computational results
presented in Table 5.2. Confirmation from the computational results of participation
factors is given in the last column of Table 5.2, Figs. 5.15 and 5.16. Figures 5.17 and
5.18 are the confirmation from the results of non-linear simulation.
From Table 5.2 and Figs. 5.15, 5.16, 5.17 and 5.18, it can be seen that when the
active power output from PMSG1 increased, dynamic interactions introduced by
PMSG1 were stronger and the impact on the inter-area EOM of concern was larger.
Consequently, damping of the inter-area oscillation decreased.
0.35
Pw0=1 p.u.
0.3 Pw0=10 p.u.
0.25
0.2
0.15
0.1
0.05
0
1 2 3 4 5 6 7 8 9 10
Generator number
Fig. 5.15 Participation of SGs in converter oscillation mode of PMSG1
0.4
Pw0=1 p.u.
0.35
Pw0=10 p.u.
0.3
0.25
0.2
0.15
0.1
0.05
0
DVdc Dxvdc Dxid
State variables
Fig. 5.16 Participation of PMSG1 in the inter-area EOM
5.1.2.5 The OLMR of PMSG1 with Local EOMs
It was demonstrated in Sect. 5.1.2.3 that the OLMR of PMSG1 with the inter-area
EOM of concern occurred when setting of PI parameters of active power control
outer loop of PMSG1 changed. In practice, parameters tuning is often carried out for
a controller in order to obtain satisfactory control performance. Figure 5.19 shows
the results of a test that when PI parameters of active power control outer loop of
1
0.8 Pw0=1 p.u.
Active power output of PMSG1

Pw0=10 p.u.
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.17 Variation of active power exchange between PMSG1 and power system at different
levels of wind power penetration
0
Pw0=1 p.u.
-2 Pw0=10 p.u.
-4
d10 – d1
-6
-8
-10
-12
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.18 Variation of relative rotor positions between the tenth and first SG at different levels of
wind power penetration
PMSG1 were tuned, the OLMR of PMSG1 with five local EOMs of example New
England power system occurred.
In the test, PI parameters of active power control outer loop of PMSG1 were
tuned gradually from Kp_vdc ¼ 0.255, Ki_vdc ¼ 16.9 to Kp_vdc ¼ 0.224, Ki_vdc ¼ 7.5
such that the open-loop converter oscillation mode of PMSG1, λvsc, moved accord-
ingly on the complex plane. Trajectory of movement of λvsc is displayed as the dash-
9
l1
8.5
l vsc
8
l2
7.5
Imaginary axis
l3
7 lˆ vsc
lˆ i
6.5
l4
6 l5
5.5
5
-0.6 -0.58 -0.56 -0.54 -0.52 -0.5 -0.48 -0.46 -0.44 -0.42 -0.4 -0.38
Real axis
Fig. 5.19 The OLMR of PMSG1 with five local EOMs
dotted curve in Fig. 5.19. Trajectory of movement of closed-loop converter oscilla-

tion mode, ^λ vsc , along with λvsc is the dashed curve in Fig. 5.19. On the way of
movement, λvsc was close to five open-loop local EOMs of the example New
England power system, λi, i ¼ 1, 2, 3, 4, 5, which are indicated by triangles in
Fig. 5.19. Trajectories of closed-loop local EOMs are shown by the solid curves in
Fig. 5.19.
From Fig. 5.19, it can be seen that the OLMR of converter oscillation mode with
local EOMs occurred when λvsc was close to λi, i ¼ 1, 2, 3, 4, 5. Points of the OLMR
are indicated by squares and triangles in Fig. 5.19. When the OLMR occurred,
closed-loop local EOMs, ^λ i , i ¼ 1, 2, 3, 4, 5, were “driven” away from λi, i ¼ 1,
2, 3, 4, 5 to the points indicated by the filled circles. Closed-loop converter
oscillation mode, ^λ vsc , moved away from λvsc to the points indicated by the hollow
circles with corresponding numbers.
Figures 5.19 shows that four closed-loop local EOMs, ^λ i , i ¼ 1, 2, 3, 4, 5,
located on the right hand side of corresponding open-loop local EOMs, λi, i ¼ 1,
2, 3, 5. Hence, grid connection of PMSG1 caused the decrease of damping of those
four local EOMs when the OLMR happened. The fourth closed-loop local EOM, ^λ 4 ,
moved to the left hand side of λ4, indicating improvement of the damping of the
fourth local EOM when the OLMR occurred. However, the corresponding closed-
loop converter oscillation mode, ^λ vsc , was on the right hand side of λvsc, indicating
the decrease of damping as caused by the OLMR.
Table 5.3 Estimation of the OLMR of PMSG1 with five local EOMs
i λi gp(λi)hp(λi) |gp(λi)hp(λi)|
1 0.55012 þ 8.4517j 0.017556 þ 0.0088645j 0.019667
2 0.47535 þ 7.8201j 0.063834 þ 0.02335j 0.067971
3 0.52929 þ 7.1658j 0.022357–0.11377j 0.11595
4 0.4902 þ 5.8855j 0.092222–0.22458j 0.24278
5 0.51978 þ 6.3095j 0.0057484 þ 0.42818j 0.42822
Results presented in Fig. 5.19 demonstrate that when the parameters of active
power controller of the GSC of PMSG1 were tuned to cause the OLMR, power
system small-signal angular stability decreased. In real power systems with the
PMSGs, detrimental effect of the OLMR may not be noticed if such detailed analysis
of the OLMR was not carried out.
Figure 5.19 demonstrates that when the OLMR occurred, either ^λ i , i ¼ 1, 2, 3, 4
, 5 or ^λ vsc was on the right hand side of point of the OLMR where λvsc λi. This
demonstrates the correctness of analytical conclusion made from (5.11) and (5.12) in
Sect. 5.1.1.3. A further confirmation
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi is presented in Table 5.3 where computational
results of gp ðλi Þhp ðλi Þ at five points of the OLMR are given. From Table 5.3,
estimated positions of ^λ i , i ¼ 1, 2, 3, 4, 5 and ^λ vsc by using (5.11) and (5.12) were
calculated and are indicated by crosses in Fig. 5.19. From Fig. 5.19, it can be seen that
the estimation from (5.11) and (5.12) was close to the actual positions of ^λ i , i ¼ 1, 2
, 3, 4, 5 and ^λ vsc indicated by hollow and filled circles. In addition, by comparing the
third column of Table 5.3 and Fig. 5.19, it can be seen that when |gp(λi)hp(λi)| was
small for the first two local EOMs, the impact of the OLMR was small.
5.1.2.6 Dynamic Interactions of PMSG2 with Power System
In [1], type 4 model for the GSC reactive power control outer loop of the PMSG was
used instead of that shown in Fig. 5.3. Noticeable dynamic interactions were found
being related to the state variables in the GSC reactive power control outer loop of
the PMSG. Analysis in Sect. 5.1.1.3 has indicated that when type 4 model is used
instead of that shown by Fig. 5.3, Hpq(s) could have a pair of converter oscillation
modes to cause the OLMR. Reactive power control outer loop of PMSG2 in the
example New England power system adopted the type 4 model. Modal computation
presented in Fig. 5.10 has confirmed the analytical conclusion that there existed a
pair of converter oscillation modes due to the type 4 model used by PMSG2. In this
subsection, the impact of PMSG2 on the inter-area EOM of concern was examined
as follow.
Firstly, PMSG1 was connected at node 22 and parameters of active power control
outer loop of PMSG1 were carefully tuned to avoid causing any OLMR. The
Table 5.4 Impact of dynamic interactions introduced by PMSG3
λvsc b
Case C 0.032042 þ 0.31104i 0.29245 þ 3.2749j 0.034188 6.142
Case D 0.29338 þ 3.3135i 0.13382 þ 3.5714j 0.34609 47.194
a b
0.3 0.25
0.25
0.2
0.2
0.15
0.15
0.1
0.1
0.05
0.05
0 0
DVdc Dxvdc Dxv Dxq 1 2 3 4 5 6 7 8 9 10
Fig. 5.20 Computational results of participation factors in Case D. (a) Participation of PMSG2 in
inter-area oscillation mode; (b) participation of synchronous generators in converter oscillation
mode of PMSG2
open-loop inter-area EOM of concern without dynamics of PMSG2 being included

was calculated to be λi ¼ 0.32559 þ 3.2833j.
Secondly, dynamics of PMSG2 were included and typical parameters of type
4 model given in [5] were used (Kiv ¼ 5, Kiq ¼ 0.02). Results of modal computation
are presented in Table 5.4 as case C. From the second row of Table 5.4, it can be seen
that in case C, open-loop converter oscillation mode of PMSG2, λvsc, was away from
open-loop EOM of concern, λi. Impact of dynamic interactions introduced by
PMSG2 was small because jΔλi j ¼ ^λ λi was small, indicating weak dynamic
interactions brought about by PMSG2.
Thirdly, parameters of type 4 model were changed to be Kiv ¼ 11.5, Kiq ¼ 1.
Results of modal computation are presented in the third row of Table 5.4 as case D. It
can be seen that in case D, open-loop converter oscillation mode of PMSG2, λvsc,
was close to open-loop inter-area EOM of concern, λi. Damping of inter-area EOM
of concern decreased significantly due to the inclusion of PMSG2 dynamics, indi-
cating strong dynamic interactions between PMSG2 and power system. This was the
effect of the OLMR between PMSG2 and power system.
Fourthly, participation factors of sum of state variables of PMSG2, PFPMSG, were
calculated and are presented in the last column of Table 5.4. Figure 5.20 shows the
participation of synchronous generators in the converter oscillation mode and
participation of state variables related to the reactive power control outer loop
(type 4 model) of PMSG2 in the inter-area EOM of concern. They confirm the
occurrence of strong dynamic interactions between PMSG2 and power system as
caused by the OLMR.
0.2
Case C
0.15 Case D
Reactive power output of PMSG2
0.1
0.05
-0.05
-0.1
-0.15
-0.2
0 2 4 6 8 10 12 14 16 18 20
Time(s)
Fig. 5.21 Variation of reactive power output from PMSG2
-1
Case C
-2 Case D
-3
-4
d10 – d1
-5
-6
-7
-8
-9
-10
0 2 4 6 8 10 12 14 16 18 20
Time(s)
Fig. 5.22 Relative rotor positions between the tenth and first SG
Fifthly, non-linear simulation was conducted to provide further confirmation.

Figures 5.21 and 5.22 present the results of simulation. From Figs. 5.21 and 5.22,
it can be seen that significant dynamic interactions were due to the considerable
variations of reactive power exchange between PMSG2 and power system. This
caused decrease of damping of inter-area electromechanical oscillation.
Finally, in order to demonstrate the OLMR introduced by PMSG2 more clearly,
setting of PI parameters of type 4 reactive power controller of PMSG2 in case C were
changed gradually to that in case D. Trajectories of open-loop and closed-loop
4
λi – g q ( λi )hq ( λi ) λî
3.5
3
λi Case D
λi + g q ( λi )hq ( λi )
2.5
Imaginary axis
1.5 λvsc
λˆvsc
1
0.5
Case C
0
-0.65 -0.6 -0.55 -0.5 -0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0
Real axis
converter oscillation mode and the EOM involved in the OLMR with the change of
parameters setting were calculated and are displayed in Fig. 5.23. From Fig. 5.23, it
can be seen that with the change of PI parameters, open-loop converter oscillation
mode, λvsc, got close to the open-loop EOM, λi, on the complex plane. Closed-loop
inter-area EOM, ^λ i , moved away from λi towards the right on the complex plane and
the damping of inter-area power oscillation decreased. In Fig. 5.23, estimated
positions of closed-loop converter oscillation mode and the inter-area EOM, ^λ vsc
and ^λ i , by using the derived equations similar to (5.11) and (5.12) are indicated by
crosses. They are very close to the actual positions of ^λ vsc and ^λ i . This confirms the
correctness of analysis and conclusions made in Sect. 5.1.1.3 about the OLMR. In
Fig. 5.23, gq(s) ¼ (s λi)Gvq(s), hq(s) ¼ (s λvsc) Hpq(s).
5.1.2.7 Impact of Grid Connection of PMSG3
PMSG3 adopted the configuration of the GSC control system shown by Fig. 5.3 and
used typical parameters given in [4]. Firstly, a converter oscillation mode of PMSG3
was calculated from the open-loop state matrix Ap in (5.1) to be
λvsc ¼ 4.2882 þ 8.2115j. With PMSG1 and PMSG2 being connected at node
22, the EOMs of example New England power system were calculated from the
open-loop state matrix of power system Ag in (5.3). A local EOM was identified to
8.8 λi – g p ( λi )hp ( λi )
8.6 λi
8.4 λˆ vsc
Imaginary axis
8.2
The OLMR
8
λvsc
7.8
λˆ i
7.6
λi + g p ( λi )h p ( λi )
7.4
7.2
-4.6 -4.4 -4.2 -4 -3.8 -3.6 -3.4
Real axis
be close to λvsc ¼ 4.2882 þ 8.2115j on the complex plane. The local EOM was
λi ¼ 4.1952 þ 8.3573j.
Secondly, from the open-loop transfer functions of PMSG3 and power system,
approximate positions of closed-loop converter oscillation mode and local EOM
were estimated. The estimated positions are indicated by two crosses in Fig. 5.24.
Thirdly, PMSG3 was connected at node 10 and closed-loop state matrix Agp in (5.5)
was established. Closed-loop converter oscillation mode, ^λ vsc , and closed-loop
EOM, ^λ i , were calculated from Agp. They are displayed in Fig. 5.24.
From Fig. 5.24, it can be seen that the OLMR happened between the open-loop
converter oscillation mode and the local EOM such that the damping of closed-loop
local EOM decreased. The OLMR due to the grid connection of PMSG3 did not
cause lightly or negatively damped local power oscillations, because open-loop local
EOM was well damped before PMSG3 was connected. In a real power system, this
may often be the case that the damping of the EOMs decreased as caused by the
OLMR. However, grid connection of the PMSGs would not be identified as the
cause of change of the damping of the EOMs if analysis of the OLMR was not
carried out.
Figure 5.25 presents the computational results of participation factors of syn-
chronous generators and PMSG3 in the converter oscillation mode and the EOM.
They clearly indicate the participations of SG2 and SG3 in the converter oscillation
mode and the state variables associated with the active power control outer loop of
PMSG3 in local EOM, thus confirming strong dynamic interactions introduced by
PMSG3 as caused by the OLMR.
a b
0.45 1.2
0.4
1
0.35
0.3 0.8
0.25
0.6
0.2
0.15 0.4
0.1
0.2
0.05
0 0
DVdc Dxvdc Dxid 1 2 3 4 5 6 7 8 9 10
Fig. 5.25 Computational results of participation factors with PMSG3 being connected. (a) Partic-
ipation of PMSG3 in the local EOM; (b) participation of SGs in the converter oscillation mode of
PMSG3
5.2 Impact of Strong Dynamic Interactions Introduced by

a Grid-Connected DFIG
Grid connection of a DFIG for wind power generation changes the power system’s
load flow conditions and introduces dynamic interactions. Changes in the load flow
conditions and the dynamic interactions introduced by the DFIG can both affect the
power system’s small-signal angular stability. In Sect. 4.1 of Chap. 4, a method was
introduced to separately assess the impact of those two affecting factors, i.e., the
change in the load flow conditions and the dynamic interactions introduced by the
DFIG on the power system’s small-signal angular stability. Owing to the fast control
speed of the VSC, it has been recognized that the dynamic interactions between the
grid-connected DFIG and the power system are usually weak because of the
dynamic “decoupling” effect of the VSC control. Using damping torque analysis,
the reason for the low impact of the weak dynamic interactions introduced by the
DFIG on the power system’s small-signal angular stability was explained in Sect.
4.1. Based on this, it was suggested to model the DFIG as a constant power source
for pre-screening its impact on the power system’s stability.
However, the computational results of the participation factors for a real large-
scale power system with grid-connected PMSGs in [1] reported an unusual phenom-
enon, wherein, there was a considerable participation of the PMSGs in the system
EOMs. In the previous section, strong dynamic interactions introduced by the
PMSGs were examined and are attributed to the OLMR between the converter
oscillation modes of the PMSGs and power system EOMs. In this section, the strong
dynamic interactions introduced by the DFIGs are investigated, because in case of
strong dynamic interactions, the DFIG cannot be modeled as a constant power
source for pre-screening its impact on the power system small-signal stability.
Investigation on the strong dynamic interactions introduced by a grid-connected
DFIG presented in this section is based on the establishment of a closed-loop
5.2 Impact of Strong Dynamic Interactions Introduced by a Grid-Connected DFIG 173
interconnected dynamic model of the power system with the DFIG, which was
introduced in Sect. 2.4 of Chap. 2. In the established model, the power system and
DFIG are modeled as two open-loop interconnected subsystems. The impact of the
dynamic interactions introduced by the DFIG on the power system’s small-signal
angular stability is assessed as the difference between the closed-loop and open-loop
EOMs of the closed-loop interconnected model. Analysis is carried using the
damping torque analysis to show that under the condition of the OLMR when an
open-loop DFIG oscillation mode is close to an open-loop EOM on the complex
plane, strong dynamic interactions between the DFIG and power system may occur,
considerably affecting the power system’s small-signal angular stability. Further
analysis reveals that the existence of the DFIG oscillation mode, causing the OLMR
with the EOM, is related to the configuration and parameter setting of the RSC
control system of the DFIG.

by the Grid-Connected DFIG and the Impact
5.2.1.1 Multivariable Closed-Loop Interconnected Model of the Power
System with a Grid-Connected DFIG
Consider a multi-machine power system with a DFIG, as shown in Fig. 5.26.

Pw þ jQw is the complex power output from the DFIG and Vw ∠θw is the terminal
voltage of the DFIG. State-space model of the DFIG was derived in Sect. 2.2 as
(2.108), which is re-written below:
DFIG
A multi- Pw + jQw Ps + jQs
machine Rotor
Gear
power box
system
GSC RSC
GSC (grid side RSC (rotor side

converter) control converter) control
system system
Fig. 5.26 Configuration of a multi-machine power system with a DFIG

d
ΔXd ¼ Ad ΔXd þ bv ΔVw þ b f Δθw
dt
ΔPw ¼ cp T ΔXd þ dd1 ΔVw þ dd2 Δθw ð5:27Þ
ΔQw ¼ cq T ΔXd þ dd3 ΔVw þ dd4 Δθw
The state-space model of the multi-machine power system was derived in Sect.
2.3 as (2.123), which is re-written as (5.28) below.
d
ΔXg ¼ Ag ΔXg þ bP ΔPw þ bQ ΔQw
dt
ΔVw ¼ cv T ΔXg þ dg1 ΔPw þ dg2 ΔQw ð5:28Þ
Δθw ¼ c f T ΔXg þ dg3 ΔPw þ dg4 ΔQw
From (5.27) and (5.28), the linearized model of the power system with the DFIG,
shown in Fig. 5.1 is obtained to be,
d
ΔX ¼ Agd ΔX ð5:29Þ
dt
T
where ΔX ¼ ΔXg T ΔXd T .
The dynamic models of the power system and the DFIG can be expressed by
transfer function matrices. From (5.27) and (5.28), they are obtained to be
Power system :
" # " #" # " #
ΔVw g11 ðsÞ g12 ðsÞ ΔPw ΔPw
¼ ¼ GðsÞ
Δθw g21 ðsÞ g22 ðsÞ ΔQw ΔQw
ð5:30Þ
DFIG :
" # " #" # " #
ΔPw d11 ðsÞ d12 ðsÞ ΔVw ΔVw
¼ ¼ DðsÞ
ΔQw d21 ðsÞ d22 ðsÞ Δθw Δθw
where
1 1
g11 ðsÞ ¼ cv T sI Ag bP þ dg1 , g12 ðsÞ ¼ cv T sI Ag bQ þ dg2
1 1
g21 ðsÞ ¼ c f T sI Ag bP þ dg3 , g22 ðsÞ ¼ c f T sI Ag bQ þ dg4
:
d11 ðsÞ ¼ cp T ðsI Ad Þ1 bv þ dd1 , d12 ðsÞ ¼ cp T ðsI Ad Þ1 b f þ dd2
d21 ðsÞ ¼ cq T ðsI Ad Þ1 bv þ dd3 , d22 ðsÞ ¼ cq T ðsI Ad Þ1 b f þ dd4
The established model shown by (5.30) can be depicted by Fig. 5.27. It is a

multivariable closed-loop interconnected dynamic model, where the dynamics of the
power system and the DFIG are modeled separately as two open-loop interconnected
subsystems.
g11 ( s )
DPw DVw
+
g 21 ( s )
g12 ( s ) Dqw
DQw
+
g 22 ( s ) Power system
d12 ( s ) DFIG
+
d11 ( s )
d 22 ( s )
+
d 21 ( s )
Fig. 5.27 Multivariable closed-loop interconnected of the power system with the DFIG
5.2.1.2 Strong Dynamic Interactions Introduced by the DFIG
Denote ^λ i as the power system EOM of concern, of the closed-loop system shown in
Fig. 5.27. Thus, ^λ i is a complex eigenvalue of the closed-loop state matrix, Agd, in
(5.29). If ΔPw þ jΔQw ¼ 0, there are no dynamic interactions between the DFIG and
the power system. The DFIG is degraded into a constant power source, Pw0 þ jQw0.
In this case, the dynamic model of the power system shown by (5.28) becomes,
d
dt
Denote λi as the oscillation mode corresponding to ^λ i , when the DFIG is modeled
as a constant power source. Obviously, λi is an eigenvalue of the open-loop state
matrix, Ag, in (5.28) and (5.31). Hence, the difference between the closed-loop and
open-loop EOMs, Δλi ¼ ^λ i λi , measures the impact of the dynamic interactions of
the DFIG with the power system, ΔPw þ jΔQw, on the concerned EOMs.
Normally, the dynamic variations of ΔPw and ΔQw are small, indicating weak
dynamic interactions between the DFIG and power system. The impact of the
dynamic interactions
on the power system’s small-signal angular stability,
^
jΔλi j ¼ λ i λi , should generally not be significant. A derivation of the damping
torque analysis similar to that presented in Sect. 4.1.1.1 of Chap. 4 will give the
following estimation of Δλi as,
XN h i
Δλi ¼ Sik Re gpk ðλi Þd11 ðλi Þ þ gqk ðλi Þd21 ðλi Þ γkv ðλi Þ
k¼1 h i ð5:32Þ

þ gpk ðλi Þd12 ðλi Þ þ gqk ðλi Þd22 ðλi Þ γkf ðλi Þ
where gpk(s) and gqk(s) are the transfer functions of the forward paths from ΔPw and
ΔQw to the electromechanical oscillation loop of the kth synchronous generator,
respectively; γkv(s) and γkf(s) are the reconstructing transfer functions of ΔVw and
Δθw, respectively, by the rotor speed of the kth synchronous generator, Δωk, i.e.,
ΔVw ¼ γkv(s)Δωk and ΔVw ¼ γkf(s)Δωk; Sik is the sensitivity of the oscillation mode
to the addition of damping torque on the kth synchronous generator; Re{} denotes the
real part of a complex number.
Denote λd as a complex eigenvalue of the open-loop state matrix of the DFIG, Ad,
in (5.27), referred to as a DFIG oscillation mode, if its frequency is between 0.1 and
2.5 Hz. Thus, λd should be a pole of the transfer function matrix of the DFIG, D(s), in
(5.30). It should satisfy|d11(λd)| ¼ 1 , |d12(λd)| ¼ 1 , |d21(λd)| ¼ 1or/and |
d22(λd)| ¼ 1. If λd is close to the power system open-loop EOM of concern on the
complex plane, i.e., λd λi, then |d11(λi)|, |d12(λi)|, |d21(λi)|or/and |d22(λi)| should be
large. In this case, it can be seen from (5.30) that the dynamic variations of either
ΔPw or/and ΔQw can be significant at the complex oscillation frequency, λi, indi-
cating strong dynamic interactions between the DFIG and the power system. From
(5.32), it can be seen that in this case, the damping torque contribution from the
DFIG to the kth synchronous generator can increase significantly at the complex
oscillation frequency, λi. Hence, the EOM of concern may be affected considerably.
This special condition of the OLMR between the open-loop power system and the
DFIG, i.e., λd λi, may lead to possible strong dynamic interactions between the
DFIG and the power system. The impact of the OLMR introduced by the DFIG on
the small-signal stability of the power system with the DFIG is examined in the
following subsection.
5.2.1.3 The Open-Loop Modal Resonance (OLMR)
The characteristic equation of the multivariable closed-loop interconnected system

shown by Fig. 5.27 is
jI GðsÞDðsÞj ¼ 0 ð5:33Þ
As λd and λi are the poles of open-loop transfer function matrices, G(s) and D(s),
respectively, it can denote
GðsÞ1 ¼ ðs λi ÞG1 ðsÞ, DðsÞ1 ¼ ðs λd ÞD1 ðsÞ ð5:34Þ

Thus, characteristic equation of (5.33) becomes

jðs λi Þðs λd ÞG1 ðsÞD1 ðsÞ Ij ¼ 0 ð5:35Þ
Denote
" #
k11 ðsÞ k12 ðsÞ
KðsÞ ¼ G1 ðsÞD1 ðsÞ ¼ ð5:36Þ
k21 ðsÞ k22 ðsÞ
From (5.35) and (5.36),

ðs λd Þðs λi Þk11 ðsÞ 1 ðs λd Þðs λi Þk12 ðsÞ

ðs λd Þðs λi Þk21 ðsÞ ðs λd Þðs λi Þk22 ðsÞ 1 ð5:37Þ

¼ 0
When λd λi, replacing s in (5.37) by ^λ i ¼ λi þ Δλi

k11 ^λ i k22 ^λ i k12 ^λ i k21 ^λ i Δλ4i
ð5:38Þ
k11 ^λ i þ k22 ^λ i Δλ2i þ 1 0
Substitute the following first-order Taylor series expansions at λi into (5.38)

k11 ^λ i ¼ k11 ðλi Þ þ k011 ðλi ÞΔλi
ð5:39Þ
k22 ^λ i ¼ k22 ðλi Þ þ k022 ðλi ÞΔλi
By ignoring the items higher than Δλ2i , it can have
½k11 ðλi Þ þ k22 ðλi ÞΔλ2i þ 1 0 ð5:40Þ
Hence
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
Δλi ¼ k ð λi Þ ð5:41Þ
k11 ðλi Þ þ k22 ðλi Þ
Denote ^λ d as the closed-loop oscillation mode of the DFIG corresponding to λd.

^λ d is also a solution of characteristic equation of (5.37). Hence, taking the derivation
similar to that from (5.37) to (5.41), it can be obtained for ^λ d that
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
Δλd ¼ ^λ d λd ¼ k ð λd Þ ð5:42Þ
k11 ðλd Þ þ k22 ðλd Þ
From (5.41) and (5.42), it can have

^λ i λi kðλi Þ
ð5:43Þ
^λ d λd kðλd Þ λi kðλi Þ
The above equations indicate that when the OLMR occurs, i.e., λd λi, the
corresponding closed-loop oscillation modes locate at approximately opposite posi-
tions on the complex plane in respect to the positions of coupled open-loop oscil-
lation modes, λd λi. Hence, there is always a closed-loop oscillation mode being on
the right-hand side of λd λi. This implies that the OLMR degrades the small-signal
stability of the power system. When the EOM of concern, ^λ i , is on the right-hand
side of λd λi, the strong dynamic interactions caused by the OLMR degrade the
damping of the EOM of concern.
Since the impact of the OLMR is detrimental to power system small-signal
stability, it is essentially important to carefully examine the condition under which
the OLMR may possibly happen. The examination can be carried out by computing
the oscillation modes of the open-loop subsystems in Fig. 5.27. In order to obtain
some general guidelines in the examination instead of completely relying on the
modal computation, in the following subsection, existence and sources of the open-
loop DFIG oscillation modes which may cause the OLMR are investigated. The
investigation is conducted by deriving a simplified open-loop transfer function
matrix model of the DFIG.
5.2.1.4 Non-existence of the DFIG Oscillation Mode
The assumptions for deriving the simplified transfer function matrix model of the
DFIG are as follows:
1. The electromagnetic transient of the rotor windings of the DFIG is much faster
than the electromechanical dynamic transient of the power system. Hence, the
electromagnetic transient of the rotor windings can be ignored.
2. The active and reactive power outputs from the grid side converter (GSC) of the
DFIG are Pr ¼ swPs, Qr ¼ 0, respectively, [6–9], where sw is the slip of the rotor
motion of the DFIG. Hence, Pw ¼ Ps þ Pr ¼ (1 sw)Ps ¼ ωdPs, Qr ¼ 0, where ωd
is the rotor speed of the DFIG.
Configuration of the RSC control system of the DFIG is shown by Fig. 2.5, which
is redrawn as Fig. 5.28. As per assumption (1) above, the current control inner loops
shown in Fig. 5.28 can be removed to obtain [6–9].
Ird ¼ Irdref , Irq ¼ Irqref ð5:44Þ
Direction of Vw is that of the q axis of the d-q coordinate system of the DFIG.
Thus, the flux equations of the stator windings of the DFIG are (4.31) below,
Vwd ¼ 0 ¼ ψsq ¼ Xss Isq þ Xm Irq
ð5:45Þ
Vwq ¼ Vw ¼ ψsd ¼ Xss Isd Xm Ird
active power control q-axis current control X m2 X

– sw ( X rr – ) I rd + sw m Vw
outer loop inner loop X ss X ss
Psref I sqref I rqref

+ Kip X + - +
K pp + – ss Kiq ( s) Vrq
– s Xm –
Ps I rq
Qsref I sdref X ss I rdref -

+ Kiq + +
K pq + – Kid ( s) Vrd
– s Xm – – +
Qs I rd
Vw X2
reactive power d-axis current sw ( X rr – m ) I rq
X m control inner loop X ss
control outer loop
Fig. 5.28 Configuration of the RSC control system of the DFIG
With assumption (2) above and (5.45) above,

Ps ¼ Vwq Isq ¼ Vw Isq
Pw ¼ ωd Ps ð5:46Þ
Qw ¼ Qs ¼ Vwq Isd ¼ Vw Isd
where ωd is the rotational speed of the DFIG. The shaft dynamics of the DFIG
described by a two-mass model are (2.58). When a simple one-mass model is used,
the rotor motion equation of the DFIG is
dωd 1
¼ ðPm Ps Þ ð5:47Þ
dt J
where ωd ¼ 12 ðωd1 þ ωd2 Þ and J ¼ 12 ðJd1 þ Jd2 Þ is the constant of inertia of the rotor.
Variation of mechanical power input is mainly affected by the wind speed and can be
ignored to have ΔPm ¼ 0 . Thus, linearization of (5.47) is
dΔωd
J ¼ ΔPs ð5:48Þ
dt
Linearizing (5.46) and then substituting (5.48),
ΔPs ¼ Vw0 ΔIsq þ Isq0 ΔVw
ΔPw ¼ Ps0 Δωd þ ωd0 ΔPs

Ps0 Ps0
¼ ΔPs þ ωd0 ΔPs ¼ ωd0 ΔPs ð5:49Þ
Js Js
ΔQs ¼ Vw0 ΔIsd þ Isd0 ΔVw
ΔQw ¼ ΔQs
Linearization of (5.45) is
Xm Xm 1
ΔIsq ¼ ΔIrq , ΔIsd ¼ ΔIrd ΔVw ð5:50Þ
Xss Xss Xss
Using linearized (5.44) and (5.50), from Fig. 5.28

Xss ref Xss Kip
ΔIrq ¼ ΔIrqref ¼ ΔIsq ¼ Kpp þ ΔPs
Xm Xm s
Xss ref 1
ΔIrd ¼ ΔIrdref ¼ ΔIsd ΔVw ð5:51Þ
Xm Xm

Xss Kiq 1
¼ Kpq þ ΔQs ΔVw
Xm s Xm
From (5.49), (5.50) and (5.51),

Kip
ΔPs ¼ Vw0 Kpp þ ΔPs þ Isq0 ΔVw
s

ð5:52Þ
Kiq
ΔQs ¼ Vw0 Kpq þ ΔQs Isd0 ΔVw
s
From (5.49) and (5.52), the elements in the transfer function matrix model of the
DFIG are obtained to be

Ps0
ωd0 Isq0 s
ΔPw Js
d11 ðsÞ ¼ ¼ , d12 ðsÞ ¼ 0
ΔVw 1 þ Vw0 Kpp s þ Kip ð5:53Þ
ΔQs ΔQw Isd0 s
d21 ðsÞ ¼ ¼ ¼ , d22 ðsÞ ¼ 0
ΔVw ΔVw 1 þ Vw0 Kpq s þ Kiq
The transfer function matrix model of the DFIG is

2
3
Ps0
ωd0 Isq0 s
6 Js 7
6 07
6 7
DðsÞ ¼ 6 1 þ Vw0 Kpp s þ Kip 7 ð5:54Þ
6 7
4 Isd0 s 5
0
1 þ Vw0 Kpq s þ Kiq
From (5.53) or (5.54), it can be seen that the DFIG does not have any oscillation
mode, when the configuration of the RSC control system shown in Fig. 5.28 is used.
Hence, it can be concluded that when the DFIG adopts the configuration of the RSC
control system shown in Fig. 5.28, The OLMR should normally not occur and the
dynamic interactions between the DFIG and power system should be weak. This
conclusion is based on the simplified transfer function matrix model of the DFIG
with assumptions (1) and (2) presented in the beginning of this subsection. It is to be
noted that the simplification ignores the faster transient of the rotor windings of the
DFIG and the slower dynamics of the wind turbine compared to the electromechan-
ical transient of the power system. In addition, the dynamic function of the GSC
control system is represented by Pr ¼ swPs, Qr ¼ 0 [6–9]. Thus, the order of the
dynamic model of the DFIG is considerably reduced.
5.2.1.5 Existence of the DFIG Oscillation Mode
The active power control outer loop in Fig. 5.28 can adopt the rotor speed of the
DFIG as the input signal, instead of the active power [9]. In this case, the first
equation in (5.51) becomes

Xss ref Xss Kip
ΔIrq ¼ ΔIrq ¼ ΔIsq ¼
ref
Kpp Δωd ð5:55Þ
Xm Xm s
From the first equation in (5.49), (5.50) and (5.55),

Kip 1
ΔPs ¼ Vw0 Kpp þ ΔPs þ Isq0 ΔVw ð5:56Þ
s Js
Thus,
JIsq0 s2
ΔPs ¼ ΔVw ð5:57Þ
Js2 þ Vw0 Kpp s þ Vw0 Kip
d11(s) can be obtained from (5.49) and (5.57) to be

ΔPw ðωd0 Js Ps0 ÞsIsq0
d11 ðsÞ ¼ ¼ ð5:58Þ
ΔVw Js2 þ Vw0 Kpp s þ Vw0 Kip
From (5.58), it can be seen that the DFIG now may have a pair of complex poles,
i.e., oscillation modes. The existence of the DFIG oscillation modes may possibly
cause the OLMR with the EOM of concern. In this case, the strong dynamic
interactions between the DFIG and the power system may be accompanied by
considerable variations of the active power output of the DFIG, ΔPw, during the
electromechanical dynamic transient of the power system.
The open-loop DFIG oscillation modes can be estimated from (5.58) to be
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vw0 Kpp j Vw0 2 Kpp 2 4JKip
λd ð5:59Þ
2J
If the open-loop power system’s EOM of concern is known to be λi , a quick
method to determine the range of the PI parameters of the active power control outer
loop for identifying and avoiding modal coupling in this case, is to solve the
following equation:
Qsref K qi X rscQ I sdref X ss I rdref - Vrd

K qv + +
– Kid ( s)
+ s + s X Xm – – +
– – rscV
Qs Vw I rd
Vw X2
reactive power d-axis current sw ( X rr – m ) I rq
X m control inner loop X ss
control outer loop
Fig. 5.29 Configuration of the type 3 model for the RSC reactive power control system of
the DFIG
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vw0 Kpp þ j Vw0 2 Kpp 2 4JKip
λi ¼ ð5:60Þ
2J

If the solution of the above equation is K ^ pp ; K
^ ip , setting the PI parameters of

the active power control outer loop outside the range, K^ pp α; K^ pp þ α and

^ ip β; K
K ^ ip þ β , (α and β are two appropriate positive numbers), may effectively
avoid and dismiss the OLMR.
Figure 5.29 shows the configuration of the generic type 3 model which the RSC
can use. From Fig. 5.29, the second equation in (5.51) becomes
Xss ref 1
ΔIrd ¼ ΔIrdref ¼ ΔIsd ΔVw
Xm X

m ð5:61Þ
Xss Kqv Kqi 1
¼ ΔQs ΔVw ΔVw
Xm s s Xm
Using (5.61), and the third equation in (5.49) and (5.50), respectively,
ΔQw Vw0 Kqv s Isd0 s2

d21 ðsÞ ¼ ¼ 2 ð5:62Þ
ΔVw s þ Vw0 Kqv Kqi
From (5.62), it can be seen that when the type 3 reactive power control model is
used, the DFIG oscillation modes can be owing to the reactive power control outer
loop in the RSC control system also. The parameter settings of the type 3 model may
possibly lead to the OLMR. In this case, the OLMR may result in significant
dynamic variations in the reactive power output of the DFIG, ΔQw.
5.2.2 Example 5.2
5.2.2.1 Evaluation of the Conclusions Regarding the Existence

of the DFIG Oscillation Mode
The example New England power system shown by Fig. 5.7 is used in this section
for demonstrating and evaluating the analysis and conclusions made in the previous
700
500
300
100
Imaginary axis
0.2
0
-0.2
-100
-300 Detailed model
-500 (5.58) used
-700
-610 -480 -330 -180 -30 -15 -10 -5 0
Real axis
Fig. 5.30 Trajectories of the open-loop eigenvalues of the DFIG, when Kpp and Kipare varied
section. A wind farm represented by a DFIG was connected to bus 22. A detailed
dynamic model of the DFIG introduced in Chap. 2 is used.
First, the RSC control system depicted in Fig. 5.28 was adopted by the DFIG
connected at node 22. The state-space model of the DFIG as shown in (5.27) was
established and the eigenvalues of the open-loop state matrix of the DFIG, Ad, were
calculated. The PI parameters, Kpp and Kip, of the active power control outer loop
were varied with Kip ¼ 20 Kpp. The trajectories of the complex eigenvalues of Ad
are displayed the solid curves in Fig. 5.30. The PI parameters, Kpq and Kiq, of the
reactive power control outer loop were varied with Kiq ¼ 20Kpq. The trajectories of
the complex eigenvalues of Ad are displayed and highlighted in blue in Fig. 5.31.
The shadow areas in Figs. 5.30 and 5.31 indicate the location of the power system
EOMs with the oscillation frequency between 0.1 and 2.5 Hz. The entries of the
trajectories of the eigenvalues of Ad into the shadow area are also displayed in
Figs. 5.30 and 5.31. From these figures, it can be seen that there were no DFIG
oscillation modes.
In Figs. 5.30 and 5.31, the trajectories of the poles of d11(s) and d21(s) as
calculated from (5.52), are also displayed as the dashed curves which overlaid the
solid curves completely. They confirm the correctness of the assumptions made for
deriving the elements,d11(s) and d21(s), in the simplified transfer function matrix in
Sect. 5.2.1.4. Therefore, these figures confirm the analytical conclusion made in
Sect. 5.2.1.4 that a DFIG oscillation mode does not occur, when the configuration of
the RSC control system shown in Fig. 5.28 is used by the DFIG. In this case, there
was no OLMR to cause the strong dynamic interactions between the DFIG and the
power system.
700
500
300
100
Imaginary axis
0.2
0
-0.2
-100
-300
Detailed model
-500 (5.58) used
-700
-610 -480 -330 -180 -30 -15 -10 -5 0
Real axis
Fig. 5.31 Trajectories of the open-loop eigenvalues of the DFIG, when Kpq and Kiq are varied
Next, the configuration of the RSC active power control system in Fig. 5.28
adopted by the DFIG was changed by replacing the input signal of the active power
by the rotor speed of the DFIG [9]. The PI parameters,Kpp and Kip, of the active
power control outer loop of the RSC were varied with Kip ¼ 20Kpp. The trajectories
of the complex eigenvalues of Ad are displayed as the solid curves in Fig. 5.32. In
addition, the trajectories of the poles of d11(s) were calculated from (5.58), and
displayed as the dashed curves in Fig. 5.32, which almost overlaid the solid curves; it
can be seen that when the PI parameters of the RSC control system are varied, a pair
of complex open-loop eigenvalues enter the shadow area, wherein, the EOMs are
located. This confirms the analytical conclusion made in Sect. 5.2.1.5 that when the
rotor speed input signal is used for the RSC active power control, a DFIG oscillation
mode could exist, causing the OLMR with the power system EOMs.
Finally, the configuration of the RSC reactive power control system shown in
Fig. 5.29 was used by the DFIG. The PI parameters of the RSC reactive power
control system were changed. The complex open-loop eigenvalues and poles of
d21(s) were calculated from Ad and (5.62), respectively. When Kqi is varied from
0.1–6 with Kqv ¼ 13.5Kqi, the trajectories of the open-loop eigenvalues and the poles
of d21(s) are displayed as the solid and dashed curve, respectively, in Fig. 5.33. From
Fig. 5.33, it can be seen that a pair of DFIG oscillation modes existed owing to the
adoption of the RSC reactive power control system shown in Fig. 5.29 that could
possibly cause the OLMR.
700
400
100
Imaginary axis
20
-20
-100
Detailed model
-400
(5.68) used
-700
-610 -480 -330 -180 -30 -20 -15 -10 -5 0
Real axis
Fig. 5.32 Trajectories of the open-loop eigenvalues of the DFIG, when the rotor speed input signal
is used with Kpp and Kip varied
700
400
100
30
20
Imaginary axis
10
0
-10
-20
-30
-100 Detailed model
-400 (5.68) used
-700
-610 -480 -330 -180 -30 -1 -0.8 -0.6 -0.4 -0.2 0
Real axis
Fig. 5.33 Trajectories of the open-loop eigenvalues of the DFIG, when the RSC reactive power
control system in Fig. 5.4 is used with Kqi and Kqv varied
5.2.2.2 The OLMR Owing to the RSC Active Power Control System
of the DFIG
In this subsection, tests are carried out to demonstrate that the DFIG oscillation mode
from the RSC active power control system of the DFIG connected at node 22 causes
the OLMR with an EOM of the example New England power system. The example
New England power system has a total of nine EOMs; the one with the lowest
oscillation frequency at around 0.5 Hz is the EOM of concern. It is related to the
inter-area oscillation of the tenth synchronous generator against the other synchro-
nous generators. The DFIG connected at node 22 was operated with a fixed unity
power factor and its active power output was 10 p.u. The configuration of the RSC
active power control system shown in Fig. 5.28 was used. The following three cases
were examined.
Case A: The configuration of the RSC active power control system shown in
Fig. 5.28 was used. Typical PI control parameters recommended in SIMULINK [10]
were used with Kpp ¼ 1, Kip ¼ 100.
Case B: The rotor speed input signal was used by the DFIG, replacing the active
power input signal in Fig. 5.28. The PI parameters were retained at Kpp ¼ 1,
Kip ¼ 100.
Case C: Same as Case B but the values of PI gains were increased to Kpp ¼ 50,
Kip ¼ 1000.
Table 5.5 presents the results of the modal computations for the above three cases.
In Table 4.5, λd is the open-loop DFIG oscillation mode associated with the RSC
active power control system of the DFIG and was calculated from the open-loop
state matrix, Ad; λi is the open-loop inter-area EOM of concern and was calculated
from the open-loop state matrix, Ag, in (5.28); ^λ i is the closed-loop inter-area EOM
of concern corresponding to λi and was calculated from the closed-loop state matrix,
Agd , in (5.29); jΔλi j ¼ ^λ i λi measures the impact of the dynamic interactions
between the DFIG and the power system.
From Table 5.5, it can be seen that in case A, the open-loop eigenvalue of the
DFIG related to the RSC active power control is a real number. Hence, no DFIG
oscillation mode existed. The effect of the DFIG’s dynamics on the inter-area EOM
of concern was negligible. In case B, λd was close to λi and the OLMR between the
DFIG and the power system occurred. The inter-area EOM of concern was affected
significantly with a considerable decrease in the damping. In case C, λd was away
from λi because of the increase in the PI control parameters. The OLMR disappeared
Table 5.5 Computational results of the DFIG’s dynamic impact owing to the RSC active power
control system
Case λd λi b
λi |Δλi|
A 16.667 0.3144 þ j3.6009 0.31677 þ j3.6008 0.002372
B 0.3125 þ j3.5217 0.16193 þ j3.222 0.40843
C 3.125 þ j10.735 0.31191 þ j3.5947 0.006681
4.6
Case C
4.4
4.2 lˆd
Imaginary axis
li
3.8 ld
Case B
3.6
lî
3.4
3.2
-0.5 -0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15
Real axis
Fig. 5.34 The OLMR in case B
and the effect on the inter-area EOM of concern introduced by the DFIG’s dynamics
was negligible.
In order to clearly show the effect of the OLMR, as discussed in Sect. 5.2.1.4, the
PI gains of the RSC active power controller of the DFIG were decreased such that the
DFIG oscillation mode, λd, moved on the complex plane from the position in case C,
towards λi. When the PI gains were decreased to the typical values recommended in
[10], λd arrived at the position in case B such that λd λi. The trajectory of the
movement of λd is displayed as the solid curve in Fig. 5.34. The corresponding
movement of the closed-loop EOM, ^λ i , and the closed-loop DFIG oscillation mode,
^λ d , (see Sect. 5.2.1.4) are also displayed as the dashed curves in Fig. 5.34,
respectively.
From Fig. 5.34, it can be seen that when λd was away from λi , ^λ i was close to λi.
Thus, the dynamic interactions between the DFIG and the system were weak.
However, when λd moved towards λi , ^λ d , moved along with λd such that ^λ i was
“driven” away from λi towards the right on the complex plane for “avoiding” ^λ d .
Thus, in case B, the damping of ^λ i decreased considerably, when the OLMR
occurred.
The participation factors of the DFIG’s state variables for the EOMs are indica-
tions of the dynamic interactions between the DFIG and the power system. Table 5.6
presents the computational results of the sum of the participation factors, PFdfig, of
all the state variables of the DFIG connected at bus 22 for the inter-area EOM of
concern. From Table 5.6, it can be clearly seen that in case B, when the OLMR
Table 5.6 Participation factors of the sum of all the state variables of the DFIG for the inter-area
EOM of concern
Case A B C
PFdfig(%) 0.40718 62.807 0.73363
0.25
Case C
Case B
0.2
0.15
0.1
0.05
0
1 2 3 4 5 6 7 8 9 10
Generator number
Fig. 5.35 Participation of the synchronous generators in the DFIG oscillation mode
occurred, the DFIG’s state variables show considerably higher participations in the
EOM of concern.
The participation factors of the sum of the rotor position and speed of the
synchronous generators in the DFIG oscillation mode were also computed for
cases B and C. The computational results are presented in Fig. 5.35; it can be seen
that when the OLMR occurred in case B, the synchronous generators participated
considerably in the DFIG oscillation mode. Table 5.6, Fig. 5.35 indicate the signif-
icant dynamic interactions between the DFIG and the power system, when the
OLMR occurred.
Finally, to identify the source of the DFIG oscillation mode that caused the
OLMR, the participation factors of the main state variables of the DFIG in cases B
and C, respectively, are displayed in Fig. 5.36; it can be seen that when the OLMR
occurred, the state variables with the maximum participation in the EOM of concern
were Δωd and Δxd. These two state variables are related to the rotor motion equation
of (5.48) and the PI controller of the RSC active power control outer loop. Hence,
this confirms the analytical conclusion in Sect. 5.2.1.4 that the OLMR was due to the
RSC active power control of the DFIG.
Figures 5.37 and 5.38 present the results of non-linear simulation for cases B and
C. In the simulation, a 20% drop in the wind power input to the DFIG occurred at 1 s
and lasted for 100 ms. From Fig. 5.37, it can be seen that there were considerable
0.35
Case C
Case B
0.3
0.25
0.2
0.15
0.1
0.05
0
Δωd Δxd
State variables
Fig. 5.36 Participation of the main state variables of the DFIG in the EOM of concern
11
Case C
10.8 Case B
10.6
Active power outout of DFIG
10.4
10.2
10
9.8
9.6
9.4
9.2
9
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.37 Active power output, Pw, from the DFIG. (a) Relative angular position between the tenth
and first synchronous generator. (b) Relative angular position between the seventh and first
synchronous generator.
a
2
Case C
1
Case B
0
-1
-2
-3
d 10 – d 1
-4
-5
-6
-7
-8
-9
-10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
b
62
Case C
61 Case B
60
59
d7 – d1
58
57
56
55
54
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.38 Relative angular positions. (a) Relative angular position between the tenth and first
synchronous generator; (b) relative angular position between the seventh and first synchronous
generator
variations in the active power exchange between the DFIG and power system in
case B, when there were strong dynamic interactions caused by the OLMR. Conse-
quently, the damping of the inter-area low-frequency power oscillation decreased
significantly, as shown in Fig. 5.38.
Table 5.7 Impact of the DFIG’s dynamics on the EOM of concern at different levels of wind
power penetration
Pw0 λi b
λi |Δλi| PFdfig(%)
1 0.30555 þ j3.43 0.29491 þ j3.3418 0.088839 29.124
5 0.30628 þ j3.5321 0.24155 þ j3.2682 0.27172 52.855
10 0.3125 þ j3.5217 0.16193 þ j3.222 0.40843 62.807
5.2.2.3 Impact of the DFIG’s Dynamics at Different Levels of Wind

