(Research Topics in Wind Energy 5) Ke Ma (Auth.) - Power Electronics For The Next Generation Wind Turbine System-Springer International Publishing (2015) PDF

RESEARCH
TOPICS
IN WIND
ENERGY 5
Ke Ma
Power Electronics
for the Next
Generation Wind
Turbine System
123
Research Topics in Wind Energy
Volume 5
Series editor
Joachim Peinke, University of Oldenburg, Oldenburg, Germany
e-mail: peinke@uni-oldenburg.de
About this Series
The series Research Topics in Wind Energy publishes new developments and
advances in the fields of Wind Energy Research and Technology, rapidly and
informally but with a high quality. Wind Energy is a new emerging research field
characterized by a high degree of interdisciplinarity. The intent is to cover all the
technical contents, applications, and multidisciplinary aspects of Wind Energy,
embedded in the fields of Mechanical and Electrical Engineering, Physics,
Turbulence, Energy Technology, Control, Meteorology and Long-Term Wind
Forecasts, Wind Turbine Technology, System Integration and Energy Economics,
as well as the methodologies behind them. Within the scope of the series are
monographs, lecture notes, selected contributions from specialized conferences and
workshops, as well as selected PhD theses. Of particular value to both the
contributors and the readership are the short publication timeframe and the
worldwide distribution, which enable both wide and rapid dissemination of research
output. The series is promoted under the auspices of the European Academy of
Wind Energy.
More information about this series at http://www.springer.com/series/11859

Ke Ma
Power Electronics
for the Next Generation Wind
Turbine System
123
Ke Ma
Department of Energy Technology
Aalborg University
Aalborg
Denmark
ISSN 2196-7806 ISSN 2196-7814 (electronic)

Research Topics in Wind Energy
ISBN 978-3-319-21247-0 ISBN 978-3-319-21248-7 (eBook)
DOI 10.1007/978-3-319-21248-7
Library of Congress Control Number: 2015944177
Springer Cham Heidelberg New York Dordrecht London

© Springer International Publishing Switzerland 2015
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
Frontmatter 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.
Printed on acid-free paper
Springer International Publishing AG Switzerland is part of Springer Science+Business Media

(www.springer.com)
Preface
The study for this book was carried out during my Ph.D. in the period between June
2010 and April 2013, at Department of Energy Technology in Aalborg University,
Denmark. Sophisticated industry and long-term academic focus on wind power is
one of the reasons I came here to do this research. After 3 years of unforgettable
researches and experiences, I start to realize that the large-scale utilization of wind
energy could be far more challenging than I expected. And more importantly, many
of the problems as well as the technology potentials may have not been uncovered
yet in this field.
The purpose of this work is to study the power electronics used for the next
generation wind turbine system. Some criteria and tools for evaluating and
improving the critical performances of wind power converters have been proposed
and established. It is the hope of the author that this book can address some
emerging problems as well as possibilities for wind power conversion, and become
an inspired reference for researchers in this field.
I would like to show grateful thanks to Prof. Frede Blaabjerg for the impressive
and fruitful discussion during this study. The constructive discussions, patient
corrections, and also continuous encouragements not only contribute to this work,
but also have great influences on my researching, networking, managing, and
supervising. Furthermore, I would like to sincerely acknowledge Prof. Marco
Liserre from Kiel University, Germany, for his inspired suggestions and invaluable
help during this work. I also want to show regard to Prof. Dehong Xu from
Zhejiang University, China for his supports and concerns, which are precious for
my staying in Denmark.
Aalborg Ke Ma
March 2015
v
Contents
Part I Monograph
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 State-of-the-Art for Wind Power Generation. . . . . . . . . . . . . . 3
1.2 Development of Wind Power Technologies . . . . . . . . . . . . . . 5
1.2.1 Evolution of Wind Turbine Concepts . . . . . . . . . . . . . 5
1.2.2 Evolution of Power Electronics for Wind
Turbines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Emerging Challenges for Wind Power Converter . . . . . . . . . . 8
1.3.1 More Grid Supports . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Higher Reliability. . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.3 Special Cost Considerations. . . . . . . . . . . . . . . . . . . . 13
1.3.4 Formulation of Overall Requirements . . . . . . . . . . . . . 14
1.4 Scopes of the Book. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2 Promising Topologies and Power Devices for Wind Power

Converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1 Promising Converter Topologies . . . . . . . . . . . . . . . . . . . . . . 19
2.1.1 Traditional Two-Level Converters . . . . . . . . . . . . . . . 19
2.1.2 Multilevel Converters . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1.3 Multi-cell Converters . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Potential Power Semiconductor Devices . . . . . . . . . . . . . . . . 26
2.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3 Criteria and Tools for Evaluating Wind Power Converter . . . . . . 31

3.1 Importance of Thermal Stress in Wind Power Converter . . . . . 31
3.1.1 Thermal Stress Versus Reliability. . . . . . . . . . . . . . . . 32
3.1.2 Thermal Stress Versus Cost. . . . . . . . . . . . . . . . . . . . 34
vii
viii Contents
3.2 Classification and Approach for the Thermal Stress

Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 37
3.2.1 Classification of Thermal Stress in Wind Power
Converter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.2 Methods and Models for Stress Analysis. . . . . . . . . . . 38
3.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4 Thermal Stress of 10-MW Wind Power Converter

Under Normal Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1 Requirements and Conditions Under Normal Operation . . . . . . 45
4.2 Stress of Converter Imposed by Wind Speeds . . . . . . . . . . . . 47
4.2.1 Thermal Stress Under Steady-State Wind Speeds . . . . . 47
4.2.2 Thermal Stress Under Wind Speed Variations . . . . . . . 50
4.3 Stress of Converter Imposed by Grid Codes . . . . . . . . . . . . . . 51
4.3.1 Converter Efficiency Considering Reactive
Power Demands by Grid Codes . . . . . . . . . . . . . .... 51
4.3.2 Thermal Stress Considering Reactive Power
Demands by Grid Codes . . . . . . . . . . . . . . . . . . . . . . 53
4.4 A Thermal Control Method Utilizing Reactive Power . . . . . . . 55
4.4.1 Control Idea and Diagram . . . . . . . . . . . . . . . . . . . . . 55
4.4.2 Idea to Overcome the Reactive Power Limits . . . . . . . 56
4.4.3 Thermal Stress Considering Extended Q Ranges
in Paralleled Converters . . . . . . . . . . . . . . . . . . . . . . 57
4.4.4 Thermal Control Results . . . . . . . . . . . . . . . . . . . . . . 57
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5 Stress Analysis of 3L-NPC Wind Power Converter Under

Fault Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.1 Requirements and Conditions Under Fault Operation. . . . . . . . 63
5.2 Stress Analysis of Converter Under LVRT. . . . . . . . . . . . . . . 67
5.2.1 Electrical Behaviors . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.2.2 Thermal Behaviors . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.3 Thermal Redistributed Modulations Under LVRT . . . . . . . . . . 71
5.3.1 Basic Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
5.3.2 A Group of Modulation Methods . . . . . . . . . . . . . . . . 74
5.3.3 Loss and Thermal Improvements . . . . . . . . . . . . . . . . 77
5.3.4 Neutral Point Potential Control and Total
Harmonic Distortion . . . . . . . . . . . . . . . . . . . ...... 79
Contents ix
5.4 New Power Control Methods Under Unbalanced

AC Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.4.1 Applicable Conditions and Control Structure . . . . . . . . 81
5.4.2 Control Ideas and Methods . . . . . . . . . . . . . . . . . . . . 82
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6 Conclusions and Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . 95

6.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.2 Proposals for Future Research Topics . . . . . . . . . . . . . . . . . . 97
7 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
7.1 Used Models for Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 99
7.1.1 Wind Speed Generator . . . . . . . . . . . . . . . . . . . . . . . 99
7.1.2 Wind Turbine Model . . . . . . . . . . . . . . . . . . . . . . . . 99
7.1.3 Generator Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7.1.4 Parameter for Thermal Impedance
of Used IGCT . . . . . . . . . . . . . . ............... 101
7.2 Experimental Setup . . . . . . . . . . . . . . . . ............... 103
Part II Specially Selected Topics
8 The Impacts of Power Switching Devices to the Thermal

Performances of 10 MW Wind Power NPC Converter . . ....... 107
8.1 Wind Power Converter for Case Study . . . . . . . . . . ....... 107
8.2 Thermal-Related Characteristics of Different
Power Switching Devices . . . . . . . . . . . . . . . . . . . . . . . . . . 108
8.2.1 Switching Loss . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.2.2 Conduction Voltage and Loss . . . . . . . . . . . . . . . . . . 110
8.2.3 Thermal Resistance . . . . . . . . . . . . . . . . . . . . . . . . . 112
8.3 Thermal Analysis of Different Device Solutions . . . . . . . . . . . 112
8.3.1 Normal Operation . . . . . . . . . . . . . . . . . . . . . . . . . . 113
8.3.2 Low-Voltage-Ride-Through Operation . . . . . . . . . . . . 115
8.3.3 Wind Gust Operation . . . . . . . . . . . . . . . . . . . . . . . . 120
8.3.4 Summary of Thermal Performances
Under Different Operation Modes . . . . . . . . ....... 121
8.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 121
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 122
9 Reliability-Cost Models for the Power Switching Devices

of Wind Power Converters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
9.1 Loss Model with Chip Number Information . . . . . . . . . . . . . . 124
9.2 Thermal Impedance Model with Chip Number
Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
x Contents
9.3 Analytical Solution of Junction Temperature with Chip

Number Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
9.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
10 Electro-Thermal Model of Power Semiconductors Dedicated

for Both Case and Junction Temperature Estimation . . . . . . . . . . 139
10.1 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
11 Reactive Power Influence on the Thermal Cycling of Multi-MW

Wind Power Inverter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
11.1 Effect of Reactive Power in Case of Single Converter . . . . . . . 146
11.2 Effect of Reactive Power in Case of Paralleled Converters . . . . 152
11.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
12 Thermal Loading of Several Multilevel Converter Topologies

for 10 MW Wind Turbines Under Low Voltage
Ride Through. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
12.1 Promising Topologies and Basic Design . . . . . . . . . . . . . . . . 159
12.2 Operation Status Under Balanced LVRT . . . . . . . . . . . . . . . . 161
12.3 Loss Distribution Under Balanced LVRT. . . . . . . . . . . . . . . . 164
12.4 Thermal Distribution Under Balanced LVRT . . . . . . . . . . . . . 166
12.5 Unbalanced LVRT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
12.6 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
13 Another Groups of Thermal Optimized Modulation Methods

of Three-Level Neutral-Point-Clamped Inverter Under
Low Voltage Ride Through . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
13.1 Basic Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
13.2 Neutral Point Potential Control Method . . . . . . . . . . . . . . . . . 183
13.3 Loss and Thermal Performances . . . . . . . . . . . . . . . . . . . . . . 185
13.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
14 Limits of the Power Controllability of Three-Phase