Power Penetration
At steady state, the active power output from the DFIG is,
Pw0 ¼ ð1 sw0 ÞPs0 ¼ ωd0 Ps0 ¼ ωd0 Vw0 Isq0 ð5:63Þ
Hence, at a higher level of the wind power penetration, Pw0 and Isq0 are bigger.
From (5.53) and (5.58), it can be seen that with an increased Isq0, |d11(λi)| increases
accordingly.
From (5.32), it can be observed that this will usually lead to the increase
of Δλi . Therefore, with the increase of the level of the wind power penetration, the
impact of the DFIG’s dynamics on the power system’s EOMs is expected to
increase.
To demonstrate and confirm the conclusion made above regarding the impact of
the DFIG’s dynamics at different levels of the wind power penetration, Table 5.7
presents the test results of the modal computations for the example New England
power system with the DFIG at node 22, when the active power output, Pw0, was
varied. The RSC control system of the DFIG was as same as that in case B in
presented in Table 5.5. From Table 5.7, it can be seen that with the increase of the
wind power penetration, the inter-area EOM of concern was affected more by the
DFIG’s dynamics with an increased |Δλi|. The DFIG participated more in the EOM
of concern with an increased participation factor, PFdfig.
The computational confirmation regarding the participation factors of the syn-
chronous generators in the DFIG oscillation mode is presented in Fig. 5.39, indicat-
ing the increased involvement of the synchronous generators in the DFIG oscillation
mode, when the wind power penetration increased. Figure 5.40 depicts the partici-
pation factors of the state variables of the DFIG which had the highest participation
in the EOM of concern, demonstrating a higher participation of the RSC active
power control in the EOM of concern, when the wind penetration was higher.
Further confirmation from the non-linear simulation is given in Figs. 5.41 and
5.42, showing the increased impact of the DFIG’s dynamics, when the wind
power penetration increased.
0.25
Pw0=1 p.u.
Pw0=10 p.u.
0.2
0.15
0.1
0.05
0
1 2 3 4 5 6 7 8 9 10
Generator number
Fig. 5.39 Participation of the synchronous generators in the DFIG oscillation mode at different
0.35
Pw0=1 p.u.
Pw0=10 p.u.
0.3
0.25
0.2
0.15
0.1
0.05
0
Δωd Δxd
State variables
Fig. 5.40 Participation of the main state variables of the DFIG in the EOM of concern at different
0.6
0.5 Pw0=1 p.u.
Pw0=10 p.u.
Active power variation of DFIG 0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
-0.7
-0.8
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.41 Active power output, ΔPw, from the DFIG at different levels of wind power penetration
6
Pw0=1 p.u.
5
Pw0=10 p.u.
4
3
2
1
d10 – d1
0
-1
-2
-3
-4
-5
-6
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 5.42 Relative angular position between the tenth and first synchronous generator at different
5.2.2.4 The OLMR Owing to the RSC Reactive Power Control System
of the DFIG
The analysis in Sect. 5.2.1.5 has concluded that when the type 3 model of the RSC
reactive power control is used, the DFIG oscillation mode may originate from the
Table 5.8 Computational results of the DFIG’s dynamic impact owing to the RSC reactive power
control system
Case λd b
λi |Δλi| PFdfig (%)
D 0.23355 þ j0.85 0.31704 þ j3.5999 0.002823 0.43015
E 0.2823 þ j3.7341 0.11453 þ j3.8062 0.28652 70.491
a b
0.4 0.35
0.3
0.3
0.25
0.2
0.2
0.15
0.1
0.1
0.05
0 0
ΔxQ ΔxV 1 2 3 4 5 6 7 8 9 10
Fig. 5.43 Participation of the DFIG and SGs in case E. (a) Participation of the DFIG’s main state
variables in the EOM; (b) participation of SGs in the DFIG oscillation mode
complex pole of d21(s) (see (5.62)), leading to the OLMR with the power system
EOMs. This possibility has been confirmed by Fig. 5.33. In this subsection, the
DFIG connected at bus 22 adopted the configuration of the RSC reactive power
control system shown in Fig. 5.29 (type 3 model). The following two tests were
carried out.
Case D: Typical parameters of the type 3 model recommended in SIMULINK
[11] were used with Kqi ¼ 0.05, Kqv ¼ 20.
Case E: The values of the PI parameters of the type 3 model of the RSC reactive
power control were changed to Kqi ¼ 1, Kqv ¼ 13.5.
Table 5.8 presents the results of the modal computation for the above two cases.
In case D, λd was away from λi ¼ 0.3144 þ j3.6009. The dynamic interactions
between the DFIG and power system were weak, imposing negligible impact on the
inter-area EOM of concern. Hence ^λ i was close to λi and the damping of the inter-
area EOM did not change considerably, when the dynamics of the DFIG were
included in power system. However, in case E, λd was close to λi ¼ 0.3144
þ j3.6009. The OLMR occurred and the damping of ^λ i decreased considerably,
indicating the occurrence of poorly damped inter-area oscillations in the example
New England power system, caused by the grid connection of the DFIG at node 22.
The participation factors of the sum of all the state variables of the DFIG, PFdfig,
were calculated and the computational results are given in the last column of
Table 5.8. Figure 5.43 presents the computational results of the participation factors
of the main state variables of the DFIG for the EOM of concern and those of the
synchronous generators for the DFIG oscillation mode in case E. The computational
results reveal the significant participation of the synchronous generators and the
4
Case E
li
3.5 lfi
3
Imaginary axis
2.5
1.5 lfd
ld
1
Case D
0.5
-0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1
Real axis
Fig. 5.44 The OLMR in case E
DFIG in the DFIG oscillation mode and EOM of concern, respectively, confirming
the considerable dynamic interactions between the DFIG and the synchronous
generators, caused by the OLMR.
For a clearer demonstration of the modal coupling, Fig. 5.44 presents the move-
ment of the open-loop and closed-loop EOMs of concern and the DFIG oscillation
mode on the complex plane, with the change of the parameters of the RSC reactive
power controller of the DFIG from the typical values in case D to those in case E,
when the OLMR occurred. From Fig. 5.44, it can be seen that when the open-loop
DFIG oscillation mode, λd, (the solid curve) moves towards λi (hollow circle), the
movement of ^λ d along with λd “drove” the closed-loop EOM, ^λ i , away from λi
towards the right on the complex plane. Subsequently, the damping of the inter-area
EOM of concern decreased owing to the OLMR. The pattern of the OLMR shown in
Fig. 5.44 is as same as that shown in Fig. 5.34, validating the analysis made
previously regarding the OLMR between the DFIG and the power system.
Further validation from the non-linear simulation is presented in Figs. 5.45 and
5.46. In the simulation, an 80% load was lost at node 8 at 1 s of the simulation and
the load loss was recovered in 100 ms. From Fig. 5.45, it can be seen that in case E,
the dynamic variations of the reactive power output from the DFIG were obvious.
This confirms that the OLMR was due to the RSC reactive power control, leading to
the considerable increase of ΔQw. In this case, the damping of the inter-area EOM of
concern was very small to be about 0.03, causing the reactive power output from the
DFIG to oscillate longer.
0.2
Case D
0.15
Case E
Reactive output of DFIG 0.1
0.05
-0.05
-0.1
-0.15
-0.2
0 1 2 3 4 5 6 7 8 9 10
Time(s)
Fig. 5.45 Variation of the reactive power output, ΔQw, from the DFIG
-4.2
Case D
-4.3 Case E
-4.4
d10 – d1
-4.5
-4.6
-4.7
-4.8
0 1 2 3 4 5 6 7 8 9 10
Time(s)
Fig. 5.46 Relative angular position between the tenth and first synchronous generator
5.2.2.5 OLMR Between Multiple DFIGs
In this subsection, a test is carried out to demonstrate that in the OLMR, the open-
loop oscillation mode of the power system can be a DFIG oscillation mode of
another DFIG in the power system. Thus, the OLMR occurs between two DFIGs.
First, the test assumes that the wind farm connected at node 22 in the example
New England power system was represented by two DFIGs. This is possible in
practice, when the DFIGs in a wind farm distribute in two areas. The DFIGs in each
area are aggregated and represented by a DFIG. Those two DFIGs connected at node
4.2
lfd2
3.8
Imaginary axis
3.6 ld1
3.4
3.2 ld2
3
lfd1
2.8
-1.25 -1 -0.75 -0.5 -0.25 0
Real axis
Fig. 5.47 The OLMR between DFIG1 and DFIG2
22 were named as DFIG1 and DFIG2, respectively, for the convenience of discus-
sion. The active power output from each of the DFIGs was 5 p.u. The configuration
of the RSC active power control system shown in Fig. 5.28 with the rotor speed input
signal being adopted by both the DFIGs. Typical PI control parameters
recommended in SIMULINK [10] were used. The dynamic interactions of DFIG2
with the rest of the power system were examined. In this test, the open-loop state-
space model of the power system described by (5.28) included the dynamics of
DFIG1. The open-loop state-space model of the DFIG depicted by (5.27) included
only the state variables of DFIG2.
From the open-loop state matrix, Ag, in (5.28), a DFIG oscillation mode of
DFIG1 was calculated and identified. This was the open-loop DFIG oscillation
mode of the power system and indicated as λd1 in Fig. 5.47 by the hollow circle.
From the open-loop state matrix, Ad, in (5.27), the open-loop DFIG oscillation mode
of DFIG2 was calculated and indicated as λd2 in Fig. 5.47 by the hollow diamond.
The corresponding closed-loop DFIG oscillation modes were calculated to be ^λ d1
and ^λ d2 , respectively. They are indicated by solid circle and diamond, respectively,
in Fig. 5.47. Since the parameters of the RSC active power control systems of DFIG1
and DFIG2 were same, λd1 and λd2 were close to each other on the complex plane.
Subsequently, the OLMR occurred between two DFIGs, causing both ^λ d1 and ^λ d2 to
move away from λd1 λd2 on the complex plane, as shown in Fig. 5.47.
The computational results of the participation factors are presented in Fig. 5.48,
indicating considerable dynamic interactions between DFIG1 and DFIG2 when the
OLMR occurred.
SG10 in the example New England power system is an aggregated synchronous
generator, representing an external large-scale power system [2]. With the
a b 0.4
0.7
DFIG1 DFIG1
0.35
0.6 DFIG2 DFIG2
0.3
0.5
0.25
0.4
0.2
0.3
0.15
0.2
0.1
0.1 0.05
0 0
Δωdfig ΔXω Δωd Δxd
State variables State variables
Fig. 5.48 Participation factors of DFIG1 and DFIG2. (a) Participation of the DFIGs in ^λ d1 ;
(b) Participation of the DFIGs in ^λ d2
4.5
lî
A lˆ d2
4
3.5
Imaginary axis
3
ld1
ld2
2.5
1.5
lˆd1
1
-2 -1.5 -1 -0.5 0 0.5
Real axis
Fig. 5.49 The OLMR between DFIG1, DFIG2 and the power system
synchronous generators in the external power system withdrawing or participating in

the operation, the constant of inertia of SG10 varies. Hence, in the second stage of
the test, the inertia of SG10 was varied such that the inter-area EOM of concern
moved on the complex plane. The movement of the EOM is displayed in Fig. 5.49 as
the solid curve with the filled circles. Subsequently, the closed-loop DFIG oscillation
modes, ^λ d1 and ^λ d2 , moved away from each other to “avoid” getting close to each
other on the complex plane, as shown in Fig. 5.49. When the power system operated
around point A, ^λ d1 moved into the right half of the complex plane. Thus, the power
system lost the small-signal stability.
The OLMR shown in Fig. 5.49 can also be explained as follows: The closeness of
the open-loop DFIG oscillation modes of DFIG1 and DFIG2 cause both ^λ d1 and ^λ d2
to move away from the position, where λd1 λd2 on the complex plane. In order to
2
After DFIG2 was connected
Before DFIG2 was connected
0
-2
-4
d10 – d1
-6
-8
-10
-12
0 2 4 6 8 10 12 14 16 18 20
Time(s)
Fig. 5.50 Non-linear simulation when the OLMR occurred between the DFIGs
avoid ^λ d2 moving closer to ^λ d1 , ^λ d2 , “drove” ^λ d1 further towards the right.

Eventually, the OLMR between DFIG1 and DFIG2 caused the loss of the system
small-signal stability. Confirmation from the simulation results is presented in
Fig. 5.50.
Data of New England Power System
They are given in Appendix 4.2.1.
Data of DFIG
Data of PMSG
Basic data
Xpd ¼ 0.25, Xpq ¼ 0.15, Xpf ¼ 0.05,
Cp ¼ 30, Jpr ¼ 8s, ψpm ¼ 1.1
Control parameters of MSC

Kpp1 ¼ 5, Kpi1 ¼ 20, Kpp2 ¼ 1, Kpi2 ¼ 100, Kpp3 ¼ 1, Kpi3 ¼ 100
Control parameters of GSC

Kpp4 ¼ 0.8, Kpi4 ¼ 20, Kpp5 ¼ 0.15, Kpi5 ¼ 84.9, Kpp6 ¼ 0.28, Kpi6 ¼ 12,
Kpp7 ¼ 0.15, Kpi7 ¼ 84.9
References
Syst 29(5):2248–2256
2. Rogers G (2000) Power system oscillations. MA Kluwer, Norwell
4. Kima HW, Kimb SS, Koa HS (2010) Modeling and control of PMSG-based variable-speed
wind turbine. Electr Power Syst Res:46–52
5. MATLAB Simulink Wind Farm—Synchronous Generator and Full Scale Converter (Type 4)
Detailed Model. Website: http://uk.mathworks.com/help/physmod/sps/examples/wind-farm-
synchronous-generator-and-full-scale-converter-type-4-detailed-odel.html?
requestedDomain¼www.mathworks.com
6. Ekanayake JB, Holdsworth L, Jenkins N (2003) Comparison of 5th order and 3rd order machine
models for doubly fed induction generator (DFIG) wind turbines. Elect Power Syst Res 67
(3):207–215
7. Feijóo A, Cidrás J, Carrillo C (2000) A third order model for the doubly-fed induction machine.
Elect Power Syst Res 56(2):121
8. Ko HS, Yoon GG, GG KNH, Hong WP (2008) Modeling and control of DFIG-based variable-
speed wind-turbine. Elect Power Syst Res 78(11):1841–1849
9. Pena R, Clare JC, Asher GM (1996) Doubly fed induction generator using back-to-back PWM
converters and its application to variable-speed wind-energy generation. IEEE Proc Electr
Power Appl 143(3):231–241
10. MATLAB Simulink Wind Farm (DFIG Phasor Model)
11. MATLAB Simulink Wind Farm-DFIG Detailed Model
Chapter 6
Small-Signal Stability of a Power System
with a VSWG Affected by the PLL
Grid connection of a VSWG to a power system is realized by the converter control,

which normally adopts the current vector control algorithm. This has been intro-
duced in Chap. 2. Implementation of current vector control needs to track the relative
positions between the d q coordinate of the converter and the common x y
coordinate of the power system in order to determine the d q coordinate of the
converter. As being illustrated by Fig. 2.8, the direction of the terminal voltage (PCC
voltage) of a PMSG in x y coordinate of the power system is normally taken as
that of d axis of d q coordinate of the GSC of the PMSG. Figure 2.10 shows that
the direction of the terminal voltage (PCC voltage) of a DFIG in x y coordinate of
the power system is often taken as that of d axis of d q coordinate of the RSC of the
DFIG. Hence, by tracking the phase of the terminal voltage of the VSWG, d q
coordinate of the converter can be determined for implementing the current vector
control. This task of phase tracking is normally fulfilled by a phase locked loop
(PLL).
So far, various types of PLLs have been proposed. The core of majority of
proposed PLLs have been a closed-loop control system for the phase tracking. In
the PLL design, parameter setting for the closed-loop control is critical and needs to
consider carefully the speed, accuracy, and robustness of the phase tracking, under
variable power system operating conditions. For example, when high-bandwidth
PLL parameters are set, a fast phase tracking performance can be obtained. How-
ever, a high-bandwidth may reduce the phase tracking ability of the PLL under
unbalanced operating conditions. In addition to considering the phase tracking
performance of the PLL itself, the impact of dynamic interactions brought about
by the PLL on the power system stability should also be taken into account in the
design.
The PLL functions to link the dynamics of the VSWG and the power system. The
impact of the dynamic interactions introduced by the PLL is in two folds: The first-
fold impact caused by the dynamic interactions introduced from the PLL is on the
stability of the VSWG. Investigation up-to-date has indicated that when a VSWG is

202 6 Small-Signal Stability of a Power System with a VSWG Affected by the PLL
weakly connected to an external power system, the PLL may cause the small-signal
instability in the VSWG system. In addition, when the bandwidth of the PLL is close
to that of DC voltage control or current control inner loops of the GSC of the PMSG,
dynamic interactions introduced by the PLL became strong which may affect the
small-signal stability of the PMSG negatively. In the investigation, the PLL param-
eters (mainly the PI gain) were varied in order to demonstrate and identify the
potential risk brought about by the PLL. The strategy of scanning the PLL parameter
space by varying the PI gains of the PLL is effective in determining the constraints
imposed by the stability requirement on the parameter setting of the PLL. In Sect.
1.2.1, this first-fold impact of the PLL has been reviewed.
The second-fold impact of the PLL is on the power system stability. When the
impact of a grid-connected VSWG on power system small-signal angular stability is
investigated, often it was assumed that the PLL has a perfect phase tracking
performance with zero phase-tracking error. Thus, the impact of the PLL was not
considered in the investigation. However, results of examination in [1, 2] indicated
that the PLL dynamics may affect the power system electromechanical oscillation
modes (EOMs) considerably. In [1], the PI gains of the PLL were varied in a wide
range for scanning the potential risk introduced by the PLL to the power system
stability. With the variation of PI gains of the PLL, it was found from the EOM
trajectories that, when the bandwidth of the PLL was reduced, the EOMs may be
affected. Results of modal analysis by varying the PI gains of the PLL in [2]
indicated that when a PLL oscillation mode (POM) was close to a power system
EOM, the PLL affected the EOM considerably. However, it was not known whether
the results of the case-by-case modal analysis in [1, 2] were special cases or not.
In this chapter, the small-signal stability of a power system with a wind farm
affected by the PLL is examined. The rest of the chapter is organized as follows. In
the next section, a closed-loop interconnected linearized model of the power system
with a PMSG is established, where the PLL for the PMSG and the rest of the power
system (ROPS) are modeled as two separate open-loop interconnected subsystems.
The established model clearly indicates that the effect of the PLL phase-tracking
error on the power system small-signal stability can be assessed as the difference
between the open-loop and closed-loop oscillation modes. Based on the established
model, analysis is carried out to indicate that, when an open-loop oscillation mode of
the PLL subsystem is close to an open-loop oscillation mode of the ROPS subsystem
on the complex plane, the open-loop modal resonance may lead to the strong
dynamic interactions between the PLL and the ROPS. It is very likely that open-
loop modal resonance may lead to a damping degradation of either the closed-loop
oscillation mode of the PLL or the ROPS such that the small-signal stability of the
power system with the VSWG decreases.
In Sect. 6.2, the small-signal stability of example PMSG systems affected by the
PLL is examined by using the theory of open-loop modal resonance presented in
Sect. 6.1 of the chapter. In this example PMSG system, dynamics of the AC power
system are not considered and hence the AC power system is modelled as an infinite
6.1 Impact of Open-Loop Modal Resonance Caused by the PLL. . . 203
Ppmsg Pvsc GSC

Vcd + jVcq V pcc q pcc
External
PMSG C Vdc power
Xf system
I d + jI q Pvsc +jQvsc
Rest of power system
PLL
q pll q pcc
PLL
Fig. 6.1 A PLL connected to an external power system
busbar. In Sect. 6.3, the small-signal angular stability of an example multi-machine

power system with PMSGs is examined to demonstrate and validate the analysis and
conclusions made about the open-loop modal resonance. Case studies show that
open-loop modal resonance occurs when the power system operating conditions
change or when the PLL parameters are varied. Subsequently, open-loop modal
resonance causes poorly damped LEPOs in the example multi-machine power
system.
6.1 Impact of Open-Loop Modal Resonance Caused by

the PLL for a Grid-Connected PMSG
6.1.1 Closed-Loop Interconnected Model of a Power System

with a PMSG
6.1.1.1 Function of a PLL
Figure 6.1 shows the configuration of a PMSG being connected to an external power
system. The external power system may be a multi-machine power system with
multiple grid-connected VSWGs. Function of the PLL is to track the phase of the
terminal voltage of the PMSG at the point of common coupling (PCC), θpcc. The
tracked phase, θpll, by the PLL is taken as the direction of the d-axis of the d q
coordinates of the grid side converter (GSC), as shown by Fig. 2.10, which is
redrawn as Fig. 6.2. An ideal PLL has a perfect phase-tracking performance with
zero phase-tracking error, θerror ¼ θpcc θpll ¼ 0. In this case, the direction of the
PMSG terminal voltage, Vpcc ∠ θpcc, coincides with that of daxis such that Vq 0.
Strictly speaking, θerror ¼ θpcc θpll 6¼ 0, although a well-designed PLL may
have a negligible phase-tracking error, i.e.,θerror ¼ θpcc θpll 0. When the
negligible phase-tracking error is ignored, it can be assumed that the PLL is ideal
Fig. 6.2 Phase tracking x

of the PMSG’s terminal
voltage by the PLL to q
determine the relative
position between
d qcoordinate of the GSC V pcc θ pcc
of the PMSG and common d
x y coordinate of the
power system θerror
θ pll
y
θ pcc
with zero phase-tracking error. Consider the PLL as a closed-loop control system
with the pair of input-output signals to be, θpcc θpll. During the transient of the
power system, the output, θpll, may not exactly follow the input, θpcc, at any instant
of time such thatθpcc 6¼ θpll.
Since the direction of d-axis direction of the d q coordinate of the GSC is
determined by θpll, as depicted in Fig. 6.2,
Vd ¼ Vpcc cosθerror , Vq ¼ Vpcc sinθerror ð6:1Þ
At steady state, θerror0 ¼ θpcc0 θpll0 ¼ 0; the linearization of (6.1) is,
ΔVd ¼ ΔVpcc , ΔVq ¼ Vpcc0 Δθerror ¼ Vd0 Δθerror ð6:2Þ
From (6.2), it can be seen that when the phase-tracking error of the PLL is
considered such that θpcc 6¼ θpll, ΔVq 6¼ 0.
6.1.1.2 Closed-Loop Model
Denote hpll(s) as the transfer function of the PLL, i.e., Δθpll ¼ hpll(s)Δθpcc. Hence,
the phase-tracking error of the PLL is,

θerror ¼ θpcc θpll ¼ 1 hpll ðsÞ Δθpcc ¼ Hpll ðsÞΔθpcc ð6:3Þ
Let the state-space realization of the transfer function, Hpll(s), be,

d
ΔXpll ¼ Apll ΔXpll þ bθ Δθpcc
dt ð6:4Þ
Δθerror ¼ cθ T ΔXpll þ dθ Δθpcc
where ΔXpll is the vector of all the PLL state variables.

Fig. 6.3 Closed-loop Dθerror Dθpcc

model of the power system Gg (s)
with the PMSG Rest of the
power system
PLL
Hpll (s)
There are various schemes for building and designing a control system to fulfil a
function for θpll to track θpcc. These schemes are often referred to as different PLLs.
However, as long as the core of majority of the PLLs is a closed-loop control system
with the pair of input-output signals and transfer function to be θpcc θpll and
hpll(s) respectively, the function of the PLL can be described by the state-space
model of (6.4) with Hpll(s) ¼ cθT(sI Apll)1bθ þ dθ. The pair of input-output
signals for the PLL is converted to θpcc θerror. This implies that for the rest of the
power system excluding the PLL, the pair of input-output signals is θerror θpcc.
Hence, the following state-space model of the rest of the power system can be
established.
8
< d ΔX ¼ A ΔX þ b Δθ
g g g g error
dt ð6:5Þ
:
Δθpcc ¼ cg ΔXg þ dg Δθerror
T
where ΔXg is the vector of all the state variables of the rest of the power system.
The transfer function model of the rest of power system can be obtained from
(6.5) as,
Δθpcc ¼ Gg ðsÞΔθerror ð6:6Þ
where Gg(s) ¼ cgT(sI Ag)1bg þ dg.

From (6.3) and (6.6), a closed-loop interconnected model of the power system
with the PMSG is established, as displayed in Fig. 6.3. In the established model, the
PLL and the rest of the power system are two separate open-loop interconnected
subsystems, as illustrated in Fig. 6.1. The transfer functions of those two open-loop
subsystems are Hpll(s) and Gg(s), respectively, as shown in Fig. 6.3. Their state-space
representations are (6.4) and (6.5), respectively. The state-space model of the closed-
loop interconnected system shown by Fig. 6.3 is,
d
ΔX ¼ AΔX ð6:7Þ
dt
where
T
ΔX ¼ ΔXg T ΔXpll T
" #
A11 A12
A¼
A21 A22
d1 bP cg T þ d2 bQ cg T
A11 ¼ Ag þ
1 d1 dP d2 d Q
ðd1 dP bP þ d2 dP bQ Þcp1 T þ ðd1 dQ bP þ d2 dQ bQ Þcp2 T
A12 ¼ bP cp1 T þ bQ cp2 T þ
1 d1 dP d2 dQ
bθ cg T
A21 ¼
1 dP d1 dQ d2
dP bθ cp1 T þ dQ bθ cp2 T
A22 ¼ Apll þ
1 dP d1 dQ d2
6.1.2 Open-Loop Modal Resonance Caused by the PLL
6.1.2.1 The Condition of Open-Loop Modal Resonance (OLMR)
For an ideal PLL, the phase-tracking error is zero, i.e., Δθerror ¼ 0. In this case, the
closed-loop interconnected model in Fig. 6.3 is open and the PLL does not introduce
any dynamic interactions with the rest of power system. The state-space model of the
rest of the power system (ROPS) becomes,
d
dt
Denote λi as an oscillation mode of the ROPS subsystem in Fig. 6.3. When an
ideal PLL is considered (Δθerror ¼ 0), λi is an eigenvalue of the open-loop state
matrix, Ag, in (6.5) or (6.8). Denote ^λ i as the oscillation mode corresponding to λi
when the PLL is non-ideal with a phase-tracking error (Δθerror 6¼ 0). Obviously, ^λ i is
an eigenvalue of the closed-loop state matrix, A, in (6.7). Hence, the impact of the
PLL phase-tracking error on the oscillation mode of the power system is the
difference between the open-loop and closed-loop oscillation mode Δλi ¼ ^λ i λi .
If ^λ i is on the right-hand side of λi on the complex plane, it implies that the PLL
phase-tracking error degrades the damping of the oscillation mode and hence, is
detrimental to the power system small-signal stability.
A well-designed PLL should have a negligible phase-tracking error, i.e.,
Δθerror 0. Thus, the closed-loop system shown by Fig. 6.3 is “approximately”
open. The difference between the open-loop and closed-loop EOMs of concern,
Δλi ¼ ^λ i λi , is negligible. In this case, the dynamic interactions between the PLL
and the rest of power system are weak. This explains why an ideal PLL can normally
be assumed, when the impact of the PMSG on the power system small-signal angular
stability is examined. However, under the special condition of open-loop modal
resonance (OLMR), this normality may change as to be elaborated as follow.
Denote λpll as a complex pole of the PLL’s transfer function, Hpll(s), which is
referred to as the PLL oscillation mode (POM). λpll is a complex eigenvalue of the
open-loop state matrix, Apll, in (6.4). An open-loop modal resonance is the special
condition that an open-loop POM is near the open-loop oscillation mode of the
power system on the complex plane, i.e., λpll λi. As |Hpll(λpll)| ¼ 1, |Hpll(λi)| is
significant, when λpll λi. From (6.3) and Fig. 6.3, it can be seen that under the
condition of the OLMR, Δθerror may become significant around the complex fre-
quency, λi. In this case, the dynamic interactions between the PLL and the rest of the
power system may be strong. Considerable dynamic interactions are introduced by
the PLL which exhibit as the considerable dynamic variations of Δθerror.
It would be extremely difficult to establish the mathematical expression,
Δλi ¼ ^λ i λi , for determining the exact impact of the strong dynamic interactions
introduced by the PLL under the condition of the OLMR. However, an estimate of
Δλi ¼ ^λ i λi in the neighborhood of λi ( ^λ i ! λi ) can be derived. The derived
estimation indicates the location of ^λ i in respective from λi on the complex plane
when the OLMR occurs. That is the impact of the OLMR caused by the PLL on the
small-signal stability of the power system, which is discussed in the next subsection.
6.1.2.2 Impact of OLMR
From Fig. 6.3, the characteristic equation of the closed-loop interconnected system
can be obtained as,
1 ¼ Gg ðλÞHpll ðλÞ ð6:9Þ
Express the PLL open-loop transfer function and that of the rest of the power
system in Fig. 6.3, respectively, as,
X
n
Rgk Xm
R
G g ðsÞ ¼ þ dg , Hpll ðsÞ ¼ pllk þ dθ ð6:10Þ
k¼1
ð s λk Þ k¼1
s λpllk
where λk, k ¼ 1, 2, . . .n are the eigenvalues of the open-loop state matrix, Ag, in
(6.5); Rgk, k ¼ 1, 2, . . .n are the residues corresponding to λk, k ¼ 1, 2, . . .n; λpllk,
k ¼ 1, 2, . . .m are the eigenvalues of the open-loop state matrix, Apll, in (6.4) and
Rpllk, k ¼ 1, 2, . . .m are the residues corresponding to λpllk, k ¼ 1, 2, . . .m.
The closed-loop oscillation mode of the power system, ^λ i , is a solution of (6.9).
Hence, from (6.9) and (6.10),
" #" #
X n
Rgk Xm
Rpllk
1¼ þ dg þ dθ ð6:11Þ
k¼1
ðλ fi λk Þ k¼1
λ fi λpllk
2
Multiplying both sides of (6.11) by ^λ i λi and under the condition of the
OLMR, i.e., λpll λi,
8 2 39
>
> >
>
>
> 6 n 7>>
2 < 6 X Rgk 7 =
^λ i λi Rgi þ ^λ i λi 6 7
þ dg 7
6
>
>
> 4 k ¼ 1 ^λ i λk 5>>
>
>
: >
;
8 2 k ¼
6 i 39 ð6:12Þ
>
> >
>
>
> 6 7 >
>
< 6 X m
Rpllk 7=
Rpll þ ^λ i λi 6
6 þ d θ
7
7>
>
> 4 ^λ i λpllk 5>
>
> k ¼ 1 >
>
: ;
k 6¼ i
where Rgi and Rpll are the residues corresponding to λi and λpll, respectively. Hence,
in the neighborhood of λi (^λ i ! λi ),
2 2
Δλi 2 ¼ ^λ i λi lim ^λ i λi Rgi Rpll ð6:13Þ
^λ i !λi
Denote ^λ pll as the closed-loop POM corresponding to λpll. The above derived
equation, (6.13), will be more meaningful, when the estimation of ^λ pll is derived.
^
λ pll is also a solution of (6.9). With a derivation similar to that of (6.9) to (6.12) for
^λ pll ,
2 2
Δλpll 2 ¼ ^λ pll λpll lim ^λ pll λpll Rgi Rpll ð6:14Þ
^λ pll !λpll
From (6.13) and (6.14),

pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
^λ i λi Rgi Rpll , ^λ pll λpll Rgi Rpll ð6:15Þ
The approximate estimation of the closed-loop POM and the oscillation mode of
power system by (6.15) indicates that when the OLMR (λpll λi) occurs, the closed-
loop POM and oscillation mode of power system intend to locate at the opposite
positions with respect to the point of modal resonance, λpll λi, on the complex
plane. Hence, it is very likely that the OLMR may degrade the power system small-
signal stability.
From (6.15), it can be seen that,
pffiffiffiffiffiffiffiffiffiffiffiffiffi
if Real part of Rgi Rpll > Real part of λi or λpll ð6:16Þ
either the closed-loop POM or the closed-loop oscillation mode of power system
may locate on the right half of the complex plane. In this case, considerable dynamic
interactions are introduced by the PLL to cause the small-signal instability of the
power system.
6.2 Small-Signal Stability of a Grid-Connected PMSG Wind Farm Affected by the PLL 209
The OLMR can be examined by calculating the complex eigenvalues of the PLL
open-loop state matrix, Apll, in (6.4) and that of the rest of the power system, Ag, in
(6.5). The proximity of an open-loop POM, λpll, to an open-loop EOM, λi, on the
complex plane indicates a case of OLMR, i.e., λpll λi. Then, further examination
should be followed by the calculation of the open-loop residues, Rgi and Rpll, to
estimate the impact of the OLMR on the power system small-signal stability using
(6.16).
6.2 Small-Signal Stability of a Grid-Connected PMSG

Wind Farm Affected by the PLL
For the recent years, a great effort has been spent by many researchers to examine the
small-signal stability of a grid-connected wind farm affected by the dynamic inter-
actions introduced from the PLL. The investigation was to scan the degree of
dynamic interactions by tuning the PI gains of the PLL. It was found that the
dynamic interactions between the PLL and the DC voltage control of the converter
control system of a PMSG were strongest when the bandwidth of the PLL was close
to that of the DC voltage control system. When the bandwidth of the PLL got close to
that of current control inner loops of the converter control systems, dynamic
interactions between the PLL and current control inner loops increased. In Sect.
1.2.1 of Chap. 1, main results of the examination mentioned above have been
reviewed.
According to the control theory of a linear system, the bandwidth of a dynamic
component in the system is normally close to the frequency of the open-loop
oscillation mode of the dynamic component. Hence, if the bandwidth of PLL is
ωp, an open-loop POM is λpll ¼ ξp jωp. If the bandwidth of a dynamic
component, such as the DC voltage control outer loop of the GSC of the PMSG is
around ωg, it is possible that the frequency of an open-loop oscillation mode of the
ROPS subsystem is around ωg. Hence, closeness of the bandwidth of PLL to the
bandwidth of the dynamic component of the ROPS, such as the DC voltage control
outer loop, implies that ωg ωp. Then, it is possible that the OLMR may happen
between λpll ¼ ξp jωp and λg ¼ ξg jωg. This explains why the closeness of
bandwidths may result in strong dynamic interactions introduced by the PLL and
increase the small-signal instability risk brought about by the PLL from the perspec-
tive of the OLMR.
In this subsection, the first example power system with the grid-connected PMSG
is presented to demonstrate how the OLMR explains the phenomenon of strong
dynamic interactions introduced by the PLL when its bandwidth is close to that of
converter control of the PMSG.
The second example power system presented in the subsection demonstrates how
the small-signal stability of a wind farm is affected by the PLL under the condition
of OLMR.
6.2.1 Example 6.1: A PMSG Connected to an External Power

System
Configuration of an example power system is shown by Fig. 6.4, where a PMSG is

connected to an infinite AC busbar via a transmission line represented by an
equivalent lumped reactance, XL. Model and parameters of the PMSG and its
converter control systems given in [3] were used. The PMSG was equipped with a
synchronous reference frame (SRF) PLL [4–8]. Model and parameters of the SRF
PLL given in [4–8] were adopted. Parameters of example power system are given in
Appendix 6.1.
The SRF PLL is the simplest and most commonly-used PLL scheme. Figure 6.5
shows the block diagram of linearized model of the SRF PLL [4–8]. The PI feedback
control configuration in the SRF PLL is often the core of a more complicated PLL
scheme. Hence, in order to demonstrate and evaluate the analysis and conclusions
made in the previous section regarding the open-loop modal resonance, it considered
that PMSG in Fig. 6.4 was equipped with a SRF PLL.
Comparing Figs. 6.1 and 6.4, it can be seen that in this example power system, the
external power system is represented by XL plus the infinite AC busbar. Hence, only
the dynamics of the transmission line connecting the PMSG with the power system
are considered. Other dynamics of the external power system are ignored. Therefore,
in the closed-loop model shown by Fig. 6.3 established for the example power
system, main dynamics of the ROPS subsystem include the control systems of the
GSC of the PMSG and the transmission line represented by XL. Impact of dynamic
interactions between the SRF PLL and those main dynamic components on the
small-signal stability is examined under the condition of OLMR in this subsection as
follows.
Fig. 6.4 A PMSG Vpcc θpcc Vb

connected to an infinite AC
Infinite
busbar PMSG
AC busbar
XL
Low-pass Filter
K pp
Δθ pcc + ΔVq + Δθ pll
1
Vd0 F (s)
– K pi + s
s Δx pll
Fig. 6.5 Block diagram of an SRF PLL

Fig. 6.6 Modal positions 90

and trajectories when Kpi
varied
70
lˆ pll lpll
Imaginary
lg1
50
C
B
30 lˆ g1
10
A Real
-40 -30 -20 -10 0
6.2.1.1 The OLMR Between the PLL and the DC Voltage Control Outer
Loop of the GSC
Firstly, an open-loop oscillation mode of the ROPS subsystem of example power

system was calculated to be λg1 ¼ 18.8716 þ j40.6902, which was found being
associated with the DC voltage control outer loop of the GSC. PI gains of the PLL
were initially set to be Kpp ¼ 0.18 and Kpi ¼ 1.2. The open-loop POM was
calculated to be λpll ¼ 24.3549 þ j8.1797. Position of λpll on the complex
plane is indicated by hollow rectangle at point A in Fig. 6.6. Position of λg1 is
indicated by hollow rectangle at point C.
Secondly, in order to demonstrate and evaluate the small-signal instability risk
caused by the OLMR between the PLL and the DC voltage control outer loop, the
integral gain of the PLL was varied from the original value, Kpi ¼ 1.2 to Kpi ¼ 23.
With the variation of Kpi, λpll moved from the initial position A towards λg1 on the
complex plane as shown by the dashed curve in Fig. 6.6. At point B , λpll λg1
where
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi the OLMR occurred and it was calculated that
Rg1 Rpll ¼ 14:3527 þ j4:8796. Then, the closed-loop oscillation modes, ^λ pll and
^λ g1 , were estimated by using (6.15). The estimated positions are indicated by crosses
in Fig. 6.6. It can be seen that under the condition of the OLMR, damping of ^λ g1 was
poor; but the grid-connected PMSG system was still stable.
Thirdly, to confirm the modal computation and estimation presented above,
closed-loop oscillation modes, ^λ pll and ^λ g1 , were calculated from the closed-loop
state matrix, A in (6.7). Trajectories of ^λ pll and ^λ g1 with the variation of Kpi are
displayed in Fig. 6.6 by solid curves. Positions of ^λ pll and ^λ g1 when the OLMR
occurred at point B are indicated by filled circles. They confirmed the correctness of
the estimation made by using (6.15) that ^λ pll and ^λ g1 located at the opposite positions
in respect to λpll λg1. The PLL degraded the damping of ^λ g1 under the condition of
the OLMR.
Fourthly, the participation factors (PFs) for ^λ pll and ^λ g1 were computed to
evaluate the dynamic interactions between the PLL and the DC voltage control
outer loop. Computational results are presented in Fig. 6.7, where the dashed curves
are the sum of the PFs of all the state variables of the PLL and the solid curves are the
sum of the PFs of all the state variables of the DC voltage control outer loop. From
Fig. 6.7, it can be seen that initially, the PLL and the DC voltage control outer loop
took part in only its own oscillation modes, ^λ pll and ^λ g1 , separately. When λpll moved
towards λg1, the PLL and the DC voltage control outer loop gradually participated
more and more in both ^λ pll and ^λ g1 , indicating the increase of their dynamic
interactions. Under the condition of the OLMR (point B) when Kpi ¼ 9.47, both
^λ pll and ^λ g1 were participated by the PLL and the DC voltage control outer loop
considerably, confirming their strong dynamic interactions.
Figure 6.8 shows the bandwidth of the PLL and the DC voltage control loop
respectively. It can be seen that their bandwidths were close to each other when the
OLMR happened, confirming the explanation in the previous section that the
closeness of the bandwidths in fact indicated the condition of the OLMR. This
explained the finding that when the bandwidth of the PLL was close to that of the Dc
voltage control loop, their dynamic interactions increased.
a b
0.8 0.8
Participation factor
PLL
DC
0.6 0.6
0.4 0.4
DC PLL
0.2
0.2
A B A B
Fig. 6.7 Computational results of the PFs. (a) The PFs for ^λ p1 . (b) The PFs for ^λ g1
Fig. 6.8 Bandwidth of the

PLL and DC voltage control
outer loop 20
PLL
Bandwidth (Hz)
15
DC voltage control
10
A B
Fig. 6.9 Modal positions

and trajectories when Kpi 250
lpll
varied
200 lˆ pll
lg2
Imaginary
150
E
100 lˆ g2
50 D
Real
-120 -100 -80 -60 -40 -20 0
0.8 0.8
PLL current control

0.6 0.6
0.4 0.4
current control
0.2 PLL
0.2
D E D E
Fig. 6.10 Computational results of the PFs for b

λ pll and b
λ g1 when Kpi varied
6.2.1.2 The OLMR Between the PLL and the Current Control Inner
Loops of the GSC
Another open-loop oscillation mode of the ROPS subsystem was

λg2 ¼ 44.5767 þ j147.6429, which was found being associated with the current
inner loop of the GSC. The test similar to that presented in 6.2.1.1 above was carried
out to demonstrate and evaluate the OLMR between the PLL and the current inner
loops of the GSC by varying the PI gains of the PLL from Kpp ¼ 0.35, Kpi ¼ 13 to
Kpp ¼ 0.35, Kpi ¼ 220. Modal positions and trajectories with the variations of the PI
gains are shown in Fig. 6.9. It can be seen that the OLMR occurred at point E
(Kpp ¼ 0.35, Kpi ¼ 91), where estimation on ^λ pll and ^λ g2 made by using (6.15)
(crosses in Fig. 6.9) was correct that ^λ pll and ^λ g2 located at the opposite positions in
respect to the resonant open-loop oscillation modes. Damping of ^λ g2 was degraded
considerably by the PLL under the condition of the OLMR at point E.
Figures 6.10 and 6.11 presents the computational results of the PFs and band-
width of the PLL and the current control inner loop respectively. It can be seen that
when λpll moved close to λg2 with the variations of the PI gains of the PLL,
bandwidth of the PLL got close to that of the current control inner loops and their
dynamic interactions increased. When the OLMR at point E happened, their
Fig. 6.11 Bandwidth of the

PLL and current control 80
inner loops PLL
Bandwidth (Hz)
60
current control
40
20
A C
bandwidths were very close and dynamic interactions were strong. This explained
the finding that when the bandwidth of the PLL was close to that of the current
control inner loops, their dynamic interactions increased.
6.2.1.3 Weak Network Connection
As it was reviewed in 1.2, it has been found that weak network connection may cause
the small-signal instability of the grid-connected PMSG system. The weakness of
the network connection is defined by the short circuit ratio (SCR). When the SCR is
below 2, the network connection is considered to be weak.
In the examination in 6.2.1.1 and 6.2.1.2 above, SCR ¼ 2.1. To evaluate the
condition of the OLMR as affected by the weak network connection, the equivalent
transmission reactance in Fig. 6.4 was creased from XL ¼ 0.45 pu to XL ¼ 0.6 pu
such that the SCR was reduced to SCR ¼ 1.7. Then, the tests as same as those
conducted in 6.2.1.1 and 6.2.1.2 were carried out under the condition of reduced
SCR (increased XL). Modal positions and trajectories were calculated and are
presented in Fig. 6.12.
From Fig. 6.12, it can be seen that when the OLMR occurred at point F and G, the
closed-loop oscillation modes, ^λ g1 and ^λ g2 , moved into the right half of the complex
plane. The grid-connected PMSG system became unstable. In Table 6.1, computa-
tional
results of R gi Rpll , i ¼ 1, 2 are presented, confirming that
Real part of Rgi Rpll increased considerably when the SCR decreased such that
the grid-connected PMSG system lost the small-signal stability.
In fact, the magnitude of Rgi Rpll , i ¼ 1, 2 is proportional to XL. Hence, the
bigger XL is, the bigger the impact of OLMR is. This is can be proved as follows.
Dynamics of the ROPS subsystem of the example power system shown by
Fig. 6.4 are comprised of the following components (For detailed linearized model
of the PMSG, see Sect. 2.1 in Chap. 2).
Fig. 6.12 Modal positions a

and trajectories when Kpi of 80
PLL varied with reduced
SCR (SCR ¼ 1.7). (a) The
OLMR between the PLL
and DC voltage control 60
Imaginary
outer loop; (b) the OLMR
between the PLL and
current control inner loop 40 F
20
Real
-40 -30 -20 -10 0 8
b
250
200
Imaginary
150
G
100
50
Real
-120 -80 -40 0 20
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Table 6.1 Computational results of Rgi Rpll , i ¼ 1, 2 and closed-loop oscillation modes for study
case (3)
The OLMR in Fig. 6.6 Rg1 Rpll ¼ 14:352 j3:979 ^λ g1 ¼ 4:191 þ j41:804
The OLMR in Fig. 6.12a Rg1 Rpll ¼ 18:361 j4:931 ^λ g1 ¼ 1:528 þ j38:385
The OLMR in Fig. 6.9 Rg2 Rpll ¼ 41:609 þ j29:081 ^λ g2 ¼ 4:376 þ j153:561
The OLMR in Fig. 6.12b Rg2 Rpll ¼ 58:675 þ j23:189 ^λ g2 ¼ 5:163 þ j157:447
1. Filter reactance:
8
>
> dΔId
¼ ω0 ðΔVcd ΔVd Þ þ ω0 X f ΔIq
<Xf
dt
ð6:17Þ
>
> dΔIq
:Xf ¼ ω0 ΔVcq ΔVq ω0 X f ΔId
dt
where Vd þ jVq is the PCC voltage of the PMSG, Vpcc ∠ θpcc, expressed in the d q
coordinate of the GSC, Id þ jIq is the output current of the GSC expressed in the
d q coordinate of the GSC.
2. DC capacitor (variation from the machine side Converter of the PMSG is ignored,
as Ppmsg is not affected by the dynamic interactions between the PMSG and the
power system; thus, it can be assumed that ΔPpmsg ¼ 0, see (2.28))
dΔVdc
CVdc0 ¼ ΔPvsc ð6:18Þ
dt
3. Control systems of the GSC. Configuration of the control system of the GSC is
shown by Fig. 2.3, which is redrawn as Fig. 6.13. Hence, linearized dynamic
equations of DC voltage control and d-axis current control are
8
>
> dΔxdc
¼ ki dc ΔVdc
<
dt
ð6:19Þ
>
> dΔxid ¼ k ΔI þ k k
: i d d i d p dc ΔVdc þ ki d Δxdc
dt
Linearized equations of reactive power control and q-axis current control are
8
>
> dΔxQ
¼ ki Q ΔQ
<
dt
ð6:20Þ
>
> dΔxiq
: ¼ ki q ΔIq þ ki q kp Q ΔQ þ ki q ΔxQ
dt
Ignoring the transient of the PMW algorithm such that in Fig. 6.13, Vcd ¼ Vcd∗ and
Vcq ¼ Vcq∗. From Fig. 6.13,
(
ΔVcd ¼ kp d ΔId þ kp d kp dc ΔVdc þ kp d Δxdc þ Δxid X f ΔIq þ ΔVd
ΔVcq ¼ kp q ΔIq þ kp q kp Q ΔQ þ kp q ΔxQ þ Δxiq þ X f ΔId þ ΔVq
ð6:21Þ
K p_dc K p_d Vd
*
V ref
dc – + I dref+ + + + Vcd
– –
+ Ki_dc + Ki_d +
Vdc Id X f iq
s xdc s xid
PWM
K p_Q K p_q Vq
Qwref – I qref + Vcq*
+ + + +
+ –
Ki_Q + Ki_q + +
Qw Iq X f id
s xQ s xiq
Outer loop Inner loop
Fig. 6.13 Configuration of control system of the GSC of the PMSG

4. Transmission line represented by XL. Denote Vx þ jVy as the PCC voltage of the
PMSG, Vpcc ∠ θpcc, expressed in the common x y coordinate of the power
system. Denote Ix þ jIy as the output current of the GSC expressed in the common
x y coordinate of the power system. Voltage at the infinite busbar is constant.
Hence,
(
ΔVx ¼ XL ΔIy
ð6:22Þ
ΔVy ¼ XL ΔIx
5. Output power from the GSC is P þ jQ. Following linearized equations can be
obtained (The phase-tracking error of the PLL at the steady state is zero. Hence,
Vd0 ¼ Vpcc0 and Vq0 ¼ 0)
(
ΔP ¼ Vpcc0 ΔId þ Id0 ΔVd þ Iq0 ΔVq
ð6:23Þ
ΔQ ¼ Vpcc0 ΔIq þ Id0 ΔVq Iq0 ΔVd
6. Linearized equations of transformation between d q and x y coordinate.

" # " #" # " #
ΔIx cos θpll0 sin θpll0 ΔId Iy0
¼ þ Δθpll ð6:24Þ
ΔIy sin θpll0 cos θpll0 ΔIq Ix0
" # " #" # " #

ΔVd cos θpll0 sin θpll0 ΔVx Vq0
¼ þ Δθpll ð6:25Þ
ΔVq sin θpll0 cos θpll0 ΔVy Vd0
From (6.17) to (6.25), the state-space model of the ROPS subsystem of (6.5) is
derived for the example power system shown by Fig. 6.4. It can have
ΔXg ¼ ½ ΔId ΔIq ΔVdc Δxdc Δxid ΔxQ Δxiq T
2 3
a11 0 a13 a14 a15 0 0
6 7
6 a21 a22 a23 0 0 a26 a27 7
6 7
6 7
6 a31 a32 0 0 0 0 0 7
6 7
6 7
Ag ¼ 6 0 0 a43 0 0 0 0 7 ð6:26Þ
6 7
6a 0 7
6 51 0 a53 a54 0 0 7
6 7
6a 0 7
4 61 a62 0 0 0 0 5
a71 a72 0 0 0 a76 0
2 3
0
6 2V2p0 kp q kp Q Id0 Iq07
6 7
6 X L 7 2 3
6 V 2
þ Q X 7 0
6 p0 0 L 7
6 7 6 7
6 7 6
Vp0
XL 7
6 V2p0 I2d0 I2q0 7 6 7
6 XL 7 6 Vp0 þ Q0 XL 7
2
6 7 6 7
6 CVdc0 V2 þ Q0 XL 7 6 7
6 p0 7 6 0 7
6 7 6 7
bg ¼ 6 7 ¼ X b ,
L g1 gc ¼ 6 7
6 0 7 6 0 7
6 7 6 7
6 0 7 6 7
6 7 6
0
7
6 7 6 7
6 2V2p0 ki Q Id0 Iq0 7 4 0 5
6 XL 7
6 V 2
þ Q X 7
6 p0 0 L 7 0
6 7
6 2V2 k k I I 7
4 p0 i q p Q d0 q0 5
XL
Vp0 þ Q0 XL
2
¼ XL cg1 ð6:27Þ
Q0 XL
dg ¼
þ Q0 XL
V2p0
where
ω0 ω0 ω0 ω0
a11 ¼ kp d , a13 ¼ kp d kp dc , a14 ¼ kp d , a15 ¼
Xf Xf Xf Xf
ω0
a21 ¼ kp q kp Q Id0 XL , a22 ¼ ki Q Iq0 XL Vg0
Xf
ω0 2Vp0 kp q kp Q Id0 Iq0 X2L
a22 ¼ kp q kp Q Iq0 XL Vp0 kp q þ
Xf V2p0 þ Q0 XL

Vp0 X2L I2d0 I2q0 ω 2V k I I X2
a23 ¼ , a26 ¼ 0 kp q þ p02 i Q d0 q0 L
CVdc0 V2 þ Q XL Xf Vp0 þ Q0 XL
p0 0
ω0 2Vp0 ki q kp Q Id0 Iq0 X2L

a27 ¼ þ
Xf V2p0 þ Q0 XL
Vp0 þ Iq0 XL Id0 XL
a31 ¼ , a32 ¼
CVdc0 CVdc0
a43 ¼ ki dc , a51 ¼ ki d , a53 ¼ ki d kp dc , a54 ¼ ki d ,
a61 ¼ ki Q Id0 XL ,

a71 ¼ ki q kp Q Id0 XL , a72 ¼ ki q kp Q Iq0 XL Vp0 ki q , a76 ¼ ki q
0.02 0.03
Point A Point D
Point B Point E
PLL tracking error (pu)

0.01 Point F 0.02 Point G
0.01
0 0
-0.01
-0.01
-0.02
time/s time/s
-0.02 -0.03
0 0.2 0.4 0.6 0.8 1 0 0.1 0.2 0.3 0.4 0.5 0.6
Fig. 6.14 Confirmation of non-linear simulation—Example 6.1
Denote υgi and wgi the left and right eigenvectors of state matrix Ag in (6.26) for
open-loop oscillation mode of the ROPS subsystem, λgi, i ¼ 1, 2. The residue Rgi for λgi
can be calculated from the state-space model of the ROPS subsystem of (6.27) to be
Rgi ¼ cg T υgi wgi bg ¼ X2L cg1

T
υgi wgi bg1 ¼ X2L Rgi1 ð6:28Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
From (6.28), Rgi Rpll ¼ XL Rgi1 Rpll . Hence, it is proved that the magnitude of
Rgi Rpll , i ¼ 1, 2 is proportional to XL.
Figure 6.14 presents the confirmation of non-linear simulation. At 0.1 s of
simulation, the reactive power output of the PMSG dropped by 0.02 p.u. which
was recovered in 0.11 s. Two simulation results are shown in Fig. 6.13: (1) λpll was at
point A in Fig. 6.6, point D in Fig. 6.9 (no OLMR occurred); (2) λpll was at point B in
Fig. 6.6, point E in Fig. 6.9, point F and G in Fig. 6.12, which were under the
condition of the OLMR. From Fig. 6.14, it can be seen that under the condition of the
OLMR, considerable variations of phase-tracking error of the PLL were observed
and poorly or negatively damped oscillations occurred.
6.2.2 Example 6.2: A PMSG Wind Farm Connected

to an External Power System
Configuration of a PMSG wind farm connected to an external power system is

shown by Fig. 6.15. Purpose of study is for the demonstration and evaluation of the
analysis and conclusions made in Sect. 6.1, the example PMSG wind farm was
simply comprised of only three PMSGs. They were aggregated models representing
the PMSGs in three different areas in the wind farm. All the three PMSGs were
equipped with the SRF PLLs. Model and parameters of the PMSG given in [6] were
used. The parameters of the GSC control systems and the PLL of the second PMSG
(PMSG2) and the third PMSG (PMSG3) were decreased and increased by 15%
respectively to reflect the fact that they represented the PMSGs in different areas in
the wind farm. Parameters of the example are given in Appendix 6.2. When a
GSC Vpcc θpcc

Vcd + jVcq
PMSG1 C Vdc
I1 Xf Xt1
P1+jQ1
Inner loop Outer loop

θerror Vdc Vb
PLL PLL1 Infinite AC
subsystem θpcc busbar
XL
PMSG 2
Xt2
ROPS
subsystem
PMSG 3
Xt3
Fig. 6.15 Configuration of a grid-connected PMSG wind farm system

and trajectories when Kpi of
PLL1 varied (SCR ¼ 2.15) 100
80
Imaginary
L
60
H
40
20
P Real
-20 -15 -10 -5 0
practical grid-connected wind farm with more PMSGs is studied, the procedure to
conduct the following demonstration and evaluation will be same.
The PLL (PLL1) for the first PMSG (PMSG1) was considered. Firstly, a closed-
loop model shown by Fig. 6.3 was established, where the open-loop PLL subsystem
in the feedback loop was PLL1; the open-loop ROPS subsystem included PMSG2,
PMSG3, the rest of PMSG1 with PLL1 being excluded and the power network. The
open-loop oscillation modes of the ROPS subsystem were calculated from the open-
loop state matrix, Ag in (6.5). Two open-loop oscillation modes, λg3 and λg4, were
found being related with the DC voltage control loop (DC2) of PMSG2 and the PLL
(PLL3) of PMSG3. Their positions on the complex plane are indicated by hollow
rectangles at point H and L in Fig. 6.16. PI gains of PLL1 were Kpp ¼ 0.12, Kpi ¼ 3
initially. The open-loop oscillation mode of PLL1 was λpll ¼ 13.6975 þ j17.8116
and its position is indicated by hollow circle at point P in Fig. 6.16.
Table 6.2 Computational results of Rgi Rpll , i ¼ 3, 4 and closed-loop oscillation modes for the
study cases of wind farm system
Point H in Fig. 6.16 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
when SCR ¼ 2.15 Rg3 Rpll ¼ 4:8969 j3:9149 ^
λ g3 ¼ 5:9845 þ j45:9711
Point H in Fig. 6.17 Rg3 Rpll ¼ 8:0309 j5:4493 ^
λ g3 ¼ 1:2377 þ j45:1659
when SCR ¼ 1.62
Point L in Fig. 6.16 Rg4 Rpll , 1 ¼ 3, 4 ¼ 8:4176 j5:1637 ^
λ g4 ¼ 2:4381 þ j73:8261
when SCR ¼ 2.15
Point L in Fig. 6.17 Rg4 Rpll ¼ 11:8275 j9:2929 ^
λ g4 ¼ 1:9755 þ j68:1097
when SCR ¼ 1.62
Secondly, the integral gain of PLL1 was varied from Kpi ¼ 3 to Kpi ¼ 40.
Trajectory of λpll with the variations of Kpi is displayed in Fig. 6.16 by dashed
curve. It can be seen that at point H, λpll λg3 such that the OLMR between PLL1
and DC2 happened. At point L, λpll λg4 where the OLMR between PLL1 and
PLL3 occurred.pffiffiffiffiffiffiffiffiffiffiffiffiffi
Thirdly, Rgi Rpll , i ¼ 3, 4 were computed for the OLMR at point H and L
respectively. Computational results are given in the first and third row of
Table 6.2. By using (6.15), the closed-loop oscillation modes, ^λ pll , ^λ g3 and ^λ g4 ,
under the condition of the OLMR at point H and L were estimated. The estimated
positions of ^λ pll , ^λ g3 and ^λ g4 are indicated by crosses. It can be seen that the dynamic
interactions between PLL1 and DC2 or PLL3 caused the stability degradation of the
example grid-connected wind farm under the condition of the OLMR. However, the
wind farm system was still stable.
Finally, the equivalent transmission reactance of the example wind farm was
increased from XL ¼ 0.31 pu to XL ¼ 0.42 pu such that the SCR decreased from
SCR ¼ 2.15 to SCR ¼ 1.62. Under the condition of weak network connection,
modal positions and trajectories were computed when Kpi varied. The computational
results are displayed in Fig. 6.17. Computational results of Rgi Rpll , i ¼ 3, 4 are
given in the second and fourth row of Table 6.2. It can be seen that with reduced
SCR, the wind farm system lost the small-signal stability under the condition of the
OLMR, because the magnitude of real part of Rgi Rpll , i ¼ 3, 4 increased
considerably.
Fourthly, ^λ pll , ^λ g3 and ^λ g4 were calculated from the closed-loop state matrix, A in
(6.7). Trajectories of ^λ pll , ^λ g3 and ^λ g4 with the variation of Kpi are displayed by solid
curves in Fig. 6.17, confirming the correctness of the estimation made by using
(6.15).
Confirmation of non-linear simulation is presented in Fig. 6.18. At 0.1 s of
simulation, the reactive power output of the PMSG dropped by 0.02 p.u. which
was recovered in 0.11 s. From Fig. 6.18, it can be seen that when the SCR was 2.15,
the poorly damped oscillation happened as caused by the dynamic interactions
introduced by PLL1 under the condition of the OLMR. When the network connec-
tion was weak (SCR ¼ 1.62), the dynamic interactions became stronger and the wind
farm system lost the small-signal stability under the condition of the OLMR.

and trajectories when Kpi of
PLL1 varied (SCR ¼ 1.62) 100 lp1
lˆp1
80 lg4
Imaginary
L
60
lg3 lˆg4
40 H
lˆg3
20
P Real
-20 -15 -10 -5 0 5
a b
0.014 PMSG1 PMSG1
PMSG2 0.06
PMSG2
PMSG3 PMSG3
0.007
0.03
0 0
-0.007 -0.03
-0.014 -0.06
time/s time/s
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fig. 6.18 Confirmation of non-linear simulation when the OLMR occurred. (a) SCR ¼ 2.15. (b)
SCR ¼ 1.62
6.3 Example 6.3: Electromechanical Oscillation Modes of a

Power System Affected by the PLL
Figure 6.19 shows the configuration of the New England test power system (NEPS)
extensively used for studying the power system electromechanical oscillations [9–
11]. The parameters of the network and the synchronous generators (SGs) given in
[9] were used. There were nine electromechanical oscillation modes (EOMs) in the
NEPS. An inter-area EOM, between the tenth SG and the rest of SGs, was the EOM
of main concern, with the smallest damping and lowest oscillation frequency. Two
PMSGs were connected at node 22, in the NEPS and operated with unity power
factors, each. The detailed 15th-order dynamic model of the PMSG, including its
control systems recommended in [6], was used. Parameters of the test power system
are given in Appendix 6.3.
The first PMSG (PMSG1) connected at node 22 was equipped with an SRF PLL.
The second PMSG was equipped with a more complicated PLL, the duel second-
order generalized integrator phase locked loop (DSOGI) PLL [12] designed to track
6.3 Example 6.3: Electromechanical Oscillation Modes of a Power System. . . 223
8
1 37
30
26 28 29
25
2
27 24
18 38
1 17 9
16
10 6
3 35
39 15
14 21
4 22
5 19
6 12 23
9 PMSG
20
11 13 33 36
31
4 7
7 2 34
10
5
8
32
3
Fig. 6.19 Configuration of the New England test power system (NEPS) with two PMSGs
ω0
vα′ + v +
α
v+
q + θ+
vα
SOGI qvα′ +
1/ 2 PI ∫
– Park +
1/ 2
+ v+β vd+ ω ff
vabc
Clark
v′β
vβ
SOGI qv′β
+ v–α
1/ 2
– + + v–β
1/ 2
Fig. 6.20 Configuration of a DSOGI PLL
the phase angle of the positive sequence voltage. The configuration of the DSOGI
PLL is depicted in Fig. 6.20. The impact of the dynamic interactions introduced by
the PMSGs connected at node 22, caused by the PLLs, on the inter-area EOM was
examined.
The PI gains of the PLLs were varied in a wide range to scan the possibility of
OLMR between the PLLs and the NEPS. The PI gain values given in MATLAB [13]
for the SRF PLL were used to start the scan; the PI gain of the DSOGI PLL was
initially set near the highest frequency EOM. This scanning strategy, by varying the
PI gains of the PLLs, has been used to study the impact of the PLL on the stability of
a VSC-based system. It was adopted in the following tests for investigating the
potential danger posed by the OLMR. Thus, in PLL design, the identified danger can
be avoided.
When the PI gains of the PMSG PLLs were varied, the positions of the open-loop
POMs on the complex plane changed. Hence, OLMR may be determined by varying
the PI gains of the PLLs. When the operating conditions of the power system
changed, the open-loop EOMs moved on the complex plane; this may also lead to
open-loop modal resonance. In the following subsections, test results are presented
to demonstrate the possibility that changes in the power system operating conditions
and the PI gains of the PLLs may lead to OLMR.
6.3.1 Test 1: Changes of Power System Operating Conditions
For the PLL equipped by PMSG1, the closed-loop model in Fig. 6.3 was derived,
where the open-loop transfer function of the SRF PLL is (see Fig. 6.5 of the
linearized model of the SRF PLL and (6.3)),
Δθerror s2
Hpll ðsÞ ¼ ¼ 2 ð6:29Þ
Δθpcc s þ Vd0 FðsÞKpp s þ Vd0 FðsÞKpi
The typical PI gains of the SRF PLL, given in [14], were Kpp ¼ 4.4 and Kpi ¼
39.27. The open-loop and closed-loop POMs were calculated to be λpll ¼ 2.1975 þ
j5.8688 and ^λ pll ¼ 1:2655 þ j5:7057, respectively. The open-loop and closed-
loop EOMs of the power system were computed from the open-loop and closed-loop
state matrices, Ag in (6.5) and A in (6.7), respectively. In Table 6.3, λi, i ¼ 1, 2, . . .9
were the open-loop EOMs of the NEPS, when the PLL phase-tracking error was
assumed to be zero. ^λ i , i ¼ 1, 2, . . . 9 were the closed-loop EOMs, when the PLL
phase-tracking error was considered. Hence, Δλi ¼ ^λ i λi , i ¼ 1, 2, . . . 9 were
caused by the dynamic interactions between the SRF PLL and the rest of the
power system. In Table 6.3, λ9 and ^λ 9 were the open-loop and closed-loop inter-
area EOMs of main concern, respectively. From the first three columns of Table 6.3,
it can be seen that the open-loop POM, λpll ¼ 2.1975 þ j5.8688, was closer to λ6
and λ7, than to the other EOMs. Hence, the SRF PLL affected λ6 and λ7 more, and
Δλ6 and Δλ7 were slightly larger.
In order to examine whether a larger Δλ6 and Δλ7 were caused by the proximity
of λpll to λ6 and λ7, the PI gains of the PLL were changed to Kpp ¼ 7.54, Kpi ¼ 377
such that the open-loop POM became λpll ¼ 3.7699 þ j19.047 and was far from
all the open-loop EOMs of the NEPS. The computational results of ^λ i , i ¼ 1, 2, . . . 9
and Δλi, i ¼ 1, 2, . . .9, when λpll ¼ 3.7699 þ j19.047, are presented in the last two
columns of Table 6.3. It can be seen that, when λpll moved far away from all the
EOMs, the impact of the PLL on all the EOMs, Δλi, i ¼ 1, 2, . . .9, was less. This
Table 6.3 Computational results of the open-loop and closed-loop EOMs of the NEPS
i λi Typical PLL parameters were used PLL parameters were changed
^λ i Δλi ^
λi Δλi
1 0.55144þ 0.5516þ 0.0002þ 0.5514þ 3e-05þ
8.4521j 8.4522j 0.0001j 8.4522j 0.0001j
2 0.4689 0.47178 0.0029 0.46993 0.00103
þ7.8529j þ7.8663j þ0.0134j þ7.8515j 0.0014j
3 0.32196 0.32533 0.0034 0.3226 0.00064
þ6.9978j þ6.9988j þ0.0010j þ6.9972j 0.0006j
4 0.53018 0.53992 0.0097 0.53094 0.00076
þ7.1673j þ7.1705j þ0.0032j þ7.1649j 0.0024j
5 0.45682 0.45623 0.0006 0.45686 4e-05
þ7.1209j þ7.1211j þ0.0002j þ7.121j þ0.0001j
6 0.49977 0.69915 0.1994 0.4975 0.00218
þ5.9183j þ6.5069j þ0.5886j þ5.9148j 0.0035j
7 0.48828 0.52733 0.0390 0.49539 0.00711
þ6.3816j þ5.8967j 0.4849j þ6.3584j 0.0232j
8 0.46338 0.45379 0.0096 0.4665 0.00314
þ6.1885j þ6.1997j þ0.0112j þ 6.1894i þ0.0009j
9 0.30609 0.31961 0.0135 0.30175 0.00434
þ3.5525j þ3.2819j 0.2706j þ3.525j 0.0275j
confirmed that when the typical PI gains recommended in [14] were used, a slight
OLMR of λpll with λ6 and λ7 occurred such that Δλ6 and Δλ7 were larger. This
explained why the closed-loop POM, ^λ pll ¼ 1:2655 þ j5:7057, was on the right of
the open-loop POM, λpll ¼ 2.1975 þ j5.8688, on the complex plane. The OLMR
of λpll with λ6 and λ7 caused the damping degradation in^λ pll ¼ 1:2655 þ j5:7057.
In the test presented above, the OLMR was not strong and the closed-loop EOMs
and POM were slightly affected. The PI gains were varied to demonstrate that, when
the OLMR was eliminated, the impact of the PLL became negligible. In the
following test, it is shown that when the operating conditions of the NEPS changed,
the open-loop POM moved closer to the EOM, λ7. Subsequently, the OLMR became
stronger, causing a greater PLL impact on the power system EOMs.
In the test, the change in the operating conditions of the NEPS was the gradual
increase in the seventh SG constant of inertia from 69.6–80s. This replicated the
changes in the power system operating conditions, when there was more generation
as the spinning reserve at the seventh SG site. With the gradual increase in inertia,
the seventh open-loop EOM, λ7, moved closer to the open-loop POM,
λpll ¼ 2.1975 þ j5.8688, on the complex plane, as shown by the dashed curve
in Fig. 6.21. The movement ended at λ7 ¼ 2.0761 þ j5.6494, which was closer to
λpll ¼ 2.1975 þ j5.8688 than it was before the constant of inertia was changed.
The OLMR points, λpll λi, are indicated by a hollow rectangle and circle,
respectively, in Fig. 6.21. The trajectories of the corresponding movement of the
seventh closed-loop EOM, ^λ 7 and the closed-loop POM, ^λ pll , are also displayed in
Fig. 6.21 as solid curves. Their positions corresponding to the open-loop modal
6.6
6.4
λˆ7
6.2
Imaginary axis
λ7 Inertia gradually
6
increased
λpll
5.8
A λˆpll
5.6
5.4
-3 -2.75 -2.5 -2.25 -2 -1.75 -1.5 -1.25 -1 -0.75 -0.5 -0.25
Real axis
Fig. 6.21 Modal resonance caused by the gradual increase in inertia of the seventh SG
resonance points are indicated by the filled rectangle and circle, respectively. From
Fig. 6.21, it can be seen that, when λpll and λ7 moved closer to each other, ^λ pll and ^λ 7
moved away from the corresponding open-loop modes λpll and λ7, indicating a larger
difference between the closed-loop and open-loop modes, i.e., a greater PLL impact.
From Fig. 6.21, it can be seen that ^λ pll and ^λ 7 (indicated by filled rectangle and
circle) located approximately at opposite positions, with respect to the open-loop
modal resonance points, indicated by a hollow rectangle and circle, respectively. The
positions of ^λ pll and ^λ 7 were also estimated using (6.15) and are indicated by crosses
in Fig. 6.21; it can be seen that the estimation is close to the actual positions indicated
by the filled rectangle and circle. Hence, the analytical conclusion made in Sect. 6.1,
from (6.15), was confirmed. In this case study, λpll was well damped initially and the
movement of ^λ pll towards the right on the complex plane was not significant. Hence,
although OLMR degraded the damping of the closed-loop POM, the power system
small-signal stability was still maintained.
At the OLMR points, the participation factors (PFs) were computed for the
closed-loop POM, ^λ pll and the closed-loop EOM, ^λ 7 . The computational results
are presented in Fig. 6.22, depicting a considerable participation of the SGs in ^λ pll
and the PLL in ^λ 7 . This confirmed that, when OLMR occurred, the dynamic
interactions between the PLL and the SGs were considerable.
0.25 0.16
0.14
0.2
0.12
0.15 0.1
0.08
0.1 0.06
0.04
0.05
0.02
0 0
1 2 3 4 5 6 7 8 9 10 Δθpll Δxpll-vq
Generator number
(a) (b)
Fig. 6.22 Computational results of the participation factors. (a) Participation of the SGs in the
POM; (b) participation of the PLL in the EOM
6.3.2 Test 2: Variation of PLL Gains (SRF PLL for PMSG1)
The second test was to show that when the PI gains of the PLL were changed, the
open-loop POM moved on the complex plane to cause OLMR with different EOMs
consecutively. The variation of the PI gains, in the test, was to identify certain
selections of the PLL PI gains, which introduced the detrimental effect of OLMR.
Thus, these selections should be avoided, when setting the PI gains of the PLL.
In the test, the PI gains of the PLL were changed gradually from Kpp ¼ 0.90,
Kpi ¼ 64 to Kpp ¼ 0.57, Kpi ¼ 13.2. With this change, the trajectory of the open-loop
POM, λpll, was depicted by dashed curves and the movement of the closed-loop
POM, ^λ pll , by solid curves in Figs. 6.23 and 6.24; it can be seen that while moving,
λpll came close to the four open-loop EOMs, λ2, λ4,λ6, and λ9. The OLMR points of
λpll with these four open-loop EOMs are indicated by hollow rectangles and circles.
In Figs. 6.23 and 6.24, the movement of the closed-loop EOMs is also displayed by
solid curves. The filled rectangles and circles indicate the positions of the closed-
loop EOMs and POMs, when the OLMR occurred.
First, from Figs. 6.23 and 6.24, it can be seen that with respect to the OLMR
points indicated by hollow rectangles and circles, the closed-loop modes, ^λ pll and^λ i ,
indicated by filled rectangles and circles, located at opposite positions. This con-
firmed the analytical conclusion made in Sect. 6.1 from (6.15). The estimated
positions of ^ λ pll and ^λ i obtained from (6.15) are depicted by crosses. It can be
seen that the estimation approximately indicated the actual positions of ^λ pll and ^λ i .
Next, from these figures, it can be seen that the level of the PLL impact caused by
the OLMR was different. From (6.15), it can be observed that, when the OLMR
occurs, the impact of the PLL is affected by Real part of Rgi Rpll . Table 6.4
presents the computational results of Real part of Rgi Rpll at four points of the
OLMR; Real part of Rgi Rpll for the OLMR between λpll and the open-loop inter-
area EOM of main concern, λ9, was the largest. From Fig. 6.23, it can be seen that the
movement of the closed-loop inter-area EOM of main concern from the point of
Enlarged view is
7 presented by Fig.9
6
Imaginary axis
lpll
4 l̂pll
lˆ 9
3 l9
2
-0.6 -0.5 -0.4 -0.3 -0.2 -0.1
Real axis
Fig. 6.23 Modal resonance caused by PLL parameter tuning
l2 lˆ2
7.5
lpll
lˆpll
7
Imaginary axis
lˆ4 l4
6.5
l6
6
lˆ6
5.5
5
-0.6 -0.55 -0.5 -0.45 -0.4
Real axis
Fig. 6.24 Enlarged view of the modal resonance caused by PLL parameter tuning
modal resonance was the farthest towards the right, confirming the computational
results presented in Table 6.4.
Figure 6.23 also shows that the OLMR of the PLL with the inter-area EOM
caused a considerable damping degradation of the inter-area EOM, posing a serious
threat to the power system small-signal angular stability. Further examination of this
Table 6.4 Estimation of the pffiffiffiffiffiffiffiffiffiffiffiffiffiffi

λi Real part of Rgi Rpll
impact of the OLMR for using
(6.15) i¼2 0.059689
i¼4 0.017552
i¼6 0.080128
i¼9 0.12531
0.35 0.16
0.3 0.14
0.12
0.25
0.1
0.2
0.08
0.15
0.06
0.1
0.04
0.05 0.02
0 0
ΔVdc Δxvdc Δθpll Δxpll-vq 1 2 3 4 5 6 7 8 9 10
(a) (b)
Fig. 6.25 Computational results of the PFs, when the OLMR of the PLL with the inter-area EOM
of concern occurred. (a) Participation of the PMSG in ^
λ 9 ; (b) participation of the SGs in ^
λ pll
particularly dangerous case of open-loop modal resonance, by calculating the PFs, is

presented in Fig. 6.25; it can be seen that, when the OLMR of the PLL with the inter-
area EOM occurred, both the PMSG and the SGs participated in the inter-area EOM
and POM considerably, indicating strong dynamic interactions between the PMSG
and the SGs.
Simulation tests were carried out for further confirmation. An 80% load drop
occurred at node 8 at 1.0 s of simulation. The lost load was restored in 100 ms. For
comparison, the simulation results for two cases are presented in Figs. 6.26 and 6.27,
respectively. In case A, the typical PLL parameters recommended in [14] are used
(Kpp ¼ 4.4 and Kpi ¼ 39.27). In case B, the OLMR of the PLL with the inter-area
EOM occurred, as shown in Fig. 6.23. From Fig. 6.26, it can be seen that, when the
OLMR occurred, there were considerable variations in the reactive power output of
the first PMSG. The damping degradation of the inter-area EOM of main concern
was confirmed by the simulation results in Fig. 6.27, where poorly damped power
oscillations were observed.
Figure 6.26 shows that considerable dynamic interactions between the first
PMSG and the power system were caused by the OLMR. The dynamic interactions
were exhibited as variations in the reactive power output from the first PMSG. This
can be elaborated as follows.
1
0.8 Case A
Reactive power output of PMSG1

Case B
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0 2 4 6 8 10 12 14 16 18 20
Time(s)
Fig. 6.26 Reactive power output from the PMSG1
-54
Case A
-56 Case B
-58
-60
d10 – d7(deg.)
-62
-64
-66
-68
-70
-72
0 2 4 6 8 10 12 14 16 18 20
Time(s)
Fig. 6.27 Relative rotor position between the tenth and seventh SGs
The active and reactive power outputs of a PMSG are (see Fig. 6.1),
Pvsc ¼ Vd Id þ Vq Iq , Qvsc ¼ Vq Id Vd Iq ð6:30Þ
At steady state, the PLL phase-tracking error is zero. Hence, Vq0 ¼ 0. From
(6.30), Qvsc0 ¼ Vd0Iq0. Usually, the PMSG operates with a high or even unity
power factor such that Qvsc0 ¼ Vd0Iq0 0. Hence, Iq0 0. The linearization of
(6.30) is,
Vd αf
Voltage Forward Feedback
s+α f
Vdcref – I dref+ + + Vcd
ref
K dc ( s ) K id ( s )
+ – –
Vdc Id X f Iq
Vq αf PWM
Voltage Forward Feedback
s+α f
Q pref + I qref + +
+ Vcqref
K q ( s) K iq ( s )
– – +
Qp Iq
X f Id
Fig. 6.28 Configuration of the GSC vector control system
ΔPvsc ¼ Id0 ΔVd þ Vd0 ΔId , ΔQvsc ¼ Id0 ΔVq Vd0 ΔIq ð6:31Þ
The linearized dynamic equation of the DC voltage across the GSC capacitor in
Fig. 6.1 is, (Ppmsg is not affected by the dynamic interactions between the PMSG and
the power system; thus, it can be assumed that ΔPpmsg ¼ 0),
dΔVdc
CVdc0 ¼ ΔPvsc ð6:32Þ
dt
From Fig. 6.1 (see 6.17),
dΔId
Xf ¼ ω0 ðΔVcd ΔVd Þ þ ω0 X f ΔIq
dt
ð6:33Þ
dΔIq
Xf ¼ ω0 ΔVcq ΔVq ω0 X f ΔId
dt
Figure 6.28 shows the configuration of the current vector control system of the
GSC (see Fig. 6.13). From Fig. 6.28, (6.32), and (6.33), the following equations are
obtained:
sω0
ΔId ¼ Gd ðsÞΔVdc ΔVd ,
ðs þ α f Þ½X f s þ ω0 Kid ðsÞ
sω0 ð6:34Þ
ΔIq ¼ Gq ðsÞΔQvsc ΔVq ,
ðs þ α f Þ X f s þ ω0 Kiq ðsÞ
where,
ω0 Kdc ðsÞKid ðsÞ ω0 Kq ðsÞKiq ðsÞ
G d ðsÞ ¼ , G q ðsÞ ¼
X f s þ ω0 Kid ðsÞ X f s þ ω0 Kiq ðsÞ
From (6.31), (6.32), (6.33), and (6.34),

CVdc0 Id0 s
ΔPvsc ¼
CVdc0 s þ Vd0 Gd ðsÞ

CVd0 ω0 Vdc0 s2
ΔVd ,
ðs þ α f Þ½CVdc0 s þ Vd0 Gd ðsÞ½X f s þ ω0 Kid ðsÞ

ð6:35Þ
Id0
ΔQvsc ¼
1 Vd0 Gq ðsÞ

ω0 Vd0 s
ΔVq
ðs þ α f Þ 1 Vd0 Gq ðsÞ X f s þ ω0 Kiq ðsÞ
Substituting (6.2) in (6.35),

CVdc0 Id0 s
ΔPvsc ¼

CVd0 ω0 Vdc0 s2
ΔVpcc ,

ð6:36Þ
CVdc0 Vd0 Id0 s
ΔQvsc ¼

CV2d0 ω0 Vdc0 s2
Δθerror
The derived equations of (6.36) indicate that the phase-tracking error of the PLL
mainly exhibits as dynamic variations of the reactive power output of the PMSG,
ΔQvsc. Hence, in the simulation results in Fig. 6.26, considerable variations in the
reactive power output of the first PMSG are observed.
6.3.3 Test 3: Variation of SCR and Wind Power Penetration

(SRF PLL for PMSG1)
It is reviewed and introduced in Sect. 1.1 of Chap. 1 that a short circuit ratio (SCR)
measures the strength of the connection between a VSC-based system, such as the
PMSG, and the power system. Normally, if the SCR is lower than 3.0, the connec-
tion is considered weak. Recent study has indicated that, when the grid connection of
the VSC-based system is weak, the PLL may degrade its stability. The following test
examined the effect of the SCR on the OLMR.
This test follows test 2 in the previous subsection, when the OLMR occurred
between the POM, λpll and the inter-area EOM of the power system, λ9. In the test,
the impedance of the transmission line connected PMSG1 and node 22 in the NEPS,
6
λˆpll
5
4 λpll – Rgi Rpll λpll

Imaginary axis
λˆ9
3
λ9
2
SCR varied from 10 to 2.5
1
λ9 + Rgi Rpll
0
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 0.1
Real axis
Fig. 6.29 Open-loop modal resonance affected by SCR variation
0.25 0.2
SCR =10 SCR =10
SCR =2.5 SCR =2.5
0.2 0.15
0.15
0.1
0.1
0.05
0.05
0 0
Δθpll Δxpll 1 2 3 4 5 6 7 8 9 10
State variables of the PLL Generator number
Fig. 6.30 PF computational results, when SCR ¼ 2.5 and SCR ¼ 10
XL, was varied such that the SCR varied between SCR ¼ 10and SCR ¼ 2.5.
Figure 6.29 shows the trajectories of the closed-loop POM, ^λ pll and the inter-area
EOM, ^λ 9 , on the complex plane; the arrows indicate the direction of the SCR
decrease. The trajectories of the estimated positionspof ^λ pll and ^λ 9 , using (6.15),
ffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffi
are also displayed. They are indicated as λpll þ Rgi Rpll and λ9 Rgi Rpll ,
respectively, in Fig. 6.29.
From Fig. 6.29, it can be seen that the OLMR degraded the damping of the inter-
area EOM for both a weak (SCR ¼ 2.5) and strong (SCR ¼ 10) grid connection of
PMSG1. In both the cases, the dynamic interactions between PMSG1 and the power
system were strong, as indicated by the PF computational results in Fig. 6.30. The
weaker the grid connection, the greater was the damping degradation of the inter-
area EOM caused by the OLMR.
From (6.30) and the discussions following it in the previous subsection, with
Iq0 0, Pvsc0 Vd0Id0. Hence, when the wind power penetration, i.e., Pvsc0,
increased, Id0 increased accordingly. From (6.36) and the discussions following it,
it can be seen that, when the OLMR occurred at a higher level of wind power
penetration, the variations of ΔQvsc were more. This implied that at a higher level of
wind power penetration, the OLMR may cause greater damping degradation of the
power system EOMs. The following test evaluated the impact of the OLMR, when
wind power penetration increased.
This test followed test 2 in Sect. 6.3.2, when the OLMR occurred between λpll and
λ9. The active power output from PMSG1 increased from 0.5 to 10 p. u. The
trajectories of the closed-loop POM, ^λ pll and the inter-area EOM, ^λ 9 , on the complex
plane are displayed in Fig. 6.31; the arrows indicate the direction of the increase in
the PMSG1 active power output. In Fig. 6.31, the trajectories of p theffiffiffiffiffiffiffiffiffiffiffiffiffi
estimated
positions of ^λ pll and ^λ 9 , using (6.15), are also displayed as λpll þ Rgi Rpll and
λ9 Rgi Rpll , respectively. The PF computational results are displayed in
Fig. 6.32.
From Figs. 6.31 and 6.32, it can be seen that, when the wind power penetration
from PMSG1 increased, the dynamic interactions between the PLL and the rest of
the power system increased and the OLMR caused a greater damping degradation of
the inter-area EOM.
5
λˆ pll
4.5
4 λpll
Imaginary axis
3.5 λpll – Rgi Rpll λˆ 9

λ9
3
The active power
2.5 output of the PMSG λ9 + Rgi Rpll
increased from 0.5 to 10
2
-0.5 -0.45 -0.4 -0.35 -0.3 -0.25 -0.2 -0.15 -0.1
Real axis
Fig. 6.31 The OLMR when the wind power penetration from PMSG1 increased
0.25 0.2
Ppmsg 0 =0.5
Ppmsg 0 =0.5
0.2 Ppmsg 0 =10
0.15
0.15 Ppmsg 0 =10
0.1
0.1
0.05 0.05
0 0
Δθpll Δxvq-pll 1 2 3 4 5 6 7 8 9 10
Fig. 6.32 PF Computational results, when the wind power penetration from PMSG1 increased
6.3.4 Test 4: DSOGI PLL for PMSG
In the above three subsections, the SRF PLL for PMSG1 was examined. In this
subsection, the DSOGI PLL for PMSG2 was considered for demonstrating and
evaluating the OLMR. The configuration of the DSOGI PLL is shown in
Fig. 6.20. Differing from the SRF PLL, the DSOGI PLL can track the phase angle
of the positive sequence voltage more accurately.
The open-loop oscillation modes of the DSOGI PLL were not in the frequency
range of the electromechanical oscillations of the NEPS, when the parameters of the
DSOGI PLL given in [12] were adopted. However, this does not imply that there is
no possibility of modal resonance of the DSOGI PLL with the power system EOMs,
in practice. For example, due to the existence of modeling errors, and variations in
the power system and PMSG operating conditions, the PLL parameters often need to
be tuned in the field. Tuning is generally a trial-and-error procedure, and guidance on
the tuning is useful for avoiding detrimental occurrences. Hence, by varying the PI
gains of the DSOGI PLL in a wide range to scan the potential dangers posed by
open-loop modal resonance, on one hand, the analysis about the OLMR can be
demonstrated and evaluated for the DSOGI PLL; on the other hand, the scanning
results provide guidance for tuning the PI gains, as they approximately indicate the
PI gain ranges that should be avoided.
Therefore, in this test, the PI gains of the DSOGI PLL were varied for scanning, in
the frequency range of the NEPS EOMs. The scanning results of the DSOGI PL,
similar to Fig. 6.23 for the SRF PLL, are shown in Fig. 6.33; it can be seen that, when
the open-loop POM of the DSOGI PLL, indicated by circles, was in the neighbor-
hood of the open-loop EOMs, indicated by the squares, the OLMR occurred. The
corresponding closed-loop POM and EOMs were located approximately on the
opposite side of the open-loop oscillation modes. The predicted positions of the
closed-loop POM and EOMs were calculated using (6.15) and indicated by the
crosses. Obviously, the evaluation results of the OLMR for the DSOGI PLL depicted
in Fig. 6.33 are as same as those for the SRF PLL depicted in Fig. 6.23.
7
Imaginary axis
5
λ9
4
λ̂pll λˆ9
3
λpll
2
-0.6 -0.5 -0.4 -0.3 -0.2 -0.1
Real axis
Fig. 6.33 Modal resonance caused by DSOGI PLL parameter tuning
a b
0.25 0.16
0.14
0.2
0.12
0.15 0.1
0.08
0.1 0.06
0.04
0.05
0.02
0 0
Δθpll Δxvq-pll 1 2 3 4 5 6 7 8 9 10
Fig. 6.34 PF Computational results for the DSOGI PLL. (a) Participation of the PMSG in ^λ 9 ; (b)
participation of the SGs in ^λ pll
For the OLMR between λpll and λ9, the PF computational results for ^λ pll and ^λ 9
are presented in Fig. 6.34. Strong interactions between the DSOGI PLL and the SGs
were confirmed.
6.3.5 Test 5: The Case of DFIG
In the NEPS shown by Fig. 6.19, a wind farm represented by a DFIG, instead of two
PMSGs, was connected to bus 22. A detailed 15th-order dynamic model of the DFIG
(see Chap. 3 for the model of the DFIG) was used that included the dynamics of the
RSC and GSC control systems (seventh-order), the rotor motion (first-order), a SRF
PLL (second-order), the induction generator (second-order), a grid-side converter
impedance, and a capacitor (third-order). The dynamics associated with the pitch
control and wind variations were not considered.
The multivariable closed-loop state-space model shown by Fig. 5.27 for the
NEPS with the DFIG connected to bus 22 was established with the dynamics of
the PLL being included. The DFIG adopted the configuration of the RSC control
system shown in Fig. 5.28 and the active power output of the DFIG was 10 p.u.
Typical PI gains recommended in SIMULINK for the PLL were used with
Kpp ¼ 4.3929, Kpi ¼ 39.2712.
First, from the open-loop state matrix, Ad, in (5.33), a DFIG oscillation mode
(DOM) was calculated as λd ¼ 2.1974 þ j5.8687. It was identified to be
associated with the SRF PLL from the computation of the participation factors of
the state variables of the DFIG. From the open-loop state matrix of the power
system, Ag, in (5.34), an open-loop EOM was calculated to be
λi ¼ 2.1005 þ j5.9133. It was identified to be a local EOM associated with
SG4 and SG5 in the NEPS. The OLMR occurred between λd ¼ 2.1974 þ j5.8687
and λi ¼ 2.1005 þ j5.9133, as shown in Fig. 6.35, where ^λ i and ^λ d are the closed-
loop local EOM and DOM corresponding to λi and λd, respectively. From Fig. 6.35,
it can be seen that the OLMR caused a considerable decrease in the damping of the
local EOM, when the DFIG was connected at node 22. The computational results of
the participation factors are presented in Fig. 6.36, confirming significant dynamic
interactions between SG5 and the DFIG.
6.4
6.3 λ̂i
6.2
Imaginary axis
6.1
6
ˆλd
5.9 λi
λd
5.8
5.7
-2.8 -2.6 -2.4 -2.2 -2 -1.8 -1.6 -1.4
Real axis
Fig. 6.35 The OLMR between the local EOM and the PLL
a b
0.9 0.4
0.8 0.35
0.7 0.3
0.6
0.25
0.5
0.2
0.4
0.15
0.3
0.1
0.2
0.1 0.05
0 0
1 2 3 4 5 6 7 8 9 10 Δxp Δφp
Genereator number State variables
Fig. 6.36 Participation factors of the DFIG and SGs. (a) Participation of SGs in the DOM; (b)
participation of the PLL in the local EOM
5.5
lˆd
5
4.5
Imaginary axis
ld li
4
3.5
lî
3
2.5
-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1
Real axis
Fig. 6.37 The OLMR between the inter-area EOM and PLL
Next, the PI gains of the PLL were decreased to move λd away from λi for
avoiding the OLMR. However, when the PI parameters were changed to
Kpp ¼ 0.6409, Kpi ¼ 13.1947, λd moved close to the inter-area EOM of the
NEPS. The OLMR between the PLL and the inter-area EOM occurred, as shown
in Fig. 6.37. Consequently, the damping of the inter-area EOM decreased consider-
ably. The OLMR between the PLL and the inter-area EOM was confirmed by the
computational results of the participation factors shown in Fig. 6.38. Further con-
firmation from the non-linear simulation is presented in Fig. 6.39.
a b
0.14 0.24
0.22
0.12
0.2
0.1 0.18
0.16
0.08 0.14
0.12
0.06 0.1
0.08
0.04
0.06
0.02 0.04
0.02
0 0
1 2 3 4 5 6 7 8 9 10 Δxpll Δθpll
Generator number State variables of the PLL
Fig. 6.38 Participation factors of the DFIG and SGs. (a) Participation of SGs in the DOM; (b)
participation of the PLL in the inter-area EOM
-2.5
PLL was of typical parameters
-3 PLL parameters were decreased
-3.5
-4
d10 – d1
-4.5
-5
-5.5
-6
-6.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time(s)
Fig. 6.39 Non-linear simulation, when the OLMR occurred between the DFIG and the inter-area
EOM, caused by the PLL
Appendix 6.1: Data of Example 6.1 (Tables 6.5, 6.6 and 6.7)
Table 6.5 Basic data of Xpd Xpq ψpm ωpref Jpr Vpdc Xpf Cp
PMSG [3, 6]
0.2 0.2 1 0.8 8s 1 0.02 30
Table 6.6 PI gains of MSC Kpp1 Kpi1 Kpp2 Kpi2 Kpp3 Kpi3
of PMSG [3, 6]
5 20 1 100 1 100
Table 6.7 PI gains of GSC Kpp4 Kpi4 Kpp5 Kpi5 Kpp6 Kpi6 Kpp7 Kpi7
of PMSG [3, 6]
2 20 0.5 100 1 10 0.5 100
Table 6.8 Basic data of Xpd Xpq ψpm ωpref Jpr Vpdc Xpf Cp
PMSG
0.25 0.15 1.1 0.8 8s 1 0.05 30
Table 6.9 PI gains of MSC Kpp1 Kpi1 Kpp2 Kpi2 Kpp3 Kpi3
of PMSG
5 20 1 100 1 100
Table 6.10 PI gains of GSC Kpp4 Kpi4 Kpp5 Kpi5 Kpp6 Kpi6 Kpp7 Kpi7
of PMSG
0.8 20 0.15 84.9 0.28 12 0.15 84.9
Data of PMSG1
They are given in Appendix 6.1.
Line Data
Xt2 ¼ 0.02, Xt3 ¼ 0.02, XL ¼ 0.02.
Data of New England Power System
Data of PMSG (Tables 6.8, 6.9 and 6.10)
Data of SRF PLL