Converter with Unbalanced AC Source . . . . . . . . . . . . . . . . . . . . 189
14.1 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
Abstract
Wind power generation has been steadily growing both for the total installed
capacity and for the individual turbine size. Due to much more significant impacts
on the power grid, the power electronics, which can change the behavior of wind
turbines from an unregulated power source to an active generation unit, are
becoming crucial in the wind turbine system. The objective of this project is to
study the power electronics technology used for next generation wind turbines.
Some emerging challenges as well as potentials like the cost of energy and reli-
ability are addressed.
First, several potential converter topologies and power semiconductor devices
for the future wind power applications are presented in respect to the
advantages/drawbacks. Then the criteria for evaluating the wind power converter
are generally discussed, where the importance of thermal stress in the power
semiconductors is emphasized and a multidisciplinary approach for stress analysis
is introduced. Based on the proposed criteria and tools, the electrical and thermal
behaviors of wind power converters are investigated under both normal and fault
conditions, where the factors of wind speeds, grid codes, converter controls, and
grid conditions are taken into account.
In order to relieve the electrical and thermal stress of the converter in wind
turbine system, some new control methods and concepts are thereby proposed. In
Chap. 4 a thermal control concept which utilizes the reactive power is used to
stabilize the thermal excursion under wind gust. In Chap. 5 a series of special
modulation methods which can achieve better thermal loading of power devices
under grid faults are introduced. Also in Chap. 5 a series of power control strategies
utilizing the zero sequence current are presented to achieve better control perfor-
mance under the unbalanced AC source.
It is concluded that power electronics will play a more important role and
regulate all the generated power in the next generation wind turbine system. In this
case, the stress in the converter components becomes more critical because power
conversion is pushed to multi-MW level with high power density requirement. It
has also been revealed that thermal stress in power semiconductors is closely related
xi
xii Abstract
to many determining factors in wind power applications such as reliability, cost,

power density, etc., therefore it is an important performance for the next generation
wind power converter. It is found that the thermal behaviors of wind power con-
verter could be rather adverse under some required operating conditions. On the
other hand it is also possible to improve the thermal behaviors by many aspects like
smart control, special modulation, advanced modeling, as well as new converter
designs.
Outlines of this Book
The book consists of two parts—the general monograph in Part I and the specially
selected topics in Part II. The monograph is divided into 7 chapters, and 7 special
topics are attached to detail and back up the analysis. The structure of the book is
organized as follows:
Chapter 1 presents the introduction and motivation of the whole work, where the
background, objectives, and structure are addressed.
In Chap. 2 several promising converter topologies for the next generation wind
power converter are first presented and discussed in respect to the
advantages/drawbacks. Afterwards, three potential power semiconductor devices
for wind power application are highlighted and evaluated.
Chapter 3 discusses the criteria for evaluating the next generation wind power
converter. The importance of thermal stress in power semiconductors is emphasized
by relating it with the reliability and cost of converter. Then a multidisciplinary
approach for stress analysis of wind power converter is introduced, in which the
factors of converter design, converter control, wind speed, and grid codes are taken
into account.
Chapter 4 gives the stress analysis of wind power converter under normal
operation based on a 10 MW wind turbine. The junction temperature profiles in the
power semiconductors are presented under both steady-state wind speeds and speed
variations. Then the converter efficiency and thermal distribution modified by grid
codes are investigated. Finally, a thermal control concept which utilizes the reactive
power circuited among paralleled converters is proposed to relieve the thermal
excursion in power devices under wind gust.
Chapter 5 investigates the thermal stress of wind power converter when suffering
grid faults. Comprehensive analysis for electrical and thermal loading of power
semiconductor devices is conducted on the three-level Neutral-Point-Clamped
(3L-NPC) wind power converter undergoing various grid faults. Afterwards a series
of thermal-redistributed modulation methods which can achieve better thermal
loading of power devices under this extreme operation are proposed. Finally, a new
power control strategy which utilizes the zero sequence current is presented to
xiii
xiv Outlines of this Book
achieve better control performance and current loading under the unbalanced AC
source condition.
In Chap. 6 the conclusions and contributions of this work and some potential
proposals for the future research are discussed.
The titles of the 7 special topics are listed as follows:
Chapter 8: “The Impact of Power Switching Devices on the Thermal
Performance of a 10 MW Wind Power NPC Converter”
Chapter 9: “Reliability-Cost Models for the Power Switching Devices of Wind
Power Converters”
Chapter 10: “Electro-Thermal Model of Power Semiconductors Dedicated for
both Case and Junction Temperature Estimation”
Chapter 11: “Reactive Power Influence on the Thermal Cycling of Multi-MW
Wind Power Inverter”
Chapter 12: “Thermal Loading of Several Multilevel Converter Topologies for
10 MW Wind Turbines Under Low Voltage Ride Through”
Chapter 13: “Another Groups of Thermal Optimized Modulation Methods of
Three-Level Neutral-Point-Clamped Inverter Under Low Voltage Ride Through”
Chapter 14: “Limits of the Power Controllability of Three-Phase Converter with
Unbalanced AC Source”
Part I
Monograph
Chapter 1
Introduction
This chapter gives the background, motivation, and organization of this work. The
state-of-the-art for wind power generation, development of power electronic tech-
nology, as well as some emerging challenges for the next generation wind power
converters are presented. Then the objectives and structure of this book are
outlined.
1.1 State-of-the-Art for Wind Power Generation
Wind Turbine System (WTS) is still the most promising renewable energy tech-
nology. It started in the 1980s with a few tens of kW power production per unit,
while nowadays multi-MW wind turbines are being installed. There is a
wide-spread use of wind turbines in the distribution networks and more and more
wind farms start to be connected with the transmission networks [1].
The cumulative wind power capacity from 1996 to 2012 is shown in Fig. 1.1; it
can be seen that the wind power has grown fast to a capacity of 282 GW with
around 45 GW installed only in 2012—this is more than any other renewable
energy sources [2]. In 2011, the global electric power installation was around
208 GW; this number indicates that the wind power is really an important player in
the modern energy supply system. As an extreme example, Denmark has a high
penetration by wind power, and today more than 30 % of the electric power
consumption is covered by wind. This country even has the ambition to achieve
100 % non-fossil-based power supply by 2050 [3].
Regarding to the markets and manufacturers of wind power, China has the
largest market with over 17.6 GW capacity installed in 2011, together with the EU
(9.6 GW) and USA (6.8 GW) sharing around 85 % of the global market. The
Danish company Vestas was still on the top position among the largest manufac-
turers, closely followed by the GE and Goldwind. Figure 1.2 summarizes the
worldwide top suppliers of wind turbines in 2011. It is interesting to see that there
are four Chinese companies in the Top 10 manufacturers with total market share of
26 % [2].
© Springer International Publishing Switzerland 2015 3

K. Ma, Power Electronics for the Next Generation Wind Turbine System,
Research Topics in Wind Energy 5, DOI 10.1007/978-3-319-21248-7_1
4 1 Introduction
Worldwide wind power capacity
(Giga Watts)
Fig. 1.1 Global cumulative installed wind power capacity from 1996 to 2012
Fig. 1.2 Distribution of wind

turbine market share by
manufacturers in 2011 [2]
Besides the quick growth in the total installed capacity, the size of individual
wind turbine is also increasing dramatically in order to reduce the price per gen-
erated kWh. In 2011, the average turbine size delivered to the market is 1.7 MW,
among which the average offshore turbine size achieves 3.6 MW. The growing
trends of emerging turbine size between 1980 and 2018 are shown in Fig. 1.3; it is
noted that the cutting-edge 8 MW wind turbines with diameter of 164 m have
already shown up in 2012 [4]. Right now most manufacturers are developing
products in the range of 4.5–8 MW, and it is expected that more and more large
wind turbines with multi-MW power level, even up to 10 MW, will be present in
the next decade—driven mainly by the considerations to lower down the cost of
energy [5].
1.2 Development of Wind Power Technologies 5
10 MW
D 190 m
7~8 MW
D 164 m
5 MW
D 124 m
2 MW
D 80 m
600 kW
500 kW D 50 m
D 40 m
100 kW
50 kW
D 20 m
D 15 m
1980 1985 1990 1995 2000 2005 2011 2018 (E)