Kppll ¼ 4.4,Kipll ¼ 39.27
References 241
References
1. Wang YJ, Hu JB, Zhang DL, Ye C, Li Q (2015) DFIG WT Electromechanical transient

behavior influenced by PLL: modelling and analysis. In: International conference of renewable
power generation, Beijing, pp 1–5
2. Wang ZW, Shen C, Liu F (2014) Impact of DFIG with phase lock loop dynamics on power
systems small signal stability. In: IEEE PES general meeting, National Harbor, MD, pp 1–5
3. Li S, Haskew TA, Swatloski RP, Gathings W (2012) Optimal and direct-current vector control
of direct-driven PMSG wind turbines. IEEE Trans Power Electron 27(5):2325–2337
4. Kaura V, Blasko V (1997) Operation of a phase locked loop system under distorted utility
conditions. IEEE Trans Ind Appl 33(1):58–63
5. Chung SK (2000) A phase tracking system for three phase utility interface inverters. IEEE Trans
Power Electron 15(3):431–438
6. Shi L, Crow ML (2008) A novel PLL system based on adaptive resonant filter. In: 40th North
American Power Symposium, Calgary, AB, pp 1–8
7. Timbus A, Liserre M, Teodorescu R, Blaabjerg F (2005) Synchronization methods for three
phase distributed power generation systems—An overview and evaluation. In: IEEE 36th
power electronics specialists conference, Recife, pp 2474–2481
8. Kesler M, Ozdemir E (2011) Synchronous-reference-frame-based control method for UPQC
under unbalanced and distorted load conditions. IEEE Trans Ind Electron 58(9):3967–3975
11. Wang HF, Du WJ (2016) Analysis and damping control of power system low-frequency
oscillations. Springer, New York
12. Golestan S, Monfared M, Freijedo FD (2013) Design-oriented study of advanced synchronous
reference frame phase-locked loops. IEEE Trans Power Electron 28(2):765–778
13. MATLAB Simulink wind farm-DFIG detailed model. https://uk.mathworks.com/help/
physmod/sps/examples/wind-farm-dfig-detailed-model.html
14. Autom R, Co AB (1997) Operation of a phase locked loop system under distorted utility
conditions. IEEE Trans Ind Appl 33(1):58–63
Chapter 7
Small-Signal Stability of a Power System
Integrated with an MTDC Network
for the Wind Power Transmission
As it is introduced in Chap. 1, small-signal stability of an AC power system

integrated with a multi-terminal DC (MTDC) network for the wind power transmis-
sion is determined by the dynamic interactions between the VSCs and synchronous
generators (SGs). The dynamic interactions are through both the MTDC network
and AC grid. Hence, the small-signal stability of an MTDC/AC power system is a
complicated issue, which is addressed in this chapter by focusing on the following
three particular aspects.
The first is the small-signal angular stability of an AC power system affected by
the dynamic interactions brought about by the MTDC network. The MTDC network
is integrated with the AC power system by VSC control. Control speed of the VSCs
is much faster than the electromechanical transient of the AC power system. Hence,
it has been recognized that dynamic interactions between the MTDC network and
the AC power system are normally weak. Usual inertia-less response of the VSCs
implies very small impact of the MTDC’s dynamics on the small-signal angular
stability of the AC power system. Can the small-signal stability of a MTDC/AC
power system be simply examined by modelling the MTDC network as constant
power injections into the AC power system? Obviously, if there is a possibility that
the dynamic interactions between the MTDC and AC system are strong under a
special condition, modelling the MTDC as constant power injections may possibly
miss out the potential threat imposed by the strong dynamic interactions on power
system small-signal angular stability. In the first subsection of this chapter, the
damping torque analysis (DTA) is applied to examine the special condition—the
open-loop modal resonance (OLMR), under which the MTDC network may intro-
duce strong dynamic interactions with the AC power system to affect the small-
signal angular stability of the MTDC/AC power system considerably. The OLMR
has been studied in Chap. 5 and 6 for the VSWGs and PLL respectively.
The second aspect is the small-signal stability of an independent MTDC network
affected by the dynamic interactions between the VSCs via the MTDC network.
When this particular aspect of problem is examined, the dynamic interactions
between the SGs through the AC grid are not considered. The examination can be

244 7 Small-Signal Stability of a Power System Integrated with an MTDC. . .
carried out for an MTDC power system, where each of the VSCs is connected to a
SG and there is no AC grid connecting all the SGs. In the second subsection of this
chapter, this aspect of small-signal stability problem is investigated.
The third aspect is the impact of a particular type of AC/DC dynamic interactions
between the VSCs and SGs—the dynamic interactions introduced by a selected
control system of a VSC in the MTDC/AC power system. Selection of the control
system of the VSC enables the examination to be focused and simplified. The small-
signal stability of MTDC/AC power system as affected by this particular type of
AC/DC dynamic interactions is studied in the third subsection of the chapter.
7.1 Small-Signal Angular Stability of an AC Power System

Affected by the Integration of an MTDC Network
for the Wind Power Transmission
7.1.1 The Open-Loop Modal Resonance (OLMR)
in the MTDC/AC Power System
7.1.1.1 Linearized Model of AC Power System

VSC-based MTDC network for wind power transmission. In order to examine the
small-signal angular stability of the AC power system affected by the dynamic
interactions introduced by the MTDC network, a closed-loop linearized model of
the MTDC/AC power system is established in this subsection.
P1 + jQ1
Wind farms V1 Vg1
I MTDC1 I dc1
X f1
Vdc1 I d 1 + jI q1
C1 VSC-1 SG-1
M
...
Pj + jQ j
T I MTDCj I dcj X fj
Vj A Vgk
D C
Cj Vdcj I dj + jI qj
C VSC-j Network SG-k
...
Network PM + jQM
VM Vgn
I MTDCM I dcM X fM
CM VdcM I dM + jI qM
VSC-M SG-n
Fig. 7.1 An AC power system integrated with a MTDC network for the wind power transmission
7.1 Small-Signal Angular Stability of an AC Power System Affected by the. . . 245
Denote Pj þ jQj, j ¼ 1, 2, . . .M as the complex power output from the jth VSC to
the AC power system. From Fig. 7.1, the active and reactive power injected from the
jth VSC to the power system can be obtained to be
P j ¼ Vxj Ixj þ Vyj Iyj , Q j ¼ Vyj Ixj Vxj Iyj , j ¼ 1, 2, . . . M ð7:1Þ
where subscript x and y denotes the x and y component of a phasor under system
common reference coordinate respectively; Vj ¼ Vxj þ jVyj, j ¼ 1, 2, . . .M and
Ij ¼ Ixj þ jIyj, j ¼ 1, 2, . . .M are the terminal voltage of and the output current from
the jth VSC respectively. Linearization of (7.1) is
ΔP j ¼ Ixj0 ΔVxj þ Iyj0 ΔVyj þ Vxj0 ΔIxj þ Vyj0 ΔIyj
ð7:2Þ
ΔQ j ¼ Iyj0 ΔVyj þ Ixj0 ΔVyj þ Vyj0 ΔIxj Vxj0 ΔIyj
where subscript 0 denotes the value of a variable at the steady state. In matrix form,
the above equations are
" #
ΔP
¼ F1 j ΔI þ F2 j ΔV ð7:3Þ
ΔQ
where
ΔP ¼ ½ΔP1 ΔP2 ΔPM T , ΔQ ¼ ½ ΔQ1 ΔQ2 ΔQM T ,

ΔI ¼ ½ ΔI1 ΔI2 ΔIM T , ΔV ¼ ½ ΔV1 ΔV2 ΔVM T ,
ΔV j ¼ ½ ΔVxj ΔVyj T , ΔI j ¼ ½ ΔIxj ΔIyj T
Following linearized network equation with the PCC nodes of the VSCs and
nodes of the synchronous generators (SGs) being kept can be established [1]
" # " #" #
ΔIg Ygg Ygj ΔVg
¼ ð7:4Þ
ΔI Y jg Y jj ΔV
where
T T
ΔVg ¼ ΔVg1 T ΔVgn T , ΔIg ¼ ΔIg1 T ΔIgn T ,
ΔVgj ¼ ½ ΔVgxj ΔVgyj T , ΔIgj ¼ ½ ΔIgxj ΔIgyj T
Vgxj þ jVgyj and Igxj þ jIgyj is the terminal voltage and output current of the jth SG
in the power system respectively. The following linearized model of the SGs can be
established [1].
sΔXAC ¼Ag ΔXAC þBg ΔVg , ΔIg ¼Cg ΔXAC þDg ΔVg ð7:5Þ
Substituting the second equation of (7.4) into (7.3) gives

" #
2 1 ΔP 21
ΔV¼ F2 j þF1 j Y jj F2 j þF1 j Y jj F1 j Y jg ΔVg ð7:6Þ
ΔQ
From the second equation of (7.5) and the first equation of (7.4), it can have

Ygg Dg ΔVg þYgj ΔV¼Cg ΔXAC ð7:7Þ
Substituting (7.6) into (7.7)

" #
ΔP
ΔVg ¼F3 ΔXAC þF4 ð7:8Þ
ΔQ
where
h 1 i1
F3 ¼ Ygg Dg Ygj F2 j þF1 j Y jj F1 j Y jg Cg
h 1 i1 1
F4 ¼ Ygg Dg Ygj F2 j þF1 j Y jj F1 j Y jg Y jg F2 j þF1 j Y jj
Substituting (7.8) into the first equation of (7.5), following state equations are
obtained.
d
ΔXAC ¼ AAC ΔXAC þ BAC ΔU ð7:9Þ
dt
where
ΔU¼½ ΔP ΔQ ,
ΔP ¼ ½ ΔP1 ΔP2 ΔPM T , ΔQ ¼ ½ ΔQ1 ΔQ2 ΔQM T ,

" #
ΔP
ΔV¼CAC ΔXAC þDAC ¼CAC ΔXAC þDAC ΔU ð7:10Þ
ΔQ
where
1 1
CAC ¼ F2 j þF1 j Y jj F1 j Y jg F3 , DAC ¼ F2 j þF1 j Y jj
1
F2 j þF1 j Y jj F1 j Y jg F4
Writing (7.9) and (7.10) together, following state-space model of the AC power
system is established:
d
ΔXAC ¼ AAC ΔXAC þ BAC ΔU
dt ð7:11Þ
ΔV ¼ CAC ΔXAC þ DAC ΔU
where ΔXAC is the vector of all the state variables of the AC power system.
From (7.11), the transfer function matrix model of the AC power system can be
obtained to be
GðsÞ¼CAC ðsI AAC Þ1 BAC þDAC ð7:12Þ
Hence,
ΔV¼GðsÞΔU ð7:13Þ
7.1.1.2 Linearized Model of the MTDC Network
The pair of input-output signal vectors of the MTDC network are ΔV ΔU. Thus,
following transfer function matrix model of the MTDC network can be derived.
ΔU¼HðsÞΔV ð7:14Þ
The detailed derivation of (7.14) is as follows.

Assume that L VSCs are connected to L wind farms and M VSCs are connected
to the AC power system in the MTDC network of Fig. 7.1. For a VSC being
connected to a wind farm, active wind power output from the wind farm is
Pwj ¼ VdcwjIdcwj, j ¼ 1, 2, . . .L. Since the wind power output is only determined
by the factors, such as wind speed, which are irrelevant with the disturbances on the
side of AC power system, it can be assumed to be a constant. Thus, for the jth VSC
being connected to the wind farm, it can have
ΔPwj ¼ Vdcwj0 ΔIdcwj þ Idcwj0 ΔVdcwj ¼ 0, j ¼ 1, 2, . . . L ð7:15Þ
Following network equation for the MTDC network can be established
2 3 2 0 0 0
32 3 1st VSC
ΔIMTDC1 y11 ðsÞ y12 ðsÞ y13 ðsÞ ΔVdc1
6 7 6 0 76 7 ðM 1Þ VSCs
4 ΔIMTDC 5 ¼ 6 Y23 ðsÞ 7
0 0
4 y21 ðsÞ Y22 ðsÞ 54 ΔVdc 5 VSCs being
0 0 0
ΔIdcw y31 ðsÞ Y32 ðsÞ Y33 ðsÞ ΔVdcw
connectedto wind farms
ð7:16Þ
where
ΔIMTDC ¼ ½ ΔIMTDC2 ΔIMTDC3 ΔIMTDCM T , ΔVdc ¼ ½ ΔVdc2 ΔVdc3 ΔVdcM T
In the following derivation, Laplace operator s in (7.16) is omitted for the

simplicity of expressions. The jth element of ΔIdcw and ΔVdcw is ΔIdcwj and ΔVdcwj
as given in (7.15). Using (7.15) to delete the last row in (7.16), it can have
" # " #" #

ΔIMTDC1 y11 y12 ΔVdc1
¼ ð7:17Þ
ΔIMTDC y21 Y22 ΔVdc
For the kth VSC being connected with the AC power system, it can have
dΔIdk
X fk ¼ ω0 ðΔVcdk ΔVdk Þ þ ω0 X fk ΔIqk
dt
ð7:18Þ
dΔIqk
X fk ¼ ω0 ΔVcqk ΔVqk ω0 X fk ΔIdk
dt
where Xfk is the reactance of the output filter of the kth VSC, ω0 is the synchronous
frequency.
Without the loss of generality, assume that the first VSC uses the DC voltage
control and the rest of VSCs implement the active power control. Figures 7.2 and 7.3
show the configuration of the vector control system of the kth VSC being connected
to the AC system. Ignoring the very fast transient of the PWM algorithm,
Vcdk ¼ Vcdk
ref
, Vcqk ¼ Vcqk
ref
, k ¼ 1, 2, . . . M. Express the transfer function of various
PI controllers in Figs. 7.2 and 7.3 as
Kdci Kqi1
Kdc ðsÞ ¼ Kdcp þ , Kq ðsÞ ¼ Kqp1 þ ,
s s
Kidi1 Kiqi1
Kid ðsÞ ¼ Kidp1 þ , Kiq ðsÞ ¼ Kiqp1 þ
s s
ð7:19Þ
Kpik Kqik
Kpk ðsÞ ¼ Kppk þ , Kqk ðsÞ ¼ Kqpk þ ,
s s
Kidik Kiqik
Kidk ðsÞ ¼ Kidpk þ , Kiqk ðsÞ ¼ Kiqpk þ
s s
DC voltage control outer loop Current control inner loop

DC voltage control loop Vd1
ref
+ + Vcd1
ref ref
V dc1 + I d1 +
Kdc (s ) Kid (s )
– – +
Vdc1 Id1 X f1 Iq1
PWM
Reactive power control loop Vq1
ref
+ + Vcq1
ref
Q1ref + I q1 +
K q (s ) K iq (s )
– – +
Q1 Iq1 X f1 Id1
Fig. 7.2 Converter control system of the first VSC with DC voltage control
Power control outer loop Current control inner loop

Active power control loop
Vdk
ref
+ + Vcdk
ref ref
P +
k I dk +
K pk (s ) Kidk (s )
– – –
Pk I dk X fk Iqk
PWM
Reactive power control loop
Vqk ref
Qkref + I qkref + + + Vcqk
K qk (s ) Kiqk (s )
– – +
Qk I qk X fk Idk
Fig. 7.3 Converter control system of the kth VSC with active power control
From Figs. 7.2 and 7.3, (7.18) and (7.19),

Kdcp s þ Kdci
ΔId1
ref
¼ ΔVdc1
s
Kppk s þ Kpik
ΔIdk
ref
¼ ΔPk , k ¼ 2, 3, . . . M
s
Kqpk s þ Kqik
ΔIqk
ref
¼ ΔQk , k ¼ 1, 2, . . . M ð7:20Þ
s
ω0 Kidpk s þ Kidik ω0
ΔIdk ¼ ΔI ref , k ¼ 1, 2, . . . M
X fk s2 þ ω0 Kidpk s þ Kidik ω0 dk
ω0 Kiqpk s þ Kiqik ω0
ΔIdk ¼ ΔI ref , k ¼ 1, 2, . . . M
X fk s2 þ ω0 Kiqpk s þ Kiqik ω0 qk
Matrix form of (7.20) is

ΔId ¼ diagGpk ðsÞΔP ¼ Gp ðsÞΔP
ð7:21Þ
ΔIq ¼ diag Gqk ðsÞ ΔQ ¼ Gq ðsÞΔQ
where
T
ΔId ¼ ½ ΔId2 ΔId3 ΔIdM T , ΔIq ¼ ½ ΔIq1 ΔIq2 ΔIqM ,
T T
ΔP ¼ ½ ΔP2 ΔP3 ΔPM , ΔQ ¼ ½ ΔQ1 ΔQ2 ΔQM
Take the direction of PCC voltage, Vk, k ¼ 1, 2, . . .M, as that of d axis of the
d q coordinate of the kth VSC. It can have Vdk ¼ Vk, Vqk ¼ 0, k ¼ 1, 2, . . .M.
Hence linearized active and reactive power output from the VSCs can be written as
ΔP1 ¼ Id10 ΔV1 þ V10 ΔId1 ¼ Vdc10 ΔIdc1 þ Idc10 ΔVdc1
ΔP ¼ Id0 ΔVk þV0 ΔId ¼Vdc0 ΔIDC þIdc0 ΔVdc ð7:22Þ
ΔQ ¼ Iq0 ΔV V0 ΔIq
where
ΔVk ¼ ½ ΔV2 ΔV3 ΔVM T , ΔV ¼ ½ ΔV1 ΔV2 ΔVM T ,

ΔIDC ¼ ½ ΔIdc2 ΔIdc3 ΔIdcM T , Id0 ¼ diag½Idk0 , V0 ¼ diag½Vk0 ,

Vdc0 ¼¼ diag½Vdck0 , Idc0 ¼ diag½Idck0 , Iq0 ¼ diag Iqk0
From (7.21) and (7.22),

1
ΔP¼ I V0 Gp ðsÞ Id0 ΔVk
1 ð7:23Þ
ΔQ¼ IþV0 Gq ðsÞ Iq0 ΔV
On the DC side of the VSC, the linearized model is
dΔVdc1
C1 þ ΔIdc1 ¼ ΔIMTDC1 ,
dt
ð7:24Þ
dΔVdc
C þ ΔIDC ¼ ΔIMTDC , C ¼ diag C j , j ¼ 2, 3, . . . M
dt
From (7.17) and (7.24),
ΔVdc ¼ ðsC Y22 Þ1 y21 ΔVdc1 ðsC Y22 Þ1 ΔIDC ð7:25Þ
From (7.21), (7.22) and (7.25),

1
ΔP ¼ I V0 Gp ðsÞ Id0 ΔVk ¼Vdc0 ΔIDC þ Idc0 ΔVdc
ð7:26Þ
¼Vdc0 ΔIDC þIdc0 ðsC Y22 Þ1 y21 ΔVdc1 Idc0 ðsC Y22 Þ1 ΔIDC
Thus
h i1 1
ΔIDC ¼ Vdc0 Idc0 ðsC Y22 Þ1 I V0 Gp ðsÞ Id0 ΔVk
h i1 ð7:27Þ
Vdc0 Idc0 ðsC Y22 Þ1 Idc0 ðsC Y22 Þ1 y21 ΔVdc1
From (7.17), (7.25) and (7.27),
ΔIdc1 ¼ y11 ΔVdc1 þ y12 ΔVdc

¼ y11 ΔVdc1 þ y12 ðsC Y22 Þ1 y21 ΔVdc1 y12 ðsC Y22 Þ1 ΔIDC
¼ y11 ΔVdc1 þ y12 ðsC Y22 Þ1 y21 ΔVdc1
h i1 1
y12 ðsC Y22 Þ1 Vdc0 Idc0 ðsC Y22 Þ1 I V0 Gp ðsÞ Id0 ΔVk
h i1
þy12 ðsC Y22 Þ1 Vdc0 Idc0 ðsC Y22 Þ1 Idc0 ðsC Y22 Þ1 y21 ΔVdc1
¼ y1 ðsÞΔVdc1 þ yðsÞΔVk
ð7:28Þ
From (7.23) and (7.28),

ΔIdc1 ¼ ½y1 ðsÞ C1 sΔVdc1 þ yðsÞΔVk ð7:29Þ
From (7.20), (7.22) and (7.28)

ΔP1 ¼ Id10 ΔV1 þ V10 ΔId1 ¼ Vdc10 ΔIdc1 þ Idc10 ΔVdc1
¼ Id10 ΔV1 þ V10 Gp1 ðsÞΔVdc1 ð7:30Þ
¼ Vdc10 ½y1 ðsÞ C1 sΔVdc1 þ Vdc10 yðsÞΔVk þ Idc10 ΔVdc1
From the above equation,

1
ΔVdcl ¼ V10 Gp1 ðsÞ Vdc10 y1 ðsÞ þ Vdc10 C1 S Idc10 Vdc10 yðsÞΔVk
1 ð7:31Þ
Id10 V10 Gp1 ðsÞ Vdc10 y‘ 1 ðsÞ þ Vdc10 C1 s Idc10 ΔV1
Substituting (7.31) into (7.30),

ΔP1 ¼ gp1 ðsÞΔV1 þ gpk ðsÞΔVk ð7:32Þ
where
gp1 ðsÞ ¼ fVdc10 ½y1 ðsÞ C1 s þ Idc10 gId10
1
V10 Gp1 ðsÞ Vdc10 y‘1 ðsÞ þ Vdc10 C1 s Idc10
gpk ðsÞ ¼ fVdc10 ½y1 ðsÞ C1 s þ Idc10 g
1
V10 Gp1 ðsÞ Vdc10 y1 ðsÞ þ Vdc10 C1 s Idc10 Vdc10 yðsÞ þ Vdc10 yðsÞ
Writing (7.23) and (7.32) together, the transfer function matrix model of the
MTDC network of (7.14) is obtained.
Let the state-space realization of the MTDC network’s transfer function matrix
model of (7.14) be
d
ΔXDC ¼ ADC ΔXDC þ BDC ΔV
dt ð7:33Þ
ΔU ¼ CDC ΔXDC þ DDC ΔV
where ΔXDC is the vector of all the state variables of the MTDC network and
HðsÞ¼CDC ðsI ADC Þ1 BDC þDDC ð7:34Þ
7.1.1.3 The Closed-Loop Model of the MTDC/AC Power System
From (7.13) and (7.14), the closed-loop model of the MTDC/AC power system is
obtained and shown by Fig. 7.4. The state-space model of the MTDC/AC power
system is obtained from (7.11) and (7.34) to be
Fig. 7.4 Closed-loop

model of the AC/MTDC G(s) AC power
power system ΔV system
ΔU MTDC
H(s) network
d
ΔX¼AΔX ð7:35Þ
dt
where
T
ΔX ¼ ΔXAC T ΔXDC T ,
" #
AAC þ BAC ðI DDC DAC Þ1 DDC CAC BAC ðI DDC DAC Þ1 CDC
A¼
BDC ðI DAC DDC Þ1 CAC ADC þ BDC ðI DAC DDC Þ1 DAC CDC
Dynamic interactions between the MTDC network and the AC power system
physically are the dynamic variations of complex power output from the MTDC,
ΔPj þ jΔQj, j ¼ 1, 2, . . .M. If ΔPj þ jΔQj ¼ 0, j ¼ 1, 2, . . .M, there are no dynamic
interactions between the MTDC network and the AC power system. In this case,
small-signal angular stability of the AC power system is not affected by dynamics of
the MTDC network at all. State-space model of the AC power system shown in
(7.11) is degraded to
d
ΔXAC ¼AAC ΔXAC ð7:36Þ
dt
Denote λACi as an electromechanical oscillation mode (EOM) of concern in the
open-loop AC power system when ΔPk þ jΔQk ¼ 0, k ¼ 1, 2, . . .M. λACi is a
complex eigenvalue of open-loop state matrix, AAC. Denote b λ ACi as the EOM
corresponding to λACi when ΔPk þ jΔQk 6¼ 0. b λ ACi is an eigenvalue of closed-
loop state matrix, A in (7.35). Obviously, difference between b λ ACi and λACi,
b
ΔλACi ¼ λ ACi λACi , is caused by the dynamic interactions between the MTDC
network and the AC power system. This difference is the impact of dynamic
interactions introduced by the MTDC network on the small-signal angular stability
of AC power system. Based on the closed-loop model, dynamic interactions between
the MTDC and the AC power system, ΔPk þ jΔQk, are linked with the difference of
closed-loop and open-loop EOM of concern, which is a measurement of the impact
of the dynamic interactions.
7.1.1.4 Damping Torque Analysis (DTA)
Re-write state equation in (7.11) as

2 3 2 32 3 2 3 2 3
Δδ 0 ω0 I 0 Δδ 0 0
7 X6 X6
M M
d6 7 6 76 7 7
4 Δω 5¼4 A21 A22 A23 54 Δω 5þ 4 bP2k 5ΔPk þ 4 bQ2k 5ΔQk
dt k¼1 k¼1
Δz A31 A32 A33 Δz bP3k bQ3k
ð7:37Þ
where Δδ and Δω are the vectors of deviations of angular positions and speeds of the
SGs respectively, Δz is the vector of all other state variables in the AC power system.
Express the transfer function matrix model of the MTDC network of (7.14) as
" # " #
ΔP H P ðsÞ
¼ HðsÞΔV ¼ ΔV ð7:38Þ
ΔQ HQ ðsÞ
From (7.37) and (7.38), the electric torque provided from the MTDC network via
M VSCs to the SGs is obtained
X
M
ΔTDC ¼ gpk ðsÞΔPk þ gqk ðsÞΔQk
k¼1
ð7:39Þ
X
M X
M X
M
¼ gpk ðsÞ hpkj ðsÞΔV j þ gqk ðsÞ hqkj ðsÞΔV j
k¼1 j¼1 j¼1
where, gpk(s) ¼ bP2k + A23(sI A33)1bP3k, gqk(s) ¼ bQ2k + A23(sI A33)1bQ3k,

hpkj(s) and hqkj(s), k, j ¼ 1, 2, . . .M are the elements of HP(s) and HQ(s) respectively.
It is proved in [2] that in the AC power system, following equation can always be
established:
ΔV j ¼ γ jm ðλACi ÞΔωm ð7:40Þ
From (7.39) and (7.40), damping torque contribution from the MTDC network to
the electromechanical oscillation loop of the mth SG is obtained to be
X
M X
M
ΔTdm ¼ Re gpkm ðλACi Þ hpkj ðλACi Þγ jm ðλACi Þ
k¼1 j¼1
ð7:41Þ
X
M X
M
þ gqkm ðλACi Þ hqkj ðλACi Þγ jm ðλACi Þ Δωm ¼ dm Δωm
k¼1 j¼1
where gpkm(s) and gqkm(s) is the mth element of gpk(s) and gqk(s) respectively, Re
[] denotes the real part of a complex number.
Denote Sim ¼ ∂λ ACi

∂dm
as the sensitivity of the EOM of concern to the damping torque
coefficient of the mth SG. The impact of dynamic interactions between the MTDC
network and the AC power system on the EOM of concern is obtained from (7.41)
to be
X
N
ΔλACi ¼ b
λ ACi λACi ¼ Sim dm ð7:42Þ
m¼1
where N is the total number of the SGs in the AC power system.

Denote λDCi as an oscillation mode of the MTDC subsystem. It is a complex
eigenvalue of open-loop state matrix of the MTDC subsystem, ADC in (7.33). λDCi
should be at least a complex pole of one element of transfer function matrix H(s),
hpkj(s) and hqkj(s), k, j ¼ 1, 2, . . .M, defined in (7.38). Without loss of generality,
assume λDCi is a complex pole of hp11(s); thus |hp11(λDCi)| ¼ 1. Open-loop modal
resonance (OLMR) is the special condition that open-loop oscillation mode of the
MTDC subsystem, λDCi, is closed to the EOM of concern of the open-loop AC
subsystem, λACi, on the complex plane, i.e., λDCi λACi. Obviously, when
λDCi λACi, |hp11(λACi)| is significantly large. From (7.38), it can be seen that ΔP1
may be large at the complex frequency λACi. This means that the first VSC may
respond considerably to disturbances in AC power system. Hence, strong dynamic
interactions between the MTDC and the AC power system may occur when the
OLMR, λDCi λACi, happens.
From (7.41), it can be seen that when λDCi λACi such that |hp11(λACi)| is
significantly large, it is possible that the damping torque contribution from the first
VSC to some of the SGs in the AC power system is significant. Consequently,
corresponding damping coefficients, dm, may be big; thus ΔλACi may not be small
any more. This explains why the OLMR may cause strong dynamic interactions
between the MTDC and the AC power system to affect the EOM of concern
considerably.
7.1.1.5 Impact of Strong Dynamic Interactions Under the Condition

of OLMR
Denote b λ DCi as the closed-loop oscillation mode corresponding to. It is a pair of

conjugate complex eigenvalues of closed-loop state matrix, A in (7.35). Similarly,
ΔλDCi ¼ b λ DCi λDCi is also caused by dynamic interactions between the MTDC
network and the AC power system. Hence, the relative positions of the closed-loop
modes, bλ ACi and b
λ DCi , in respect to the positions of the open-loop modes, λDCi λACi,
indicate how much the dynamic interactions between the MTDC network and the
AC power system affect the small-signal stability of MTDC/AC power system under
the condition of OLMR. Using the theory on modal resonance, the indication can be
further examined as follows.
a b
area A
area B lˆDCi
lDCi
area B
l ACi
lÂCi lÂCi
lDCi
l ACi
lˆDCi
area A
Real axis Real axis
Fig. 7.5 Illustration of near strong modal resonance around point of the OLMR
Modal resonance is the coincidence of eigenvalues of state matrix of a linear

system. There are two types of modal resonance mathematically defined, strong
modal resonance and weak modal resonance [3]. Only near strong modal resonance
may occur in power systems when certain system parameters vary [4–6]. It has been
mathematical proved that in the neighborhood of near strong modal resonance, two
resonant eigenvalues intend to move towards opposite directions [3–6]. Near strong
modal resonance of two power system EOMs with variations of some system
parameters was investigated in [4–6]. Results of investigation showed that two
EOMs moved close to each other and then towards opposite directions on the
complex plane. Subsequently, one EOM moved towards the right on the complex
plane to lead to damping degradation of power system low-frequency oscillations
[4–6].
For the MTDC/AC power system shown in Fig. 7.1, when the dynamic interac-
tions between the MTDC network and the AC power system are weak, b λ DCi is close
to λDCi and bλ ACi is close to λACi respectively on the complex plane. Hence, when λDCi
moves from a position away from λACi with variations of some parameters of open-
loop MTDC network, b λ DCi should move along with λDCi. Meanwhile, b λ ACi should
remain close to λACi owing to the weak dynamic interactions between the MTDC
network and the AC power system. This is illustrated by shadow areas A in Fig. 7.5a.
However, when λDCi moves close to the point of the OLMR, λDCi λACi, on the
complex plane, b λ DCi moves along with λDCi to get close to b λ ACi if dynamic
interactions between the MTDC network and the AC power system remain being
weak. Closeness of two eigenvalues of closed-loop state matrix, A in (7.35), b λ DCi and
b
λ ACi , implies that a near strong modal resonance is destined to happen in the
neighborhood of point of the OLMR, λDCi λACi. In this case, b λ ACi and b
λ DCi should
move towards the opposite directions away from the point of the OLMR as illus-
trated by circled areas B in Fig. 7.5a. This is the case that both ΔλACi ¼ b λ ACi λACi
and ΔλDCi ¼ b λ DCi λDCi are not small any more, indicating increased dynamic
interactions between the MTDC network and the AC power system. Hence, in the
neighborhood of point of the OLMR, λDCi λACi, strong dynamic interactions may
occur and their impact may be considerable. In addition, since b λ ACi and b
λ DCi move
away towards the opposite directions from the point of the OLMR, λDCi λACi, it is
λ ACi or b
very likely that either b λ DCi may locate on the right hand side of λDCi λACi,
indicating negative impact of dynamic interactions between the MTDC network and
the AC power system on the small-signal stability of the MTDC/AC power system.
On the other hand, when some parameters of open-loop AC subsystem change,
λACi moves from a position away from λDCi towards the point of the OLMR,
λDCi λACi. A similar analysis to that presented above can be illustrated by
Fig. 7.5b. It is likely that near strong modal resonance may occur in the neighbor-
hood of point of the OLMR, λDCi λACi. Consequently, the small-signal stability of
the MTDC/AC power system may possibly decrease under the condition of
the OLMR.
From the examination presented above and illustration in Fig. 7.5, it can be seen
that the point of the OLMR, in fact, indicates the location, around which the near
strong modal resonance of closed-loop MTDC/AC power system may occur. It can
be concluded that the impact of strong dynamic interactions caused by the OLMR
may very likely reduce the small-signal stability of the MTDC/AC power system.
7.1.2 Example MTDC/AC Power Systems
7.1.2.1 Example 7.1—New England Power System Integrated

with a MTDC Network for the Wind Power Transmission
The New England test system (NETS) has been used to study power system small-
signal angular stability in many occasions [7] and also as the example power systems
in the previous chapters. Figure 7.6 shows the configuration of modified NETS.
Modifications were the integration of a ten-terminal DC network for wind power
transmission. Five VSCs were connected with the wind farms. Each of wind farms
was modelled as a PMSG. Detailed dynamic model of the PMSGs and parameters
given in [8] were used. Other five VSCs were connected to the NETS. Parameters of
SGs, loads and network of the NETS given in [7] were used. Appendix 7.1 presents
the data of the example power system with the MTDC network.
There were in total nine EOMs in the NETS. How the EOMs were affected by the
integration of the MTDC was examined. An inter-area EOM with lowest frequency
is the mode of concern, λACi, in the examination. It was most lightly damped EOM.
Low-frequency oscillations related to λACi is between G10 and rest of the SGs.
Two types of MTDC control strategies were examined. The first one was the
master-slave control [9] with the VSC1 adopting DC voltage control and other VSCs
using active power control. The second type of control strategy was the DC voltage
droop control [9, 10] implemented by the VSC1 and VSC2. Other VSCs adopted
26
5 28 29
25 27 4
34 24 38
2 18 33
37 9
17 6
30 8 20 35
1 19 7
1 16 36
3
22
15
10
14 21 23
2
4 40 44 43 42 41
39 31 12
5 11 13
6 10 VSC-1 VSC-5 VSC-4 VSC-3 VSC-2
9 7
r MTDC NET r
32
8 3 WIND
NETS
FARMs
Fig. 7.6 Configuration of the NETS integrated with a ten-terminal DC network
active power control. Wind active power delivery to the NETS was shared equally
between VSC1 and VSC2. The droop coefficient was 0.1%. Parameters of the
ten-terminal DC network (MTDC) recommended in [9, 10], including control
parameters of the VSCs, were used.
7.1.2.2 Strong Dynamic Interactions When the Master-Slave Control

Was Implemented
First, total wind power penetration to the NETS via the MTDC network gradually
increased from Pw0 ¼ 10 p.u to a higher level, Pw0 ¼ 16 p.u. It was found that a
complex eigenvalue of open-loop state matrix of the MTDC subsystem, λDCi, moved
close to the EOM of concern, λACi, as shown by dashed curves in Fig. 7.7, where
arrows indicate the direction of increase of wind power penetration to the NETS via
the MTDC network. In Fig. 7.7, trajectories of complex eigenvalues of state matrix
of closed-loop MTDC/AC system corresponding to λACi and λDCi are shown by solid
curves. They are b λ ACi and b
λ DCi .
From Fig. 7.7, it can be seen that when Pw0 ¼ 10 p.u, λDCi and λACi position away
from each other on the complex plane. b λ DCi is close to λDCi and b
λ ACi is close to λACi
respectively. This was the case of weak dynamic interactions between the MTDC
and the NETS. With increase of wind power penetration, λDCi and λACi moved close
to each other on the complex plane. Around point A when Pw0 ¼ 16 p. u, λDCi λACi
Fig. 7.7 Modal trajectories Imaginary axis (rad/s)

when the OLMR occurred 3.8
(Master-slave control) Master-slave control
λDCi λÂCi
3.6
λ̂DCi
A
λACi
3.4
-0.5 -0.3 -0.1 0.1
Real axis
0.5 1.0
Pwo=10 pu Pwo=10 pu
Pwo=16 pu Pwo=16 pu
0.4 0.8
0.3 0.6
0.2 0.4
0.1 0.2
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5
Generator number VSC number Generator number VSC number
Fig. 7.8 Participation factors (PFs) for b

λ ACi and b
λ DCi (Master-slave control)
and bλ ACi moved away into the right half of complex plane, causing small-signal
angular instability of the NETS. From Fig. 7.7, it can be seen that when λDCi λACi,
ΔλACi ¼ b λ ACi λACi was not small, indicating significant dynamic interactions
between the MTDC and the NETS.
Computational results of participation factors (PFs) of the SGs and the VSCs for
b
λ ACi and bλ DCi are presented in Fig. 7.8 when Pw0 ¼ 10 p.u and Pw0 ¼ 16 p. u
respectively. From Fig. 7.8, it can be seen that when Pw0 ¼ 10 p.u (λDCi was away
from λACi on the complex plane), b λ ACi was mainly participated by the SGs and b
λ DCi
was dominantly participated by VSC1. This indicated weak dynamic interactions
between the MTDC network and the NETS. However, when Pw0 ¼ 16 p. u
(λDCi λACi), both b λ ACi and b
λ DCi were participated considerably by the SGs and
VSC1, confirming strong dynamic interactions between the MTDC network and the
Fig. 7.9 Non-linear 0.6

simulation (Master-slave Pwo=10 pu
Active power output from VSC1 (pu)

control) Pwo=16 pu
0.4
0.2
-0.2
-0.4
-0.6
-0.8
0 2 4 6 8 10
Time(s)
8
Pwo=10 pu
Pwo=16 pu
4
Dd1 – Dd10 (deg.)
-4
-8
0 2 4 6 8 10
Time(s)
NETS. That was why under the condition of the OLMR, λDCi λACi, the inter-area
EOM of concern was affected considerably by the dynamic interactions.
To further confirm the results of modal computation presented above, Fig. 7.9
shows the results of non-linear simulation. Simulation was carried out with a sudden
loss of 80% load at node 10 in the NETS at 0.1 s of simulation. The lost load was
recovered in 0.1 s. From Fig. 7.9, it can be seen that when Pw0 ¼ 10 p.u, dynamic
variations of active power output from the VSC1 were very small, indicating weak
dynamic interactions between the MTDC network and the NETS. The integrated
MTDC/NETS was stable. However, when the wind penetration increased to
Pw0 ¼ 16 p. u, dynamic variations of active power output from VSC1 were
Fig. 7.10 Modal Imaginary axis (rad/s)

trajectories to dismiss the 6
OLMR (Master-slave Master-slave control
control)
5.5
4.5
λDCi
4 λˆDCi λÂCi
λACi
3.5
-0.6 -0.5 -0.3 -0.1 0.1
Real axis
significant, indicating strong dynamic interactions. The integrated MTDC/NETS

became unstable.
Participation factors (PFs) of various state variables of VSC1 for b λ ACi were
calculated under the condition of the OLMR. It was found that the state variables
associated with the DC voltage control system of VSC1 participated in b λ ACi mostly.
Hence, in order to examine if the strong dynamic interactions may be dismissed by
tuning control parameters of the MTDC, parameters of the DC voltage controller of
VSC1 were changed from Kdcp ¼ 0.3, Kdci ¼ 2.4 to Kdcp ¼ 0.5, Kdci ¼ 20.
Trajectory of λDCi on the complex plane with parameters tuning of the DC voltage
controller of VSC1 are displayed in Fig. 7.10, when the wind power penetration was
Pw0 ¼ 16 p. u. In Fig. 7.10, hollow circles are the positions of open-loop EOMs of
the NETS and solid curves are the trajectories of the corresponding closed-loop
EOMs.
From Fig. 7.10, it can be seen that when parameters of the DC voltage controller
were tuned, λDCi moved away from the open-loop inter-area EOM of concern, λACi.
Closed-loop EOM of concern, b λ ACi , moved back towards the open-loop EOM of
concern, λACi. On the way that λDCi moved away from λACi, three closed-loop local
EOMs were affected when λDCi was close to the corresponding open-loop local
EOMs. It can be observed that the impact of dynamic interactions brought about by
the MTDC network on those three local EOMs of the NETS was not very significant,
though the closed-loop local EOMs were “driven” towards the right slightly by the
closeness of λDCi. Results shown in Fig. 7.10 confirmed that the damping degrada-
tion of inter-area EOM of concern was caused by the OLMR, λDCi λACi. In
addition, condition of the OLMR can be dismissed by tuning the parameters of
DC voltage controller of VSC1.
7.1.2.3 Numerical Evaluation of Connections Between the OLMR

and Near Strong Modal Resonance
Nearness of two closed-loop oscillation modes can be described by the mode-to-

parameter sensitivity: When variation of a parameter causes two closed-loop modes
moving towards each other, the mode-to-parameter sensitivity increases towards the
infinity [3]. As infinite mode-to-parameter sensitivity is not practical, two closed-
loop modes cannot be identical in any practical system, rather than get close to each
other, but never be in coincidence on the complex plane. Two closed-loop modes
getting close to each other will move away from each other at a point of their
proximity to each other. This is the point on the complex plane when the near strong
modal resonance occurs [3]. Hence, at the point of near strong modal resonance, the
mode-to-parameter sensitivity reaches the maximum.
In order to numerically demonstrate and evaluate the connections between the
near strong modal resonance and the OLMR, PI gains of the DC voltage controller of
VSC1 were initially set to be Kdcp ¼ 0.21, Kdci ¼ 1.95. Then, integral gain was
changed from Kdci ¼ 1.95 to Kdci ¼ 3.5. Trajectory of λDCi on the complex plane
with changes of Kdci is displayed by dashed curve in Fig. 7.11, when the wind power
penetration was Pw0 ¼ 16 p. u. In Fig. 7.11, solid curves are the trajectories of
closed-loop oscillation modes, b λ ACi and b
λ DCi . From Fig. 7.11, it can be seen that with
the increase of Kdci, λDCi moved towards λACi firstly. Around the points indicated by
hollow circles, λDCi almost overlapped λACi and then moved away from λACi.
Between point A and E on the dashed trajectory of λDCi, closed-loop modes started
moving away from open-loop modes, as being caused by the closeness of λDCi and
λACi. Hence, it was identified that the OLMR occurred between point A and
E. Strongest OLMR happened at point C on the dashed trajectory of λDCi when
two open-loop modes almost overlapped. Corresponding positions of closed-loop
modes, bλ ACi and bλ DCi , are indicated by filled circles on the solid curves. They were
when λ ACi and b
b λ DCi were furthest from each other on the complex plane as caused by
Fig. 7.11 Trajectories of Imaginary axis (rad/s)

open-loop and closed-loop 4.5
oscillation modes E
Ki_vdc1 =2.7
D
4
C D
λˆDCi λDCi
λACi λÂCi
3.5 B
B
Ki_vdc1 =2.15
A
3
-0.6 -0.5 -0.3 -0.1 0.1
Real axis
Fig. 7.12 Variations of | 110

hp11(λACi)|
⏐hpll(lACi)⏐
60
10
AB C D 3 E 4 Kdci
Fig. 7.13 Variations of the 0.5

PFs of sum of state variables
of the MTDC network
Participation factors of MTDC
0.25
VSC-1
Others
0
A B C D 3 Kdci E
the OLMR. For convenience of discussion, point C is called the point of strongest
OLMR.
Figures 7.12 and 7.13 respectively show the variation of |hp11(λACi)| and the PFs
of summation of all the state variables of VSC1 and rest of the MTDC network for
the closed-loop EOM, b λ ACi . From Figs. 7.12 and 7.13, it can be seen that the closer
the λDCi was to λACi, the bigger |hp11(λACi)| and the PFs were; thus the stronger the
dynamic interactions between the MTDC and AC system were. At point C of
strongest OLMR, |hp11(λACi)| and the PFs reached the peak, indicating strongest
dynamic interactions.
Fig. 7.14 Variations of 1.5

mode-to-parameter
sensitivity
⏐ΔlÂCi⏐⁄ΔKdci
1
0.5
0
A B C D 3 Kdci

b
Δλ ACi
Variations of mode-to-parameter sensitivity, ,with increase of Kdci are
ΔKdci
displayed in Fig. 7.14. It can be observed that the sensitivity achieved the peaks at
point B and D, which are the points when the closed-loop EOM, b λ ACi , was closest to
b
λ DCi . They were the points of near strong modal resonance as shown in Fig. 7.11.
7.1.2.4 Strong Dynamic Interactions When the DC Voltage Droop

Control Was Implemented
When the DC droop voltage control was implemented by VSC1 and VSC2, increase
of wind power penetration from Pw0 ¼ 10 p.u to Pw0 ¼ 16 p.u to cause strong
dynamic interactions between the MTDC network and the NETS is shown by
Fig. 7.15. In Fig. 7.15, arrows indicate the direction of increase of wind power
penetration. From Fig. 7.15, it can be seen that the closeness of open-loop mode of
the MTDC network, λDCi, to the open-loop inter-area EOM of concern, λACi, led to
the small-signal angular instability of the NETS.
Computational results of participation factors for b λ ACi and b
λ DCi are presented in
Fig. 7.16 when Pw0 ¼ 10 p.u and Pw0 ¼ 16 p.u respectively. From Fig. 7.16, it can be
seen that strong dynamic interactions occurred between VSC1, VSC2 and the NETS
when Pw0 ¼ 16 p.u. Further confirmation from non-linear simulation is presented in
Fig. 7.17. Significant variations of active power output from VSC1 and VSC2 were
observable. The integrated MTDC/NETS became unstable as caused by strong
dynamic interactions between the MTDC network and the NETS.
Participation factors of all the state variables of VSC1 and VSC2 for b λ ACi were
calculated and compared. It was found that state variables of the DC voltage control
systems of both VSC1 and VSC2 participated in b λ ACi mostly when the OLMR
occurred. In order to dismiss the condition of the OLMR, parameters of the DC
voltage controllers of VSC1 and VSC2 were tuned respectively when Pw0 ¼ 16 p.u.
Fig. 7.15 The OLMR Imaginary axis (rad/s)

occurred at a high level of 3.7
wind penetration (DC DC voltage droop control
voltage droop control)
3.6 λDCi
λÂCi
λ̂DCi
3.5 B
λACi
3.4
-0.6 -0.5 -0.3 -0.1 0.1
Real axis
0.5 1.0
Pwo=10 pu Pwo=10 pu
Pwo=16 pu Pwo=16 pu
0.4 0.8
0.3 0.6
0.2 0.4
0.1 0.2
0 0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5
Generator number VSC number Generator number VSC number
ˆ ACi
PFs for λ ˆ DCi
PFs for λ
Fig. 7.16 Participation factors (PFs) for b

λ ACi and b
λ DCi (DC voltage droop control)
Movements of λDCi on the complex plane with the parameters tuning are shown in
Fig. 7.18.
From Fig. 7.18, it can be seen that when the parameters of the DC voltage
controller of either VSC1 or VSC2 were tuned, λDCi moved away from the open-
loop inter-area EOM of concern, λACi. Subsequently, the closed-loop inter-area
EOM moved back towards λACi and its damping was improved. This confirmed
that damping degradation of inter-area EOM of concern caused by the OLMR can be
dismissed by tuning parameters of the DC voltage controllers of the VSCs which
took part in the droop voltage control. From Fig. 7.18, it can be seen that the OLMR
occurred between λDCi and other local EOMs of the NETS when the parameters of
the DC voltage controller of either VSC1 or VSC2 were tuned. However, impact of
the OLMR between λACi and those local EOMs was not significant.
0.4 0.15
Pwo=10 pu Pwo=10 pu
Pwo=16 pu Pwo=16 pu

0.1
0.2
0.05
0 0
-0.05
-0.2
-0.1
-0.4 -0.15
0 2 4 6 8 10 0 2 4 6 8 10
Time(s) Time(s)
Active power output from VSC1 and VSC2 (DC voltage droop control)
6
Pwo=10 pu
Pwo=16 pu
4
2
Dd1–Dd10 (deg.)
-2
-4
-6
0 2 4 6 8 10
Time(s)
Relative rotor angle position between SG1 and SG10
Fig. 7.17 Non-linear simulation (DC voltage droop control)
a b
Imaginary axis (rad/s) DC voltage droop control Imaginary axis (rad/s) DC voltage droop control
6 6
Trajectories of closed- Trajectories of closed-
loop local EOMs loop local EOMs
5.5 5.5
5 5
λDCi
λDCi
4.5 4.5
λÂCi
4 4
λˆDCi λˆDCi λÂCi
λACi λACi
3.5 3.5
-0.6 -0.5 -0.3 -0.1 0.1 -1.1 -0.7 -0.3 0.1
Real axis Real axis
Fig. 7.18 Modal trajectories when parameters of VSC1 and VSC2 were tuned (DC voltage droop
control). (a) Parameters of the DC voltage controller of VSC1 were tuned. (b) Parameters of the DC
voltage controller of VSC2 were tuned
7.1.2.5 Example 7.2—Test with the CIGRE Five–terminal DC Network
Figure 7.19 shows the configuration of the NETS integrated with the CIGRE five-
terminal DC network [11]. Data of the CIGRE MTDC network is presented in
Appendix 7.2. The DC network adopted the master-slave control scheme with
VSC1 using the DC voltage control, VSC2 and VSC3 using the active power
control. VSC4 and VSC5 were connected separately to two wind farms. The total
wind power penetration was 16 p.u. In this subsection, test results to numerically
evaluate the connections between the near strong modal resonance and the OLMR
are presented.
In the test, PI gains of the DC voltage controller of VSC1 were varied with
Kdcp ¼ KpiKdci, Kpi ¼ 0.15. The integral gain was changed from Kdci ¼ 2 to Kdci ¼ 5.
Trajectory of λDCi on the complex plane is displayed by dashed curve in Fig. 7.20,
where solid curves are the corresponding trajectories of closed-loop oscillation
modes, b λ ACi and b
λ DCi . Figures 7.21 and 7.22 are the variations of |hp11(λACi)| and
the PFs of all the state variables of VSC1 and the rest of the MTDC network for the
closed-loop EOM, b λ ACi . Figure 7.23 shows the variations of mode-to-parameter
b
Δλ ACi
sensitivity, .
ΔKdci
From Fig. 7.20, it can be seen that the OLMR occurred between point A and E on
the trajectory of λDCi. Whilst at point F on the trajectory of λDCi, dynamic interac-
tions between the CIGRE five-terminal DC network and AC system were obviously
weak. The near strong modal resonance happened at point B and D (see Fig. 7.23).
26
5 28 29
25 27 4
34 24 38
2 18 33
37 9
17 6
30 8 20 35
1 19 7
1 16 36
3
22
15
10 14 21 23
2
4 41 42
39 31 12
VSC-2 VSC-3
5 11 13
6 10
MTDC WIND
9 7 NET FARMs
32
8 VSC-1
3
NETS 40
Fig. 7.19 New England power system integrated with the CIGRE five-terminal DC system
Fig. 7.20 Trajectories Imaginary axis (rad/s)

of open-loop and 6
closed-loop oscillation
modes with variation of Kdci
F
D
4
D λÂCi
λˆDCi λDCi λACi
C
B
B
3
A
-1.0 -0.8 -0.5 -0.2 0.1
Real axis
Fig. 7.21 Variations 110

of |hp11(λACi)| with changes
of Kdci
⏐hpll(lACi)⏐
60
10
AB C D 3 E 4 Kdci
Fig. 7.22 Variations of the 0.5

PFs of the MTDC with
Participation factors of MTDC
changes of Kdci
0.25
VSC-1
Others
0
A B C D 3 Kdci E
Fig. 7.23 Variations of 1.5

mode-to-parameter
sensitivity with changes of
Kdci
⏐ΔlÂCi⏐⁄ΔKdci
1
0.5
0
A B C D 3 Kdci E
The strongest point of OLMR was C, when the damping degradation caused by the
OLMR was the most. The closer the λDCi was to λACi, the stronger the dynamic
interactions between the CIGRE five-terminal DC network and AC system were. At
point C, the dynamic interactions were strongest.
Confirmation from non-linear simulation is presented in Fig. 7.24 when the
CIGRE five-terminal DC network/NETS operated at point C of strongest OLMR
and point F of weak modal coupling. At 0.1 s of simulation, 80% load at node 10 in
the NETS was lost. The lost load was recovered in 0.1 s. From Fig. 7.24, it can be
observed that when the OLMR occurred, the active power output from VSC1 varied
considerably, indicating strong dynamic interactions. The strong dynamic interac-
tions caused growing low-frequency inter-area power oscillations.
7.2 Small-Signal Stability of an MTDC Network

for the Wind Power Transmission Affected by
the Dynamic Interactions Between the VSCs
7.2.1 Small-Signal Stability Analysis
7.2.1.1 Linearized Model When a VSC Implementing Active Power

Control Is Considered
Figure 7.25 shows the configuration of an MTDC network for wind power trans-
mission being connected to multiple SGs. The configuration is different to that of the
MTDC/AC power system shown by Fig. 7.1 because in Fig. 7.25, there is no AC
transmission grid connecting M SGs. Each SG is connected to one of the VSCs
which are connected by the MTDC network. Hence, there are two types of dynamic
interactions in the MTDC/AC power system shown by Fig. 7.25: (1) AC/DC
dynamic interactions between a VSC and a SG; (2) DC/DC dynamic interactions
7.2 Small-Signal Stability of an MTDC Network for the Wind Power. . . 269
Fig. 7.24 Confirmation 1

from non-linear simulation Weak dynamic interactions at F

of the NETS with the Open-loop modal coupling at C
CIGRE five-terminal DC
system 0.5
-0.5
-1
0 2 4 6 8 10
Time(s)
Weak dynamic interactions at F

8 Open-loop modal coupling at C
4
Dd1–Dd10 (deg.)
-4
-8
0 2 4 6 8 10
Time(s)
between the VSCs via the MTDC network. For the simplicity of discussion, the
MTDC/AC power system shown by Fig. 7.25 is referred to as the MTDC power
system. In this subsection, the small-signal stability of the MTDC power system
affected by the dynamic interactions between the VSCs is examined. In order to
conduct the examination, closed-loop model of the MTDC power system needs to be
established firstly.
Consider the kth VSC (VSC-k) connected to the SG in the MTDC power system.
Denote Pk þ jQk and Vk as the power output from and the magnitude of terminal
voltage of VSC-k. State-space model of the kth SG (SG-k) can be written as
P1 + jQ1
Wind farms V1
I MTDC1 I dc1
X f1
Vdc1 I d 1 + jI q1
C1 VSC-1 SG-1
... M Pk + jQk
T IMTDCk I dck X fk
Vk
D
Ck Vdck I dk + jI qk
C VSC-k SG-k
...
Network PM + jQM
VM
I MTDCM I dcM X fM
CM VdcM I dM + jI qM
VSC-M SG-M
Fig. 7.25 Configuration of a MTDC/AC power system
d
ΔXk ¼ Ak ΔXk þ bpk ΔPk þ bqk ΔQk
dt ð7:43Þ
ΔVk ¼ ck T ΔXk þ dpk ΔPk þ dqk ΔQk
where ΔXk is the vector of all the state variables of SG-k. From (7.43), transfer
function model of SG-k is obtained to be
ΔVk ¼ Gpk ðsÞΔPk þ Gqk ðsÞΔQk ð7:44Þ
where Gpk(s) ¼ ckT(sI Ak)1bpk þ dpk, Gqk(s) ¼ ckT(sI Ak)1bqk þ dqk.

Take the direction of the terminal voltage of VSC-k as that of d axis of d q
coordinate of VSC-k. Thus,
Vdk ¼ Vk , Vqk ¼ 0 and ΔVdk ¼ ΔVk , ΔVqk ¼ 0 ð7:45Þ
where subscript d and q respectively indicates d and q component of a variable in the

d q coordinate of VSC-k. From (7.44), linearized active and reactive power output
from VSC-k can be obtained to be
ΔPk ¼ Idk0 ΔVk þ Vk0 ΔIdk ¼ Vdck0 ΔIdck þ Idck0 ΔVdck
ð7:46Þ
ΔQk ¼ Iqk0 ΔVk Vk0 ΔIqk
where subscript 0 denotes the value of a variable at the steady state. Linearized line
current equations at the terminal of VSC-k are (7.18).
Consider that VSC-k adopts the active power control. Configuration of vector
control system of VSC-k is shown by Fig. 7.3. From (7.18) and Fig. 7.3,
ΔIdk ¼ hpk ðsÞΔPk , ΔIqk ¼ hqk ðsÞΔQk ð7:47Þ
2 D (s)
G pk (s)
+ ΔVk Δ Pk 1 +
H pk (s) 1
+ 3 – Vdck 0
I dck 0 Δ I dck
VSC control –
closed loop H qk (s) 1/ Ck s
+
5 MTDC 6
Gqk (s) ΔVdck closed loop ΔI MTDCk
4 ΔQk C (s)
M (s)
Fig. 7.26 Closed-loop interconnected model when VSC-k uses the active power control
ω K ðsÞK ðsÞ ω K ðsÞK ðsÞ

where hpk ðsÞ ¼ X0fk sþω
idk pk
0 Kidk ðsÞ
, hqk ðsÞ ¼ X0fk sþω
iqk qk
0 Kiqk ðsÞ
.
ΔPk ¼ Hpk ðsÞΔVk , ΔQk ¼ Hqk ðsÞΔVk ð7:48Þ
I
where Hpk ðsÞ ¼ 1VIk0dk0hpk ðsÞ , Hqk ðsÞ ¼ 1þVk0qk0
hqk ðsÞ.
Linearized DC voltage equation of VSC-k is
dΔVdck
Ck ¼ ΔIMTDCk ΔIdck ð7:49Þ
dt
From (7.44), (7.46), (7.48) and (7.49), the closed-loop model of the MTDC power
system is obtained and shown by block diagram of Fig. 7.26. In Fig. 7.26, the VSC
control closed loop on the left is depicted by (7.44) and (7.48). From (7.46),
ΔPk Idck0ΔVdck ¼ Vdck0ΔIdck, which together with (7.49) describes the upper
loop D(s) on the right in Fig. 7.26; whilst dynamics of the rest of the MTDC power
system excluding the dynamics of VSC-k and SG-k are expressed by transfer
function M(s) as ΔIMTDCk ¼ M(s)ΔVdck.
The closed-loop model of Fig. 7.26 is derived for the VSC which adopts the
active power control. In the derivation, the entire MTDC power system is divided
into two parts: (1) VSC-k and SG-k; (2) the rest of MTDC power system. The
division is reflected at point ⑤ and ⑥ in Fig. 7.26. The part of rest of MTDC power
system is modelled below point ⑤ and ⑥, above which is the part of SC-k and
SG-k.
From Fig. 7.26, it can be seen that there are two closed loops in the model. The
VSC control closed loop reflects the AC/DC dynamic interactions between SG-k and
the control system of VSC-k. Dynamic interactions between the dynamics of VSC-k
and the rest of MTDC power system are through the MTDC closed loop in Fig. 7.26.
It can be noted that the dynamic connection between those two closed loops is
unidirectional at point ①.
7.2.1.2 Linearized Model When a VSC Implementing DC Voltage

Control or DC Voltage Droop Control Is Considered
Consider that VSC-k adopts the DC voltage control. Configuration of DC voltage

control is shown by Fig. 7.2. Comparing Fig. 7.2 and Fig. 7.3, it can be seen that in
two types of converter control configurations, only the input to the DC voltage and
active power control outer loop is different. However, this difference changes the
configuration of closed-loop model as to be demonstrated as follow.
From (7.18) and Fig. 7.2,
ΔIdk ¼ hpk ðsÞΔVdck , ΔIqk ¼ hqk ðsÞΔQk ð7:50Þ
Substituting (7.50) into (7.46),

ΔPk ¼ Idk0 ΔVk þ Vk0 hpk ðsÞΔVdck , ΔQk ¼ Hqk ðsÞΔVk ð7:51Þ
From (7.44), (7.46), (7.49) and (7.51), the closed-loop model of the MTDC power
system can be derived and is shown by the block diagram of Fig. 7.27. Comparing
Figs. 7.26 and 7.27, it can be seen that in Fig. 7.27, there is an additional VSC closed
loop, which reflects the dynamic interactions between the control system of VSC-k
and the rest of MTDC power system via the MTDC network. The additional VSC
closed loop is formed by (7.51), which indicates that when the DC voltage control is
implemented by VSC-k, variation of active power output from VSC-k is affected not
only by ΔVk as it is in the case of active power control, but also by ΔVdck.
The closed-loop model of Fig. 7.27 is derived for the VSC which adopts the DC
voltage control. Figures 7.26 and 7.27 show two models for the same MTDC power
system shown by Fig. 7.25. Up to the VSC being considered, the model can be that
shown by either 7.26 or Fig. 7.27. For example, consider that the MTDC network
implements the master-slave control where VSC-1 adopts the DC voltage control
and all the other VSCs use the active power control. When VSC-1 is selected to
derive the closed-loop model of the MTDC power system, the derived model is
shown by Fig. 7.27. If any of other VSCs using the active power control is chosen,
the derived model is shown by Fig. 7.26.
2 D (s)
G pk (s)
+ ΔVk ΔPk 1 +
I dk 0 1
+ 3 + – Vdck 0
Δ I dck
VSC control I dck 0
closed loop
VSC –
H qk (s) closed loop 1/ Ck s
+
5 MTDC 6
Gqk (s) Vk 0h pk (s) closed loop ΔI MTDCk
4 ΔQk C (s) ΔVdck
M (s)
Fig. 7.27 Closed-loop model when VSC-k uses DC voltage control

Power control outer loop Current control inner loop

DC voltage droop control
PWM
Reactive power control loop
Fig. 7.28 Converter control system of VSC-k with DC voltage droop control
Finally, consider the case that the DC voltage droop control is implemented by
VSC-k. In this case, input to the
DC voltage
control outer loop in Fig. 7.2 is changed
from Vdckref
Vdck to Kdpk Pkref Pk þ Vdckref
Vdck , where Kdpk is the droop
coefficient. Configuration of the reactive power control of VSC-k remains
unchanged. Figure 7.28 shows the configuration of the DC voltage droop control
implemented by VSC-k.
In case of DC voltage droop control, (7.50) becomes

ΔIdk ¼ hpk ðsÞ Kdpk ΔPk þ ΔVdck , ΔIqk ¼ hqk ðsÞ ΔQk ð7:52Þ

Idk0 ΔVk þ Vk0 hpk ðsÞΔVdck
ΔPk ¼
1 Vk0 hpk ðsÞKdpk ð7:53Þ
¼ Dk s Idk0 ΔVk þ Vk0 hpk ðsÞΔVdck
ð Þ
ΔQk ¼ Hqk ðsÞΔVk
Similar to the case of VSC-k with the DC voltage control, the closed-loop model
of the MTDC power system when VSC-k with the DC voltage droop control is
obtained and shown by the block diagram of Fig. 7.29, where
Dk(s) ¼ (1 KdpkVk0 gpk(s))1. The VSC closed loop is formed by (7.53).
7.2.1.3 Small-Signal Stability of the MTDC Network Affected by

the OLMR Between the VSCs
A simpler representation of block diagram for Fig. 7.26 is shown by Fig. 7.30. From
Fig. 7.30, it can be seen that there are three open-loop subsystems in the closed-loop
model. Transfer functions of those three open-loop subsystems are C(s), D(s) and M
2 D (s)
G pk (s)
+ ΔVk 1 +
I dk 0 Dk (s) 1
+ 3 + Δ Pk – Vdck 0
Δ I dck
VSC control I dck 0
closed loop
VSC –
H qk (s) closed loop 1/ Ck s
+
5 6
Vk 0h pk (s) MTDC
Gqk (s) ΔVdck closed loop Δ I MTDCk
4 ΔQk C (s)
M (s)
Fig. 7.29 Closed-loop interconnected model when VSC-k uses the DC voltage droop control
Fig. 7.30 Simpler

representations of closed- C (s) D (s)
loop interconnected model
5
MTDC
6 C (s) VM (s)
for Fig. 7.26 closed loop
ΔVdck Δ I MTDCk
M (s)
VM (s)
a b
C (s) D (s) V (s)
V (s) MTDC MTDC
5 6
5 closed loop 6 closed loop
ΔVdck ΔVdck Δ I MTDCk
VSC ΔI MTDCk
closed loop M (s) M (s)
VM (s)
Fig. 7.31 Simple representation of closed-loop interconnected model for Fig. 7.27 and 7.29
(s) respectively. Open-loop subsystems, D(s) and M(s), form a closed-loop

subsystem VM(s) as shown in Fig. 7.30b. For the convenience of discussion, D
(s) is referred to as the open-loop VSC subsystem and M(s) is named as the open-
loop subsystem of the RDCN (rest of MTDC network). If and only if both the open-
loop subsystem C(s) and the closed-loop subsystem VM(s) in Fig. 7.30b are stable,
the MTDC power system is stable. Obviously, stability of the closed-loop subsystem
VM(s) is determined by the dynamic interactions between the VSCs through the
MTDC closed loop.
A simpler representation of the block diagram for Figs. 7.27 and 7.29 is shown by
Fig. 7.31. In Fig. 7.31b, transfer function of the open-loop VSC subsystem in the
forward loop is V(s), which consists of C(s) and D(s). The open-loop subsystem of
the RDCN in the feedback loop is still M(s). For the convenience of discussions, the
transfer function of closed-loop system in Fig. 7.31b is still denoted as VM(s), which
consists of V(s) and M(s). Hence, in both cases of Fig. 7.30 and 2.31, the impact of
dynamic interactions between the VSCs on the small-signal stability of the MTDC
power system can be studied by examining the stability of the closed-loop system
VM(s).
According to the theory of stability of a closed-loop system, if poles of two
interconnected open-loop subsystems move close to each other on the complex
plane, dynamic interactions between two open-loop subsystems increase and the
stability of the closed-loop interconnected system decreases. This has been demon-
strated in Chaps. 5 and 6 when the VSWGs and the PLL are examined. For the
convenience of discussions, the closeness of complex poles of two open-loop sub-
systems in Fig. 7.30 and 7.31 is still referred to as open-loop modal resonance
(OLMR). Impact of OLMR on the stability of closed-loop system VM(s) in Fig. 7.30
and 7.31, i.e., the small-signal stability of the power system with the MTDC
network, is examined as follows.
Denote λm as an oscillation mode of the open-loop subsystem of the RDCN. It is a
complex pole of M(s) in Fig. 7.30 and 7.31. Denote λv as an oscillation mode of
open-loop VSC subsystem. It is a complex pole of either D(s) in Fig. 7.30 or V(s) in
Fig. 7.31. The OLMR is when λm is close to λv on the complex plane, i.e., λv λm.
Since λm is related with the dynamics of other VSCs in the MTDC power system,
λv λm may be the OLMR between the VSCs. Denote b λ v and b
λ m as the oscillation
modes of the closed-loop system VM(s) in Figs. 7.30 and 7.31 corresponding to λv
and λm respectively. Impact of OLMR between the VSCs on the small-signal
stability of MTDC power system can be examined by estimating b λ v and b
λ m when
the OLMR occurs.
λ v and b
b λ m are the solutions of the following characteristic equations of closed-loop
system VM(s) shown by Figs. 7.30 or 7.31.
1 ¼ DðsÞMðsÞ or 1 ¼ VðsÞMðsÞ ð7:54Þ
Denote
dð s Þ vð s Þ m ðsÞ
D ðsÞ ¼ or VðsÞ ¼ , M ðsÞ ¼ ð7:55Þ
ðs λ v Þ ðs λ v Þ ð s λm Þ
Under the condition of open-loop modal coupling, i.e., λv λm, for b

λ v it can have
2
bλ v λv b λ v λm ¼ d b
λv m b
λv b
λ v λv or
2 ð7:56Þ
bλ v λv b λ v λm ¼ v b
λv m b
λv b
λ v λv
Thus, in the neighborhood of λv, i.e., b λ v ! λv

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
λ v λv dðλv Þmðλv Þ or b
b λ v λv vðλv Þmðλv Þ ð7:57Þ
Transfer functions of open-loop subsystems can be written as

Rv X Ri
DðsÞ ¼ or VðsÞ ¼ þ
ðs λv Þ ðs λ i Þ
i¼1
i6¼v
Rm X Rj ð7:58Þ
M ðsÞ ¼ þ
ð s λm Þ s λj
j¼1
i6¼v
where λi are the poles of open-loop subsystem D(s) in Fig. 7.30 or V(s) in Fig. 7.31,
λj are the poles of the open-loop RDCN subsystem M(s), Ri and Rj are the residues
corresponding to λi and λj; Rv and Rm are the residues corresponding to λv and λm
respectively. From (7.55) and (7.58), it can have v(λv)m(λv) ¼ RvRm. Hence, (7.57)
becomes
pffiffiffiffiffiffiffiffiffiffiffiffi
b
λ v λv Rv Rm ð7:59Þ
Taking the derivation similar to that from (7.56) to (7.59) for b

λ m , it can have
pffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffi
b
λ m λm Rv Rm λv Rv Rm ð7:60Þ
The above derived equations of (7.59) and (7.60) indicate that when the OLMR
occurs, the corresponding closed-loop complex poles, b λ v and bλ m , locate at positions
approximately opposite to each other in respect to the positions where λv λm.
Hence, it is very likely that one of the closed-loop oscillation modes, either b λ v or b
λm,
is on the right hand side of corresponding open-loop oscillation mode. This implies
that the OLMR between the VSCs degrades the closed-loop stability of the MTDC
power system.
From (7.59) and (7.60), it can be seen that either b
λ v or bλ m may be in the right half
of complex plane if
pffiffiffiffiffiffiffiffiffiffiffiffi

real part of Rv Rm > jreal part of λv or λm j ð7:61Þ
That is the condition under which the OLMR between the VSCs may lead to the
instability of the MTDC network.
From Fig. 7.26, it can be seen that for the VSC adopting the active power control,
the open-loop VSC subsystem D(s) in the closed-loop system VM(s) in Fig. 7.30 is a
first-order system and is not of any oscillation mode. Thus, there is no possibility that
the VSC takes part in the OLMR to cause the instability of the MTDC power system.
Hence, when the MTDC network adopts the master-slave control, one VSC uses the
DC voltage control and all the other VSCs adopt the active power control. There is
no OLMR between two or more VSCs to happen. Thus, it can be concluded that in
the MTDC network with the master-slave control, no OLMR between the VSCs can
occur to cause the instability of the MTDC power system.
If the DC voltage droop control is implemented by the MTDC network, there are
at least two VSCs adopting the DC voltage droop control. From Fig. 7.31, it can be
seen that the open-loop VSC subsystem, V(s), may have an oscillation mode because
the VSC closed loop, C(s), is a part of subsystem V(s). Hence, it is possible that the
OLMR may occur between the VSCs adopting the DC voltage droop control to lead
to the instability of the MTDC network. Therefore, it can be concluded that when the
DC voltage droop control is implemented by the MTDC network, instability risk of
MTDC power system increases because of the possibility of OLMR between
the VSCs.
7.2.2 Example MTDC Power Systems

7.2.2.1 Example 7.3—An Eight-Terminal DC Network
Fig. 7.32 shows the configuration of an example MTDC power system. Parameters
of the MTDC and AC system given in [4, 12–14] were used. The MTDC network is
of eight terminals. VSC-k (k ¼ 1,2,3,4) operated with unity power factor and their
active power output at steady state was 3.43 p.u, 1 p.u, 0.5 p.u and 0.5 p.u
respectively. Each of other four VSCs is connected to a wind farm represented
by a PMSG. Detailed 14th-order model and parameters of the PMSG recommended
in [8] were used. Data of the example MTDC network is given in Appendix 7.3.
Tests were conducted to demonstrate and evaluate the impact of OLMR between
the VSCs on the small-signal stability of example MTDC power system. In the tests,
the master-slave control and the DC voltage droop control were implemented
respectively by the MTDC network. When the master-slave control was
implemented, VSC-1 adopted the DC voltage control and other three VSCs used
V1
r1 x1 200km X f1
r8 r2 C1 P + jQ1
30km x r7 200km 200kmr6 VSC-1 1 SG-1
8 x2 30km V2
x7 x6 A Xf2
r9 r3 C2 P2 + jQ2
40km x SG-2
9 x3 40km VSC-2 V3
X f3
x6 x7
r8 r6 200km 200km r7 r4 C3 P3 + jQ3 SG-3
30km x
8 x4 30km VSC-3 V4
r5 x5 130km Xf4
C4 P4 + jQ4
VSC-4 SG-4
Fig. 7.32 Configuration of an example AC/MTDC power system

the active power control. When the DC voltage droop control was implemented,
VSC-1 and VSC-2 adopted the DC voltage droop control. Droop coefficient was
0.01. VSC-3 and VSC-4 still used the active power control.
With a partition at point A in Fig. 7.30, the closed-loop models of the example
MTDC power system were established. The open-loop VSC subsystem in the
forward loop in Figs. 7.30 and 7.31 was comprised of VSC-2 and SG-2 with
ΔIMTDC2 and ΔVdc2 being the input and output variable respectively. Hence, the
open-loop state-space model of the open-loop VSC subsystem was
d
ΔXv ¼ Av ΔXv þ bv ΔIMTDC2
dt ð7:62Þ
ΔVdc2 ¼ cv T ΔXv þ dv ΔIMTDC2
where ΔXv is the vector of all the state variables of VSC-2 and SG-2. When the
MTDC network implemented the master-slave control, VSC-2 used the active power
control. Thus, transfer function of the VSC subsystem was
Vs ðsÞ ¼ cv T ðsI Av Þ1 bv þ dv ð7:63Þ
When the MTDC network implemented the DC voltage droop control, VSC-2
used the DC voltage droop control. Hence, transfer function of the VSC subsystem
was
VðsÞ ¼ cv T ðsI Av Þ1 bv þ dv ð7:64Þ
The open-loop RDCN subsystem in the feedback loop in Figs. 7.30 and 7.31 was
comprised of the rest of example MTDC power system with ΔVdc2 and ΔIMTDC2
being the input and output variable respectively. Following state-space model of the
open-loop RDCN subsystem was established
d
ΔXm ¼ Am ΔXm þ bm ΔVdc2
dt ð7:65Þ
ΔIMTDC2 ¼ cm T ΔXm þ dm ΔVdc2
where ΔXm is the vector of all the state variables of the rest of example MTDC
power system excluding those of VSC-2 and SG-2. Thus, transfer function of the
open-loop RDCN subsystem is
MðsÞ ¼ cm T ðsI Am Þ1 bm þ dm ð7:66Þ
Complex poles of the open-loop VSC and RDCN subsystems were calculated
respectively from the open-loop state matrix As in (7.62) and Am in (7.65). They are
the complex poles of transfer functions, Vs(s) in (7.63) or (7.64) and M(s) in (7.66).
Tests were focused on the OLMR between λv and λm. λv was an open-loop
oscillation mode of the VSC subsystem, Vs(s) or V(s). λm was an open-loop complex
pole of RDCN subsystem M(s). The participation factors (PFs) for λv and λm were
calculated from (7.62) and (7.65). Computational results of the PFs indicated that λm
and λv were associated with the current control inner loops of VSC-1 and VSC-2
respectively. Initially, parameters of control systems of VSC-1 and VSC-2 were all
same except the integral gains of current control inner loop, Kiik, k ¼ 1, 2. For
VSC-1, Kii1 ¼ 0.13 p. u; and for VSC-2, Kii2 ¼ 0.1p. u.
To find the OLMR between VSC-1 and VSC-2, value of integral gain of current
control inner loop of VSC-1 was varied from Kii1 ¼ 0.13 p. u to Kii1 ¼ 0.7 p.
u. With the variation of Kii1, open-loop oscillation mode, λm, moved on the complex
plane. Trajectories of movement are displayed in Figs. 7.33 and 7.34 as dashed
curves. In Figs. 7.33 and 7.34, positions of λv on the complex plane are indicated by
hollow circles. From Figs. 7.33 and 7.34, it can be seen that at point ① and ② on the
dashed curves indicated by hollow circles when Kii1 ¼ 0.1 p. u, λm was close to λv.
Thus, ① and ② may be the points of OLMR.

and trajectories of the
example MTDC system
(Master-slave control)
Imaginary axis (rad./s.)
225
220 λm λv
1
215
210
-6.35 -6.33 -6.31
Real axis (rad./s.)

example MTDC system
(DC voltage droop control)
240
λv
220 λm ˆλv
ˆλm
2
200
180
-9.5 -7 -4.5 -2 0.5
Real axis (rad./s.)
To examine whether the OLMR occurred at point ① and ②, residues

corresponding to λv and λm were computed fromp(7.62) ffiffiffiffiffiffiffiffiffiffiffiffi and (7.65). When the
MTDC network used the master-slave control, Rv Rm ¼ 0. This means that
although at point ① in Fig. 7.33, λv λm, closed-loop oscillation modes were
estimated being equal to the open-loop oscillation modes. The closeness of open-
loop modes, i.e., λv λm, affected little the stability of the example MTDC power
system. This implied that λv in fact was not a complex pole of open-loop subsystem
D(s) in the closed-loop system VM(s) in Fig. 7.30 and thus, the OLMR did not occur.
This confirmed the analysis made in Subsect. 7.2.1.3 that λv was a complex pole of C
(s) because VSC-2 adopted the active power control.
When the DC voltage droop control was implemented by the MTDCpnetwork, ffiffiffiffiffiffiffiffiffiffiffiffi
computational results of residues from (7.62) and (7.65) p were
ffiffiffiffiffiffiffiffiffiffiffi
ffi Rv Rm ¼
2:11 j0:73. Thus, at point ② in Fig. 7.34, real part of Rv Rm ¼ 2:11 >|real
part of λv| ¼ 1.81. This indicated that at point ②, the OLMR happened. According to
the estimation by (7.61), the OLMR at point ② may cause the small-signal insta-
bility of the example/MTDC power system.
In order to confirm the conclusion obtained above from the modal analysis for the
open-loop subsystems of (7.62) and (7.65), the state-space model of closed-loop
MTDC power system was derived from (7.62) and (7.65) to be
d
ΔX¼AΔX ð7:67Þ
dt
where
T
ΔX ¼ ΔXv T ΔXm T
2 3
bv dm cv T bv dm dv cm T
6 Av þ 1 d d bv cm þ
T
6 v m 1 dv dm 7 7
A¼6 7
4 bm v dm c v
d T
bm dv cm T 5
bm cv T þ Am þ
1 dm dv 1 dm dv
Closed-loop oscillation modes corresponding to λv and λm were calculated from
the closed-loop state matrix, A in (7.67), to be b λ v and bλ m . When the master-slave
control was implemented by the MTDC network, trajectory of b λ v with variation of
Kii1 was found to coincide completely that of λv in Fig. 7.33. b λ m was also in
coincidence with λm in Fig. 7.33. Thus, trajectory of b λ v and position of bλ m are not
b b
displayed in Fig. 7.33. Therefore, computational results of λ v and λ m confirmed that
the closeness of λv to λm on the complex plane at point ① in Fig. 7.33 caused no
change of b λ v and b
λ m , not affecting the small-signal stability of the example MTDC
power system at all. Hence, point ① in Fig. 7.33 was not a point of OLMR between
VSC-1 and VSC-2 when the master-slave control was implemented by the MTDC
network with VSC-2 adopting the active power control.
When the MTDC network implemented the DC voltage droop control, trajecto-
ries of closed-loop poles, b λ v and b
λ m , with variation of Kii1 were calculated from the
closed-loop state matrix A in (7.67) and are displayed by solid curves in Fig. 7.34.
Master-slave Master-slave
1.2 Voltage droop 1.2 Voltage droop
0.6 0.6
0 0
PFv PFm PFv PFm
The PFs for ˆλv The PFs for ˆλm
Fig. 7.35 Computational results of the PFs
Positions of bλ v and b
λ m corresponding to λv and λm at point ② are indicated by solid
circles. Obviously, the OLMR between VSC-1 and VSC-2 at point ② caused the
small-signal instability of the example MTDC system, confirming the estimation
made previously by using (7.61).
Computational results of the PFs for b λ v and b
λ m corresponding to λv and λm at point
① and ② in Figs. 7.33 and 7.34 are presented in Fig. 7.35. In Fig. 7.35, PFv is the
sum of the PFs of all the state variables of VSC-2 and PFm is the sum of the PFs of all
the state variables of VSC-1. From Fig. 7.35, it can be seen that when the MTDC
network implemented the master-slave control, VSC-1 participated only in b λ m and
VSC-2 took part in only b λ v . Hence, there were no dynamic interactions between
VSC-1 and VSC-2. When the MTDC network adopted the DC voltage droop
control, both VSC-1 and VSC-2 participated in both b λ v and b
λ m . This indicated the
strong dynamic interactions between VSC-1 and VSC-2 when the OLMR occurred.
Results of non-linear simulation are displayed in Fig. 7.36. At 0.1 s of simulation,
80% of the load at the AC terminal of VSC-1 was lost. The lost load was recovered
in 0.1 s. In Fig. 7.36, solid curves are the simulation results when the MTDC network
implemented the master-slave control. Dashed curves are the simulation results
when the DC voltage droop control was implemented by the MTDC network.
Parameters of control systems of VSC-2 were set such that the example MTDC
power system operated at point ① and ② indicated in Figs. 7.33 and 7.34.
From Fig. 7.36, it can be seen that when the master-slave control was used by the
MTDC network, the example MTDC power system was stable. However, when the
DC voltage control was implemented by the MTDC network, the oscillations
occurred and the example MTDC power system was unstable. The instability was
caused by the OLMR between VSC-1 and VSC-2 which used the DC voltage droop
control. The oscillation frequency was 34.7 Hz, This was in consistence with the
results of open-loop modal analysis presented in Fig. 7.34, where the OLMR
occurred at point ② and the negatively damped closed-loop SSO mode was
b
λ v ¼ 0:17 þ j217:6ðrad:=s:Þ.
0.6
Variation of IMTDC1 (p.u)

Master slave control
Time (second)
-0.6
0 0.5 1 1.5 2
0.1
Variation of Vdc1 (p.u)
Time (second)
-0.1
0 0.5 1 1.5 2
0.2
Time (second)
-0.2
0 0.5 1 1.5 2
0.1
Time (second)
-0.1
0 0.5 1 1.5 2
Fig. 7.36 Non-linear simulation

Tests presented above demonstrated and confirmed the following analytical

conclusions made in Subsect. 7.2.1: (1) When the MTDC network uses the
master-slave control, no OLMR between the VSCs may occur to degrade the
small-signal stability of MTDC power system; (2) When the DC voltage droop
control is used by the MTDC network, it is possible that the OLMR between the
VSCs may occur to lead to the stability degradation of MTDC power system. In the
worst case, instability may occur as caused by the dynamic interactions between the
VSCs. The instability risk can be identified by using (7.61).
Study cases of the example MTDC power system presented above indicated that
when two VSCs adopting the DC voltage droop control are of the same control
system parameters, open-loop oscillation modes associated with the VSCs’ control
systems are very possibly close to each other on the complex plane. This was the
case demonstrated in Fig. 7.34. Hence, same settings of control system parameters of
the VSCs in the MTDC network should be avoided if those VSCs take part in the DC
voltage droop control, because it may likely bring about the risk of OLMR between
the VSCs to degrade the small-signal stability of the MTDC power system.
7.2.2.2 Example 7.4—The CIGRE Five-Terminal DC Network
Figure 7.37 shows the configuration of the CIGRE five-terminal DC network [11]
connecting the wind farms with the SGs. The master-slave control and the DC
voltage droop control were implemented respectively by the five-terminal DC
network. When the master-slave control was used, VSC-1 adopted the DC voltage
control and other two VSCs connected with the SGs used the active power control.
When the DC voltage droop control was implemented, VSC-1 and VSC-2 adopted
the DC voltage droop control. Droop coefficient was 0.01. VSC-3 still used the
active power control. Data of example MTDC network is given in Appendix 7.2.
200km V1
r1 x1 X f1
C1 P1 + jQ1
VSC-1 SG-1
r2 500km
r6 r4 x2
500km V2
400km Xf2
x6 x4
x3 C2 P2 + jQ2
r3 VSC-2 SG-2
200km
V3
r5 x5 X f3
200km C3 P3 + jQ3
VSC-3 SG-3
Fig. 7.37 Configuration of power system with CIGRE five-terminal DC network

Initially, parameters of control systems of VSC-1 and VSC-2 were set to be same
except the integral gains of current control inner loops, Kiik, k ¼ 1, 2. For VSC-1,
Kii1 ¼ 0.191 p. u; and for VSC-2, Kii2 ¼ 0.159 p. u. To examine the existence of the
OLMR between VSC-1 and VSC-2, value of integral gain of current control inner
loop of VSC-1 was varied from Kii1 ¼ 0.191 p. u to Kii1 ¼ 0.127 p. u. Modal
trajectories similar to those presented in Figs. 7.33 and 7.34 were calculated and are
shown by Figs. 7.38 and 7.39. From 7.38, it can be seen that when the master-slave
control was used, there was no OLMR to cause the stability degradation of the
CIGRE five-terminal DC power system. Figure 7.39 indicated the occurrence of
OLMR when the DC voltage droop control was implemented by the CIGRE five-
terminal DC network. The OLMR led to the loss of the stability of the CIGRE five-
terminal DC power system. The negatively damped closed-loop SSO mode was
b
λ v ¼ 0:06 þ j213:7ðrad:=s:Þ.

CIGRE five-terminal DC
power system (Master-slave
control) 220
λm λv
210
200
190
-9.27 -9.17 -9.07
Real axis (rad./s.)

AC/CIGRE five-terminal
DC power system
(DC voltage droop control) 220

λm λv
ˆ
λ v
210 ˆλ
m
200
190
-11 -8 -5 -2 1
Real axis (rad./s.)
To confirm the results of modal computation presented in Figs. 7.38 and 7.39,
non-linear simulation was conducted. Results of non-linear simulation are displayed
in Fig. 7.40. At 0.1 s of simulation, 80% of the load at the AC terminal of VSC-1 was
lost. The lost load was recovered in 0.1 s. It can be seen that growing oscillations
occurred when the DC voltage droop control was implemented.
0.2
Time(second)
-0.2
0 0.75 1.5 1.75 3
0.1
Time(second)
-0.1
0 0.75 1.5 1.75 3
0.3
Time(second)
-0.3
0 0.75 1.5 1.75 3
0.1
Time(second)
-0.1
0 0.75 1.5 1.75 3
Fig. 7.40 Non-linear simulation (AC/five-terminal DC power system)

7.3 Method of Open-loop Modal Analysis to Examine

the Impact of Dynamic Interactions Introduced by
a Selected VSC Control on the Small-Signal Stability
of an MTDC/AC Power System
7.3.1 Method of Open-Loop Modal Analysis
7.3.1.1 Closed-Loop Model When the DC Voltage Control Is Selected
Consider the MTDC/AC power system shown by Fig. 7.1, where VSC-1 adopts the
DC voltage control and VSC-k uses the active power control, with k ¼ 2, 3, . . .M.
Figure 7.2 shows the configuration of the control system of VSC-1, which is
comprised of the DC voltage control loop and reactive power control loop. In this
subsection, the DC voltage control loop of VSC-1 is selected to derive a closed-loop
model of the MTDC/AC power system.
As indicated in Fig. 7.1, the power output from VSC-1 is P1 þ jQ1. The terminal
voltage of VSC-1 expressed in its d q coordinate is Vd1 þ jVq1. Normally, the
direction of the terminal voltage of the VSC is chosen as that of d axis of the d q
coordinate, such that ΔVd1 ¼ ΔV1 and ΔVq1 ¼ 0, where V1 is the magnitude of the
terminal voltage of VSC-1.
The dynamic equations of the AC output current from VSC-k expressed in the
d q coordinate are given in (7.18).
From Fig. 7.2,

ref
Id1 ¼ Kdc ðsÞ Vdc1 Vdc1
ref
ref ð7:68Þ
ref
Vcd1 Vd1 þ X f1 Iq1 ¼ Kid ðsÞ Id1 Id1
where Vdc1 is the DC voltage of VSC-1, C1 is the DC capacitance of VSC-1, Idc1 is

the DC current of VSC-1.
Combining (7.68) and (7.18) and ignoring the very fast transient of the PWM
algorithm such that Vcd1 ¼ Vcd1
ref
,
dId1 ref
X f1 ¼ ω0 Kid ðsÞ Id1 Id1 ð7:69Þ
dt
From (7.69)
ω0 Kid ðsÞ
Id1 ¼ I ref ð7:70Þ
X f1 s þ ω0 Kid ðsÞ d1
Active power balance equation on the DC and AC side of VSC-1 is

dVdc1
Vdc1 IMTDC1 C1 Vdc1 ¼ P1 ð7:71Þ
dt
7.3 Method of Open-loop Modal Analysis to Examine the Impact of Dynamic. . . 287
Linearization of the first equation of (7.68), (7.70), and (7.71) is

dΔVdc1
C1 Vdc10 ¼ IMTDC10 ΔVdc1 þ Vdc10 ΔIMTDC1 ΔP1
dt
ð7:72Þ
ω0 Kid ðsÞ
ΔId1
ref
¼ Kdc ðsÞΔVdc1 , ΔId1 ¼ ΔI ref
X f1 s þ ω0 Kid ðsÞ d1
From (7.72)
ΔId1 ¼ H0 ðsÞΔVdc1
Vdc10 1 ð7:73Þ
ΔVdc1 ¼ ΔIMTDC1 ΔP1
C1 Vdc10 s IMTDC10 C1 Vdc10 s IMTDC10
where, H0 ðsÞ ¼ Xω0f1Ksþω

id ðsÞKdc ðsÞ
0 Kid ðsÞ
.
Since Vq1 ¼ 0, Vd1 ¼ V1,
ΔP1 ¼ Id10 ΔV1 þ Vd10 ΔId1 ð7:74Þ
From (7.73) and (7.74),

ΔP1 ¼ Hv1 ðsÞΔV1 þ Hi1 ðsÞΔIMTDC1 ð7:75Þ
where
Id10 ðsC1 Vdc10 IMTDC10 Þ
Hv1 ðsÞ ¼ ,
ðsC1 Vdc10 IMTDC10 Þ þ Vd10 H0 ðsÞ
Vd10 Vdc10 H0 ðsÞ
Hi1 ðsÞ ¼
ðsC1 Vdc10 IMTDC10 Þ þ Vd10 H0 ðsÞ
From (7.75), it can be seen that ΔV1 and ΔIMTDC1 are the input variables to the
selected DC voltage control of VSC-1 and ΔP1 is the output variable from the
selected DC voltage control of VSC-1. This implies that to the rest of MTDC/AC
power system, excluding the selected DC voltage control system of VSC-1, ΔV1 and
ΔIMTDC1 are the output variables and ΔP1 is the input variable.
From (7.75), the state-space model of the DC voltage control system of VSC-1
can be written as
d
ΔXp ¼ Ap ΔXp þ bd ΔIMTDC1 þ bv ΔV1
dt ð7:76Þ
ΔP1 ¼ cp T ΔXp þ dd ΔIMTDC1 þ dv ΔV1
where ΔXp is the vector of all the state variables of the DC voltage control system of
VSC-1.
Since the input variable is ΔP1 and output variables are ΔV1 and ΔIMTDC1 for the
rest of the MTDC/AC power system, excluding the DC voltage control system of
VSC-1, the state-space model of the rest of the MTDC/AC power system can be
written as
d
ΔXs ¼ As ΔXs þ bp ΔP1
dt ð7:77Þ
ΔIMTDC1 ¼ cd T ΔXs þ d1 ΔP1 , ΔV1 ¼ cv T ΔXs þ d2 ΔP1
where ΔXs is the vector of all the state variables of the rest of the MTDC/AC system,
excluding those in the DC voltage control system of VSC-1. From (7.77), the
following transfer function model of the rest of the MTDC/AC system can be
obtained
ΔIMTDC1 ¼ Gi1 ðsÞΔP1 , ΔV1 ¼ Gv1 ðsÞΔP1 ð7:78Þ
where Gi1(s) ¼ cdT(sI As)‐1bp þ d1, Gv1(s) ¼ cvT(sI As)1bp þ d2.