Power Rating : 0% 10% 30% 100%
Electronics Role : Soft starter Rotor Rotor Full generator
resistance power power
Fig. 1.3 Evolution of wind turbine size and the power electronics seen from 1980 to 2018
(Estimated), blue circle indicates the power coverage by power electronics
1.2 Development of Wind Power Technologies
1.2.1 Evolution of Wind Turbine Concepts
The technologies used for Wind Turbine System (WTS) have also changed dra-
matically for the last 30 years with four to five generations emerged [6–8]. Until
now the existed or existing wind turbine configurations can be generally catego-
rized into four concepts [8]. The main differences between these concepts locate on
the types of generator, power electronics, speed controllability, and the way in
which the aerodynamic power is limited.
A. Fixed Speed Wind Turbines (Type A)
As shown in Fig. 1.4, this configuration corresponds to the so called “Danish
concept” that was very popular in 80s. The wind turbine is equipped with
Bypass switch
Squirrel Cage
Induction Generator
Transformer
Grid
Gear
Soft starter
Capacitor
bank
Fig. 1.4 Fixed speed wind turbine with direct grid connection
6 1 Introduction
Bypass switch
Wound Rotor
Induction Generator
Transformer
Grid
Gear
Soft starter
Rotor Capacitor
resistance bank
Fig. 1.5 Partial variable speed wind turbine with variable rotor resistance
asynchronous Squirrel Cage Induction Generator (SCIG), and smoother grid con-
nection can be achieved by incorporating a soft-starter.
The disadvantages of this early concept are as follows: a reactive power com-
pensator (e.g., capacitor bank) is required to compensate the reactive power demand
by the asynchronous generator. Because the rotational speed is fixed without any
controllability, the mechanical parts must be strong enough to withstand adverse
mechanical torque, and the wind speed fluctuations are directly transferred into the
electrical power pulsations which could yield to instable output voltage in case of
week power grid.
B. Partial Variable Speed Wind Turbine with Variable Rotor Resistance (Type B)
As presented in Fig. 1.5, this concept is also known as OptiSlip (VestasTM)
emerged in the mid 1990s [9]. It introduces the variable rotor resistance and thus
limited speed controllability of wind turbines. The Wound Rotor Induction
Generator (WRIG) and corresponding capacitor compensator are typically used,
and the generator is directly connected to the grid by a soft-starter.
A technology improvement of this concept is that the rotational speed of the
wind turbine can be partially adjusted by altering the rotor resistance. This feature
will contribute to the mechanical stress relief and make a more smooth electrical
power output. However, the power loss dissipating constantly in the rotor resistors
is a significant drawback for this concept.
C. Variable Speed Wind Turbine with Partial-Scale Power Converter (Type C)
This concept is the most established solution nowadays and it has been used since
2000s. As shown in Fig. 1.6, a back-to-back power electronics converter is adopted
in conjunction with the Doubly-Fed Induction Generator (DFIG). The stator
windings of DFIG are directly connected to the power grid, while the rotor
windings are connected to the power grid by the power electronics converter with
normally 30 % capacity of the wind turbine [10, 11].
By the use of power electronic converter, the frequency and current in the rotor
can be flexibly regulated and thus the variable speed range can be further extended
to a satisfactory level. Meanwhile the power converter can partially regulate the
1.2 Development of Wind Power Technologies 7
Double-fed Transformer
induction generator Grid
Gear
AC DC
DC AC Filter
1/3 scale power converter
Fig. 1.6 Variable speed wind turbine with partial-scale power converter
output power of the generator, improving the power quality and providing limited
grid support. The smaller converter capacity makes this concept attractive from a
cost point of view. However, its main drawbacks are the use of slip rings and the
insufficient power controllability in case of grid faults—these disadvantages may
comprise the reliability performance and are hard to satisfy the future grid
requirements as claimed in [12, 13].
D. Variable Speed Wind Turbine with Full-Scale Power Converter (Type D)
Another promising concept that is becoming popular for the newly installed wind
turbines is shown in Fig. 1.7. It introduces a full-scale power converter to inter-
connect the power grid and stator windings of generator; thus all the generated
power by the wind turbine can be regulated. The asynchronous generator, Wound
Rotor Synchronous Generator (WRSG), or Permanent Magnet Synchronous
Generator (PMSG) have been reported to be used in this concept.
The elimination of slip rings, simpler or even eliminated gearbox, full power and
speed controllability as well as better grid support ability are the main advantages of
this solution compared to the DFIG-based concept. Thanks to the use of full-scale
power converter, the voltage level of the power conversion stage can be rather
flexible; in the future the voltage might be high enough to directly connect to the
power grid without the bulky low-frequency-transformer, which is an attractive
feature for the future wind turbine system. However, more stressed and expensive
power electronics are expected and the price for Permanent Magnet (PM) materials
may raise some uncertainties for this concept to be further commercialized—
leading to the development of generator technology with less or even no PM in the
future.
Transformer
AC DC Grid
Filter DC AC Filter
Gear
Asynchronous/
Synchronous Full scale power converter
generator
Fig. 1.7 Variable speed wind turbine with full-scale power converter
8 1 Introduction
1.2.2 Evolution of Power Electronics for Wind Turbines
Figure 1.3 summarizes the role and capacity changes of power electronics in the
wind turbine system. For each of the development stage the corresponding power
electronic components are also highlighted in Figs. 1.4, 1.5, 1.6 and 1.7, respec-
tively. In the 1980s, the power electronics for wind turbines is just a soft-starter;
only some simple thyristors were applied and they do not need to constantly carry
the full power. In the 1990s, the power electronic was mainly used for the rotor
resistance control; more advanced diode bridges and semi-controlled thyristors
were used to constantly alter the rotor resistance. Since 2000, even more advanced
back-to-back power converters were introduced in partial-scale capacity and started
to regulate the power in the rotor; the full-controlled devices like IGBT or GCT
were applied. Nowadays thanks to the more powerful semiconductor devices,
full-scale power converters are adopted to regulate all the generated power by wind
turbines; it is possible to be flexible and to fully control the current/voltage in the
wind turbine system.
As it can be seen, in the last few decades, power electronics gradually become
more and more advanced and bring significant performance improvements for the
wind turbine system—not only to reduce the mechanical stress and increase the
energy capturing efficiency, but also to enable the whole wind turbine system to act
as a controllable generation unit in order to be better integrated with the power grid
[6, 7]. With the continuous technology improvement and going down price of
power semiconductor devices, it can be predicted that the variable speed wind
turbines equipped with full-scale power converter at multi-MW level will be the
dominant solution for the next generation wind turbine system.
1.3 Emerging Challenges for Wind Power Converter
Accompanying with the advancements of technology and power capacity, the

requirements for WTS are also getting tougher. In this section, some emerging
technology challenges for the next generation wind power converters are going to
be addressed and summarized.
1.3.1 More Grid Supports
The fluctuated and unpredictable features of wind energy are un-preferred for the
power grid operation. Most countries have strict requirements for the behaviors of
wind turbines, known as “grid codes” which are updated regularly [14–18].
Basically, the grid codes are always trying to make WTS to act as a conventional
power plant from the point view of electrical network. That means the WTS should
1.3 Emerging Challenges for Wind Power Converter 9
not only be a passive power source simply injecting available power, but also
behave like an active generation unit which can provide frequency/voltage supports
for the power grid. Examples of state-of-the-art grid supporting requirements are
given in the following—they are specified either for individual wind turbine or for
the whole wind farm.
A. Support Under Normal Operation of Grid
According to most grid codes, the individual wind turbines must be able to control
the active power at the Point-of-Common-Coupling (PCC). Normally, the active
power has to be regulated based on the grid frequency, e.g., in Denmark, Ireland,
and Germany, so that the grid frequency can be somehow maintained. As an
example, the characteristic of frequency supports in the Danish grid codes is shown
in Fig. 1.8, where the active power should be decreased when the frequency rises
above 48.7 or 50.15 Hz depending on the power reserving strategy [16].
Similarly, the reactive power delivered by the WTS also has to be regulated in a
certain range. As shown in Fig. 1.9, both the Danish and German grid codes give a
range of the reactive power delivered by WTS against to the active power output.
Also the Transmission System Operator (TSO) will normally specify the reactive
power range of wind turbine system according to the grid voltage levels; this
reactive power control should be realized slowly under the time constant of minutes
[14].
B. Supports Under Grid Faults
Besides the normal operation, the TSOs in different countries have issued strict
supporting requirements for the WTS under grid faults. As shown in Fig. 1.10
[14, 15], in which the boundaries with various grid voltage dipping amplitudes as
well as the allowable disturbing time are defined for a wind farm. It is becoming a
need that the WTS should meanwhile provide reactive power (up to 100 % current
capacity) to contribute to the voltage recovery, when the grid voltage sags are
present. Figure 1.11 shows the required amount of reactive current against to the
Fig. 1.8 Frequency control Available power

profiles for the wind turbines
100%
connected to the Danish grid
[16]
75% With full
production
50%
With reduced
production
25%
fg (Hz)
48 49 50 51 52
48.7 49.85 50.15 51.3
10 1 Introduction
Fig. 1.9 Reactive power P/Prated (p.u.)

ranges under different
generating powers for a wind 1.0
farm by German grid codes Underexcited Overexcited
0.8
[18] Boundary Boundary
0.6
0.4
0.2
Q/Prated (p.u.)
-0.3 0.4
Underexcited Overexcited
Fig. 1.10 Voltage profile for

Voltage(%)
low voltage ride-through 100
capability of wind turbines by Germany
90
different countries [14] Spain
Denmark
75
US
25 Keep connected
above the curves
Time (ms)
0
150 500 750 1000 1500
Fig. 1.11 Reactive current Dead band

requirements for a wind farm Iq /Irated
100%
during grid sags by the
German and Danish grid code
[17, 18]
Vg (p.u.)
0
0.5 0.9 1.0
grid voltage amplitude by German [18] and Danish grid codes [17]. This demand is
relatively difficult to be met by some of the old wind turbine concepts like Type A,
type B, or even type C, and other power quality units like STACOM may probably
be introduced to help the wind turbine system to achieve this tough requirement.
The requirements for more grid supports by wind turbines have on one hand
increased the cost per produced kWh, but on the other hand made the wind energy
more suitable to be largely utilized and integrated into the power grid. It can be
predicted that the stricter grid codes in the future will keep challenging the wind
turbine system and pushing forward the power electronics technology.
1.3.2 Higher Reliability
Besides the demands for more grid supports, reliability performance is another
emerging problem for the wind turbine system. The dramatic growth in the total
installation and individual capacity makes the failures of wind turbines harmful or
even unacceptable. The failures of wind turbine system will not only cause stability
problems of the power grid due to the sudden absence of large amount of power,
but also result in very high cost for repairing and maintenance especially for those
large and remotely located wind turbines, which nowadays are becoming popular
and dominant [1].
Unfortunately, former field feedbacks have shown that the larger wind turbines
seem to be more easily having failures, as indicated in [19]. When looking at the
failure rates and down time distribution in individual wind turbine system, as shown
in Fig. 1.12, it is obvious that the control and power electronic parts tend to have
higher failure rate than the other subsystems with a factor of 2–4 [20]. It is noted
that although the generator and gearbox have the largest downtime (i.e., time
needed for repair), their failure rates are lower than the electrical and control parts.
As a result, further understanding and improving the reliability of the power
electronics converter will be crucial needs for the future wind turbine systems,
especially for those large ones at multi-MW level.
Turbines Gearbox Generator Electrical Control

(days) failure rate
Down time Annual
1/2
Hydraulic Blades
1/4
2
4
6
Fig. 1.12 Failure rate and down time for different parts in the wind turbine system [20]
12 1 Introduction
Stress Analysis Strength Modeling

. .
Cycles to failure
Mission profile Failure mechanism
. Converter design . Accelerating test
. Load estimation . Field feedback
. ... . ...
10 100
? Tj (K)
Mission profile to component stress. Design tools for Lifetime model