From (7.75) and (7.78), a closed-loop model with the DC voltage control system
of VSC-1 in the feedback loop can be established, and is shown in Fig. 7.41. The
state-space model of the closed-loop interconnected system can be obtained from
(7.76) and (7.77) as
d
ΔX ¼ AΔX ð7:79Þ
dt
where.
T
ΔX ¼ ΔXp T ΔXs T
2 3
ðd1 bd þ d2 bv Þcp T ðd1 bd þ d2 bv Þðdd cd T þ dv cv T Þ
6 Ap þ 1 d d d d bd cd T þ bv cv T þ
1 dd d1 dv d2 7
6 d 1 v 2 7
A¼6 7
4 b p cp T dd bp cd T þ dv bp cv T 5
As þ
1 dd d 1 dv d2 1 dd d1 dv d2
Fig. 7.41 Closed-loop Subsystem

interconnected model with
: rest of Gi1 (s) ΔIMTDC1
the DC voltage control loop ΔP1
power
in the feedback loop
system
Gv1 (s) ΔV1
Hv1 (s) Subsystem:

DC voltage
control of
Hi1 (s) VSC-1
7.3.1.2 Closed-Loop Model When the Active or Reactive Power Control

Is Selected
VSC-k, k ¼ 2, 3, . . .M use the active power control, configuration of which is shown

by Fig. 7.3. The active power control system of VSC-k can be selected to derive a
closed-loop model similar to that shown by Fig. 7.41 as follows.
When the active power control system of VSC-k is considered, (7.68) becomes

ref
Idk ¼ Kpk ðsÞ Pk Pkref
ref ð7:80Þ
ref
Vcdk Vdk þ X fk Iqk ¼ Kidk ðsÞ Idk Idk
From Linearized (7.18) and (7.80),

ω0 Kidk ðsÞ
ΔIdk ¼ ΔI ref , ΔIdk
ref
¼ Kpk ðsÞΔPk ð7:81Þ
X fk s þ ω0 Kidk ðsÞ dk
From (7.81),
ΔIdk ¼ H0k ðsÞΔPk ð7:82Þ
where H0k ðsÞ ¼ X fkωsþω

0 Kidk ðsÞ
0 Kidk ðsÞ
Kpk ðsÞ.
Since Vqk ¼ 0, Vdk ¼ Vk,
ΔPk ¼ Idk0 ΔVk þ Vdk0 ΔIdk ð7:83Þ
From (7.82) and (7.83),

Idk0
ΔPk ¼ Hpk ðsÞΔVk , Hpk ðsÞ ¼ ð7:84Þ
1 Vdk0 H0k ðsÞ
Transfer function model of the selected active power control system of VSC-k is
(7.84). From (7.84), it can be seen that ΔVk is the input signal of the selected active
power control system of VSC-k and the output variable of the rest of the MTDC/AC
system, and ΔPk is the output variable of the selected active power control system of
VSC-k and the input variable of the rest of MTDC/AC system. Hence, following
state-space model of the selected active control system of VSC-k can be obtained.
d
ΔXp ¼ Ap ΔXp þ bpv ΔVk
dt ð7:85Þ
ΔPk ¼ cp T ΔXp þ dpv ΔVk
where ΔXp is the vector of all the state variables of the selected active control system
of VSC-k.
The state-space model for the rest of the MTDC/AC system in Fig. 7.1, excluding
the active control system of VSC-k, can be written as
d
ΔXsp ¼ Asp ΔXsp þ bsp ΔPk
dt ð7:86Þ
ΔVk ¼ cvp T ΔXsp þ dsp ΔPk
where ΔXsp is the vector of all the state variables of the rest of the MTDC/AC system
excluding those in the selected active control system of VSC-k. From (7.884), the
following transfer function model of the rest of the MTDC/AC system can be
obtained
1
ΔVk ¼ Gpk ðsÞΔPk , Gpk ðsÞ ¼ cvp T sI Asp bsp þ dsp ð7:87Þ
From (7.84) and (7.87), the closed-loop model is obtained, and shown in
Fig. 7.42.
The reactive power control of VSC-k, k ¼ 1, 2, 3, . . .M can also be selected to
derive a closed-loop model of the MTDC/AC power system similar to that shown by
Fig. 7.42. The derivation is as follows.
From Fig. 7.3,

ref
Iqk ¼ Kqk ðsÞ Qk Qkref
ð7:88Þ
ref
Vcqk Vqk X fk Idk ¼ Kiqk ðsÞ Iqk
ref
Iqk
From linearized (7.18) and (7.88),

ω0 Kiqk ðsÞ
ΔIqk
ref
¼ Kqk ðsÞΔQk , ΔIqk ¼ ΔI ref ð7:89Þ
X fk s þ ω0 Kiqk ðsÞ qk
From (7.89),
ω0 Kiqk ðsÞKqk ðsÞ
ΔIqk ¼ H1k ðsÞΔQk , H1k ðsÞ ¼ ð7:90Þ
X fk s þ ω0 Kiqk ðsÞ
Since Vqk ¼ 0, Vdk ¼ Vk, linearized reactive power output equation from VSC-k
can be obtained as
Fig. 7.42 Closed-loop Subsystem: rest of the system

model with active power
control system of VSC-k in ΔPk ΔVk
the feedback loop Gpk (s)
Hpk (s)
Subsystem: active power control

of VSC-k
ΔQk ¼ Iqk0 ΔVk Vdk0 ΔIqk ð7:91Þ
From (7.90) and (7.91),

Iqk0
ΔQk ¼ Hqk ðsÞΔVk , Hqk ðsÞ ¼ ð7:92Þ
1 þ Vdk0 H1k ðsÞ
The transfer function model of the selected reactive power control system of VSC-k
is (7.92). From (7.92), it can be seen that ΔVk is the input variable of the selected reactive
power control system of VSC-k and the output variable of the rest of the MTDC/AC
system, ΔQk is the output variable of the selected reactive power control system of
VSC-k and the input variable of the rest of the MTDC/AC system. Thus, the following
state-space model of the selected reactive control system of VSC-k is obtained.
d
ΔXq ¼ Aq ΔXq þ bqv ΔVk
dt ð7:93Þ
ΔQk ¼ cq T ΔXq þ dqv ΔVk
where ΔXq is the vector of all the state variables of the selected reactive control
system of VSC-k.
The state-space model for the rest of MTDC/AC system in Fig. 7.1, excluding the
selected reactive control loop of VSC-k, can be written as
d
ΔXsq ¼ Asq ΔXsq þ bq ΔQk
dt ð7:94Þ
ΔVk ¼ cvq T ΔXsq þ dq2 ΔQk
where ΔXsq is the vector of all the state variables of the rest of the MTDC/AC system,
excluding those in the selected reactive control loop of VSC-k. From (7.94), the
following transfer function model of the rest of the MTDC/AC system can be obtained
1
ΔVk ¼ Gqk ðsÞΔQk , Gqk ðsÞ ¼ cvq T sI Asq bq þ dq2 ð7:95Þ
From (7.92) and (7.95), the closed-loop model is derived, and is shown in
Fig. 7.43.
Fig. 7.43 Closed-loop Subsystem: rest of the system

model with reactive power
control system of VSC-k in ΔQk ΔVk
the feedback loop Gqk (s)
Hqk (s)
Subsystem: reactive power control

of VSC-k
7.3.1.3 Open-Loop Modal Analysis
Denote λti as an oscillation mode of the open-loop subsystem of the rest of the power
system and λpk as a complex pole (oscillation mode) of the open-loop subsystem of
the DC voltage control system of VSC-1. λti is a complex eigenvalue of the open-
loop state matrix As in (7.77). λpk is a complex eigenvalue of open-loop state matrix
Ap in (7.76). Denote b λ ti and b λ pk as the closed-loop oscillation modes corresponding to
λti and λpk, respectively. b λ ti and b λ pk are the complex eigenvalues of the closed-loop
state matrix A in (7.79).
When ΔP1 ¼ 0, the closed-loop system shown in Fig. 7.39 is open. Closed-loop
and corresponding open-loop eigenvalues are the same. Hence, the differences
between the closed-loop and corresponding open-loop oscillation modes, Δλti ¼ b λ ti
λti and Δλpk ¼ b λ pk λpk , are caused by ΔP1 6¼ 0, which is the physical exhibition
of the AC/DC dynamic interactions between the DC voltage control system of
VSC-1 and the rest of the MTDC/AC power system. Normally, the dynamic
interactions are weak, such that ΔP1 0. The system shown in Fig. 7.41 should
approximately be open such that Δλti ¼ b λ ti λti 0 and Δλpk ¼ b
λ pk λpk 0.
This implies that the weak dynamic interactions caused by the DC voltage control
system of VSC-1 should normally affect little the oscillation modes in the MTDC/
AC system. However, under a special condition of open-loop modal resonance
(OLMR), ΔP1 may become considerable, which is elaborated as follows.
The special condition of OLMR is when λpk is close to λti on the complex plane,
i.e., λpk λti. Since λpk is a complex pole of Hi1(s) or/and Hv1(s) in Fig. 7.41, |
Hi1(λpk)| ¼ 1 or/and |Hv1(λpk)| ¼ 1; thus |Hi1(λti)| or/and |Hv1(λti)| are significant
when λpk λti. From (7.75), it can be seen that ΔP1 may become considerable
around the oscillation complex frequency λti. Hence, when λpk λti, the dynamic
interactions introduced by the DC voltage control system of VSC-1 may become
strong. The impact of strong dynamic interactions caused by the OLMR is Δλti ¼ b λ ti
λti and Δλpk ¼ b λ pk λpk . The derivation of analytical expressions of Δλti and Δλpk
may be very difficult to determine the exact impact of the dynamic interactions
introduced by the DC voltage control system of VSC-1. However, an approximate
estimation of Δλti and Δλpk can be derived as follows.
From Fig. 7.41, the characteristic equation of the closed-loop system is
obtained as
1 ¼ Hi1 ðsÞGi1 ðsÞ þ Hv1 ðsÞGv1 ðsÞ ð7:96Þ
The transfer functions of open-loop subsystems in Fig. 7.41 can be written as
R1p Xnh
R R Xnh
R
Hi1 ðsÞ ¼ þ dd þ 1pn , Hv1 ðsÞ ¼ 2p þ dv þ 2pn
s λpk n¼1 s λ pn s λ pk n¼1 s λpn
R1ti Xng
R1k R2ti Xng
R2k
Gi1 ðsÞ ¼ þ dp1 þ , Gv1 ðsÞ ¼ þ dp2 þ
ðs λti Þ k¼1
ð s λk Þ ðs λti Þ k¼1
ðs λ k Þ
ð7:97Þ
where λpk and λpn, n ¼ 1, 2, . . .nh are the eigenvalues of Ap in (7.76), Rmp and Rmpn
(m ¼ 1, 2, n ¼ 1, 2, . . .nh) are the corresponding residues; λti and λk, ( k ¼ 1, 2, . . .ng)
are the eigenvalues of As in (7.77), Rmti and Rmk, m ¼ 1, 2, k ¼ 1, 2, . . .ng are the
corresponding residues.
Substitute (7.97) into (7.96); replace s by b λ ti and then multiply both sides of the
2
b
equation by λ ti λti . With λpk λti, it can have
( " #)
2 Xnh
R
bλ ti λti R1p þ b λ ti λti dd þ 1pn

n¼1
s λpn
( " #)
Xng
R1k
b
R1ti þ λ ti λti dp1 þ þ
k¼1
ð s λk Þ
( " #) ð7:98Þ
X nh
R2pn
b
R2p þ λ ti λti dv þ
k¼1
s λpn
( " #)
Xng
R2k
b
R2ti þ λ ti λti dp2 þ
k¼1
ð s λk Þ
In the neighborhood of λpk λti (b λ ti ! λti), the following estimation of Δλti ¼ b λ ti

λti is obtained from (7.98)
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
lim Δλti ¼ lim b λ ti λti R1p R1ti þ R2p R2ti ð7:99Þ
bλ ti !λti bλ ti !λti
The above estimation will become more meaningful when the estimation of Δλpk
¼b λ pk λpk is established. b λ pk is also a solution of (7.96). By taking the derivation
similar to that from (7.96) to (7.99) for b λ pk , it can have
lim Δλpk ¼ lim b λ pk λpk R1p R1ti þ R2p R2ti ð7:100Þ
bλ pk !λpk bλ pk !λpk
From (7.99) and (7.100)

b
λ ti λti Rki , b
λ pk λpk Rki λti Rki ð7:101Þ
where, for the convenience of discussion, Rki ¼ R1p R1ti þ R2p R2ti is called the
open-loop modal residue.
From (7.101), it can be seen that when the OLMR occurs, in the neighborhood of
λpk λti, b λ pk and bλ ti are located at the positions approximately opposite to each other
on the complex plane with respect to the positions of λpk λti. Thus, it is likely that
either b λ pk or bλ ti may be on the right hand side of λpk λti. Therefore, strong dynamic
interactions caused by the OLMR may possibly degrade the damping of either b λ pk or
b
λ ti . In addition, if |Real part of Rki| >|Real part of λpk or λti|, it is possible that either
b
λ pk or b
λ ti may become negatively damped such that growing oscillations occur in the
MTDC/AC power system.
From the above discussions, it can be seen that to examine the risk caused by the
AC/DC dynamic interactions introduced by the DC voltage control system of the
VSC, the OLMR first needs to be identified. Secondly, the degree of stability
degradation caused by the identified OLMR is estimated by computing the open-
loop modal residue Rki. The identification and estimation can be achieved from the
results of modal analysis for the open-loop subsystems. The state-space model of the
open-loop MTDC/AC power system can be obtained by assuming ΔP1 ¼ 0, i.e.,
assuming the active power output from VSC-1 to be a constant P10.
Similarly, when the active power control system or reactive power control system
is selected rather than the DC voltage control system, the active power and reactive
power outputs from the VSC are assumed to be constant. With the assumption, the
state-space model of the open-loop MTDC/AC power system can be derived. From
the modal analysis of the derived open-loop subsystems, the OLMR can be identified
and the degree of stability degradation caused by the identified OLMR can be
estimated.
Therefore, the following method of open-loop modal analysis to examine the
impact of AC/DC dynamic interactions introduced from a selected VSC control
system on the small-signal stability of the MTDC/AC power system can be
proposed.
1. Modelling all the VSCs connecting the MTDC network and AC power system as
constant power sources, Pk0 þ jQk0, k ¼ 1, 2, . . .M; the direction of terminal
voltage of each VSC is taken as that of the d axis of the d q coordinate of the
VSC; establish the state-space model of the AC power system. From the
established open-loop state matrix, calculate and identify the open-loop oscilla-
tion modes λti, i ¼ 1, 2, . . ., of the AC system.
2. Calculate the open-loop oscillation modes λpk, k ¼ 1, 2, . . ., of the control
systems of the VSCs
3. If any λpk λti is found, the corresponding open-loop modal residue Rki is
calculated. By applying the criterion |Real part of Rki| > |Real part of λpk or λti|,
the potential risk of growing oscillations is identified. In addition, the trouble-
making control systems of the VSCs and the dynamic components, such as the
SGs, in the AC power system responsible for the growing oscillations are
identified.
4. The parameters of the identified control systems of the VSCs are tuned to move
λpk away from λti on the complex plane. Thus, the identified OLMR is dismissed
and the risk of growing oscillations is eliminated.
7.3.2 An Example MTDC/AC Power System for Applying

the Open-Loop Modal Analysis
7.3.2.1 Example 7.5—Small-Signal Instability Risk Caused by

the OLMR
Figure 7.44 shows the configuration of an example MTDC/AC power system. The
part of AC power system is the well-known New England power system (NEPS).
The parameters of loads and network of the NEPS given in [4] were used. The model
of the SGs given in [15] for studying the sub-synchronous resonance (SSR) was used
and their parameters were adjusted according to the capacity of the SGs. The MTDC
network had eight terminals. Four terminals were connected by VSC-k, k ¼ 1, 2, 3,
4 to the nodes of the NEPS. VSC-1 used the DC voltage control and VSC-k (k ¼ 2,
3, 4) adopted the active power control. The active power output from VSC-k (k ¼ 1,
2, 3, 4) at the steady state was 3.43 p.u, 1 p.u, 0.5 p.u, and 0.5 p.u, respectively.
These four VSCs operated with a unity power factor. Each of the other four VSCs
was connected to a wind farm represented by a PMSG. In addition, a wind farm
represented by a PMSG was connected at node 39. The detailed 14th-order model
and parameters of the PMSG recommended in [8] were used. Model and parameters
of the MTDC network given in [14] were used. Data of the example power system is
given in Appendix 7.4.
First, VSC-k (k ¼ 1, 2, 3, 4) were modeled as constant power sources. The open-
loop state-space model of the NEPS was established. Open-loop oscillation modes of
9
25 26
28 38
27
2 18 16
37 29
19 24
30 8 6
1 20
17 35
1 33 7
3
34 36
15 4
10 4 5 22
14
PMSG 21 23
39 12 42 40 43
GSC MSC 11 13 VSC-3 VSC-1 VSC-4
5 10
6
9 7 r MTDC NET r
32
31
8 3 VSC-2
2 NEPS
41
WIND FARMs
Fig. 7.44 Configuration of a test MTDC/AC power system

Table 7.1 Selected oscillation modes of the MTDC/NEPS

Open-loop SSO Closed-loop SSO
modes Related dynamic components modes
λt1 ¼ 0:69 Mass of the second low pressure cylinder (LPC) b
λ t1 ¼ 0:02
þj91:51 of SG1 þj91:12
λt2 ¼ 0:84 Mass of the second LPC of SG6 b
λ t2 ¼ 0:83
þj97:75 þj97:31
λt3 ¼ 0:71 Mass of the second LPC of SG5 b
λ t3 ¼ 0:71
þj100:8 þj100:8
λt4 ¼ 5:13 The grid side converter (GSC) of the PMSG at b
λ t4 ¼ 5:18
þj90:03 node 39 þj89:98
λp1 ¼ 1:98 DC voltage control system of VSC-1 b
λ p1 ¼ 2:58
þj90:98 þj90:12
the NEPS were calculated. In the first column of Table 7.1, four selected open-loop
sub-synchronous oscillation (SSO) modes of the NEPS are listed as λti, i ¼ 1, 2, 3,
4. The dynamic components related to λti, i ¼ 1, 2, 3, 4 were identified by calculating
the PFs from the open-loop state-space model of the NEPS. Results of identification
are given in the second column of Table 7.1.
Secondly, the open-loop oscillation modes of the MTDC network were calcu-
lated. λp1 related with the DC voltage control system of VSC-1 was found to be close
to the torsional SSO mode of SG1, λt1 (see the last row in the first column of
Table 7.1).
Thirdly, the open-loop modal residue for the OLMR, λp1 λt1, was calculated
and found to be R11 ¼ 1.96 j0.45. Hence, |Real part of R11| > |Real part of λt1 |,
indicating that the OLMR may cause poorly or negatively damped torsional SSOs in
the MTDC/NEPS.
To confirm that the OLMR, λp1 λt1, may have caused the damping degradation
of the SSOs in the MTDC/NEPS, the closed-loop oscillation modes were calculated,
and are presented in the last column of Table 7.1. It can be seen that the closed-loop
oscillation modes, b λ t1 and b
λ p1 , were located at the approximately opposite positions
on the complex plane with respect to λp1 λt1, confirming the conclusion made from
the open-loop modal analysis. The closed-loop torsional SSO mode, b λ t1 , of SG1 was
poorly damped, confirming that poorly damped torsional SSOs occur, caused by the
DC voltage control of VSC-1.
The OLMR is different from the closeness of frequency of the open-loop oscil-
lation modes. For example, in Table 7.1, the frequency (imaginary part) of λt4 is
close to that of λp1. However, on the complex plane, λt4 is not in the proximity of λp1
because the real parts of λt4 and λp1 are very different.
7.3.2.2 Parameter Tuning of the VSC Control
Tuning parameters of the DC voltage control system of VSC-1 can move the open-
loop oscillation mode, λp1, away from the open-loop torsional SSO mode, λt1, to
eliminate the OLMR, λp1 λt1. With the OLMR being eliminated, the damping of
the closed-loop torsional SSO mode, b λ t1 , can be improved. Hence, the integral gain
of the DC voltage control Kdci was tuned within the range from Kdci ¼ 0.93 to
Kdci ¼ 1.54. With Kpi being tuned, the trajectory of λp1 on the complex plane is
displayed as the dashed curve in Fig. 7.45. The positions of the open-loop torsional
SSO modes, λti, i ¼ 1, 2, 3, listed in the first column of Table 7.1, are indicated by
hollow circles in Fig. 7.45. It can be observed that when Kdci was tuned, λp1 was in
the proximity of λti, i ¼ 1, 2, 3 consecutively at points ①, ②, and ③. Point ① was
the case of OLMR, λp1 λt1, identified above, while points ② and ③ were the new
cases of OLMR of the DC voltage control system of VSC-1 with the open-loop
torsional SSO modes of SG5 and SG6, respectively.
The computational results of the open-loop modal residues for the new cases of
open-loop SSO modal proximity at points ② and ③ were R12 ¼ 1.76 þ j0.06 and
R13 ¼ 1.61 j0.03, respectively. Since |Real part of R1k| > |Real part of λtk |, k ¼ 2,
3, it can be predicted that the OLMR at points ② and ③ may probably lead to the
poorly or negatively damped SSOs in the MTDC/NEPS. From the open-loop modal
residues, the estimated positions of the closed-loop oscillation modes were obtained
by using (7.101) and indicated by crosses in Fig. 7.45.
To confirm the above prediction made from the open-loop modal analysis, the
trajectories of closed-loop oscillation modes, b λ ti , i ¼ 1, 2, 3 and b
λ p1 , are displayed as
solid curves in Fig. 7.45. The positions of closed-loop SSO modes, when the OLMR
occurred at point ①, ②, and ③, are indicated by filled circles. They confirmed that
the OLMR degraded the damping of closed-loop torsional SSO modes of the related
SGs. When λp1 was away from λti, i ¼ 1, 2, 3 by tuning Kpi, the damping of the
closed-loop torsional SSO modes was improved as they moved toward where the
corresponding open-loop torsional SSO modes were located on the complex plane.
Imag axis (rad/s)

102
3 lt3 lˆt3
2
lt2 lˆt2
92
1 lt1 lˆt1
lˆ p1 lp1
Real axis
82
-4 -3 -2 -1 0 1
Fig. 7.45 OLMR caused by tuning Kpi

Imag axis (rad/s)
92 λˆp1 λˆt1
λˆt4 λt4
λt1
4 λp1 1
Real axis
87
-6.5 -4.5 -2.5 -0.5
Fig. 7.46 Trajectories of oscillation modes when Kpp is tuned
Three SGs involved in the open-loop modal analysis were at different geograph-
ical locations in the NEPS. Their electric distances to VSC-1 were different. How-
ever, the results presented in Fig. 7.45 indicated that the strong dynamic interactions
can occur between VSC-1 and the torsional dynamics of those SGs, leading to
poorly damped torsional oscillations. The impact of OLMR was determined by the
open-loop modal residue defined in (7.101).
Figure 7.45 shows that tuning Kpi may improve the damping of related closed-
loop torsional SSO modes. However, it may also cause new cases of OLMR. Hence,
the proportional gain Kdcp of the DC voltage control system of VSC-1 was tuned
within the range of Kdcp ¼ 0.16 to Kdcp ¼ 0.36. The trajectory of λp1 on the complex
plane with Kdcp being tuned is displayed as the dashed curve in Fig. 7.46. The solid
curves are the trajectories of the related closed-loop oscillation modes. From
Fig. 7.46, it can be observed that when Kdcp was tuned, λp1 moved away from λt1
toward the left on the complex plane such that the damping of closed-loop torsional
modes b λ t1 and b
λ p1 was improved. However, the OLMR occurred between λp1 and λt4
at point ④, indicating the strong dynamic interactions between the VSC-1 and the
PMSG at node 39. The open-loop modal residue was calculated and found to
be R14 ¼ 2.25 þ j1.04. Since |Real part of R14| << |Real part of λp1 or λt4|, the
OLMR at point ④ did not cause poorly damped SSOs in the MTDC/NEPS.
Results presented in Fig. 7.46 indicated that tuning the proportional gain of the
DC voltage control system of VSC-1 was a better solution for eliminating the OLMR
at point ① in Fig. 7.46, although a new case of OLMR occurred at point ④, because
all the related closed-loop oscillation modes were sufficiently damped. This test
shows that when the parameters of VSC control are set, it is beneficial to assign more
damping to the open-loop oscillation modes of the VSCs. In this case, even if the
OLMR occurs, the related closed-loop oscillation modes may probably have had
enough damping initially, such that the OLMR would not cause poorly or negatively
damped oscillations in the MTDC/AC system.
Confirmation from the non-linear simulation is presented in Fig. 7.47. At 0.1 s of
the simulation, the load at node 10 was lost by 80%. The lost load was recovered in
100 ms. From Fig. 7.47, it can be observed that when the OLMR occurred at point ①
in Fig. 7.46, the poorly damped torsional SSOs of SG1 occurred. After Kpp was
tuned to move λp1 to point ④, the poorly damped torsional SSOs disappeared.
Open-loop modal proximity at a No open-loop modal proximity at b

3.63 1.04
Variation of active power output

Variation of active power output
from VSC-2 (p.u.)

3.53 1.02
from VSC-1 (p.u.)
3.43 1.0
3.33 0.98
Time(s) Time(s)
3.23 0.96
0 0.5 1 1.5 2 0 0.5 1 1.5 2
0.38
Variation of torque on the
2nd LPC of SG1 (p.u.)
0.33
0.28
0.23
Time(s)
0.18
0 0.5 1 1.5 2
Fig. 7.47 Confirmation from non-linear simulation
Table 7.2 Data of the Parameters Values

VSC-MTDC system in
VSC rated power 200 (MW)
example 7.1
VSC rated direct voltage 150 (kV)
AC voltage 220 (kV)
DC impedance base 225(Ω)
DC capacitance base 14.15 (uF)
Converter capacitor C ¼ 33p. u.
The DC network is cyclic with r ¼ 0.002 p.u
Table 7.3 Data of VSC control system in case A

DC voltage control loop Kdc(s) ¼ 1 þ 30/s
Active power control loop: Kp(s) ¼ 0.3 þ 2.4/s
Inner current control loop: Kid(s) ¼ Kiq(s) ¼ 0.15 þ 88.48/s
Data of New England power system is given in Appendix 4.2 (Tables 7.2, 7.3 and
7.4).
Table 7.4 Data of VSC control system in case B

DC voltage control loop Kdc(s) ¼ 2 þ 2.8/s
Table 7.5 Data of the Parameters Values

VSC-MTDC system in
Converter capacitor C ¼ 33 p. u.
examples 7.2 and 7.4
The DC system is CIREG network with r ¼ 0.002 p.u
Table 7.6 Data of VSC control system (master slave control)

DC voltage control loop Kdc(s) ¼ 0.75 þ 5/s
Table 7.7 Data of the VSC-MTDC network in example 7.3

Parameters Values
Converter capacitor 5.65 (p.u).
DC cable’s resistance 0.038 Ω/km
DC cable’s inductance 0.189 mH/km
DC cable’s capacitance 0.27uF/km
DC cable’s length Indicated in Fig. 7.3
PI parameters of the active power control loop 0.5 þ 0.25/s (p.u)
PI parameters of the reactive power control loop 1.0 þ 0.13/s (p.u)
PI parameters of the inner current control loop 2.0 þ 0.23/s (p.u)
Data of New England power system is given in Appendix 4.2 (Tables 7.5 and 7.6).
Appendix 7.3: Data of Example 7.3 (Table 7.7)
Data of New England power system is given in Appendix 4.2 (Table 7.8).
References 301
Table 7.8 Data of the VSC-MTDC system in Example 7.1

Parameters Values
Converter capacitor 5.65 (p.u)
DC resistance R 0.05 (p.u)
Proportional parameter of the DC voltage control loop (VSC-1) 0.2 (p.u)
PI parameters of the active power control loop 0.5 þ 0.25/s (p.u)
PI parameters of the reactive power control loop 1.0 þ 0.13/s (p.u)
PI parameters of the current control loop 2.0 þ 0.23/s (p.u)
References
1. Wang XF, Song YH, Irving M (2011) Modern power systems analysis. Springer, Berlin
2. Wang HF, Du WJ (2016) Analysis and damping control of power system low-frequency
oscillations. Springer, New York
3. Seyranian AP (1993) Sensitivity analysis of multiple eigenvalues. Mech Struct Mach 21
(2):261–284
5. Dobson I, Zhang J, Greene S, Engdahl H, Sauer PW (2001) Is strong resonance a precursor to
power system oscillations. IEEE Trans Circuits Syst I Fundam Theory Appl 48(3):340–349
6. Padiyar KR, SaiKumar HV (2006) Investigations on strong resonance in multimachine power
systems with STATCOM supplementary modulation controller. IEEE Trans Power Syst 21
(2):754–762
8. Li SH, Haskew TA (2012) Optimal and direct-current vector control of direct-driven PMSG
wind turbines. IEEE Trans Power Electron 27(5):2325–2337
9. Lu W, Ooi BT (2003) Optimal acquisition and aggregation of offshore wind power by multi
terminal voltage-source HVDC. IEEE Trans Power Del 18(1):201–206
10. Dierckxsens C, Srivastava K, Reza M, Cole S, Beerten J, Belmans R (2012) A distributed DC
voltage control method for VSC MTDC systems. Elect Power Syst Res 82:54–58
11. Rouzbehi K, Miranian A, Candela JI, Luna A, Rodriguez P (2015) A generalized voltage droop
strategy for control of multiterminal DC grids. IEEE Trans Ind Appl 51(1):607–618
12. Anderson PM, Agrawal BL, Van Ness JE (1990) Subsynchronous proximity in power systems.
IEEE Press, New York
13. Haileselassie TM, Uhlen K (2012) Impact of DC line voltage drops on power flow of MTDC
using droop control. IEEE Trans Power Syst 27(3):1441–1449
14. Chaudhuri NR, Majumder R, Chaudhuri B (2013) System frequency support through multi-
terminal DC (MTDC) grids. IEEE Trans Power Syst 28(1):347–356
15. Anderson PM, Agrawal BL, Van Ness JE (1999) Subsynchronous resonance in power systems.
Wiley, IEEE Press, New York
Chapter 8
Probabilistic Analysis of Small-Signal
Stability of a Power System Affected
by Grid-Connected Wind Power Generation
The operation of power system is stochastic in nature, and the uncertainties can be
brought about by many aspects of the system, such as the stochastic variation of load
conditions and intermittence of renewable energy generations (e.g., wind and solar).
Analysis of power system small-signal stability introduced in this book so far (i.e.,
modal analysis and damping torque analysis) as well as system time-domain simu-
lation are based on the deterministic system operating conditions with specific
loading situation and constant network configuration. Hence, they are not able to
deal with stochastic fluctuations of random variables in power systems and may
bring impractical results to the system stability analysis. This limitation of deter-
ministic analysis motivates the research of probabilistic analysis into small-signal
stability issues, in which the uncertainty and randomness of power system behaviors
can be fully considered and discussed.
Probabilistic analysis of small-signal stability is able to take into account the
effects of the inherent stochastic nature of the power system and hence it is regarded
as an advanced analysis of power system small-signal stability. The rest of the
chapter is organized as follows: In Sect. 8.1, an introduction on the probabilistic
analysis of power system small-signal stability in terms of its framework and
methodologies is presented. Then the probabilistic analysis of power system small-
signal stability considering the impact of directly grid-connected onshore wind
power generation is carried out. An efficient cumulant theory based analytical
method is proposed for the probabilistic analysis to replace the time-consuming
Monte Carlo simulation. In Sect. 8.3, to examine the impact of offshore wind power
generation and multi-terminal DC (MTDC) network which connects the wind power
to the main grid, two probabilistic indices of small-signal stability margin of a hybrid
AC/DC power system are proposed. In addition, a novel variance indices analysis of
probabilistic small-signal stability of a power system as a derivation from the
cumulant theory is implemented.

304 8 Probabilistic Analysis of Small-Signal Stability of a Power System. . .
8.1 Probabilistic Analysis of Small-Signal Stability of a

Power System: An Introduction
During the last a few decades, many research efforts have been devoted to the
development of probabilistic analysis in small-signal stability. A number of proba-
bilistic analysis methodologies have been applied and the probabilistic analysis of
small-signal stability has been widely used in power system planning, assessing and
stabilizer designing etc. This section aims to give a brief introduction on the
framework and methodologies of the probabilistic analysis of small-signal stability.
8.1.1 Framework of Probabilistic Small-Signal Stability

Analysis
The principle task for probabilistic analysis of power system small-signal stability is
to determine the probability distribution of critical eigenvalue of the system
according to the distribution of stochastic uncertainty sources and hence the prob-
ability of the overall system small-signal stability margin can be computed. The
framework of probabilistic analysis of small-signal stability is displayed in Fig. 8.1.
Figure 8.1 is briefly explained below:
1. The model of stochastic uncertainty sources is usually described by Probabilistic
Density Function (PDF) or Cumulative Distribution Function (CDF) of consid-
ered stochastic variables [1, 2].
2. The sensitivity of the critical eigenvalue with respect to stochastic uncertainty
sources is required by every analytical method of probabilistic analysis, which
can be computed in either an analytical way [3] or a numerical way [4] after the
critical eigenvalue is determined.
3. The various probabilistic methodologies can be roughly classified into three
categories (i.e. numerical method, analytical method and combination of the
former two).
4. The probabilistic index employed in the analysis can be computed by
ð0
Pðξk < 0Þ ¼ Fξk ð0Þ ¼ f ξk ðxÞdx ð8:1Þ
1
where ξk is the real part of the kth eigenvalue λk (critical eigenvalue) of the power
system, F() and f() denote CDF and PDF of corresponding stochastic variables
respectively. P(ξk < 0) is the probabilistic index of overall system stability margin.
8.1 Probabilistic Analysis of Small–Signal Stability of a Power System. . . 305
Modeling of
Deterministic Power
System
Linearizing Established
System Model
a
Deterministic Modeling of Stochastic
Small-signal Stability Uncertainty Sources
Analysis
b
Critical Eigenvalue
(and Its Sensitivity with
Respect to Uncertainty
Sources)
c
Various Probabilistic Analysis Methodologies
Probability Distribution
of Critical Eigenvalue
d
Probabilistic Index of
System Small-signal
Stability Margin
Fig. 8.1 Framework of probabilistic analysis of small-signal stability of power systems
8.1.2 Methodologies of Probabilistic Small-Signal Stability

Analysis
There are mainly three types of methodologies used in the probabilistic analysis of
small-signal stability as shown in Fig. 8.1, namely numerical method, analytical
method and the combination of the numerical and analytical method.
8.1.2.1 Numerical Method
The numerical method to determine probabilistic small-signal stability refers to

usually so-called Monte Carlo simulation [5]. It is one of the most commonly-used
probabilistic method and successfully employed in [6, 7] to study the effect of

uncertainty of both generation and demand on power system small-signal stability.
Monte Carlo simulation is to generate a large number of random computational
scenarios according to the distribution density of the stochastic sources such as load,
wind and solar power in a power system. Accumulation of computational results of
deterministic power system small-signal stability (the critical eigenvalue) from all
the scenarios forms the distribution density of the critical eigenvalue, hence to
determine the power system probabilistic stability.
Obviously, Monte Carlo simulation is quite simple to perform and can provide
accurate results. However, it is a method of extremely time-consuming. Therefore, it
has been well accepted that the analytical method of probabilistic analysis is more
applicable, especially for the study of probabilistic stability of large-scale power
systems. Monte Carlo simulation can be used as a reference to evaluate and confirm
the correctness of probabilistic analysis by using analytical method.
8.1.2.2 Analytical Method
The first analytical method to study probabilistic small-signal stability of a power

system is introduced by Burchett and Heydt in 1978 [8]. Generalized tetrachoric
series is used to determine the stochastic distribution density of system critical
eigenvalue, and hence the probability of small-signal stability of the power system.
The influence of uncertainty in system parameters from several stochastic uncer-
tainty sources subject to the multivariate normal distribution is considered. This
analytical method is later developed to accommodate random variables with any
type of distribution by using a moment method in [9].
A series of work later [10–12] has further improved the various aspects of the
analytical methods of probabilistic small-signal stability. Since the research focus
are mainly the uncertainties of system parameters and loads, methods proposed in
[10–12] are only used to approximate Normal distribution. Hence, the mean and
variance of system eigenvalues are calculated in these studies. First-order Taylor
series expansion based method is applied to calculate the expectation value and
variance of system eigenvalues considering the multiple stochastic system parame-
ters in [10]. The stochastic uncertainty of system load is studied in [11, 12]. Covari-
ance matrix of eigenvalues is calculated according to the correlations of load by
using eigenvalue sensitivity with respect to multiple system operation modes in
[11]. Reference [12] proposes a generalized tetrachoric series based method to
compute the probabilistic small-signal stability region, in which all the system
eigenvalues are observed.
A cumulant theory based (or named Gram-Charlier series expansion based)
method of probabilistic small-signal stability analysis is proposed to investigate
the stochastic impact of load and generation on system small-signal stability in
[13]. This analytical method removes the assumption of Normal distribution in
contrast to the methods presented above and can approximate any type of distribu-
tions with comparatively high accuracy. Moreover, it is computational effective and
8.2 Probabilistic Analysis of Small-Signal Stability of a Power System. . . 307
capable of handling large-scale power systems with high efficiency. Hence, the
method is employed to investigate the stochastic impact of grid-connected wind
power generation in the next section.
8.1.2.3 Combination of Numerical and Analytical Method
Although the problem with great amount of computation for Monte Carlo simulation
can be avoided by using analytical method, approximation and complicated math-
ematical analysis are usually needed by analytical method for the consideration of
the convergence within reasonable time. Therefore, there is a need for developing
new methods which can alleviate the calculation burden while holding satisfactory
computing accuracy.
Point estimate method, as a combination of numerical and analytical method, is
proposed to tackle the problem mentioned above. A Two Point Estimate (TPE)
method is presented in [14, 15] to analyze probabilistic stability. For a system with
m stochastic variables, only 2m times of deterministic eigenvalue calculations are
needed to estimate the distribution of the critical eigenvalue according to the
distribution of stochastic variables by using TPE based method. In contrast, huge
amount of numerical calculation may be required by Monte Carlo simulation
especially when n is large. Other approximate techniques combining numerical
and analytical method also include Probabilistic Collocation Method (PCM) and
First-Order Second Moment Method (FOSMM) etc. PCM is applied to consider the
uncertainty brought about by stochastic load powers and load models in [16].

of a Power System Affected by Directly Grid-Connected
Onshore Wind Power Generation
This section presents the probabilistic analysis of small-signal stability of a power

system as affected by directly grid-connected onshore wind power generation so as
to demonstrate certain aspects of probabilistic analysis of power system small-signal
stability. The cumulant theory based method is employed to derive the PDF of
system critical eigenvalues from the well-known Weibull distribution of wind speed.
The method of probabilistic analysis presented in this section can determine the
probabilistic small–signal stability of a power system penetrated by multiple onshore
wind power sources by just performing the step-by-step computation proposed once.
Hence it successfully avoids the complex convolution calculation of some other
analytical methods and high computational burden of the Monte Carlo simulation.

of Probabilistic Small-Signal Stability Analysis of a
Conventional AC Power System
A simple case without considering the spatial correlations of grid-connected wind

power generation is given firstly in order to give a clear presentation of the cumulant
theory based analytical method. Then the case of taking account of the correlations
between different wind power sources is further demonstrated. The analytical
method is modified to accommodate the more realistic case that the grid-connected
wind generation could be spatially correlated.
8.2.1.1 Distribution Function and Associated Parameters of Stochastic

Variation of a Grid-Connected Onshore Wind Power
Generation
Weibull (or normal skew) distribution is considered to be one of the most applicable
description of stochastic fluctuation of wind power generation, which is presented as
follows
8
>
> 1 Fwspeedi ðv fi Þ Fwspeedi ðvci Þ δðPwi Þ,
>
>
>
>
>
> for Pwi ¼ 0
>
>
>
> k0i -1 " k0i #
>
> k P k P k
>
>
0i wi 2i
exp
wi 2i
,
< k1i c0i k1i c0i k1i c0i
f wpoweri ðPwi Þ ¼ ð8:2Þ
>
> for 0 < Pwi < Pri
>
>
>
>
>
> Fwspeedi ðv fi Þ Fwspeedi ðvri Þ δðPwi Pri Þ,
>
>
>
>
>
> for Pwi ¼ Pri
>
>
:
0, for Pwi < 0 or Pwi > Pri
8
k
>
> Pwi k2i 0i
>
> 1 exp þ 1 Fwspeedi ðv fi Þ,
>
>
k1i c0i
>
>
>
< for 0 Pwi < Pri
Fwpoweri ðPwi Þ ¼ ð8:3Þ
>
> 1, for Pwi ¼ Pri
>
>
>
> Pwi > Pri
>
> 1, for
>
:
0, for Pwi < 0
where Pwi is the active power supplied by the ith wind power generation source (wind
farm) connected to a multi-machine power system, fwpoweri() and Fwpoweri() is the
PDF and CDF of the wind power, vci the cut-in wind speed, vri the rated wind speed,
vfi the furling wind speed, Fwspeedi() the CDF of the Weibull distribution of wind
speed, δ() the impulse function and Pri the rated wind power. Parameters in (8.2) and
(8.3) are given by the following equations
k0i ¼ ðσi =μi Þ1:086 , c0i ¼ μi =Γð1 þ 1=k0i Þ,

ð8:4Þ
k1i ¼ Pri =ðvri vci Þ, k2i ¼ k1i vci
where Γ() is a Γ function, μi and σi is the mean and standard deviation of wind speed
respectively.
The wind power variation can be defined to be ΔPwi ¼ Pwi Pw0i, where Pw0i is
the deterministic wind power generation. According to the probability theory, the nth
moment αnΔPwi of the wind power variation, ΔPwi, can be computed from (8.2) or
(8.3) as follows
Ð P P Ð P P
αn ΔPwi ¼ Pri w0i w0i xn dFwpoweri ðxÞ ¼ Pri w0i w0i xn f wpoweri ðxÞdx
Ð Pw0i n
¼ Pw0i x 1 Fwspeedi ðv fi Þ Fwspeedi ðvci Þ δðx þ Pw0i Þdx

k
k0i 1 xk2i þPw0i 0i
Ð P P k0i x k2i þ Pw0i k1i c0i
þ Pri w0i w0i xn e dx
k1i c0i k1i c0i
Ð P Pw0i n ð8:5Þ
þ PririPw0i x Fwspeedi ðv fi Þ Fwspeedi ðvri Þ δ½x ðPri Pw0i Þdx

¼ 1 Fwspeedi ðv fi Þ Fwspeedi ðvci Þ ðPw0i Þn

þ Fwspeedi ðv fi Þ Fwspeedi ðvri Þ ðPri Pw0i Þn

k
k0i 1 xk2i þPw0i 0i
Ð P P k0i x k2i þ Pw0i k1i c0i
þ Pri w0i w0i xn e dx
k1i c0i k1i c0i
By performing the variable transformation twice, i.e. t ¼ x k2i þ Pw0i and

k0i
τ ¼ k1itc0i , the above equation becomes

αn ¼ 1 Fwspeedi ðv fi Þ Fwspeedi ðvci Þ ðPw0i Þn
ΔPwi

k0i
P k
Ð kri1i c0i2i h 1
in
þ k0i k1i c0i τk0i þ ðk2i Pw0i Þ eτ dτ
-k1ik2ic0i
ð8:6Þ
¼ 1 Fwspeedi ðv fi Þ Fwspeedi ðvci Þ ðPw0i Þn

k0i
X n ð Pri k2i
k1i c0i k
þ Cnk ðk1i c0i Þk ðk2i Pw0i Þnk k0i τk0i eτ dτ
k2i
k¼0 k
1i c0i
k0i
ð Pri k2i
n! k1i c0i k
where Cnk ¼ and k0i τk0i eτ dτ is an incomplete Γ function.
k!ðn kÞ! k
k2i
1i c0i
The nth order cumulant, γnΔPwi , also known as the semi-invariant, is the polyno-
mial in α1ΔPwi , α2ΔPwi , . . . , αnΔPwi , for example
γ1 ΔPwi ¼ α1 ΔPwi
γ2 ΔPwi ¼ α2 ΔPwi α21 ΔPwi ð8:7Þ

γ3 ΔPwi ¼ α3 ΔPwi 3α1 ΔPwi α2 ΔPwi þ 2α31 ΔPwi
Hence by using (8.7), the nth order cumulant, γnΔPwi , can be computed from the
various-order moment.
8.2.1.2 Construction of Distribution Function of Critical Eigenvalue
According to the probability theory in [1, 2], if the relationship between a random
variable ρ and m other independent random variables ηj, j ¼ 1, 2, . . .m is linear, that
is ρ ¼ a1η1 þ a2η2 þ þ amηm, their nth order cumulants satisfy the following
equation
γn ρ ¼ a1 n γ n η1 þ a2 n γn η2 þ þ am n γn ηm ð8:8Þ
If there are m grid-connected wind generation sources (wind farms) in the multi-
machine power system and λk ¼ ξk þ jωk is the kth eigenvalue (critical eigenvalue) of
the power system, the following relationship between the critical eigenvalue and the
wind power generation can be established for power system small-signal stability
analysis
X
m
Δλk ¼ Δξk þ jΔωk ¼ ð∂λk =∂Pwi ÞΔPwi
i¼1
ð8:9Þ
X
m
¼ Re½ð∂λk =∂Pwi ÞΔPwi þ jIm½ð∂λk =∂Pwi ÞΔPwi
i¼1
where Re() and Im() denote the real and imaginary part of a complex variable
respectively. The sensitivity of the critical eigenvalue with respect to m wind power
sources at an equilibrium point in (8.9) can be computed in an analytical way [17] or
conveniently in a numerical way given by the following equation
∂λk λk ðPwi þ ΔPwi Þ λk ðPwi Þ
¼ , i ¼ 1, 2, . . . m ð8:10Þ
∂Pwi ΔPwi
The assumption of linearity is implied in deriving (8.9). For small-signal stability
analysis, linearized model of power systems is used and the nonlinearities are not
considered. Hence, the assumption of linearity in (8.9) is tenable. From (8.8) and
(8.9) it can have
Xm
n
∂λk
γn Δξk ¼ Re γn ΔPwi ð8:11Þ
i¼1
∂Pwi
where γnΔξk is the nth order cumulant of the stochastic variation of the real part of the
critical eigenvalue, Δξk. The mean of Δξk is μΔξk ¼ γ1Δξk . It is noted that (8.11) is
derived under the assumption that all the wind power sources are independent. In
practice this is true if wind farms locate far away to each other geographically. The
spatial correlations between different wind farms will be considered in Sect. 8.2.1.4.
The nth order central moment, βnΔξk , of Δξk is calculated from its cumulants by
using the following equation
β1 Δξk ¼0
β2 Δξk ¼ γ2 Δξk ¼ σΔξk 2
ð8:12Þ
β3 Δξk ¼ γ3 Δξk
β4 Δξk ¼ γ4 Δξk þ 3γ22 Δξk
where σΔξk is the standard deviation of Δξk.