reliability
Monitoring & Control Probability & Statistics

. Lifetime prediction . Six sigma
. Control for reliability . Devices variation
. Efficient maintenance . Production robustness
IGBT
Module
Infrared
camera Thermal . ... . ...
picture
Temperature measurement Reliability engineering
Fig. 1.13 Multi-discipline approaches for more reliable power electronics
The reliability research in power electronics has been carried out for decades and
now is moving from a solely statistical approach that has been proven to be
unsatisfactory in the automotive industry, to a more physical-based approach which
involves not only the statistics but also the investigation and modeling of the root
cause behind the failures [21, 22]. As shown in Fig. 1.13, in order to achieve more
cost-efficient and reliable power electronics, multi-disciplinary approaches are
necessary which involve stress analysis, strength modeling, statistical consider-
ations, and also the online monitoring/control of the converter system. It is noted
that a series of reliability design tools may be necessary to transfer, organize, and
combine the relevant analysis to more insight reliability performances.
The stress analysis may focus on complete mission profile definition, converter
design, stress estimation, and measurement. This group of discipline will target to
accurately establish the converter’s loading profile which can trigger the failure
mechanisms of critical components, like the thermal cycling in power devices [23],
voltage increase in the DC bus [24], vibration, and humidity [25, 26].
The strength modeling may involve the identification, modeling, and acceler-
ating test of failure mechanisms, e.g., the bond wire lift-off and soldering crack
inside the power devices [23]. The goal of this group discipline is to seek corre-
lations between the established/measured stresses and the quantified
fatigues/failures of the critical components.
The monitoring and control approach may relate to the lifetime monitoring [27,
28], stress-relieved controls [29, 30], and intelligent maintenance. This group of
discipline will target to monitor and control the converter lifetime during operation.
For example, the Collector-Emitter voltage VCE of an IGBT, which is subjected to
an accelerated test, experiences a sudden increase just before the IGBT failure [27]
and can be used for predictive maintenance in the wind turbine system.
The probability and statistics may add the statistical distribution and correlation
to the acquired stress, strength, and component configuration. This group of dis-
cipline will target to enhance the robustness of designed converter and take into
account the severe usage, six sigma strategies, and quality variations of components
in order to give estimation about failure rate in the next 10 years of the product.
It can be seen that the higher reliability requirements will enable many new
possibilities and methods to analyze, evaluate, and improve the performances of
power electronics, which have not yet been well addressed especially in the wind
turbine system.
1.3.3 Special Cost Considerations
The cost issues are the most important considerations which will determine the
feasibility of certain energy technologies to be widely utilized in the future. As a
typical example, the competitive cost advantage is the main reason why the wind
power showed much more significant growth in the last few decades compared to
the other renewable energy sources. In order to quantify and compare the cost for
different energy technologies, Levelized Cost of Energy (LCOE) index is generally
used [31]. LCOE represents the price at which the electricity is generated from a
specific energy source over the whole lifetime of the project. It is an economic
assessment of the cost of the energy-generating system including all the costs like
initial investment, operations and maintenance, cost of fuel, and capital cost. LCOE
be defined in a single formula as
CCap þ CO&M
LCOE ¼ ð1:1Þ
EAnnual
where the Ccap represents the capital cost of the whole generation system, CO&M
represents the cost for operation and maintenance during the operating lifetime, and
Eannual represents the annual energy production. According to (1.1), it is clear that in
order to reduce the energy cost, one effective way is to reduce the capital cost and
cost for operation and maintenance; the other effective way is to extend the annual
energy production or lifetime of the generation system. Figure 1.14 shows an
example of estimated LCOE in the U.S. for several promising renewable energy
technologies in 2018 [32]; it can be seen that the cost distribution of different
technologies varies a lot and the wind power (onshore) still shows cost advantages
against to the other renewable energy sources. It can be also concluded that the
capital cost may still be dominant for most of the renewable energy technologies in
the next decade.
As more power electronics are introduced in the energy system to improve the
performances of power generation, the cost for the power electronics is becoming
more important and it depends a lot on the specific applications. In the wind turbine
system, there are also some special cost considerations which impose challenges for
the design and the selection of power electronics system.
For instance, the needs for higher power capacity and full-scale power con-
version in the wind power application will increase the cost for power semicon-
ductors, passive components, and corresponding cooling management—thereby
more efficient circuits and devices are required. Moreover, due to the limited space
14 1 Introduction
Estimated Cost of Energy in 2018

($ / MWh)
Fig. 1.14 Estimated levelized cost of energy (LCOE) for several renewable energy technologies
in 2018. (Source Energy Information Administration, Annual Energy Outlook 2013, U.S.) [32]
in the nacelle or tower, a higher power density will be emphasized and this may
lead to extra cost for insulation materials and compact design. Also the weight of
the whole converter system should be limited as much as possible because it has to
be placed in the nacelle or tower of the wind turbines. The long cable connections
between nacelle and tower-base range from dozens to hundreds of meters and may
require higher voltage level in the power conversion stage in order to reduce the
cable losses. Furthermore remote locations of the wind turbines may increase the
cost for installation and maintenance, which also demands high reliability and
modularity/redundancy design of the converter system.
It can be expected that these special cost considerations in the wind power
application will push the technology limits of power electronics and result in sig-
nificantly different approaches for analysis and design of converter compared to the
other converter applications.
1.3.4 Formulation of Overall Requirements
As it can be seen, there are many emerging challenges and design considerations for
the next generation wind power converters; these requirements can be generally
categorized into the following three groups as also shown in Fig. 1.15:
For the generator side, the current flowing in the generator rotor or stator should
be controlled to adjust torque and as a consequence the rotating speed of wind
turbine. This will contribute to the active power balance not only in normal oper-
ation when extracting the maximum power from the wind turbine, but also in case
of grid faults when the generated power needs to be quickly reduced. Moreover, the
converter should have the ability to handle variable fundamental frequency and
voltage amplitude of the generator outputs.
P P
Q Q
Generator side Wind Power Grid side

Converter
System
1. Controllable I 1. Energy balance/storage 1. Fast/long P response

2. Variable freq & U 2. High power density 2. Controllable/large Q
3. Cost effective 3. freq & U stabilization
4. Reliable 4. Low Voltage Ride Through
Fig. 1.15 Demands for the next generation wind power conversion
For the grid side, the converter must comply with the power grid requirements
regardless of the wind speed. This means it should have the ability to control the
inductive/capacitive reactive power Q delivered into the power grid, and meanwhile
perform a fast active power P response. The fundamental frequency as well as
voltage amplitude on the grid side should be maintained almost fixed under the
normal operation, and the Total Harmonic Distortion (THD) of the current must be
restrained at a low level.
Inherently, the converter system needs to be cost effective and to be easy for
maintenance, leading to design considerations for high power density, reliability,
and modularity. Furthermore, the power converter may need the ability to store
some energy, and boost up the voltage level from the generator side to the grid side.
The demands for wind power converter system on one hand reflect the signifi-
cant penetration of wind energy into the power grid. On the other hand, in order to
satisfy these tough requirements, the existing control and design philosophy for
power electronics could be changed, resulting in new criteria and methods to
evaluate and improve the performances of wind power converter.
1.4 Scopes of the Book
The objective of this project is to study the advanced power electronic technology
for the next generation wind turbine system, which is estimated to be at the power
level of 10 MW. Potential methods and tools for evaluating and improving the
critical performances of the wind power converter will be proposed and established,
where the electrical and thermal stresses of power semiconductors are the main
focus.
16 1 Introduction
References
1. Liserre M, Cardenas R, Molinas M, Rodriguez J (2011) Overview of Multi-MW wind turbines

and wind parks. IEEE Trans Ind Electron 58(4):1081–1095
2. REN21—Renewables 2012 Global Status Report. http://www.ren21.net. Accessed June 2012
3. Report of danish commission on climate change policy, green energy—the road to a Danish
energy system without fossil fuels. http://www.klimakommissionen.dk/en-US/. Accessed Sept
2010
4. Website of vestas wind power, wind turbines overview. http://www.vestas.com/. Accessed
April 2011
5. Up wind project, design limits and solutions for very large wind turbines (2011)
6. Chen Z, Guerrero JM, Blaabjerg F (2009) A review of the state of the art of power electronics
for wind turbines. IEEE Trans Power Electron 24(8):1859–1875
7. Blaabjerg F, Chen Z, Kjaer SB (2004) Power electronics as efficient interface in dispersed
power generation systems. IEEE Trans Power Electron 19(4):1184–1194
8. Hansen AD, Iov F, Blaabjerg F, Hansen LH (2004) Review of contemporary wind turbine
concepts and their market penetration. J Wind Eng 28(3):247–263
9. Wallace K, Oliver JA (1998) Variable-speed generation controlled by passive elements. Proc
ICEM’ 1998:1554–1559
10. Muller S, Deicke M, De Doncker RW (2002) Doubly fed induction generator systems for wind
turbines. IEEE Ind Appl Mag 8(3):26–33
11. Xiang D, Ran L, Tavner PJ., Yang S (2006) Control of a doubly fed induction generator in a
wind turbine during grid fault ride-through, IEEE Trans Energy Convers 21(3):652–662
12. Teodorescu R, Liserre M, Rodriguez P (2011) Grid converters for photovoltaic and wind
power systems. IEEE/Wiley
13. Blaabjerg F, Teodorescu R, Liserre M, Timbus AV (2006) Overview of control and grid
synchronization for distributed power generation systems. IEEE Trans Ind Electron 53
(5):1398–1409
14. Altin M, Goksu O, Teodorescu R, Rodriguez P, Bak-Jensen B, Helle L. (2010) Overview of
recent grid codes for wind power integration. In: Proceedings of OPTIM’2010, pp. 1152–1160
15. Tsili M (2009) A review of grid code technical requirements for wind farms. IET J Renew
Power Gener 3(3):308–332
16. Energinet—wind turbines connected to grids with voltages below 100 kV (2003)
17. Energinet—technical regulation 3.2.5 for wind power plants with a power output greater than
11 kW (2010)
18. E.ON-Netz—grid code. Requirements for offshore grid connections in the E. ON Netz
network (2008)
19. Faulstich S, Lyding P, Hahn B, Tavner P (2009) Reliability of offshore turbines–identifying
the risk by onshore experience. In: Proceedings of European offshore wind, Stockholm
20. Hahn B, Durstewitz M, Rohrig K (2007) Reliability of wind turbines—experience of 15 years
with 1500 WTs. Wind Energy, Spinger, Berlin
21. Wolfgang E, Amigues L, Seliger N, Lugert G (2005) Building-in reliability into power
electronics systems. The World of Electronic Packaging and System, Integration, pp 246–252
22. Hirschmann D, Tissen D, Schroder S, De Doncker RW (2005) Inverter design for hybrid
electrical vehicles considering mission profiles. IEEE Conf Veh Power and Propul 7–9:1–6
23. Busca C, Teodorescu R, Blaabjerg F, Munk-Nielsen S, Helle L, Abeyasekera T, Rodriguez P
(2011) An overview of the reliability prediction related aspects of high power IGBTs in wind
power applications. Microelectron Reliab 51(9–11):1903–1907
24. Kaminski N, Kopta A (2011) Failure rates of HiPak modules due to cosmic rays, ABB
application note 5SYA 2042
25. Wolfgang E (2007) Examples for failures in power electronics systems, presented at ECPE
tutorial on reliability of power electronic systems. Nuremberg, Germany
References 17
26. Yang S, Bryant AT, Mawby PA, Xiang D, Ran L, Tavner P (2011) An industry-based survey
of reliability in power electronic converters. IEEE Trans Ind Appl 47(3):1441–1451
27. Yang S, Xiang D, Bryant A, Mawby P, Ran L, Tavner P (2010) Condition monitoring for
device reliability in power electronic converters: a review. IEEE Trans Power Electron 25
(11):2734–2752
28. Due J, Munk-Nielsen S, Nielsen R (2011) Lifetime investigation of high power IGBT
modules. In: Proceedings of EPE’2011—Birmingham
29. Ma K, Blaabjerg F (2012) Thermal optimized modulation method of three-level NPC inverter
for 10 MW wind turbines under low voltage ride through. IET J Power Electron 5(6):920–927
30. Ma K, Blaabjerg F, Liserre M (2012) Reactive power control methods for improved reliability
of wind power inverters under wind speed variations. Proc ECCE’ 2012:3105–3122
31. Wikipedia cost of electricity by source. http://en.wikipedia.org/wiki/Cost_of_electricity_by_
source. Accessed April 2013
32. Report of the US Energy Information Administration (EIA) of the U.S. Department of Energy
(DOE). Levelized cost of new generation resources in the annual energy outlook 2013,
Released in spring (2013)
Chapter 2
Promising Topologies and Power Devices
for Wind Power Converter
In this chapter, several promising converter topologies for the next generation wind
power converter are first presented and discussed in respect to their
advantages/drawbacks. Later, three potential power semiconductor devices for wind
power application are highlighted and basically evaluated.
2.1 Promising Converter Topologies
In the past, there has not been much diversity for the converter topology used in
wind power applications. Pulse-width-modulation voltage source converters with
Two-Level voltage output (2L-PWM-VSC) or simpler circuits were widely used in
the low voltage level and they are capable to satisfy most of the requirements.
Nevertheless, because of the significantly increased demand for the power capacity,
cost-effectiveness, reliability, and controllability, the performance of single
2L-PWM-VSC converter seems to be not enough for the future wind turbine sys-
tem. Consequently, a number of more-powerful and advanced power converter
solutions for the next generation wind turbines are proposed. In this section, some
of the promising configurations either in the academics or in the industry are going
to be reviewed and discussed.
2.1.1 Traditional Two-Level Converters
2L-PWM-VSC is the most frequently used topology so far in wind turbine systems.
The knowledge available in this converter is extensive and well established. In the
wind turbine system, the 2L-PWM-VSC can be used in different configurations
which are introduced as follows:
A. Two-Level UNI-Directional Voltage Source Power Converter (2L-UNI)
It is becoming popular to use permanent magnet synchronous generator (PMSG) in
the WTS. Because there is no reactive power required in such generators and active

20 2 Promising Topologies and Power Devices for Wind Power Converter
power flows uni-directionally from the generator to the power grid. Only a simple
diode rectifier can be applied on the generator side to achieve cost-efficient solution,
as shown in Fig. 2.1. However, the diode rectifier even with multiphase or
12-pulses introduces low frequency pulsations which can trigger the shaft reso-
nance in a worst case [1, 2].
B. Two-Level Back-To-Back Voltage Source Power Converter (2L-BTB)
It is very popular to configure two 2L-PWM-VSCs as a back-to-back structure
(2L-BTB) in the wind power system, as shown in Fig. 2.2. A technical advantage of
the 2L-BTB solution is the full power controllability (4 quadrant operation), with
relatively simpler structure and fewer components. The 2L-BTB is the
state-of-the-art solution for DFIG-based wind turbine concept [2, 3]. Several
manufacturers also use this topology for the wind turbine concept, with full-scale
converter and squirrel-cage induction generator.
However, the 2L-PWM-VSC topology may suffer from larger switching losses
and lower efficiency when the power conversion is pushed to megawatts
(MW) level, and the available power semiconductor devices may need to be con-
nected in parallel or in series, to obtain enough current and voltage ratings. Another
problem of the 2L-PWM-VSC is the two-level voltage outputs, which introduce
large dv/dt stresses to the generator and transformer windings, requiring bulky
output filters to limit the voltage gradient and reduce the harmonics level, especially
Transformer
Filter Filter
Fig. 2.1 Two-level unidirectional voltage source converter for wind turbine (2L-UNI)
Transformer
Filter Filter
2L-VSC 2L-VSC
Fig. 2.2 Two-level back-to-back voltage source converter for wind turbine (2L-BTB)
2.1 Promising Converter Topologies 21
in the case of long cable connections. Consequently, it is becoming more difficult

for the 2L-BTB solution to achieve acceptable performance in the WTS.
2.1.2 Multilevel Converters
With the abilities of more voltage levels and larger power handling capacities,
multilevel converters are becoming promising in the future wind turbine system [4].
Generally, the multilevel topologies can be classified into three types [5, 6]:
neutral-point diode clamped, flying capacitor clamped, and cascaded converter
cells. In order to realize cost-effective design, it is ideal for multilevel converters to
be used in the environment with higher power and higher voltage conversion [7],
which are preferred features for the next generation wind power converters.
A. Three-Level Neutral Point Diode Clamped Back-to-Back Topology (3L-NPC
BTB)
Three-level neutral point diode clamped converter is one of the most commer-
cialized multilevel topologies on the market. Similar to the 2L-BTB, it is usually
configured as a back-to-back structure in the wind turbine system, as shown in
Fig. 2.3, which is called 3L-NPC BTB for convenience.
3L-NPC BTB can achieve one more voltage level output and less dv/dt stress
compared to the 2L-BTB, thereby, the filter size can be smaller under the same
power and voltage. The potential fluctuation of the mid-point in the DC-bus, used to
be a drawback of the 3L-NPC BTB. This problem has been extensively researched
and is considered improved, by the controlling of redundant switching states or by
introducing extra circuits [8]. However, it is found that the loss distribution is
unequal among the power devices in a switching arm [9, 10], and this problem
might lead to de-rated power capacity when it is practically used.
Transformer
Filter Filter
3L-NPC 3L-NPC
Fig. 2.3 Three-level neutral-point-clamped back-to-back converter for wind turbine (3L-NPC
BTB)
Transformer
open windings
Filter Filter
3L-HB 3L-HB
Fig. 2.4 Three-level H-bridge back-to-back converter for wind turbine (3L-HB BTB)
B. Three-Level H-Bridge Back-to-Back Topology (3L-HB BTB)

The 3L-HB BTB solution is composed of two 3-phase H-bridge converters con-
figured in a back-to-back structure, as shown in Fig. 2.4. It achieves the similar
output performance like the 3L-NPC BTB, but the unequal loss distribution and
clamping diodes can be avoided. Thereby, more efficient and equal loading of the
power switching devices as well as higher designed capacity can be obtained [11].
Moreover, as only half of the DC bus voltage is needed compared to the 3L-NPC
BTB, there are less series connection of capacitors and no mid-point in the DC bus.
However, the 3L-HB BTB solution needs an open winding structure both in the
generator and transformer in order to achieve isolation among each phase. This
feature has both advantages and disadvantages: On one hand, a potential fault
tolerant ability is obtained if one or even two phases of the generator are out of
operation. On the other hand, extra cost, weight, loss, and inductance/capacitance in
the cables can be the major drawbacks, especially in the wind power application.
C. Five-Level H-Bridge Back-to-Back Topology (5L-HB BTB)
The 5L-HB BTB converter is composed of two 3-phase H-bridge converters
making use of 3L-NPC switching arms, as shown in Fig. 2.5. It is an extension of
3L-HB BTB solution, and shares the same requirements for the open-winding
generator and transformer. 5L-HB BTB can achieve five-voltage-level output, and
doubled voltage amplitude compared to the 3L-HB BTB solution with the same
Transformer
(open windings)
Filter Filter
5L-HB 5L-HB
Fig. 2.5 Five-level H-bridge back-to-back converter for wind turbine (5L-HB BTB)
Table 2.1 Comparison of three multilevel power converters for wind turbines
3L-NPC 3L-HB 5L-HB
IGBT numbers 24 24 48
Diode 36 24 72
numbers1
Switch current Iph Iph Iph
Switch voltage 0.5Vdc Vdc 0.5Vdc
Max. output 0.5Vdc Vdc Vdc
voltage2
Output-Switch 1 1 2
voltage ratio
Voltage 0.84 1.15 0.73
WTHD3 (%)
Output Standard Open winding Open winding
connection
Fault-tolerant No Yes Yes
ability4
Advantages Matured Less DC link More output voltage levels,
technology capacitors, equal higher voltage utilization of
loss distribution device5
Disadvantages Unequal loss Zero-sequence Zero-sequence current path,
distribution, DC current path, more more cables and devices
bus midpoint cables
Notes
1. Include both freewheeling diodes and clamping diodes
2. Theoretical maximum amplitude of output phase voltage
3. Grid inverter, fs /fo = 21, M = 1, until 80th harmonics, line-to-line voltage
4. If one or two phase of generator side converter fails, still keep working
5. Larger output-switch voltage ratio (higher output voltage using the same voltage rating devices)
power devices. These features enable smaller output filter and less current rating in
the components as well as in the cables [12]. However, the 5L-HB BTB converter
introduces more switching devices, unequal loss distribution, and larger DC link
capacitors.
The comparisons between the three multilevel topologies used in the wind
turbine system are shown in Table 2.1, in which the number of power semicon-
ductor devices, the output performance, fault tolerant ability, as well as major
advantages and disadvantages are focused.
2.1.3 Multi-cell Converters
Besides the solution of multilevel converters, some configurations with multi-cell

converters, i.e., parallel/series connection of converter cells, are widely adopted by
the industry in order to handle the fast growing power of wind turbines.
(a) (b)
AC DC AC DC
DC AC DC AC
2L-VSC 2L-VSC Transformer Generator 2L-VSC 2L-VSC Transformer
...
Multi winding
...
...
...
...
...
generator
AC DC AC DC
DC AC DC AC
2L-VSC 2L-VSC 2L-VSC 2L-VSC
Fig. 2.6 Multi-cell converter with paralleled converter cells (MC-PCC). a Variant 1. b Variant 2
A. Multi-cell Converter with Paralleled Converter Cells (MC-PCC)