From the cumulants and central moments of Δξk, the CDF and PDF of the
Δξk μΔξk
standardized Δξk, Δξk ¼ , can be obtained by using the following well-
σΔξk
known Gram-Charlier expansion
g1 ð1Þ g g
F _ ðxÞ ¼ g0 ΦðxÞ þ Φ ðxÞ þ 2 Φð2Þ ðxÞ þ 3 Φð3Þ ðxÞ þ
Δξ k 1! 2! 3!
ð8:13Þ
g1 ð1Þ g2 ð2Þ g3 ð3Þ
f _ ð x Þ ¼ g 0 φð x Þ þ φ ð x Þ þ φ ð x Þ þ φ ð x Þ þ
Δξ k 1! 2! 3!
where F _ ðxÞ and f _ ðxÞ is the CDF and PDF of Δξk respectively, Φ(x) and φ(x) the
Δξk Δξk
CDF and PDF of standard normal distribution respectively and the superscript
number (n) denotes the nth order derivative of Φ(x) and φ(x). Coefficients in the
Gram-Charlier expansion of (8.13) are the polynomial in the central moments of Δξk
as given as follows
g0 ¼ 1
g1 ¼ g2 ¼ 0
β3 Δξk
g3 ¼ ð8:14Þ
σ3Δξk
β4 Δξk
g4 ¼ 3
σ4Δξk
Obviously, the CDF of Δξk can easily be obtained from that of Δξk to be

x ΔμΔξk
FΔξk ðxÞ ¼ F _ ð8:15Þ
Δξk σΔξk
Because Δξk ¼ ξk ξk0 (ξk0 is the deterministic value of the real part of λk), the
CDF of ξk can be obtained to be
Fξk ðxÞ ¼ FΔξk ðx ξk0 Þ ð8:16Þ
The PDF is the derivative of the CDF of ξk obtained in (8.16).

Therefore, the probability of the real part of the kth eigenvalue in the power system
with m grid-connected wind power sources to be negative can be computed by (8.1).
8.2.1.3 Treatment of Ending Values of Distribution Function of Critical

Eigenvalue
The wind power distribution is not continuous as given by (8.2). Thus f ξk ðxÞ 6¼ 0 only
exists over a certain interval of ξk, i.e., [ξk_left, ξk_right]. The CDF and PDF given by
(8.16) is for ξk within [ξk_left, ξk_right] and their value at the left end (ξk ¼ ξk_left) and
right end (ξk ¼ ξk_right) of the interval needs to be calculated separately. Hence (8.16)
needs to be modified. This modification is carried out by dividing the wind generation
sources into two groups firstly. Group A is of positive Re(∂λk/∂Pwi) and group B is of
negative Re(∂λk/∂Pwi). In deterministic eigenvalue analysis, ξk is calculated as ξk_left,
when the wind generation sources in group A is at the cut-in wind power (i.e., Pwi ¼ 0,
i 2 A) and in group B is at the furling wind power (i.e., Pwi ¼ Pri, i 2 B). ξk_right is
calculated similarly but when the wind generation sources in group A is at the furling
wind power and group B the cut-in wind power. The modified CDF and PDF are
8 ( )
> Y 1 Fwspeedi1 ðvfi1 Þ Fwspeedi1 ðvci1 Þ
>
>
>
>
>
> Fwspeedi2 ðvfi2 Þ Fwspeedi2 ðvri2 Þ
>
>
i12A, i22B
>
>
>
> δðx ξk left Þ,
>
>
>
> for x ¼ ξk left
>
>
>
>
< derivative of ð8:16Þ, for ξk left < x < ξk right
f ξ k ð xÞ ¼ ( )
> Y 1 Fwspeedi1 ðvfi1 Þ Fwspeedi1 ðvci1 Þ
>
>
>
>
>
> i12B, i22A Fwspeedi2 ðvfi2 Þ Fwspeedi2 ðvri2 Þ ð8:17Þ
>
>
>
>
>
> δ x ξk right ,
>
>
>
>
>
> for x ¼ ξk right
>
>
:
0, for x < ξk left or x > ξk right
8
>
< 0, for x ξk left
Fξ k ð x Þ ¼ ð13Þ, for ξk left < x < ξk right
>
: 1, for x ξk right
8.2.1.4 Spatial Correlations Between Wind Power Generation Sources
The complementarity of variable wind generation sources at different places effec-

tively can smooth the total wind power fluctuation in a power system. Thus the effect
of power variation from multiple wind power generation sources on power system
small-signal stability will be different if the wind complementarity is considered.
The spatial correlations between different wind power sources are an important
factor of wind complementarity. Hence the probabilistic small-signal stability of a
power system is affected by the spatial correlations of wind generation sources.
The correlation between two wind power sources is closely related to their
geographical distance [18]. Two locations over 1200 km have a correlation coeffi-
cient close to 0, whilst very close locations (less than 100 km) have a correlation
coefficient close to 1. This can be used to approximately estimate their correlation
coefficient ρij and thus to establish the correlation coefficient matrix [ρij]m m for m
grid-connected wind power sources [18].
If there are sufficient wind speed data available, the correlations between pairs of
wind speed can be calculated to be

Cov vi ; v j
ρij ¼ ð8:18Þ
σvi σv j
where vi and vj are wind speed random variables corresponding to two wind power
source locations, Cov(vi, vj) represents the covariance of the speed vi and vj, and σvi
and σv j are the standard deviations of the speeds.
For each wind speed random variable, vi and vj, a wind speed sample series can be
generated correspondingly. Each wind speed sample series should satisfy the
Weibull distribution and contain the spatial correlation. There are several methods
available to generate the eligible wind speed sample series, such as the normal
transformation method in [19, 20] and the Copulas in [21].
With the wind speed sample series available, the wind power sample series with
correlations (i.e., ½Pwi NSample 1) can be generated by employing the power-wind speed
curve of each wind power sources described in [22]. The wind power variation
sample series ½Pwi NSample 1 can be obtained to be
½ΔPwi NSample 1 ¼ ½Pwi NSample 1 ½Pw0i NSample 1 , i ¼ 1, 2, . . . m ð8:19Þ
where Nsample is the size of each wind power sample series and ½Pw0i NSample 1 is a
vector with all the samples equal to the deterministic value Pw0i.
When the correlation of the different wind power sources is considered, (8.11) is
modified to calculate the nth order cross cumulant of Δξk to be
2 3
m X
X X m 6 7
m
6 ∂λk ∂λk ∂λk 7
γn Δξk ¼ 6Re Re Re γn ΔPwi1 i2 in 7
i1 ¼1 i2 ¼1
4 ∂Pwi1 ∂Pwi2
in ¼1 |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl
∂Pwin 5
ffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
n
ð8:20Þ
where γnΔPwi i in denotes the nth order cross cumulants of the multiple wind power
12
variations. For independent wind power sources, i1 ¼ i2 ¼ ¼ in ¼ i. The cumulant
γnΔPwi i in is equal to γnΔPwi and (8.11) is replaced by (8.20).
12
Normally the higher order cross cumulants of Δξk have less impact on the
accuracy of the CDF and PDF curve of ξk. Hence it is only needed to calculate the
first several order cross cumulants of Δξk by using (8.20) to include the correlation.
The cumulants of the rest with higher orders can still be calculated by (8.11) so as to
reduce the computation cost. The 1st order cross cumulant of wind power variation is
still γ1ΔPwi . Their 2nd and 3rd order cross cumulants can be easily calculated from the
following equations [1, 2].
h i
γ2 ΔPwi i ¼ β2 ΔPwi i ¼ E ΔPwi1 μΔPwi ΔPwi2 μΔPwi
12 12 1 2
γ3 ΔPwi1 i2 i3 ¼ β3 ΔPwi i i ð8:21Þ

123
h i
¼ E ΔPwi1 μΔPwi ΔPwi2 μΔPwi ΔPwi3 μΔPwi
1 2 3
where βnΔPwi i in is the nth order cross central moment and μΔPwin is the mean of ΔPwin .
12
The expected values in (8.21) can be calculated directly when each wind power
variation sample series is obtained. Hence by using (8.20) and (8.21), first three order
cross cumulants of Δξk are obtained.
When the spatial correlations between wind power generation sources are con-
sidered, (8.17) needs to be modified as follows.
Firstly, a vector of the approximate values of ξk is calculated to be
Xm

∂λk
½ξk NSample 1 ¼ ½ξk0 NSample 1 þ Re ½ΔPwi NSample 1 ð8:22Þ
i¼1
∂Pwi
where ½ξk0 NSample 1 is a vector with all the samples equal to the deterministic value
ξk0. Secondly the maximums and minimums of the vector ½ξk NSample 1 are determined.
It is noted that there are possibilities that multiple maximums and multiple mini-
mums exist in some cases. All the wind power data sets corresponding to the
maximums and minimums of ½ξk NSample 1 are recorded (each wind power data set
[Pw1, Pw2, Pwm]1 m is consist of m wind power data which are respectively from
the same row of each ½Pwi NSample 1 as the maximum or minimum in ½ξk NSample 1 ).
Thirdly, the deterministic eigenvalue analysis is carried out to compute ξk by using
the recorded wind power data sets as output power of m wind power sources. ξk_left
and ξk_right are the smallest and largest ξk in the deterministic eigenvalue analysis
respectively. The probabilistic density of ξk_left and ξk_right are calculated by
Nξk left
δðx ξk left Þ, for x ¼ ξk left
Nsample
ð8:23Þ
Nξk right
δ x ξk right , for x ¼ ξk right
Nsample
Where Nξk_left and Nξk_right are the repeated numbers of ξk_left and ξk_right in the
deterministic eigenvalue analysis respectively. Hence, the PDF and CDF of ξk with
modified left and right ends are obtained. The form of PDF and CDF of ξk is similar
to (8.17), but the probabilistic density values at left and right ends are changed as
given by (8.23).
8.2.1.5 Procedure of Cumulant Theory Based Analytical Method
The procedure of probabilistic analysis of power system small-signal stability

introduced above can be summarized as follows:
1. If the grid-connected onshore wind power sources are spatially dependent, the
wind power variation sample series should be generated from the correlation
coefficient matrix by using (8.19). If not, go to step 2) directly.
2. Determine the stochastic distributions of grid-connected wind power sources
defined by (8.2) and (8.3).
3. Compute the nth order moment by using (8.6) and then the nth cumulant by (8.7)
of each wind power variation.
4. Compute the nth order cumulant of the variation of the real part of the critical
eigenvalue, Δξk, by using (8.11) or (8.20) (if considering correlations) and then
its nth order central moment by (8.12).
5. Compute the CDF and PDF of the standardized Δξk by using (8.13) and (8.14).
6. Convert the CDF and PDF of Δξk into the CDF and PDF of ξk by using (8.15) and
(8.16).
7. Modify the CDF and PDF of ξk by using (8.17) or (8.17) and (8.23)
(if considering correlations).
8. Determine the probability of the real part of the critical eigenvalue to be negative
according to (8.1).
9. Step 2–7 can also be achieved by applying the non-analytical method of Monte
Carlo simulation.
8.2.2 Example 8.1
Figure 8.2 shows the configuration of an example 16-machine 5-area power system
with 3 grid-connected onshore wind power sources at node 69, 70 and 71. The
network data, system load condition, synchronous generator model and parameters
are given in Appendix 8.1. For the purpose of better demonstration, no PSSs are
installed in the power system. A 70 MW doubly fed induction generator (DFIG)
model is used for all three wind farms [23] and the DFIGs’ parameters in p. u. are.
H ¼ 1.7s, D ¼ 0.0, xs ¼ 2.9, xr ¼ 2.9, xm ¼ 2.6, rs ¼ 0.0, rr ¼ 0.0013,
Pw0 ¼ 0.3333.
The controller model of the DFIG rotor-side converter used in the example
system is shown by Fig. 8.3, where KP ¼ 30, KQ ¼ 30. Parameters of the wind
speed distribution are vc ¼ 4m/s, vr ¼ 10m/s, vf ¼ 22m/s, Pr ¼ 1.0p. u., μ ¼ 6m/s,
σ ¼ 2.5.
From the deterministic modal analysis, the 29th eigenvalue is identified to be the
critical eigenvalue, i.e., λ29 ¼ 0.0106 j3.3004. Hence deterministically the
system is stable. Denote the wind farm at node 69, 70 and 71 to be the first, second
and third source of wind generation respectively. The sensitivity computation of the
critical eigenvalue with respect to three sources of wind generation, Pw1, Pw2 and
Pw3, is obtained to be
WG NYPS
14 8
69 A3
66 1
60 25 26
41 40 48 47 53 28 29
61
2
18 27
9
42 1
17
67 38 16
31 32 30 24
70 3 15
15 62 63 9
WG 21
46 4 14 22
10 11
A4 33 5 6 12
58
19 23
49 34 36 7 54 11
20 6 59
13
35 64 2 56 57
8 10 55 7
51 45
12 4 5
37 3
A5 50 39
52 A1
44
68 43 65 A2
71
16 13 NETS
WG
Fig. 8.2 Line diagram of example 16-machine 5-area power system integrated with onshore wind
power generation
X m2 X V
– s X rr – I rd + s m s – Rr I rq
X ss X ss
I rq
Ps0 + V rq
– X ss I rqref
S KP S
X m Vs + +
Qs0 –1 I rdref Vrd

Xss S S KQ S
+
X m Vs + – +
+
2 I rd
Vs
X m2
s X rr – I rq – Rr I rd
X ss
Fig. 8.3 Controller model of DFIG rotor-side converter
Table 8.1 Moments and cumulants of wind power variations

Moments of wind power variations Cumulants of wind power variations
α1ΔPw1, 2, 3 ¼ 0.0319 γ1ΔPw1, 2, 3 ¼ 0.0319
α2ΔPw1, 2, 3 ¼ 0.1080 γ2ΔPw1, 2, 3 ¼ 0.1070
α3ΔPw1, 2, 3 ¼ 0.0281 γ3ΔPw1, 2, 3 ¼ 0.0178
α4ΔPw1, 2, 3 ¼ 0.0262 γ4ΔPw1, 2, 3 ¼ 0.0111
γ5ΔPw1, 2, 3 ¼ 0.0126 γ5ΔPw1, 2, 3 ¼ 0.0104
Table 8.2 Cumulants and central moments of real part of critical eigenvalue
Cumulants of critical eigenvalue Central moments of critical eigenvalue
Mean ¼ 8.10 104 β1_Δξ29 ¼ 0.0
Variance ¼ 2.32 105 β2_Δξ29 ¼ 2.32 105
γ3_Δξ29 ¼ 3.35 108 β3_Δξ29 ¼ 3.35 108
γ4_Δξ29 ¼ 1.82 1010 β4_Δξ29 ¼ 1.44 109
γ5_Δξ29 ¼ 1.51 1012 β5_Δξ29 ¼ 6.26 1012
∂λ29 ∂λi29
¼ 0:0096 j0:0489, ¼ 0:0083 j0:0466,
∂Pw1 ∂Pw2
∂λ29
¼ 0:0075 j0:0394
∂Pw3
Table 8.1 gives the computational results of the first five orders of moments and
cumulants of three wind power variations computed by using (8.6) and (8.7).
Table 8.2 gives the computational results of the first five orders of cumulants and
the central moments of the real part of the critical eigenvalue obtained from (8.11)
and (8.12). Table 8.3 gives the computational results of the first six coefficients of the
Table 8.3 Coefficients of g0 g1 g2 g3 g4 g5

Gram-Charlier expansion
1 0 0 0.2989 0.3382 0.5794
λ Real Part PDF

140
Analytical Method 100
Monte Carlo Simu
120
Monte Carlo sample number

80
0.0117δ(x+0.0185)
probabilistic density
100
80 60
60
40
40
0.0003δ(x-0.0094) 20
20
0 0
-0.02 -0.015 -0.01 -0.005 0 0.005 0.01
λ real part
Fig. 8.4 PDF of real part of critical eigenvalue
Gram-Charlier expansion according to (8.14). The PDF curve of the real part of the
critical eigenvalue is constructed by use of (8.17) finally, which is shown by in
Fig. 8.4. The PDF obtained by the Monte Carlo simulation with 5000 samples is also
presented by Fig. 8.4, which confirms the result by using the introduced method of
probabilistic analysis.
A comparison of computation time between the Monte Carlo simulation and the
analytical method has been carried out. Based on the same computational resource
(Dell OptiPlex 745, Intel Core 2 CPUs 2.66GHz, 3GB RAM), the time of the Monte
Carlo simulation with 5000 iterations is 15236.48 s, while only 38.56 s for the
analytical method with first five-order Gram-Charlier expansion is needed. The
analytical method is about 395 times faster than the Monte Carlo simulation.
According to the PDF of Fig. 8.4 it can be obtained that
ð0
Pðξ29 < 0Þ ¼ f ξ29 ðxÞdx ¼ 0:9710
1
Hence when the stochastic variation of wind generation is considered, the critical
eigenvalue of the example power system has a probability of 97.10% to remain in the
left half-plane. Although the system is considered to be stable by the deterministic
analysis, as the critical eigenvalue is λ29 ¼ 0.0106 j3.3004, it still has a
probability of 2.90% to be unstable due to the uncertainty of wind generation.
To demonstrate the case with the correlations between three wind power sources
being considered, the correlation coefficient matrix [ρij]3 3 is constructed according
to the assumed geographical distance between the three wind farms as
2 3
1 0:8 0
h i
6 7
ρij ¼ 4 0:8 1 0 5 ð8:24Þ
33
0 0 1
(8.24) in fact suggests that the first two wind power sources are strongly corre-
lated, while the third is independent with the other two due to the long distance. By
employing the procedure presented in Sect. 8.2.1.4, the 2nd and 3rd order cross
cumulants of Δξ29 are computed by using (8.21) to be
γ2 Δξ29 ¼ 3:63 105 , γ3 Δξ29 ¼ 9:02 108
The probabilistic density of ξ29_left and ξ29_right is calculated by using (8.23) to be

Nξ29 left
δðx ξ29 left Þ ¼ 0:0396δðx þ 0:0185Þ
Nsample
Nξ29 right
δ x ξ29 right ¼ 0:0028δðx 0:0094Þ
Nsample
Finally, the PDF curve of the real part of the critical eigenvalue is obtained as
shown by Fig. 8.5. The result of Monte Carlo simulation (with 5000 samples) in
λ Real Part PDF

0.0396δ(x+0.0185) 200
Analytical Method
240 Monte Carlo Simu
160
200
160 120
120
80
80
0.0028δ(x-0.0094) 40
40
0 0
-0.02 -0.015 -0.01 -0.005 0 0.005 0.01
λ real part
Fig. 8.5 PDF of real part of critical eigenvalue

Fig. 8.5 has verified the analytical method introduced when the correlations of wind
power sources are considered.
According to the PDF of Fig. 8.5 it can be obtained that
ð0
Pðξ29 < 0Þ ¼ f ξ29 ðxÞdx ¼ 0:9334
1
This result indicates that the consideration of correlation has changed the prob-
ability of system small-signal stability.
8.3 Probabilistic Analysis of Small-Signal Stability of a

Power System Affected by MTDC–Connected Offshore
Wind Power Generation
VSC-HVDC transmission is an attractive way to connect offshore wind power

generation due to its flexibility in topology and no need for reactive compensation.
It is envisaged that an increasing number of offshore wind farms are deployed and
connected to AC grids through VSC based subsea MTDC networks. Grid integration
of offshore wind power generation through MTDC networks increases the complex-
ity of system stability analysis, as dynamic of MTDC networks may affect the
stability of the whole AC/DC power system. In this section, the probabilistic
small-signal stability of a hybrid AC/DC power system where MTDC networks
connect the variable offshore wind power to the main AC grid is carefully examined.

of Probabilistic Small-Signal Stability Analysis of a
Hybrid AC/DC Power System
As voltage stability is always a concern of HVDC networks, the cumulant theory

based analytical method is used for the probabilistic analysis of both small-signal
angular and small-signal voltage stability of an AC/DC power system with offshore
wind farms supplying power through a MTDC network in this section.
8.3.1.1 Indices for Small-Signal Angular and Voltage Stability of a

Hybrid AC/DC Power System
In the cumulant theory based analytical method for probabilistic analysis, the key
step is to identify the stability index (i.e., critical eigenvalue) and its sensitivity with
respect to the stochastic variables in the system.
In the small-signal angular stability presented in the previous sections, the

following linearized dynamic model of power system is established at an operating
point when wind power output is taken under mean wind speed
Δx_ ¼ AΔx ð8:25Þ
where Δx is the system state variable and A the system state space matrix. The
critical eigenvalue (λA_cr ¼ ξA_cr þ jωA_cr) of A is identified and then its sensitivity
to the variation of wind power output is computed to link the wind generation with
the power system small-signal angular stability. Hence the analytical method can
perform the probabilistic analysis of power system small-signal angular stability in
one-step computation from the well-known Weibull distribution of wind speed.
Small-signal voltage stability of a power system can be assessed based on
eigenvalue decomposition of reduced Jacobian matrix at a steady-state system oper-
ating point [24–26]. The linearized steady-state voltage equations can be written as
" # " #" #
ΔP JPθ JPU Δθ
¼ ð8:26Þ
ΔQ JQθ JQU ΔU
where ΔP, ΔQ, Δθ and ΔU are vectors of variation of bus active, reactive power
injection, bus voltage angle and magnitude respectively. When the focus of study is
DC network, all AC buses except the converter AC buses can be eliminated and thus
(8.26) becomes
" # " #" #
ΔPc JPθ c JPU c Δθc
¼ ð8:27Þ
ΔQc JQθ c JQU c ΔUc
where the subscript c denotes the vectors or matrices regarding the converter buses.
Set ΔPc ¼ 0 to reduce (8.27), thus giving
ΔQc ¼ JR ΔUc ð8:28Þ
where JR is the reduced Jacobian matrix of the converter buses and

JR ¼ JQUc JQθc J1 Pθc JPUc .
According to [25, 26], the DC network will be voltage-stable if all the eigenvalues
of JR are positive. Therefore, the minimal eigenvalue of JR, λJR min , can be used as an
index of AC/DC power system small-signal voltage stability. With the assumption of
ΔPc ¼ 0 in (8.28), it can be seen that λJR min indicates the margin of small-signal
voltage stability for each specific operating point (or load level) of system. The
sensitivity of λJR min with respect to m wind power sources on the given steady-state
operating point can be computed conveniently in a numerical way similar to (8.10).
On this basis, the probability of small-signal angular and voltage stability of the
AC/DC power system with m grid-connected offshore wind farms can be assessed
by the following equations
ð0
PðξA cr < 0Þ ¼ f ξA cr ðxÞdx
1
ð1 ð8:29Þ
PðλJR min > 0Þ ¼ f λJR min
ðxÞdx
0
The assessment of both probabilistic small-signal angular and voltage stability

can be derived from the complete differential-algebraic model. However, as the two
types of stability correspond to different matrices of the complete model (i.e., system
state space matrix and reduced Jacobian matrix), the probability values calculated in
(8.29) are not directly correlated, which can be also easily explained from their
essential physical nature. Moreover, these two types of stability may have different
shapes of probability distribution due to different sensitivities and their changing
rates. Nevertheless, the increase of wind speed variation range, numbers of wind
power sources and spatial correlation of wind farms in the system may deteriorate
both stability issues. Therefore, the procedure to examine the probabilistic small-
signal angular and voltage stability of a hybrid AC/DC power system as affected by
the stochastic fluctuations of grid-connected offshore wind power generation is
illustrated by Fig. 8.6.
8.3.1.2 Multi-Point Linearization Technique
Probabilistic analysis of small-signal stability presented in Fig. 8.6 is based on the

linearization of power system non-linear model at a single operating point. Accuracy
of analytical results based on the single-point linearization is certainly lower than
that of Monte Carlo simulation especially for a complex hybrid AC/DC power
system with a strong nonlinearity. Work in [27] indicates that if different points of
linearization can be used, the probabilistic information around these points can be
evaluated with enhanced accuracy. Hence, if multiple steady-state operating points
under different wind speed conditions are considered, the analytical method can be
improved to better handle power system non-linearity to assess the effect of grid-
connected wind power generation on power system probabilistic small-signal sta-
bility. Procedure of this improved analysis with multi-point linearization can be
explained as follows.
Take the ith offshore wind power generation source (wind farm) in the AC/DC
power system as an example. The active power output Pwi from the wind farm can
vary from 0 to Pri (the rated wind power). If Nlin-point linearization is applied (i.e.,
Nlin steady-state operating points are considered), the interval of variation of Pwi, [0,
Pri], firstly is divided equally into 2Nlin sub-intervals. The endpoints of all the
sub-intervals are recorded as a vector, ½Pwi0 ; Pwi1 ; Pwi2Nlin 1 ; Pwi2Nlin ð2Nlin þ1Þ1 . If
the sensitivity of ξA_cr (or λJR min) with respect to the ith wind output power is positive
(negative), the endpoints are recorded in a normal order (i.e., Pwi_0 ¼ 0, Pwi2Nlin ¼ Pri).
Conversely, the endpoints are recorded in a reverse order (i.e., Pwi_0 ¼ Pri, Pwi2Nlin ¼ 0).
Assume that there are m wind farms in the power system. Thus in total 2Nlin þ 1 initial
A Given System Steady- Given Wind Power

state Operating Point Distributions
Linearizing Power System

Model
If Correlated
Deterministic Yes
Small-signal Angular and
Voltage Stability Analysis
Generating Wind Power

Critical Eigenvalue Variation Sample Series
(and Its Sensitivity with
Respect to Wind Output
Power)
Angular and Voltage

Stability Analysis Computing the Moments
and Cumulants of Wind
Power Variations
Computing the Cumulants

and Central Moments of
Critical Eigenvalue
Computing Coefficients of
Gram-Charlier Series
Generating Distribution of
Standardized Eigenvalue
Probabilistic Distribution of
Critical Eigenvalue
Conventional Cumulant Theory

Based Analytical Method
Probability of Power System

Voltage Stability
Fig. 8.6 Procedure of probabilistic small-signal angular and voltage stability analysis of a hybrid
AC/DC power system
Generating N lin Steady-state Given Wind Power

Operating Points Distributions
Setting n=1
Conventional Cumulant Theory Based Analytical

Method for Probabilistic Small-signal
Angular and Voltage Stability Analysis
n=n+1
If n>Nlin No
Yes
Combining N lin piecewise

PDFs of Stability Indices
Probability of Power System

Voltage Stability
Fig. 8.7 Procedure of probabilistic stability analysis based on multi-point linearization technique
operating points will be obtained (i.e., [Pw1_j, Pw2_j, . . .Pwm 1_j, Pwm_j]m 1, j ¼ 0,
1, . . .2Nlin). Secondly, the deterministic eigenvalue analysis is carried out under the jth
initial operating point to compute ξA_cr_j and λJR min when j is even (Nlin þ 1 times). The
other Nlin initial operating points (when j is odd) are the selected Nlin steady-state
operating points. Thirdly, the method introduced by Fig. 8.6 is employed under the
selected Nlin steady-state operating points to derive the PDF for each selected point.
Finally, Nlin PDFs obtained are properly combined at each point of ξA_cr_j and λJR min
(when j is even) to give the final PDFs of multi-point linearization of ξA_cr and λJR min .
The computational procedure can be illustrated by Fig. 8.7.
In most cases, the final PDF of a stability index obtained by using this multi-point
linearization technique does not have an area under the curve exactly equal to unity.
Since the selected steady-state operating points are evenly located, the linear solu-
tions are located more closely when the sensitivity of stability index with respect to
wind output power becomes smaller, which sequentially makes the curve compar-
atively more precise. Therefore, the area under the curve can be calculated from the
side with a smaller sensitivity. When the area reaches unity, the remaining part of the
PDF can be set to zero [27].
WG1 WG2 WG3
69 70 71
~ ~ ~
= = =
DC1 DC2 DC3
NYPS
14 8
DC4
DC6 A3 DC5
66
~ ~ ~ 1
60
= = = 25 26
41 40 48 47 53 28 29
61
2
27
18 9
42 1
17
67 38 16
31 32 30
3 15 24
15 62 63 9
21
4 14
46 10 11 22
A4
33 5 6 12
58
19 23
49 34 36 7 54 11
13 20 6 59
35 64 2 10 56 57
8 55 7
51 45
12 4 5
3
A5 50 39 37
52 A1
44
68 43 65 A2
16 13
NETS
Fig. 8.8 Line diagram of example 16-machine 5-area hybrid AC/DC power system integrated with
offshore wind power generation
8.3.2 Example 8.2
Figure 8.8 shows the configuration of an example 16-machine 5-area AC/DC power
system with three grid-connected offshore wind farms at node 69, 70 and 71. A
six-terminal DC network connects the offshore wind farms to the onshore AC grid.
The AC network data, system load condition, synchronous generator model and
parameters are given in Appendix 8.1. Parameters of the VSC-HVDC network in
p.u. are R12 ¼ 0.0009, R16 ¼ 0.0006, R23 ¼ 0.0012, R25 ¼ 0.0006, R26 ¼ 0.0011,
R34 ¼ 0.0006, L12 ¼ 0.0033, L16 ¼ 0.0022, L23 ¼ 0.0044, L25 ¼ 0.0022,
L26 ¼ 0.0040, L34 ¼ 0.0022.
A 70 MW doubly fed induction generator (DFIG) model is used for all three wind
farms and the DFIGs’ parameters in p. u. are H ¼ 1.7s, D ¼ 0.0, xs ¼ 0.29, xr ¼ 0.29,
xm ¼ 2.6, rs ¼ 0.0, rr ¼ 0.0013, Pw0 ¼ 0.3333.
The controller model of the DFIG rotor-side converter used in the example system
is shown by Fig. 8.3, where KP ¼ 30, KQ ¼ 30. Parameters of the wind speed
distribution are vc ¼ 3m/s, vr ¼ 12m/s, vf ¼ 22m/s, Pr ¼ 1.0p. u., μ ¼ 6m/s, σ ¼ 2.5.
From the deterministic small-signal angular and voltage stability analysis, the 31st
eigenvalue of the system state space matrix A is identified to be the critical
eigenvalue, i.e., λA_cr ¼ 0.0113 j2.0164, and the minimal eigenvalue λJR min
of the reduced Jacobian matrix JR is 0.2821. Hence deterministically the system is

stable. The wind farms at node 69, 70 and 71 are denoted to be the 1st, 2nd and 3rd
source of wind generation respectively. The sensitivity computation of two stability
indices with respect to three sources of wind generation is
∂ξA cr ∂ξ ∂ξ
¼ 0:0092, A cr ¼ 0:0081, A cr ¼ 0:0078,
∂Pw1 ∂Pw2 ∂Pw3
∂λJR min ∂λJR min ∂λJR min
¼ 0:2539, ¼ 0:2215, ¼ 0:2186
13-point linearization technique is applied by employing the procedure presented
in Sect. 8.3.1.2. Hence 13 steady-state operating points are determined under
13 different output power conditions of wind generation. The values of stability
indices (x-axis) and their sensitivities with respect to the first wind power output
(y-axis) are calculated at each steady-state operating point by deterministic analysis
and plotted in Fig. 8.9 as an example.
It can be seen from Fig. 8.9 that the sensitivity of stability indices has a noticeable
change with the increase of wind power penetration level. This explains that the
conventional method based on one-point linearization may not provide accurate
result.
The final PDFs of ξA_cr and λJR min are computed by use of the analytical method
illustrated in Fig. 8.7 with the 13-point linearization and displayed in Figs. 8.10 and
8.11 respectively. For the purpose of comparison, results obtained by using the
0.015
1 2 3 4 5 6
dξA_cr /dP w 1
0.01 7 8 9 10 11
12
13
0.005
0
-20 -15 -10 -5 0
ξA_cr -3
x 10
-0.1
dλJr /dP w 1
-0.2
-0.3 4 3 2 1
7 6 5
10 9 8
-0.4 13 12 11
-0.5
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
λJr
Fig. 8.9 Stability indices and their sensitivities with respect to 1st wind output power at each
steady-state operating point (from 1 to 13)
ξ PDF
A_cr
100
Proposed Analytical Method
160
Conventional Analytical Method
Monte Carlo Simulation
140 80

120
60
100
80
0.0015δ(x+0.0198) 40
60
40
20
2e-6δ(x-0.0022)
20
0 0
-0.025 -0.02 -0.015 -0.01 -0.005 0 0.005
ξA_cr
Fig. 8.10 PDF of ξA_cr obtained by analytical methods and Monte Carlo simulation
λJ _min PDF
R 120
6 Proposed Analytical Method
Conventional Analytical Method
Monte Carlo Simulation 100
5
80
4
0.0015δ(x-0.4880)
60
3
2 40
1 2e-6δ(x+0.2497) 20
0 0
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
λJ _min
R
Fig. 8.11 PDF of λJR min obtained by analytical methods and Monte Carlo simulation
Table 8.4 Probability stability assessed by use of three different methods

Monte Carlo Conventional analytical Proposed analytical
Probability simulation method method
P(ξA_cr < 0) 0.9936 0.9881 0.9940
PðλJR min > 0Þ 0.9676 0.9762 0.9678
conventional cumulant theory based analytical method and Monte Carlo simulation
are also presented. From the PDFs in Figs. 8.10 and 8.11, the probability of small-
signal angular and voltage stability of the example system is calculated and the
results are given in Table 8.4. From Figs. 8.10, 8.11 and Table 8.4, it can be seen that
the multi-point analytical method proposed in Sect. 8.3 gives more accurate assess-
ment of probabilistic small-signal angular and voltage stability.
Table 8.4 also indicates that although the AC/DC system is considered to be
stable by the deterministic analysis of small-signal angular and voltage stability,
there still exists a probability of 0.64% to be unstable in terms of angular stability
and 3.24% in terms of voltage stability, when the random variations of wind
generation are considered.
8.3.3 Variance Indices Analysis of Probabilistic Small-Signal

Stability of a Hybrid AC/DC Power System
Traditional point-to-point HVDC connection is a mature technology to receive and

supply offshore wind power. On the other hand, the multi-terminal VSC-HVDC
(MTDC) technology has been rapidly developed recently, which are considered as a
more favorable way for the transmission and share of large-scale wind power among
different regions, such as European supergrid scheme in the North Sea
[28, 29]. There are broadly two control strategies for the VSC-HVDC, namely, the
constant DC voltage control scheme (or so called master-slave control) and the DC
voltage droop control scheme. Traditional constant DC voltage control has been
extended to the MTDC from point-to-point HVDC network. The DC voltage droop
control as a novel control developed for the MTDC has more advantages than the
former in the aspects of reliability and facilitation of converter outage.
Thus three main factors of offshore wind power generation that may influence the
probabilistic small-signal stability are the wind complementarity, connection tech-
nology and network control strategy. A variances indices analysis as a derivation
from cumulant theory will be introduced in this section to study the probabilistic
small-signal stability as affected by these three factors.
According to [1, 2], if the relationship between a random variable ρ and m other
random variables ηj, j ¼ 1, 2, . . .m is linear, that is ρ ¼ a1η1 þ a2η2 þ þ amηm,
their nth order cumulants satisfy the following equation
2 3
m X
X m X
m
γðρnÞ ¼ 4ai1 ai2 ain γðnÞ 5 ð8:30Þ
j1 ¼1 j2 ¼1 jn ¼1
|fflfflfflfflfflffl{zfflfflfflfflfflffl} ηj1 jn
n
where γðρnÞ denotes the nth order cumulant of ρ and γðηnj Þj denotes the nth order cross
1 n
cumulant among different ηj, j ¼ 1, 2, . . .m. For a particular case of (8.30) when the
m random variables ηj are independent, (i.e.,γðηnj Þj 6¼ 0only when j1 ¼ j2 ¼ ¼ jn ¼ j),
1 n
(8.30) becomes
γðρnÞ ¼ a1n γðηn1Þ þ a2n γðηn2Þ þ þ amn γðηnmÞ ð8:31Þ
where γðηnjÞ denotes the nth order cumulant of ηj. In this section, (8.30) and (8.31) are
employed as the foundation of the variance indices analysis.
8.3.3.1 Variance Index for Complementarity of Combined Wind Power
The complementarity level of combined wind power sources connected to a power

system can be evaluated by many ways including the variance (or standard devia-
tion) of the combined wind power, which is a commonly-used index to measure the
fluctuation of wind power [30]. If there are m grid-connected wind farms in the
power system, the combined wind power is Pwc ¼ Pw1 þ Pw2 þ þ Pwm. Hence
according to (8.30), the 2nd order cumulant of Pwc can be expressed as
ð2Þ
m X
X m
ð2Þ
γPwc ¼ γPwj ð8:32Þ
1 j2
j1 ¼1 j2 ¼1
ð2Þ
As the variance of Pwc, σ2Pwc , is equal to γPwc and the covariance of Pwj1 and Pwj2 ,
ð2Þ
cov Pwj1 ; Pwj2 , is equal to γPwj j , (8.32) can be written as
12
m X
X m
σ2Pwc ¼ cov Pwj1 ; Pwj2 ¼ σ2Pw1 þ σ2Pw2 þ þ σ2Pwm
j1 ¼1 j2 ¼1 |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
m ð8:33Þ
þ 2ρPw1 Pw2 σPw1 σPw2 þ 2ρPw1 Pw3 σPw1 σPw3 þ þ 2ρPwm1 Pwm σPwm1 σPwm
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
mðm1Þ=2
where ρPwj Pwj denotes the correlation coefficient between Pwj1 and Pwj2 and σPwj
1 2
denotes the standard deviation of Pwj. Normally, wind output power of different
wind farms is positively correlated, i.e. ρPwj Pwj 2 ½0; 1, if there is no special control
1 2
of wind power.
Therefore, the variance of combined wind power sources assessed by (8.33) can
be used as an index of their complementarity level. The smaller value of variance
suggests the higher level of complementarity. It can be seen from (8.33) that the
complementarity of combined wind power sources relies on the standard deviation

of each wind power and their spatial correlations.
8.3.3.2 Variance Index for Probabilistic Small-Signal Stability
The probability of small-signal stability of a power system is closely related to the

mean and variance of the PDF of the critical eigenvalue real part. The mean of the
PDF is mainly decided by the mean wind speed. The variance of probabilistic
distribution of the critical eigenvalue real part can be assessed as follows.
As discussed for (8.9), the relation between the real part of a critical eigenvalue
and a wind power source can be established as
X
m
Δξc ¼ Re ∂λc =∂Pwj ΔPwj ð8:34Þ
j¼1
where ξc denotes the real part of a critical eigenvalue λc, Re() represents the real part
of the sensitivity of λc with respect to each wind power Pwj and Δ is the difference
between the variable wind power Pwj and its deterministic value.
According to (8.30), the variance of Δξc can be derived as
X m X m

ð2Þ ∂λc ∂λc ð2Þ
σ2Δξc ¼ γΔξc ¼ Re Re γΔPwj j ð8:35Þ
j ¼1 j ¼1
∂Pwj1 ∂Pwj2 12
1 2
Since σ2ξc ¼ σ2Δξc and σPwj ¼ σΔPwj ,

h i2 h i2
∂λc ∂λc
σ2ξc ¼ Re ∂P w1
σP w1
þ þ Re ∂Pwm
σP wm

∂λc ∂λc
þ 2Re Re ρ σP σP þ ð8:36Þ
∂Pw1 ∂Pw2 Pw1 Pw2 w1 w2

∂λc ∂λc
þ 2Re Re ρ σP σP
∂Pwm1 ∂Pwm Pwm1 Pwm wm1 wm
It can be seen from (8.36) that the variance of ξc PDF, σ2ξc , relies on the standard
deviation of Pwj, the correlation coefficients between different wind power sources
and the sensitivities of λc with respect to each wind power source.
Normally, if the system under mean wind speed is stable (i.e., the mean of ξc is
negative), the smaller σ2ξc leads to a higher probability of small-signal stability as the
variance could affect the shape of the PDF predominantly and thus the probability
considerably. Since the standard deviation and correlation are related to the com-
plementarity of wind power and the sensitivity of the real part of the critical
eigenvalue is affected by HVDC connection topology and control strategy, σ2ξc has
roughly established a relationship between the complementarity, connection tech-
nology and control strategy of offshore wind power generation and the probabilistic
small-signal stability of a power system.
8.3.4 Example 8.3
Variance indices analysis will be demonstrated in this section by using the example
of 16-machine 5-area power system of Fig. 8.2. Offshore wind farms is connected at
node 23, 24 and 29, which is named to be the first, second and third power injection
point of offshore wind power generation respectively. The AC network data, system
load condition, synchronous generator model and parameters are given in Appendix
8.1. Two types of the VSC-HVDC network topologies for the integration of offshore
wind farms in Fig. 8.12 are discussed. The parameters of the wind generators, power-
wind speed curve and wind speed distributions are (in p.u.)
H ¼ 1:7s, D ¼ 0:0, xs ¼ 0:29, xr ¼ 0:29, xm ¼ 2:6, rs ¼ 0:0, rr ¼ 0:0013, Pw0 ¼ 0:3333

KP ¼ 30, KQ ¼ 30:
vc ¼ 4m=s, vr ¼ 10m=s, v f ¼ 22m=s, Pr ¼ 1:0p:u:,
μ ¼ 6m=s, σ ¼ 2:5m=sðCase AÞ, σ ¼ 2:0m=sðCase B; C; D; E; FÞ
2 3
1 0:9 0:3
6 7
ρ ¼ 4 0:9 1 0:5 5ðCase AÞ,
0:3 0:5 1
2 3
1 0:5 0
6 7
ρ ¼ 4 0:5 1 0 5ðCase B; C; D; E; FÞ
0 0 1
a b
WF1 WF2 WF3 WF1 WF2 WF3
69 70 71 69 70 71
~ ~ ~ ~ ~ ~
= = = = = =
DC1 DC2 DC3 Offshore DC1 DC2 DC3 Offshore
MTDC network
DC4 DC5 DC6 Onshore DC4 DC5 DC6 Onshore

= = = = = =
~ ~ ~ ~ ~ ~
1 2 3 1 2 3
Fig. 8.12 Two typical HVDC connection topologies: (a) point-to-point HVDC transmission lines;
(b) multi-terminal HVDC network
Parameters of the VSC-HVDC network are (in p.u.)