Figure 2.6a shows a multi-cell solution adopted by Gamesa in their 4.5 MW wind
turbines [13], which have several 2L-VSCs paralleled, both on the generator side
and on the grid side. Siemens also introduce the similar configuration in their
best-selling multi-MW wind turbines, as indicated in Fig. 2.6b [14]. The standard
and proven low voltage converter cells as well as redundant and modular charac-
teristics are the main advantages. This converter configuration is the state-of-the-art
industrial solution for the wind turbines higher than 3 MW.
B. Cascaded H-Bridge Converter with Medium-Frequency-Transformers (CHB-
MFT)
This configuration shares the similar idea with the next generation traction con-
verters [15, 16], and is also proposed in the European UNIFLEX-PM Project [17],
as shown in Fig. 2.7. It is based on a structure of back-to-back Cascaded H-bridge
converters, with galvanic insulated DC/DC converters as interface. The DC/DC
converters have medium frequency transformer (MFT) operating at several kHz to
dozens of kHz, thereby the transformer size can be significantly reduced. Moreover,
because of the cascaded structure, it can be directly connected to the distribution
networks (10−20 kV), with high output quality, redundancy, and filter-less design.
This solution would become more attractive if it can be placed in the nacelle of
Fig. 2.7 Cascade H-bridge

converter with Medium
Frequency Transformer for MFT
AC DC AC DC
wind turbine (CHB-MFT) Cell 1
DC AC DC AC
...
...
...
AC DC MFT AC DC
Cell N
DC AC DC AC
wind turbines because the bulky low-frequency transformer can be replaced by the
more compact and flexibly-configured power semiconductor devices.
C. Modular Multilevel Converter (MMC)
This configuration shares the similar idea with some of the emerging converters
used for High Voltage Direct Current (HVDC) Transmission [18, 19], as shown in
Fig. 2.8. It is also based on a back-to-back structure with cascaded converter cells of
2L-VSC. One advantage of this configuration is the easily scalable voltage/power
capability; therefore, it can achieve high power conversion at dozens of kV with
good modularity and redundancy. The output filter can also be eliminated because
of significantly increased voltage levels. However, the useable voltage rating in the
wind turbine system may be greatly limited by the insulation materials of the
generator. Moreover, the low fundamental frequency of the generator outputs
(which is the normal case for MW synchronous generator) may introduce large DC
voltage fluctuation in the converter cells of generator side, and thereby results in
bulky capacitors in the converter system, being un-preferable for the compact
design.
It can be seen that, generally the multi-cell converters have modular and
redundant features which can contribute to higher fault-tolerant performance. But,
on the other hand, all the three configurations have significantly increased com-
ponents count, which could compromise the system reliability and increase the cost.
Fig. 2.8 Modular multilevel

converters for wind turbine
(MMC)
AC DC
DC AC
...
...
AC DC
DC AC
AC DC
DC AC
...
...
AC DC
DC AC
The overall merits and defects of these multi-cell converters used in the wind power
application still need to be further evaluated.
More potential power electronic-topologies and technologies used in the wind
power application can be found in [20].
2.2 Potential Power Semiconductor Devices
As reported in the industry and academics, the potential power semiconductor

technologies in the wind power application are among the module packaging
Insulated Gate Bipolar Transistor (IGBT), press-pack packaging IGBT and the
press-pack packaging Integrated Gate Commutated Thyristor (IGCT). The three
power semiconductor devices have quite different characteristics and are generally
compared in Table 2.2. The module packaging technology of IGBT has a longer
track record of applications and fewer mounting regulations. However, because of
the soldering and bond-wire connection of the internal chips, module packaging
devices may suffer from larger thermal resistance, lower power density, and higher
failure rates. The press-pack packaging technology improves the connection of
chips by direct press-pack contacting, which leads to improved reliability (yet to be
scientifically proved but known from industrial experience), higher power density
Table 2.2 Dominant power semiconductor devices for wind power application [23–28]
IGBT module IGBT IGCT
Press-pack Press-pack
Power density Low High High
Reliability Moderate High High
Cost Moderate High High
Failure mode Open circuit Short circuit Short circuit
Easy maintenance + − −
Insulation of heat + − −
sink
Snubber − − +
requirement
Thermal resistance Large Small Small
Switching loss Low Moderate Moderate
Conduction loss Moderate Moderate Moderate
Gate driver Small Small Large
Major Infineon, Semikron, Mitsubishi, Westcode, ABB
manufacturers ABB, Fuji ABB
Medium voltage 3.3/4.5/6.5 kV 2.5/4.5 kV 4.5/6.5 kV
ratings
Max. current 1.5 kV /1.2 kA /750 A 2.3/2.4 kA 3.6/3.8 kA
ratings
2.2 Potential Power Semiconductor Devices 27
(easier stacking for series connection), and better cooling capability with the dis-
advantage of higher cost. Press-pack IGCTs were first introduced into the
medium-voltage motor drives in the 1990s and has already become state-of-the-art
technology in the applications of oil, gas, HVDC, power quality, etc. However,
IGCTs have not yet been mass adopted in the wind turbine system [21, 22].
In the attached special topic 1: “The Impact of Power Switching Devices on the
Thermal Performance of a 10 MW Wind Power NPC Converter,” more loss and
thermal performances of the three device solutions applied in a 10 MW wind power
converter are presented. It is noted that, because of the rating limits and paralleled
devices, in this case study the current flowing in each individual IGBT device is
only half or 1/3 of that flowing in the IGCT, therefore, the loss and thermal level
presented in the IGBT devices could be underestimated if considering the same
current load in each individual device.
2.3 Summary
It is becoming more difficult for the traditional 2L-BTB converter to achieve

acceptable performance in the wind power application. With higher power and
voltage handling abilities, some multilevel and multi-cell converters are become
promising for the next generation wind turbines. However, the tradeoffs among
performance and cost, as well as the uncertain trends of price and technology, make
the suitable configurations for the next generation wind power converter hard to be
finally concluded.
Regarding the power semiconductor devices, press-packing IGCT and IGBT
shows significant improvement in respect to the thermal resistance compared to the
module packaging devices. Due to the current rating limits on the market, paralleled
connection of devices may be needed in order to achieve the required amount of
power for next generation wind turbines—this will significantly modify the
loss/thermal behaviors of individual devices as well as the overall power density,
also the cost is another important issue to be investigated when choosing the proper
device solutions for the future wind power converter.
Related special topic:
Chapter 8.
References
1. Oliveira DS, Reis MM, Silva C, Colado LB, Antunes F, Soares BL (2010) A three-phase
high-frequency semicontrolled rectifier for PM WECS. IEEE Trans Power Electron 25
(3):677–685
2. Wu B, Lang Y, Zargari N, Kouro S (2011) Power conversion and control of wind energy
systems. Wiley, Hoboken
3. 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. Electr Power Appl
143(3):231–241
4. Carrasco JM, Franquelo LG, Bialasiewicz JT, Galvan E, Portillo R, Prats MM, Leon JI,
Moreno-Alfonso N (2006) Power-electronic systems for the grid integration of renewable
energy sources: a survey. IEEE Trans Ind Electron 53:1002–1016
5. Rodriguez J, Bernet S, Wu B, Pontt JO, Kouro S (2007) Multilevel voltage-source-converter
topologies for industrial medium-voltage drives. IEEE Trans Ind Electron 54(6):2930–2945
6. Krug D, Bernet S, Fazel SS, Jalili K, Malinowski M (2007) Comparison of 2.3-kV
medium-voltage multilevel converters for industrial medium-voltage drives. IEEE Trans Ind
Electron 54(6):2979–2992
7. Kouro S, Malinowski M, Gopakumar K, Pou J, Franquelo LG, Wu B, Rodriguez J, Perez MA,
Leon JI (2010) Recent advances and industrial applications of multilevel converters. IEEE
Trans Power Electron 57(8):2553–2580
8. Rodriguez J, Bernet S, Steimer PK, Lizama IE (2010) A survey on neutral-point-clamped
inverters. IEEE Trans Ind Electron 57(7):2219–2230
9. Teichmann R, Bernet S (2005) A comparison of three-level converters versus two-level
converters for low-voltage drives, traction, and utility applications. IEEE Trans Ind Appl 41
(3):855–865
10. Bruckner T, Bernet S, Guldner H (2005) The active NPC converter and its loss-balancing
control. IEEE Trans Ind Electron 52(3):855–868
11. Senturk OS, Helle L, Munk-Nielsen S, Rodriguez P, Teodorescu R (2009) Medium voltage
three-level converters for the grid connection of a multi-MW wind turbine. In: Proceedings of
EPE’2009, pp 1–8
12. Hosoda H, Peak S (2010) Multi-level converters for large capacity motor drive. In:
Proceedings of IPEC’10, pp 516–522
13. Andresen B, Birk J (2007) A high power density converter system for the Gamesa G10x
4.5 MW wind turbine. In: Proceedings of EPE’2007, pp 1–7
14. Jones R, Waite P (2011) Optimised power converter for multi-MW direct drive permanent
magnet wind turbines. In: Proceedings of EPE’2011, pp 1–10
15. Engel B, Victor M, Bachmann G, Falk A (2003) 15 kV/16.7 Hz energy supply system with
medium frequency transformer and 6.5 kV IGBTs in resonant operation. In: Proceedings of
EPE’2003, Toulouse, 2–4 Sep 2003
16. Inoue S, Akagi H (2007) A bidirectional isolated DC–DC converter as a core circuit of the
next-generation medium-voltage power conversion system. IEEE Trans Power Electron 22
(2):535–542
17. Iov F, Blaabjerg F, Clare J, Wheeler O, Rufer A, Hyde A (2009) UNIFLEX-PM-a
key-enabling technology for future European electricity networks. EPE J 19(4):6–16
18. Davies M, Dommaschk M, Dorn J, Lang J, Retzmann, Soerangr D (2008) HVDC PLUS—
basics and principles of operation. Siemens Technical articles
19. Lesnicar A, Marquardt R (2003) An innovative modular multilevel converter topology suitable
for a wide power range. In: Proceedings of the IEEE Bologna powertech conference, pp 1–6
20. Infineon application note: thermal resistance theory and practice, January 2000
21. Jakob R, Keller C, Gollentz B (2007) 3-Level high power converter with press pack IGBT. In:
Proceedings of EPE’2007, pp 2–5
22. Alvarez R, Filsecker F, Bernet St (2011) Comparison of press-pack IGBT at hard switching
and clamp operation for medium voltage converters. In: Proceedings of EPE’2011, pp 1–10
23. Faulstich A, Stinke JK, Wittwer F (2005) Medium voltage converter for permanent magnet
wind power generators up to 5 MW. In: Proceedings of EPE 2005, pp 1–9
24. Ma K, Blaabjerg F, Xu D (2011) Power devices loading in multilevel converters for 10 MW
wind turbines. In: Proceedings of ISIE 2011, pp 340–346
References 29
25. Ma K, Blaabjerg F (2011) Multilevel converters for 10 MW wind turbines. In: Proceedings of
Epe’2011, Birmingham pp 1–10
26. Website of ABB Semiconductors. http://www.abb.com/semiconductors
27. Website of Westcode. http://www.westcode.com/igbt.htm
28. Blaabjerg F, Liserre M, Ma K (2012) Power electronics converters for wind turbine systems.
IEEE Trans Ind Appl 48(2):708–719
Chapter 3
Criteria and Tools for Evaluating Wind
Power Converter
This chapter discusses the criteria for evaluating the next generation wind power
converter. The importance of thermal stress in the power semiconductors is
emphasized by relating it with the reliability and cost performances. Then a mul-
tidisciplinary approach for the stress analysis of wind turbine system is introduced,
where the factors of converter design, converter control, wind speed, and grid codes
are all taken into account.
3.1 Importance of Thermal Stress in Wind Power

Converter
In the past, the power electronics are not so important in the wind power application
and they just need to carry little or partial power produced by the wind turbines. As
a result when designing the converter system, the losses generated by the power
electronic components are not so significant and the voltage/current ratings were
mainly focused.
However, the power electronics nowadays are used to handle full generated
power from the wind turbines even up to 8 MW; the loading and loss of power
electronic components become much more severe. Considering the limited space in
the nacelle or tower of wind power application, it is more and more difficult to
achieve satisfactory cooling environment for the internal parts, resulting in adverse
thermal stress of the power electronic devices. Consequently, not only the power
handling ability but also the loss/thermal behaviors have to be carefully taken into
account when designing the next generation wind power converter.
As one of the most important loading indicators, the thermal stress is found to be
closely related to the reliability, cost, and power density of the converter system, as
will be illustrated in the following.