R14 ¼ R25 ¼ R36 ¼ 0:0006, X14 ¼ X25 ¼ X36 ¼ 0:0022ðCase AÞ
R17 ¼ R37 ¼ R47 ¼ R67 ¼ 0:0007, R27 ¼ R57 ¼ 0:0003ðCase B; C; DÞ
X17 ¼ X37 ¼ X47 ¼ X67 ¼ 0:0025, X27 ¼ X57 ¼ 0:0011ðCase B; C; DÞ
where the central star node of star topology (ST) MTDC network is denoted as the
DC7 bus.
The VSC-converter model in [31] is used for all the converters. The loss of the
converters is neglected. The parameters in p.u. are Rc ¼ 0.0001, Xc ¼ 0.1643.
The DC voltage droop constants for Case E and F are kdroop ¼ 0.04(Case E),
0.004(Case F).
8.3.4.1 Case A (Offshore Wind Generation of Low Complementarity

Level Connected by Point-to-Point HVDC)
In Case A, three point-to-point HVDC transmission lines shown in Fig. 8.12a are
used to connect three offshore wind farms respectively. From the deterministic
small-signal stability analysis, the 29th eigenvalue of the system state space matrix
is identified to be the critical eigenvalue, i.e., λc ¼ ξc jωc ¼ 0.0043 j3.3742.
Hence deterministically the system is stable. The offshore wind farms at node 69, 70
and 71 (see Fig. 8.12a) are denoted to be the first, second and third source of wind
generation respectively. The sensitivity computation of critical eigenvalue real part
with respect to three sources of wind generation is
∂ξc ∂ξc ∂ξc

¼ 0:0030, ¼ 0:0028, ¼ 0:0068
Low complementarity level of offshore wind generation is taken into account in
this case. Three wind speed sample series that satisfy the Weibull distribution and
spatial correlation given above are generated by normal transformation method
[19, 20] or Copulas method [21]. Then the corresponding wind power sample series
can be obtained by employing the power-wind speed curve. Thus, the standard
deviation of each wind power and their correlation coefficients can be calculated
to be
σPw1 ¼ σPw2 ¼ σPw3 ¼ 0:3271,
ρPw1 Pw2 ¼ 0:8827, ρPw1 Pw3 ¼ 0:2750, ρPw2 Pw3 ¼ 0:4643
The variance indices of complementarity level and probabilistic distribution are

computed by use of (8.33) and (8.36) to be
σ2Pwc ¼ σ2Pw1 þ σ2Pw2 þ σ2Pw3 þ 2ρPw1 Pw2 σPw1 σPw2 þ 2ρPw1 Pw3 σPw1 σPw3 þ 2ρPw2 Pw3 σPw2 σPw3
¼ 6:6825 101
h i2 h i2 h i2
σ2ξc ¼ ∂P w1
σ P w1
þ ∂Pw2
σP w2
þ ∂Pw3
σP w3
∂ξc ∂ξc ∂ξc ∂ξc

þ2 ρ σP σP þ 2 ρ σP σP
∂Pw1 ∂Pw2 Pw1 Pw2 w1 w2 ∂Pw1 ∂Pw3 Pw1 Pw3 w1 w3
∂ξc ∂ξc
þ2 ρ σP σP ¼ 1:1492 105
∂Pw2 ∂Pw3 Pw2 Pw3 w2 w3
Monte Carlo simulation (with 5000 iterations) is carried out and the PDF of ξc by
Monte Carlo simulation is displayed in Fig. 8.13. A comparison of computation time
between Monte Carlo simulation and variance index computation has been carried
out. Based on the same computational resource (Dell OptiPlex 745, Intel Core
2 CPUs 2.66 GHz, 3 GB RAM), the time of Monte Carlo simulation with 5000
iterations is 28954.39 s, while only 14.32 s for the variance calculation, which is
more than 2000 times faster than Monte Carlo simulation.
According to the PDF in Fig. 8.13, it can be obtained that
σ2ξc ¼ 1:1310 105

Ð0
Pðξc < 0Þ ¼ 1 f ξc ðxÞdx ¼ 0:8484
ξc PDF
450
400
350
300
250
200
150
100
50
0
-8 -6 -4 -2 0 2 4
ξc x 10-3
Fig. 8.13 PDF of ξc obtained by Monte Carlo simulation (Case A)

The result has verified the variance computation by using (8.36). It is indicated
that although the system is considered to be stable by the deterministic analysis of
small-signal stability, the system has a high probability of 15.16% to be unstable due
to the low complementarity level of combined wind power, when the uncertainty of
wind generation is considered.
8.3.4.2 Case B (Offshore Wind Generation of High Complementarity

Level Connected by Point-to-Point HVDC)
In this case, high complementarity level of offshore wind generation is considered

and the load condition of Case A is used. Hence, the results of deterministic small-
signal stability analysis and sensitivity computation are exactly the same as Case
A. By employing the same procedure as in Case A, the standard deviation of three
wind power and their correlation coefficients are calculated to be
σPw1 ¼ σPw2 ¼ σPw3 ¼ 0:2860,

ρPw1 Pw2 ¼ 0:4847, ρPw1 Pw3 ¼ 0:0005, ρPw2 Pw3 ¼ 0:0011
By use of (8.33) and (8.36), the variance indices of complementarity level and
probabilistic distribution of ξc are computed to be
σ2Pwc ¼ 3:2489 101 , σ2ξc ¼ 5:8724 106
which are smaller than that obtained in Case A. This is brought about by the
decrease of variation of each wind power and their comparatively weak correlations.
Monte Carlo simulation (with 5000 iterations) is carried out to confirm the
variances calculated above and the PDF of ξc by Monte Carlo simulation is displayed
in Fig. 8.14. According to the PDF in Fig. 8.14, it can be obtained that
σ2ξc ¼ 5:8222 106

Ð0
Pðξc < 0Þ ¼ 1 f ξc ðxÞdx ¼ 0:9454
The result indicates that in the case of high complementarity condition, the
probability of ξc remaining in the left half-plane increases to 94.54% due to variance
reduction when the stochastic variation of wind generation is considered. Hence, it is
successfully demonstrated by Case B that the effectiveness of complementarity of
wind generation to enhance probabilistic small-signal stability of power systems
even when the offshore wind farms are separately connected to different buses of
power system by point-to-point HVDC topology. In other words, impact of com-
plementarity does not rely on the physical combination of wind power by MTDC
network. Moreover, it can be seen from (8.36) that high complementarity level may
also lead to the reduction of probability of small-signal stability if the sensitivities of
ξc have different signs.
ξc PDF
80
70
60
50
40
30
20
10
0
-8 -6 -4 -2 0 2 4
ξc x 10 -3
Fig. 8.14 PDF of ξc obtained by Monte Carlo simulation (Case B)
8.3.4.3 Case C (Offshore Wind Generation of High Complementarity

Level Connected by MTDC with Constant DC Voltage Control)
The MTDC network in Fig. 8.12b is applied to connect the three offshore wind farms
and traditional constant DC voltage control scheme is employed. MTDC network
can be achieved by various topologies such as general ring topology (GRT), star
topology (ST), wind farm ring topology (WFRT) and substation ring topology
(SSRT) [32]. In this case, the ST MTDC network configuration is selected. The
results of deterministic small-signal stability analysis and the standard deviation of
each wind power and their correlation coefficients are the same as in Case B since the
same load condition and wind speed distribution are applied. However, due to the
change of the HVDC connection, the sensitivity of ξc with respect to three sources of
wind generation changes to be
¼ 0:0028
The variance index of probabilistic distribution of ξc are computed by using
(8.36) to be
σ2ξc ¼ 2:4999 106
Compared with Case B, σ2ξc decreases which is caused by the decrease of

sensitivities when σ2Pwc ¼ 3:2489 101 . Monte Carlo simulation (with 5000
ξc PDF
70
60
50
40
30
20
10
0
-7 -6 -5 -4 -3 -2 -1 0 1
ξc x 10-3
Fig. 8.15 PDF of ξc obtained by Monte Carlo simulation (Case C)
iterations) is carried out and the PDF of ξc by Monte Carlo simulation is displayed in
Fig. 8.15. From Fig. 8.15, it can be obtained that
σ2ξc ¼ 2:4773 106

Ð0
Pðξc < 0Þ ¼ 1 f ξc ðxÞdx ¼ 0:9944
It confirms the computational results of σ2ξc previously by using (8.36). The result
also shows that in the case of the MTDC connection, a very high probability, nearly
100%, to be stable in terms of small-signal stability is achieved owing to the
reduction of σ2ξc . Additionally, it can be concluded that in the case of the MTDC
network, high complementarity level can certainly enhance the probabilistic small-
signal stability of a power system due to the identical sensitivity.
8.3.4.4 Case D (Offshore Wind Generation Connected by MTDC

with Constant DC Voltage Control and DC Slack Bus at
Different Location)
In Case D, the results of deterministic small-signal stability analysis and the standard
deviation of each wind power and their correlation coefficients are the same as in
Case B and C. Three offshore wind farms are still connected by the ST MTDC
network with the constant DC voltage control. However, the location of DC voltage
ξc PDF
70
60
50
40
30
20
10
0
-0.01 -0.005 0 0.005 0.01
ξc
Fig. 8.16 PDF of ξc obtained by Monte Carlo simulation (Case D)
control of the MTDC has been changed from DC5 to DC6. Hence, the sensitivity
of ξc with respect to three sources of wind generation becomes

¼ 0:0068
The variance index of probabilistic distribution of ξc are computed to be
σ2ξc ¼ 1:5114 105 , which increases due to the increase of sensitivities. Monte
Carlo simulation (with 5000 iterations) is carried out and the PDF of ξc by Monte
Carlo simulation is displayed in Fig. 8.16. From Fig. 8.16, it can be obtained that
σ2ξc ¼ 1:5481 105

ð0
Pð ξ c < 0 Þ ¼ f ξc ðxÞdx ¼ 0:8402
1
It confirms the computational result of σ2ξc previously and also indicates that the
probability of ξc remaining negative drops considerably to 84.02% even the com-
plementarity level of combined wind power is high. Therefore, in order to improve
the probability of stability, the DC voltage control should be located at the AC bus
with comparatively small sensitivity of ξc with respect to wind power, so that the
stochastic variation of wind power can have less impact on the probabilistic small-
signal stability.
8.3.4.5 Case E (Offshore Wind Generation Connected by MTDC

with DC Voltage Droop Control)
In Case E, three offshore wind farms are connected by the ST MTDC network with
DC voltage droop control and other preconditions stay the same as in Case B, C and
D. there are two types of DC voltage droop control schemes as the linear droop
relation can be established either between DC bus voltage and injected DC power or
between DC bus voltage and injected DC current [33]. When the injected DC power
related voltage droop control is used, the sensitivity of ξc with respect to three
sources of wind generation becomes

¼ 0:0042
It is reasonable that the sensitivities above are almost identical due to the sym-
metry of the DC network and the value of the sensitivities (0.0042) is between 0.0028
and 0.0068 as the change of wind power is distributed on each grid-side converter.
The variance index of probabilistic distribution of ξc are computed to be
σ2ξc ¼ 5:7591 106 , which is very close to that in Case B because same comple-
mentarity level of wind generation is assumed and the change of combination of the
sensitivity related to each grid-side converter is quite moderate, which is a combi-
nation of the sensitivity related to each grid-side converter. Monte Carlo simulation
(with 5000 iterations) is carried out and the PDF of ξc by Monte Carlo simulation is
displayed in Fig. 8.17.
ξc PDF
70
60
50
40
30
20
10
0
-10 -8 -6 -4 -2 0 2 4
ξc x 10-3
Fig. 8.17 PDF of ξc obtained by Monte Carlo simulation (Case E)

From the PDF in Fig. 8.17, it can be obtained that
σ2ξc ¼ 5:6946 106

ð0
Pð ξ c < 0 Þ ¼ f ξc ðxÞdx ¼ 0:9491
1
The probability of ξc remaining negative is only a little bit higher than the
probability in Case B. Hence the impact of MTDC connection on the probabilistic
small-signal stability is not predominant when the DC voltage droop control scheme
is applied due to the moderate sensitivity.
The sensitivity and variance index of ξc are computed and Monte Carlo simula-
tion is carried out also for the injected DC current related voltage droop control
scheme as follows.
¼ 0:0042
σ2ξc ¼ 5:7653 106
σ2ξc ¼ 5:7007 106 ðMonte CarloÞ
ð0
Pð ξ c < 0 Þ ¼ f ξc ðxÞdx ¼ 0:9490
1
It can be seen that calculating results of the two types of droop control schemes
are quite similar.
8.3.4.6 Case F (Offshore Wind Generation Connected by MTDC

with DC Voltage Droop Control and Using Different DC Voltage
Droop Constant)
Compared with Case E, a smaller DC voltage droop constant is selected in Case F for
both types of the droop control schemes. The sensitivity of ξc with respect to three
sources of wind generation is computed to be

0:0044
The sensitivity increases slightly since the injected DC power of grid-side
converter becomes more sensitive to the DC voltage change brought about by the
variation of wind power. The variance index of probabilistic distribution of ξc are
computed to be σ2ξc ¼ 6:1923 106 . Monte Carlo simulation (with 5000 iterations)
is carried out and the PDF of ξc by Monte Carlo simulation is displayed in Fig. 8.18.
ξc PDF
70
60
50
40
30
20
10
0
-10 -5 0 5
ξc x 10-3
Fig. 8.18 PDF of ξc obtained by Monte Carlo simulation (Case F)
From the PDF in Fig. 8.18, it can be obtained that
σ2ξc ¼ 6:1214 106

ð0
Pð ξ c < 0 Þ ¼ f ξc ðxÞdx ¼ 0:9388
1
It can be seen that the decrease of the DC voltage droop constant has weakened
the probabilistic small-signal stability compared to Case E owing to the rise of the
sensitivity of ξc with respect to wind power. That is to say, probabilistic stability
could benefit from the MTDC connection if DC voltage droop constants are
increased.
The injected DC current related voltage droop control scheme is also examined
with the following results.
0:0047
σ2ξc ¼ 7:2149 106
σ2ξc ¼ 7:1293 106 ðMonte CarloÞ
ð0
Pð ξ c < 0 Þ ¼ f ξc ðxÞdx ¼ 0:9218
1
Appendix 8.1: Data of Examples 8.1, 8.2 and 8.3 341
It can be observed that the critical eigenvalue ξc is more sensitive to the variation
of wind generation in this case mainly because one unit change in the injected
current will cause two units change in the injected power for a bipole DC grid.
Appendix 8.1: Data of Examples 8.1, 8.2 and 8.3

Power System [34] (Tables 8.5, 8.6 and 8.7)
All the synchronous generators employ sixth-order detailed model with damping
D ¼ 0.0. The structure of first-order excitation system model is shown by Fig. 8.19.
The parameters of excitation system model are
KA ¼ 7.4, TA ¼ 0.1s, efdmax ¼ 10.0, efdmin ¼ 10.0 (Example 8.1).

Table 8.5 Bus data

Reactive
1 – – 2.5270 1.1856 – –
2 – – 0.0000 0.0000 – –
3 – – 3.2200 0.0200 – –
4 – – 5.0000 1.8400 – –
5 – – 0.0000 0.0000 – –
6 – – 0.0000 0.0000 – –
7 – – 2.3400 0.8400 – –
8 – – 5.2200 1.7700 – –
9 – – 1.0400 1.2500 – –
10 – – 0.0000 0.0000 – –
11 – – 0.0000 0.0000 – –
12 – – 0.0900 0.8800 – –
13 – – 0.0000 0.0000 – –
14 – – 0.0000 0.0000 – –
15 – – 3.2000 1.5300 – –
16 – – 3.2900 0.3200 – –
17 – – 0.0000 0.0000 – –
18 – – 1.5800 0.3000 – –
19 – – 0.0000 0.0000 – –
(continued)

Reactive
20 – – 6.8000 1.0300 – –
21 – – 2.7400 1.1500 – –
22 – – 0.0000 0.0000 – –
23 – – 2.4800 0.8500 – –
24 – – 3.0900 0.9200 – –
25 – – 2.2400 0.4700 – –
26 – – 1.3900 0.1700 – –
27 – – 2.8100 0.7600 – –
28 – – 2.0600 0.2800 – –
29 – – 2.8400 0.2700 – –
30 – – 0.0000 0.0000 – –
31 – – 0.0000 0.0000 – –
32 – – 0.0000 0.0000 – –
33 – – 1.1200 0.0000 – –
34 – – 0.0000 0.0000 – –
35 – – 0.0000 0.0000 – –
36 – – 1.0200 0.1946 – –
37 – – 60.0000 3.0000 – –
38 – – 0.0000 0.0000 – –
39 – – 2.6700 0.1260 – –
40 – – 0.6563 0.2353 – –
41 – – 10.0000 2.5000 – –
42 – – 11.5000 2.5000 – –
43 – – 0.0000 0.0000 – –
44 – – 2.6755 0.0484 – –
45 – – 2.0800 0.2100 – –
46 – – 1.5070 0.2850 – –
47 – – 2.0312 0.3259 – –
48 – – 2.4120 0.0220 – –
49 – – 1.6400 0.2900 – –
50 – – 1.0000 1.4700 – –
51 – – 3.3700 1.2200 – –
52 – – 24.7000 1.2300 – –
53 1.0450 – – – 2.5000 –
54 0.9800 – – – 5.4500 –
55 0.9830 – – – 6.5000 –
56 0.9970 – – – 6.3200 –
57 1.0110 – – – 5.0520 –
58 1.0500 – – – 7.0000 –
59 1.0630 – – – 5.6000 –
60 1.0300 – – – 5.4000 –
(continued)

Reactive
61 1.0250 – – – 8.0000 –
62 1.0100 – – – 5.0000 –
63 1.0000 – – – 10.0000 –
64 1.0156 – – – 13.5000 –
65 1.0110 0.0000 – – – –
66 1.0000 – – – 17.8500 –
67 1.0000 – – – 10.0000 –
68 1.0000 – – – 40.0000 –
Table 8.6 Line data

1 1 2 0.0035 0.0411 0.6987 1
2 1 30 0.0008 0.0074 0.4800 1
3 2 3 0.0013 0.0151 0.2572 1
4 2 25 0.0070 0.0086 0.1460 1
5 2 53 0.0000 0.0181 0.0000 1.025
6 3 4 0.0013 0.0213 0.2214 1
7 3 18 0.0011 0.0133 0.2138 1
8 4 5 0.0008 0.0128 0.1342 1
9 4 14 0.0008 0.0129 0.1382 1
10 5 6 0.0002 0.0026 0.0434 1
11 5 8 0.0008 0.0112 0.1476 1
12 6 7 0.0006 0.0092 0.1130 1
13 6 11 0.0007 0.0082 0.1389 1
14 6 54 0.0000 0.0250 0.0000 1.07
15 7 8 0.0004 0.0046 0.0780 1
16 8 9 0.0023 0.0363 0.3804 1
17 9 30 0.0019 0.0183 0.2900 1
18 10 11 0.0004 0.0043 0.0729 1
19 10 13 0.0004 0.0043 0.0729 1
20 10 55 0.0000 0.0200 0.0000 1.07
21 12 11 0.0016 0.0435 0.0000 1.06
22 12 13 0.0016 0.0435 0.0000 1.06
23 13 14 0.0009 0.0101 0.1723 1
24 14 15 0.0018 0.0217 0.3660 1
25 15 16 0.0009 0.0094 0.1710 1
26 16 17 0.0007 0.0089 0.1342 1
27 16 19 0.0016 0.0195 0.3040 1
28 16 21 0.0008 0.0135 0.2548 1
(continued)

29 16 24 0.0003 0.0059 0.0680 1
30 17 18 0.0007 0.0082 0.1319 1
31 17 27 0.0013 0.0173 0.3216 1
32 19 20 0.0007 0.0138 0.0000 1.06
33 19 56 0.0007 0.0142 0.0000 1.07
34 20 57 0.0009 0.0180 0.0000 1.009
35 21 22 0.0008 0.0140 0.2565 1
36 22 23 0.0006 0.0096 0.1846 1
37 22 58 0.0000 0.0143 0.0000 1.025
38 23 24 0.0022 0.0350 0.3610 1
39 23 59 0.0005 0.0272 0.0000 1
40 25 26 0.0032 0.0323 0.5310 1
41 25 60 0.0006 0.0232 0.0000 1.025
42 26 27 0.0014 0.0147 0.2396 1
43 26 28 0.0043 0.0474 0.7802 1
44 26 29 0.0057 0.0625 1.0290 1
45 28 29 0.0014 0.0151 0.2490 1
46 29 61 0.0008 0.0156 0.0000 1.025
47 9 30 0.0019 0.0183 0.2900 1
48 9 36 0.0022 0.0196 0.3400 1
49 9 36 0.0022 0.0196 0.3400 1
50 36 37 0.0005 0.0045 0.3200 1
51 34 36 0.0033 0.0111 1.4500 1
52 35 34 0.0001 0.0074 0.0000 0.946
53 33 34 0.0011 0.0157 0.2020 1
54 32 33 0.0008 0.0099 0.1680 1
55 30 31 0.0013 0.0187 0.3330 1
56 30 32 0.0024 0.0288 0.4880 1
57 1 31 0.0016 0.0163 0.2500 1
58 31 38 0.0011 0.0147 0.2470 1
59 33 38 0.0036 0.0444 0.6930 1
60 38 46 0.0022 0.0284 0.4300 1
61 46 49 0.0018 0.0274 0.2700 1
62 1 47 0.0013 0.0188 1.3100 1
63 47 48 0.0025 0.0268 0.4000 1
64 47 48 0.0025 0.0268 0.4000 1
65 48 40 0.0020 0.0220 1.2800 1
66 35 45 0.0007 0.0175 1.3900 1
67 37 43 0.0005 0.0276 0.0000 1
68 43 44 0.0001 0.0011 0.0000 1
69 44 45 0.0025 0.0730 0.0000 1
70 39 44 0.0000 0.0411 0.0000 1
(continued)

71 39 45 0.0000 0.0839 0.0000 1
72 45 51 0.0004 0.0105 0.7200 1
73 50 52 0.0012 0.0288 2.0600 1
74 50 51 0.0009 0.0221 1.6200 1
75 49 52 0.0076 0.1141 1.1600 1
76 52 42 0.0040 0.0600 2.2500 1
77 42 41 0.0040 0.0600 2.2500 1
78 41 40 0.0060 0.0840 3.1500 1
79 31 62 0.0000 0.0260 0.0000 1.04
80 32 63 0.0000 0.0130 0.0000 1.04
81 36 64 0.0000 0.0075 0.0000 1.04
82 37 65 0.0000 0.0033 0.0000 1.04
83 41 66 0.0000 0.0015 0.0000 1
84 42 67 0.0000 0.0015 0.0000 1
85 52 68 0.0000 0.0030 0.0000 1
86 1 27 0.0320 0.3200 0.4100 1

0 00 0
G1 B53 100 0.0125 0.1000 0.0310 0.0250 10.20
G2 B54 100 0.0350 0.2950 0.0697 0.0500 6.56
G3 B55 100 0.0304 0.2495 0.0531 0.0450 5.70
G4 B56 100 0.0295 0.2620 0.0436 0.0350 5.69
G5 B57 100 0.0270 0.3300 0.0660 0.0500 5.40
G6 B58 100 0.0224 0.2540 0.0500 0.0400 7.30
G7 B59 100 0.0322 0.2950 0.0490 0.0400 5.66
G8 B60 100 0.0280 0.2900 0.0570 0.0450 6.70
G9 B61 100 0.0298 0.2106 0.0570 0.0450 4.79
G10 B62 100 0.0199 0.1690 0.0457 0.0400 9.37
G11 B63 100 0.0103 0.1280 0.0180 0.0120 4.10
G12 B64 100 0.0220 0.1010 0.0310 0.0250 7.40
G13 B65 200 0.0030 0.0296 0.0055 0.0040 5.90
G14 B66 100 0.0017 0.0180 0.00285 0.0023 4.10
G15 B67 100 0.0017 0.0180 0.00285 0.0023 4.10
G16 B68 200 0.0041 0.0356 0.0071 0.0055 7.80
00 0 00 0 00
Generator Td0 ðsÞ xq xq xq Tq0 ðsÞ Tq0 ðsÞ M
No.
G1 0.05 0.0690 0.0280 0.0250 1.50 0.035 42.0
G2 0.05 0.2820 0.0600 0.0500 1.50 0.035 30.2
(continued)

0 00 0
00 0 00 0 00
Generator Td0 ðsÞ xq xq xq Tq0 ðsÞ Tq0 ðsÞ M
No.
G3 0.05 0.2370 0.0500 0.0450 1.50 0.035 35.8
G4 0.05 0.2580 0.0400 0.0350 1.50 0.035 28.6
G5 0.05 0.3100 0.0600 0.0500 0.44 0.035 26.0
G6 0.05 0.2410 0.0450 0.0400 0.40 0.035 34.8
G7 0.05 0.2920 0.0450 0.0400 1.50 0.035 26.4
G8 0.05 0.2800 0.0500 0.0450 0.41 0.035 24.3
G9 0.05 0.2050 0.0500 0.0450 1.96 0.035 34.5
G10 0.05 0.1150 0.0450 0.0400 1.50 0.035 31.0
G11 0.05 0.1230 0.0150 0.0120 1.50 0.035 28.2
G12 0.05 0.0950 0.0280 0.0250 1.50 0.035 92.3
G13 0.05 0.0286 0.0050 0.0040 1.50 0.035 248.0
G14 0.05 0.0173 0.0025 0.0023 1.50 0.035 300.0
G15 0.05 0.0173 0.0025 0.0023 1.50 0.035 300.0
G16 0.05 0.0334 0.0060 0.0055 1.50 0.035 225.0
e fd max
Vtref e fd 0
+ +
Vt KA Δe fd e fd
Σ Σ
– 1 + TAs +
e fd min
Fig. 8.19 The first-order excitation system model of synchronous generator
References
1. Kendall M (1987) Kendall’s advanced theory statistics. Oxford University Press, New York,
NY
2. Cramer H (1946) Numerical methods of statistics. Princeton University Press, Princeton, NJ
3. Ma J, Dong ZY, Zhang P (2006) Eigenvalue sensitivity analysis for dynamic power system. In:
International conference on power system technology, Chongqing, pp 22–26
4. Dong ZY, Pang CK, Zhang P (2005) Power system sensitivity analysis for probabilistic small
signal stability assessment in a deregulated environment. Int J Control Automat Syst 3:355–362
5. Robert CP, Casella G (2004) Monte Carlo statistical methods. Springer, New York
6. Xu Z, Dong ZY, Zhang P (2005) Probabilistic small signal analysis using Monte Carlo
simulation. In: IEEE PES SM Paper 2: San Francisco, pp 1658–1664
7. Xu Z, Ali M, Dong ZY, Li X (2006) A novel grid computing approach for probabilistic small
signal analysis. In: IEEE PES SM Paper, Montreal, Que.
8. Burchett RC, Heydt GT (1978) Probabilistic methods for power system dynamic stability
studies. IEEE Trans Power Appar Syst PAS-97:695–702
References 347
9. Burchett RC, Heydt GT (1978) A generalized method for stochastic analysis of the dynamic
stability of electric power system. IEEE PES SM Paper, A78528
10. Brucoli M, Torelli F, Trovato M (1981) Probabilistic approach for power system dynamic
stability studies. IEE Proc 128(5):295–301
11. Wang KW, Tse CT, Tsang KM (1997) Algorithm for power system dynamic stability studies
taking account the variation of load power. In: 4th International conference on advances in
power system control operation and management 2(2), Hong Kong, pp 445–450
12. Pang CK, Dong ZY, Zhang P, Yin X (2005) Probabilistic analysis of power system Small signal
stability region. In: International conference on control and automation 1(1), Budapest, pp
503–509
13. Wang KW, Chung CY, Tse CT, Tsang KM (2000) Improved probabilistic method for power
system dynamic stability studies. IEE Proc Gener Transm Distrib 147(1):37–43
14. Yi HQ, Hou YH, Cheng SJ, Zhou H, Chen GG (2007) Power system probabilistic small signal
stability analysis using two point estimation method. In: 402–407, UPEC, Brighton
15. Xu XL, Lin T, Zhang XM (2009) Probabilistic analysis of small signal stability of microgrid
using point estimate method. In: International conference on SUPERGEN, Nanjing, pp 1–6
16. Li MY, Ma J, Dong ZY (2009) Uncertainty analysis of load models in small signal stability. In:
International conference on SUPERGEN, Nanjing, pp 1–6
17. Bu SQ (2012) Probabilistic small-signal stability analysis and improved transient stability
control strategy of grid-connected doubly fed induction generators in large-scale power sys-
tems. PhD thesis.
18. Freris L, Infield D (2008) Renewable energy in power systems. John Wiley and Sons
19. Usaola J (2010) Probabilistic load flow with correlated wind power injections. Electr Power
Syst Res 80(5):528–536
20. Morales JM, Baringo L, Conejo AJ, Minguez R (2010) Probabilistic power flow with correlated
wind sources. IET Gener Transm Distrib 4(5):641–651
21. Papaefthymiou G, Kurowicka D (2009) Using copulas for modeling stochastic dependence in
power system uncertainty analysis. IEEE Trans Power Syst 24(1):40–49
22. Abouzahr I, Ramakumar R (1991) An approach to assess the performance of utility-interactive
wind electric conversion systems. IEEE Trans Energy Conver 6:627–638
23. Akagi H, Sato H (2002) Control and performance of a doubly-fed induction machine intended
for a flywheel energy storage system. IEEE Trans Power Electron 17:109–116
24. Gao B, Morison GK, Kundur P (1992) Voltage stability evaluation using modal analysis. IEEE
25. Aik D, Andersson G (1997) Voltage stability analysis of multi-infeed HVDC systems. IEEE
Trans Power Deliver 12(3):1309–1318
26. Aik D, Andersson G (1998) Use of participation factors in modal voltage stability analysis of
multi-infeed HVDC systems. IEEE Trans Power Deliver 13(1):203–211
27. Allan R, Silva A (1981) Probabilistic load flow using multilinearizations. IEE Proc 128(5)
28. Blau J (2010) Europe plans a north sea grid. IEEE Spectr 47(3):12–13
29. Rudion K, Orths A, Eriksen PB, Styczynski ZA (2010) Toward a benchmark test system for the
offshore grid in the north sea. In: IEEE power and energy society general meeting, Providence,
RI, pp 1–8.
30. Simonsen T, Stevens B (2004) Regional wind energy analysis for the contral United States. In:
Proceedings of the global wind power conference, pp 1–16
31. Beerten J, Cole S, Belmans R (2012) Generalized steady-state VSC MTDC model for sequen-
tial AC/DC power flow algorithms. IEEE Trans Power Syst 27(2):821–829
32. Gomis-Bellmunt O, Liang J, Ekanayake J, King R, Jenkins N (2010) Topologies of
multiterminal HVDC-VSC transmission for large offshore wind farms. Electr Power Syst Res
81:271–281
33. Beerten J, Belmans R (2012) Modeling and control of multi-terminal VSC HVDC systems.
Energy Procedia 24:123–130
Index
A Cumulant theory, 306, 308–315

AC network data, system load condition, 331 Cumulative Distribution Function (CDF), 304
AC/DC power system, 321, 322
Aggregate model, 69–70
Analytical and numerical comparison D
analysis, 86 Damping contributions, 75
Analytical method, 306, 308 Damping torque analysis (DTA), 7, 72, 80, 98,
Automatic voltage regulators (AVRs), 10, 104 100, 150, 243, 253–254
computation’ methods, 64
DFIG, 73
B FSIG and DFIG possess, 63, 64
Base case comparison study, 80–84 generic implementation framework, 70
Bus Data, 89, 139, 341 grid-connected wind power induction
generators, 63–65
WPIGs, 64, 69
C Damping torque contribution, 106, 107
Closed-loop and open-loop EOM, 112 DC capacitor, 216
Closed-loop and open-loop state matrix, 125 DC voltage control, 337
Closed-loop converter oscillation mode, DC voltage droop control, 338
166, 171 Decomposed modal analysis, 97–103, 105, 108
Closed-loop interconnected model, 98, 100, DFIG, 98
105, 149 DFIG oscillation mode (DOM), 178–182,
DFIG, 61–62 188, 237
PMSG, 59–60 DFIG rotor-side converter, 317
Closed-loop interconnected system, 112, Diagonal matrix/block diagonal matrix, 57
150–152 Distribution function, 312
Closed-loop power system, 113 Doubly-fed induction generators (DFIGs), 316
Closed-loop state space model, 8 active and reactive power injected, 54
Closed-loop system, 14, 15 basic data, 143
Combined wind power, 329–330 bus data, 145
Conventional modal analysis, 98 comparison analysis, 79
Converter control-based generators (CCBGs), 5 computational results, 110
Converter oscillation modes, 153–156, constant rotor voltage, 77–79
158–159 damping coefficient, 73
Copulas method, 332 damping contributions, 79

Variable Speed Wind Generators, https://doi.org/10.1007/978-3-319-94168-4
350 Index
Doubly-fed induction generators (DFIGs) Electromechanical low-frequency

(cont.) oscillation, 147
damping effectiveness, 83 Electromechanical oscillation modes (EOMs),
dynamic components, 84 2, 82, 98, 202
dynamic interactions, 103, 117, 118, DFIG, 236–239
122, 128 DSOGI PLL, 223, 235–236
eigenvalue, 81 GSC vector control system, 231
electrical function, 40 modal resonance, 226, 228
EOM, 118 NEPS, 222, 225
fourteenth-order model, 105 OLMR, 229
FSIG dynamics, 83 open-loop AC power system, 252
FSIG featured, 80 PLL gains, 227–232
FSIG, 63, 64 power system operating conditions,
gains of GSC, 143 224–226
generator-side dynamics, 77 reactive power output, 230
grid connection study, 64 relative rotor position, 230
grid connection, 97 SCR and wind power penetration,
GSC and associated control system, 49–53 232–235
induction generator and two-mass shaft VSC-based system, 224
rotational system, 41–45 Example power system, 104–105, 157
internal dynamics, 81 Excitation system model, 88, 341
line data, 144
linearized model diagram, 76, 77
machine data, 145 F
offset rotor voltage, 77 Filter reactance, 215
PQ control, 80 First-order excitation system model, 94
q-axis and d-axis current control, 121 First-Order Second Moment Method
rotor circuit, 79 (FOSMM), 307
rotor motion equations, 121 Fixed-speed induction generator (FSIG), 63
RSC and associated control system, 45–49
RSC control system model, 74
RSC controller parameters, 80 G
RSC dynamics, 75–77 General ring topology (GRT), 335
RSC vector control system, 119 Gram-Charlier expansion, 311, 317, 318
RSC, 73 Grid connection, 148
side linearized model diagram, 78 Grid side converter (GSC), 13, 33–38,
simplified transfer function model, 118–121 49–53, 203
state equations, 53 Grid-connected wind power, 312, 313
steady-state value, 78 Grid-connected wind power induction
terminal voltage control vs. reactive power generators, 63–64, 70–72
control, 122–124 Grid-side converter (GSC), 73, 148
transfer functions, 82 active power control, 155
wind power generation, 40 active power control outer loop, 156
WPIG reference, 75 active power control path, 155
Duel second-order generalized integrator phase configuration, 154, 158
locked loop PLL (DSOGI PLL), 222, converter control system, 154
223, 235–236 dynamic function, 181
Dynamic interactions, 105, 147, 163, 172–199 OLMR, 159
parameters, 158–160
PMSG1, 167
E reactive control path, 155, 156
Eigenvalue analysis, 324 reactive power control, 155, 167
Eigenvalue variation, 82 GSC control system, 154
Index 351
H constant power injections, 243

Hooke-Jeeves direct searching algorithm, 134 converter control system, 249
HVDC networks, 320 DC voltage control, 286–288
DC voltage droop control, 263–265
DTA, 253–254
I dynamic interactions, 254–256
Impedance model, 55 eight-terminal DC network, 277–283
Index of dynamic interactions (IDI) input-output signal vectors, 247
advantages, 117 Laplace operator, 247
analytical expression, 116 linearized active and reactive power
computational burden, 117 output, 249
derivation, 113–116 master-slave control, 257–260
EOM, 114, 116 NETS, 256–257
estimation error, 115 network equation, 247
grid-connected DFIG, 111–113 OLMR and near strong modal resonance,
mathematical derivation, 115 261–263
modal sensitivity, 114 open-loop modal analysis, 292–294
Induction generator parameters, 88 parameter tuning, 297–298
Input impedance matrix model, 13 PI controllers, 248
Inter-area oscillation modes, 107 small-signal instability risk, 295–296
small-signal stability, 243, 244
transfer function matrix model, 251
J VSC
Jacobian matrix, 326 active power control, 268–271
DC voltage control/voltage droop
control, 272–273
L OLMR, 273–277
Line Data, 91, 140, 343 wind power output, 247
Linearized dynamic equation, 154 wind power transmission, 243
Linearized state-space model, 97 Multi-terminal VSC-HVDC (MTDC)
Linearized voltage equation, 153 technology, 328
Load flow change, 108–109 Multivariable closed-loop interconnected
model, 173–174
M
Machine Data, 93, 142, 345 N
Machine side converter (MSC), 30–33 New England Power System (NEPS), 124, 125,
Modal resonance, 255 128, 143, 157, 166, 182, 194, 197,
Modal sensitivity, 114, 116 222, 223, 240, 341
Mode-to-parameter sensitivity, 261, 263 New England test system (NETS), 256–257
Monte Carlo simulation, 306, 307, 318, 319, Non-linear simulation, 259
322, 333–335, 338–340 Normal distribution, 306
MTDC network, 334, 336 Normal transformation method, 332
Multi-machine power system, 55, 68, 97, 98,
100, 173, 174
Multiple-input and multiple-output (MIMO), 18 O
Multi-point linearization technique, 322–324 Offshore wind farms, 331
Multi-terminal DC (MTDC) Offshore wind generation, 332–334
AC power system, 243–247 Offshore wind power generation, 328
active/reactive power control, 289–291 Open-loop and closed-loop converter
CIGRE five-terminal DC network, oscillation mode, 169
266–269, 283–285 Open-loop converter oscillation mode, 151
closed-loop model, 251–252 Open-loop DFIG oscillation modes, 181
352 Index
Open-loop eigenvalues, 185 dynamic interactions, 148–156, 160–162,

Open-loop modal residue, 293 167–170
Open-loop modal resonance (OLMR) dynamics, 168
characteristic equation, 176 EOM, 150, 171
condition, 151, 206–207 grid-connection, 148–151, 166,
consequence, 148 170–171
converter oscillation mode, 166 GSC and associated control system, 33–38
DFIG and the power system, 195 GSC, 240
DFIG1 and DFIG2, 197, 199 line data, 240
DFIGs, 176, 196 MSC, 240
dynamic interactions, 151, 161, 190 MSC and associated control system, 30–33
EOM, 162, 166, 195 multi-machine power system, 148
EOMs, 166, 171 New England power system, 157
equations, 178 OLMR, 163, 164, 166, 167, 170
estimation, 167 open-loop converter oscillation mode, 148,
GSC, 148 151, 168
GSC active power control, 159 participation, 164
impact of, 207–209 power exchange, 161
multiple DFIGs, 196–199 power output, 39
non-linear simulation, 199 power system, 149, 150, 155, 156, 160, 162,
open-loop DFIG oscillation mode, 173 169, 171
open-loop power system, 176 reactive power output, 169
PLL SG, 27–30
current control inner loops of GSC, SRF PLL, 240
213–214 state-space and transfer function
DC voltage control outer loop of GSC, models, 149
211–212 state-space model, 40
PMSG1, 164–167 total participation, 161
PMSG1 and power system, 163 transfer function model, 155
PMSG2, 169 wind power generation, 147, 148
PMSG2 and power system, 168, 170 Phase locked loop (PLL)
PMSG3, 171 AC power system, 202
PMSG3 and power system, 171 bandwidth, dynamic component, 209
PMSGs, 148, 156, 159, 172 closed-loop control system, 201
power control outer loop, 159 closed-loop model, 204–206
power system, 152 current vector control algorithm, 201
RSC active power control system, 186–190 DC voltage control system, 209
RSC reactive power control system, dynamic interactions, 201, 202
193–195 EOMs (see Electromechanical oscillation
Open-loop state matrix, 197 modes (EOMs))
Open-loop system, 15, 84 external power system, 203
external power system, PMSG
infinite AC busbar, 210
P non-linear simulation, 219, 222
Participation factors (PFs), 7, 212, 226, 227, OLMR (see Open-loop modal resonance
258, 264 (OLMR))
Permanent magnet synchronous generators SRF, 210
(PMSGs) weak network connection, 214–219
active power exchange, 165 wind farm connection, 219–221
active power output, 163 function, 203–204
configuration, 27, 28, 153 grid-connected wind farm, 209
converter oscillation modes, 147, 159 multi-machine power system, 203
DFIG, 27 OLMR, 206–209
Index 353
open-loop and closed-loop oscillation Rest of power system (ROPS), 149

modes, 202 Rest of the power system (ROPS), 98, 202,
PMSG, 202 205–207, 209, 224, 234
PMSG’s terminal voltage, 204 ROPS subsystem, 99
ROPS, 202 Rotor motion equations, 121
zero phase-tracking error, 202 Rotor-side converter (RSC), 45–49
Phillips-Heffron model, 65–69, 87 algebraic model, 76
PLL oscillation mode (POM), 202, 207 configuration diagram, 74
Point estimate method, 307 control parameters, 88
Point of common coupling (PCC), 8, 34, 203 DFIG model, 75–77
Power system stabilizers (PSSs), 10, 104 generator parameters, 80, 84
Power-wind speed curve, 331 induction generator, 75
Practical large-scale power system, 117 offset rotor voltage, 80
Probabilistic analysis P-Control algebraic model, 77
AC/DC power system, 323 PI controllers, 77
AC/DC system, 328 state variables, 73
analytical method, 306 transfer functions, 75
assumption of linearity, 311 RSC active power control system, 197
categories, 304 RSC basic control, 75
CDF and PDF curve, 314 RSC controller parameters, 75
combined wind power, 337 RSC Q-Control algebraic model, 77
cumulant theory, 306 RSC vector control system, 120
distribution function, 310–312
grid-connected onshore wind power
generation, 307 S
HVDC connection topology, 330 Short circuit ratio (SCR), 13, 214, 232–235
index, 304 Single machine infinite bus (SMIB)
methodologies, 304 system linearized model, 87
numerical method, 305 WPIG, 86–88
one-point linearization, 326 Single-input and single-output (SISO), 18
PDF and CDF, 308, 315 16-Machine 5-area AC/DC power system, 325
procedure, 315, 319 16-Machine 68-bus NYPS-NETS
semi-invariant, 310 test system, 81
small-signal stability, 305, 322 16-Machine 68-bus power system, 104
spatial correlations, 311 Small-signal angular stability, 321
variance index, 330, 335 Small-signal stability analysis, 336
wind power data, 314 Small-signal voltage stability, 321
Probabilistic Collocation Method (PCM), 307 Spatial correlations, 313
Probabilistic density function (PDF), 304 State-space and transfer function models, 149
Probabilistic small-signal stability, 304, 313 State-space models, 56–58, 98, 99, 174
Probability theory, 310 Steady-state operating points, 324, 326
Pulse width modulation (PWM), 31, 35, 154 Step-by-step model transformation analysis, 79
Substation ring topology (SSRT), 335
Sub-synchronous oscillation (SSO), 17, 296
R Sub-synchronous resonance (SSR), 295
Reactive power control outer loop, 167 Synchronous generators (SGs), 64, 88, 94, 101,
Reactive power control/terminal voltage 102, 151, 161, 222
control, 109–110 Synchronous reference frame (SRF), 210
Reactive power/voltage control, 109 Synchronous reference frame PLL (SRF PLL),
Redrawn model, 100 210, 219, 223, 224, 227–232,
Relative rotor positions, 169 235, 237
354 Index
T renewable power generation, 11

Taylor series expansion, 153 small-signal stability
Terminal voltage control, 110 configuration, 13
Terminal voltage control vs. reactive power damping torque analysis, 16
control, 137–138 dynamic interactions, PLL, 14–15
Time domain simulation, 83 frequency-domain analysis, 16
Transfer function matrix, 100 VSC-based HVDC line, 17–19
Transfer function matrix model, 180 VSC-based MTDC network, 19–22
Transfer function model, 99, 112 weak grid connection and PLL, 12–14
Transitional generator model, 79 VSC-converter model, 332
Transmission network, 56 VSC-HVDC network, 325, 332
Two Point Estimate (TPE) method, 307 VSC-HVDC transmission, 320
V W
Variable speed wind generators (VSWGs), Weibull (or normal skew) distribution, 308
110, 147 Weibull distribution, 307, 313
affecting factors, 7–8 Western Electricity Coordinating Council
DFIG, 2 (WECC) power system, 147
EOMs, 2 Wind farm ring topology (WFRT), 335
grid connection, 2 Wind generation, 338
LEPOs, 1 Wind power generation, 147, 157, 310, 322
load flow, 2 Wind power induction generators (WPIGs), 64,
PLL, 10–11 69–70
power system, 2–7, 9 algebraic interface equations, 68
reactive power/terminal voltage control, 10 closed-form solution, 72
SGs, 2–4, 8–10 damping mechanisms, 64, 72
single-machine infinite-bus power system, 6 dynamics, 72
wind power generation, 3 G5-G15 power angle curve, 83
Variance index, 330 linearized model diagram, 76
Vector control system, 119 multi-machine power system, 64
Voltage source converter (VSC) open-loop system, 84
conventional stability theory, 12 parameters, 80
DC/AC power system, 12 types, 72
dynamic interactions, 147 Wind power penetration, 126
grid connection, 12 Wind power transmission, 256–257
HVDC lines and MTDC, 11 Wind power variations, 309, 317

Du2018 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Du2018 PDF

Uploaded by

Copyright:

Available Formats

Wenjuan Du · Haifeng Wang · Siqi Bu

ISBN 978-3-319-94167-7 ISBN 978-3-319-94168-4 (eBook)

Library of Congress Control Number: 2018947164

© Springer International Publishing AG, part of Springer Nature 2018

dynamic interactions introduced by the grid-connected variable speed wind gener-

Beijing, China Wenjuan Du

2.3 Linearized Model of the Power System . . . . . . . . . . . . . . . . . . . . 55

Appendix 3.1: A Typical Example of a SMIB System

Appendix 5.1: Data of Examples 5.1 and 5.2 . . . . . . . . . . . . . . . . . . . . 199

7.2 Small-Signal Stability of an MTDC Network for the Wind

Appendix 8.1: Data of Examples 8.1, 8.2 and 8.3 . . . . . . . . . . . . . . . . . 341

1.1 Impact of Grid-Connected VSWGs on the Small-Signal

© Springer International Publishing AG, part of Springer Nature 2018 1

1.1.1 Impact of the VSWGs: A Brief Review

1.1.1.1 Representatives of Earlier Study

Representatives of earlier published study about the impact of grid-connected

Fig. 1.1 Case considered in

Fig. 1.2 Case considered in [2]: the VSWG replacing the SG

Fig. 1.3 Case considered in VSWG

1.1.1.2 Representatives Study: The VSWG Displacing the SG

Strategy of investigation via a VSWG displacing a SG in a power system is shown

1.1.1.3 Representatives Study: Adding the VSWG in the Power System

1.1.2 Impact of the VSWGs: Discussions

1.1.2.1 Strategy of Adding the VSWG in the Power System

Fig. 1.4 A single-machine Infinite bus

Fig. 1.5 Addition of a Infinite bus

1.1.2.2 Separate Examination of Affecting Factors

Fig. 1.6 Closed-loop state

1.1.2.3 Strategy of the VSWG Displacing the SG

A VSWG displacing a SG at a same location rarely happens in practical power

Fig. 1.7 Power system with Infinite bus

1.1.2.4 Impact of the VSWG Reactive Power/Terminal Voltage Control

1.1.2.5 Effect of the Phase Locked Loop (PLL)

1.2 Small-Signal Stability of the VSC-Based DC/AC Power

1.2.1 Small-Signal Stability of a Single Grid-Connected VSC

Inner loops PLL

Outer loops Power

Fig. 1.8 Conﬁguration of a single grid-connected VSC system

1.2.1.2 Impact of Dynamic Interactions Introduced by the PLL

Fig. 1.9 Closed-loop

1.2.1.3 Summary: The Single VSC System

1.2.2 Small-Signal Stability of Integrated VSC-Based DC/AC

DC/AC DC/DC interactions between DC/AC

manageable. However, the applicability of the conclusions obtained is constrained

1.2.2.1 Small-Signal Stability of AC Power System Integrated

Figure 1.10 shows the conﬁguration of an AC power system integrated with a

Interactions through the Interactions through the AC

Network ... ... Network SG-2

1.2.2.2 Small-Signal Stability of AC Power System Integrated

Figure 1.11 shows the conﬁguration of an AC power system integrated with a

1.2.2.3 Summary: The HVDC/AC and MTDC/AC Power System

Key issues in the study of small-signal stability of the VSC-based HVDC/AC or

Linearized model of a power system with a grid-connected variable speed wind

2.1 Linearized Model of a PMSG

Figure 2.1 shows the conﬁguration of a grid-connected PMSG, which is comprised

2.1.1 Linearized Model of the Permanent Magnet SG

© Springer International Publishing AG, part of Springer Nature 2018 27

Vpsd +jVpsq Ipcd

w prref Ipsdref Vpdcref Q pref

Fig. 2.1 Conﬁguration of a grid-connected PMSG

Tpe ¼ ψpsq Ipsd ψpsd Ipsq ð2:3Þ

Flux linkage equations of the permanent magnet SG are

By neglecting the variation of the mechanical torque input to the permanent