32 3 Criteria and Tools for Evaluating Wind Power Converter
3.1.1 Thermal Stress Versus Reliability
According to the statistics carried by [1], the proportion of various stresses that
contribute to the failures of power electronic components is shown in Fig. 3.1. It
can be seen that there are 55 % of the component failures which are caused by the
temperature or thermal stresses. It is noted that the stress distribution in Fig. 3.1
could be varied depending on the application; however, as reported in [2–5] it is
generally accepted that the thermal loading is an important “trouble maker” for
most of the failure mechanisms in the electronic devices, like capacitor, printed
circuit board, power semiconductors, etc.
Figure 3.2 shows a breakdown of different layers inside a IGBT module [6],
which is chosen here as an example because it has been a popular power semi-
conductor technology used in almost all power converter applications. It was found
that the main driver for the IGBT failures is the thermal cycling of different layers
of materials with mismatching expansion coefficients: The losses generated from
the silicon chips periodically propagate through a couple of materials to the heat
sink, causing temperature excursions on different layers, which then expand and
compress until cracks or disconnections are triggered. It has been widely recognized
that the three dominant failure mechanisms for IGBT modules are the bond wire
lift-off, solder joints cracking under the chip, and solder joints cracking under the
substrate, as also indicated in Fig. 3.2 [7–9]. As for the press-pack IGCT or IGBT,
although the quantified failure models are not established yet, it is found that the
thermal stress is also an important cause of problems for these devices [6].
Nevertheless, it has been proved that the thermal stress and lifetime of IGBT
modules can be mathematically correlated, as summarized in [6, 8], where a series
of lifetime models are introduced. Normally the parameters of these lifetime models
need to be acquired experimentally by accelerating tests, in which the power
devices are running at various thermal loadings and the failures are observed and
recorded. Afterwards, some mathematical functions are used to fit these recoding
points e.g., Coffin-Masson model [8]. Figure 3.3 shows an example of the lifetime
Fig. 3.1 Critical stresses

contribute to the failures of
power electronic components
3.1 Importance of Thermal Stress in Wind Power Converter 33
IGBT module bond wire
Chip Chip Chip

Solder
Copper Substrate
Base solder
Base plate
Thermal Heat sink

grease
Cooling system
Fig. 3.2 Breakdown of different layers of materials in and IGBT module fastened on a heat sink
testing results provided by Semikron, in which several fitting curves are plotted to
represent the relationship between the cycles to failure of a series of IGBT modules
and the applied thermal stresses—in this case the junction temperature excursion
ΔTj and the mean junction temperature Tjm.
In [10], the similar model as Fig. 3.3 is utilized in a wind power converter to
estimate the lifetime of the power devices subjected to a series of wind profiles. The
estimated results are listed in Table 3.1, where two thermal stress levels are applied
Fig. 3.3 Illustration of

industrial standard cycles to
failure versus dTj of IGBT
module by Semikron [20]
Cycles to failure
10E 100
Tj (K)
Table 3.1 Lifetime Failure mechanisms B10 lifetime

estimation results of [10] (year)a
Stress I Stress II
Baseplate solder joints (due to ΔTc) 358 24
IGBT chip solder joints (due to ΔTj_IGBT) 438 22
Wire bonds (due to ΔTj_IGBT) 2633 74
Overall (determined by the shortest one) 358 22
a
B10 lifetime means the number of cycles where 10 % of the
modules of a population fail [9]
and the lifetime in respect to the three failure mechanisms of IGBT module are
indicated respectively. It is obvious that the thermal stress has significant impacts to
the reliability of the power semiconductor devices and their relationship can be
analytically correlated.
3.1.2 Thermal Stress Versus Cost
The thermal stress of power semiconductors is determined by many design factors

like the rating margins, packaging technology, heat sink design, circuit topologies,
etc., which are all related to the cost of the converter. A straight forward example is
that if the power devices are used to convert the same amount of power with larger
power ratings and using a more powerful heat sink, the junction temperature will be
more or less relieved with drawback of the increased cost.
Therefore, it should be possible to correlate the thermal stress and the cost of
converter. As shown in Fig. 3.4, where the symbol in circuit, real construction, and
uncovered picture of a IGBT module are illustrated, it is noted that actually the
IGBT module is composed of many IGBT/Diode silicon chips in parallel internally,
and the number of the paralleled chips will determine the current rating and thereby
the size of IGBT module; therefore, the chip numbers can be an easier quantified
parameter for the cost of the converter.
In the special topic attached Chap. 9, a thermal-cost model which correlates the
junction temperature of IGBT module and the corresponding parallel silicon chip
numbers is established.
In the attached special topic Chap. 9, the thermal-cost model is demonstrated on
a 10-MW 3L-NPC wind power converter. One of the analyzing results is shown in
Fig. 3.5: the vertical axis represents the junction temperature in different power
devices of the 3L-NPC converter, and the horizontal axis represents the corre-
sponding paralleled chip numbers in the IGBT module. The mean junction tem-
perature Tm and the temperature fluctuation ΔTj are both shown, respectively, in
Fig. 3.5 a, b. It can be seen that the thermal stress also has significant impacts to the
rating/cost of the power semiconductor devices and their relationship can be ana-
lytically correlated.
(a) (b) C
C bond wires
IGBT Diode
G chips chips
G
bond wires
E E
(c)
Fig. 3.4 Construction of IGBT module. a Symbol in circuit. b Construction—chips in parallel.

c Picture of an uncovered IGBT module
(a) (b)
Junction temperature fluctuation ΔTj (K)
Mean Junction temperature Tm (°C)
250 40
T1
T1
200 30
Dnpc Dnpc
150 T2
T2 20
100
D1 10
D1
D2
50 D2
0
20 30 40 50 60 70 20 30 40 50 60 70
Chip numbers N Chip numbers N
Fig. 3.5 Thermal loading of each power device in 3L-NPC grid inverter versus chip numbers
(Normal operation, Po = 10 MW, fs = 800 Hz, Vll = 3.3 kVrms). a Mean junction temperature Tm.
b Junction temperature fluctuation ΔTj
Fig. 3.6 Thermal-oriented Thermal stress

design of converter
Design point
Target
thermal level
Current rating needed Chip numbers
With the established thermal-cost relationship, more advanced design and eval-
uation of wind power converter can be realized, as demonstrated in the attached topic
(Chap. 9). For example, it is possible to select the “just right” current rating of the
power device to achieve the target junction temperature, as shown in Fig. 3.6.
Thereby more cost-effective converter design can be achieved especially for the
multi-level converters with unequal loading of the power devices. Also it is possible
to unify various converters under the same thermal-cost domain, as shown in Fig. 3.7,
enabling a more sensible comparison of different converter solutions.
According to the illustrations of thermal-related models as well as their utili-
zation in the wind power converter, it can be revealed that the thermal stress is
closely related to the reliability and cost of the power electronic components.
Moreover, as shown in Figs. 1.9, 1.10 and 1.11, many grid codes have regulations
for the delivered reactive/active power by wind turbines and thereby will also
change the operating condition of converter in respect to the current, voltage,
Fig. 3.7 Thermal-cost profile Thermal stress

comparison of different
converters
Converter 1
Converter 2
Converter 3
Chip numbers
phase-angle, and the loading of power devices. As a result, the thermal stress is an
important performance to be deeply investigated for the next generation wind
power converter.
3.2 Classification and Approach for the Thermal Stress

Analysis
However, correctly transforming the mission profile of wind power converter into
the corresponding thermal stress of the power devices is not an easy task [9, 11]:
For example, many factors influence on the loading of converter which should be
taken into account, like the wind speed variations, grid disturbances, control
strategies, converter design, etc., thus multidisciplinary approaches may be
involved. Also in the case of long-term stress analysis (e.g., 1-year operation which
is necessary for reliability estimation [10]), large amount of loading data at various
time constants from seconds to months may be generated and difficult to handle.
Therefore, it is important first to identify and distinguish the causes of thermal
stresses.
3.2.1 Classification of Thermal Stress in Wind Power

Converter
According to the time intervals, the thermal behaviors of the wind power converter
can be categorized into the three types, as indicated in Fig. 3.8, in which the ranges
of the dominant time constants and the corresponding main causes are defined.
Environmental Mechanical Electrical Grid
year hour second millisec
Long term Medium term Short term

(Ambient change) (Turbine control) (Thermal design)
Fig. 3.8 Thermal stress in a wind turbine system with various time constants
A. Long-Term Thermal Behavior (Days to Years)

This group of thermal behavior is mainly caused by the environmental disturbances,
e.g. the variation of wind speeds or the ambient temperatures in a few months or
years. The corresponding thermal stress of components will change enormously but
slowly in a long term. In order to get an acceptable data size and maintain enough
information for daily-based thermal behaviors, simplified models and larger sam-
pling time may be needed for analysis.
B. Medium-Term Thermal Behavior (Seconds to Minutes)
This group thermal behavior is mainly caused by the wind speed variations and the
mechanical control of wind turbines, for example, the short-term wind gust ranging
from seconds to minutes, or the pitch or speed control of turbines. Therefore, more
complicated models especially for the mechanical parts have to be used in order to
generate enough details of the stress profiles.
C. Short-Term Thermal Behavior (Micro to Mini Seconds)
This group of thermal behavior is mainly caused by the electrical disturbances, for
example, periodical switching/alternating of the load current in the power devices,
or the electrical impacts imposed by grid faults. Therefore, the small and fast
temperature behaviors will normally be observed and detailed models and small
sampling time especially for the electrical parts are necessary.
The three types of thermal behaviors in the power semiconductors devices may
lead to significantly different impacts to the reliability of wind power converter, and
also result in various approaches for stress analysis or improvement in the wind
turbine system.
3.2.2 Methods and Models for Stress Analysis
As mentioned before, the thermal stress of power semiconductors is influenced by

many factors and may involve multidisciplinary approaches to analyze. As shown
in Fig. 3.9, a potential framework for the multidiscipline stress analysis is estab-
lished, where the determining factors that have impacts to the loadings of the
converter are divided into three groups: 1. The factors of mission profile which
includes the wind speed or ambient temperature variations as well as the grid
conditions. 2. The factors of converter design which relates to the used topologies,
power semiconductors, and passive components. 3. The factors of controls which
involve the modulation or power control strategies of converter/turbine. It is noted
that in this analysis framework, a series of multi-domain models are necessary to
translate the stress-determining factors into the corresponding stress profiles of the
converter, as highlighted by the “Stress generation” block in Fig. 3.9.
The multi-domain models used for stress generation are detailed in Fig. 3.10,
where the mechanical, electrical, and thermal behaviors of the wind turbine system
3.2 Classification and Approach for the Thermal Stress Analysis 39
Fig. 3.9 A framework for the multidiscipline stress analysis of wind power converter
Fig. 3.10 The multi-domain models used for stress generation in Fig. 3.9
are characterized, respectively. It can be seen that these models are not independent
but closely related to each other by the parameters indicated in Fig. 3.10.
The wind turbine and the generator models, which are used to translate the wind
speeds into the electrical power flowing into the power converter, are listed in
Sect. 7.1 for simplicity, while the loss and thermal models which are critical for the
stress estimation in power semiconductor devices are detailed as follows:
A. Loss Model
The loss model of the power semiconductors can transfer the electrical loading
subjected to the devices (i.e., current and voltage) into the power losses generated
on the semiconductor chips or junction. The used loss model in this monograph is a
widely accepted method which is detailed in [12–14]. The estimated losses are
based on the datasheet parameters of the selected power switching devices and they
are normally temperature dependent. Therefore in order to get a relative more
accurate loss estimation, the mean junction temperature from the outputs of the
thermal model has to be feedback for iteration, as also illustrated in Fig. 3.10.
However, sometimes for simplicity, approximated estimation or lack of information
from datasheets, the temperature dependency of the loss model is not considered
and the worst junction temperature condition (normally at 125 °C) is assumed.
B. Thermal Model
The thermal model for power semiconductors devices can transfer the generated
power losses on the chips or junction into the temperatures at different locations of
the power devices (e.g., the junction, case, or heat sink). Normally, the thermal
behavior of a certain material is represented by the thermal resistance Rth and
capacitance Cth [15]. The total thermal impedance of a power switching devices
from junction to ambient can be modeled as cascaded thermal RC networks rep-
resenting different material layers as shown in Fig. 3.2.
Based on the connection of thermal capacitance, the thermal RC networks may
have two different forms: the first one is called Foster network which is normally
provided by manufacturers on the datasheets. This thermal model is just a mathe-
matical fitting of the measured external thermal behavior and do not represent any
physical meaning for each RC lump [16–19]. The other one is called Cauer network
which is based on the material modeling of each layer inside power devices and
thereby has physical meaning for each RC lump. However, this type of Cauer
network is complicated to be accurately modeled and the Finite Element Method
simulation may be introduced.
According to [9], not only the junction temperature, but also the case and heat
sink temperatures are important for the lifetime estimation of power semiconduc-
tors. However, it is found that both of the existing thermal networks have their
limits to acquire the appropriate case and heat sink temperatures, as detailed in the
attached special topic: Chap. 10. As a result, a new thermal model which combines
the advantages of these two thermal networks is proposed. By the proposed thermal
model and simulation method, it is possible to estimate not only the junction
temperature but also the case and heat sink temperature in a relative longer
time period for the reliability evaluation. The proposed thermal model is shown in
Fig. 3.11; it can be seen that it contains two paths for the thermal flow. The first one
is used for junction temperature estimation and the second one is used for case and
heat sink temperature estimation.
When applying the proposed thermal model to a designed 1-MW grid side
converter in Chap. 10., the corresponding power loss generated inside individual
3.2 Classification and Approach for the Thermal Stress Analysis 41
Loss Tj TA
Pin
IGBT module
multi-layers Foster Zj-c
IGBT module TC TH TA
Pout Thermal
Grease
Heat sink
equivalent 1 layer Cauer Zj-c Zc-h+Zh-a
Fig. 3.11 Proposed thermal model of power devices for lifetime evaluation
IGBT Pin, power loss output from IGBT base plate Pout, as well as the IGBT
junction temperature Tj and case temperature TC are shown in Fig. 3.12a, in which
the behaviors of the equivalent Cauer thermal model are also indicated as a com-
parison. It can be seen that under the same generated power loss Pin, the proposed
new thermal model can achieve almost the same junction and case temperatures at
steady state as the Cauer thermal model, with less filtered Pout. Figure 3.12b shows
a dynamic change of the converter loading from rated 1 MW to 0.2 MW output at
Fig. 3.12 Loss and temperature responses of the new thermal model (compared to Cauer thermal
model based on a 1 MW 690 Vrms output grid side converter). a Steady state condition.
b Dynamic loss change
the 0.6 s. It can be seen that the new thermal model can achieve faster thermal
response in the Pout, TC, and Tj. The advantages and results of this thermal model
are further explained in the attached special topic Chap. 10.
3.3 Summary
Due to the growing power and limited space in the wind power application, the
thermal loading of the wind power converter is becoming significant especially at
multi-MW level. It has been proven that the thermal stress of power devices has
significant impacts to the reliability and cost of the converter and their relationship
can be analytically correlated. Therefore, thermal stress analysis is crucial important
for the next generation wind power converter system.
The loading of wind power converter is influenced by many factors which may
involve multidisciplinary approaches of analysis under various time constants.
Thereby, a framework with multi-domain models of the wind turbine system is
established for the stress analysis. In this framework, the factors of mission profile,
converter design, and converter controls are taken into account and they can be
translated into the corresponding stress profile in the power semiconductor devices.
Relevant special topics:
Chapter 9.
Chapter 10.
References
1. ZVEL, Handbook for robustness validation of automotive electrical/electronic modules, June

2008
2. Tuchband B, Vichare N, Pecht M (2006) A method for implementing prognostics to legacy
systems. In: Proceedings IMAPS military, aerospace, space and homeland security: packaging
issues and applications, 2006
3. Wolfgang E (2007) Examples for failures in power electronics systems. Paper presented at
ECPE tutorial on reliability of power electronic systems, Nuremberg, Germany, April 2007
4. Yang S, Bryant AT, Mawby PA, Xiang D, Ran L, Tavner P (2011) An industry-based survey
of reliability in power electronic converters. IEEE Trans Ind Appl 47(3):1441– 1451
5. Wolfgang E, Amigues L, Seliger N, Lugert G (2005) Building-in reliability into power
electronics systems. In: The world of electronic packaging and system integration, 2005,
pp 246–252
power applications. Microelectron Reliab 51(9–11):1903–1907
7. Wintrich A, Nicolai U, Reimann T (2011) Semikron application manual, p 128
8. Kovacevic IF, Drofenik U, Kolar JW (2010) New physical model for lifetime estimation of
power modules. In: Proceedings IPEC’10, pp 2106–2114
9. ABB Application Note, Load-cycling capability of HiPak IGBT modules, 2012
References 43
10. Wang H, Ma K, Blaabjerg F (2012) Design for reliability of power electronic systems. In:
Proceedings of IECON’ 2012, pp 33–44
11. Hirschmann D, Tissen D, Schroder S, De Doncker RW (2005) Inverter design for hybrid
electrical vehicles considering mission profiles. In: IEEE conference on vehicle power and
propulsion, pp 1–6, 7–9 Sep 2005
12. Blaabjerg F, Jaeger U, Munk-Nielsen S, Pedersen J (1995) Power losses in PWM-VSI inverter
using NPT or PT IGBT devices. IEEE Trans Power Electron 10(3):358–367
13. User manual of PLECS blockset version 3.1, March 2011
14. Graovac D, Purschel M, IGBT Power losses calculation using the data-sheet parameters,
Infineon Application Note, Jan 2009
15. Marz M, Nance P (2000) Thermal modeling of power electronic system, Infineon Application
Note
16. Infineon Application Note: Thermal resistance theory and practice, Jan 2000
17. Infineon Application Note AN2008-03: Thermal equivalent circuit models, June 2008
18. ABB Application Note 5SYA 2093-00: Thermal design and temperature ratings of IGBT
modules, 2012
20. Semikron Application Manual Power Semiconductors, 2011, p 128
Chapter 4
Thermal Stress of 10-MW Wind Power
Converter Under Normal Operation
This chapter gives the stress analysis of wind power converter under normal
operation based on a 10-MW wind turbine. The junction temperature profiles in the
power semiconductors are first presented under both steady-state and variable wind
speeds. Then the converter efficiency and thermal distribution modified by grid
codes are also investigated. Finally, a thermal control concept which utilizes the
reactive power circuited among paralleled converters is proposed to relieve the
excursion of junction temperature under wind gust.
4.1 Requirements and Conditions Under Normal

Operation
The normal operation is the most dominant status for a wind power converter.
Under this operating mode, the wind turbine system is running between the
designed cut-in and cut-out wind speeds, and the grid voltage is always maintained
at rated value with an amplitude variation below 5 %. The wind power converter is
normally injecting all the available power from wind turbines to the power grid, and
the delivered reactive power should be restrained in a certain range depending on
the active power output as well as countries. An example of German grid codes for
offshore wind farm is shown again in Fig. 4.1, where the boundaries for under-
excited and overexcited reactive powers delivered by the wind turbine system are
both specified [1, 2].
As mentioned before, the three-level neutral-point-clamped (3L-NPC) topology
seems to be a promising candidate for the next generation wind turbine system. This
topology is chosen and basically designed for a 10-MW wind turbine as a case
study in this chapter, as shown in Fig. 4.2, where the major design parameters for
the converter system are summarized in Table 4.1 [3–8].
The converter is simulated under the PLECS Blockset in Simulink environment
[9], and the detailed simulation condition is indicated as follows:
For the electrical simulation, the parameters of a 10-MW PMSG which are the
same as Table 7.1 in Sect. 7.1 are used [10–12], and the generator side filter is not

46 4 Thermal Stress of 10-MW Wind Power Converter …
P/Prated (p.u.)
1.0
Underexcited 0.8 Overexcited
Boundary Boundary
0.6
0.4
0.2
Q/Prated (p.u.)
-0.3 0.4
Fig. 4.1 German grid codes for P/Q range of wind power converter under normal operation
Generator side Grid side

Vdc
T1 D1
Generator
Vp Vg
Dnpc D2
T2
Grid
Lf
Fig. 4.2 Three-level neutral-point-clamped converter used in a WTS
Table 4.1 Parameters of Rated output active power Po 10 MW

three-level neural point
clamped inverter for case DC bus voltage Vdc 5.6 kV DC
a
study Rated primary side voltage Vp 3.3 kV rms
Rated line-to-line grid voltage Vg 20 kV rms
Rated load current Iload 1.75 kA rms
Carrier frequency fc 750 Hz
Filter inductance Lf 1.1 mH (0.25 p.u.)
a
Line-to-line voltage in the primary windings of transformer
taken into account. All of the power switching devices have the commutated
voltage of 2.8 kV in order to utilize the available 4.5 kV high-power IGCT/IGBT
on the market. The most commonly used carrier-based PD-PWM for 3L-NPC
converter and the d/q frame power control method based on PI controller are
applied [13, 14]. The equivalent switching frequency of converter is designed to be
4.1 Requirements and Conditions Under Normal Operation 47
750 Hz for the grid side and 412.5 Hz for the generator side (27.5 Hz fundamental
frequency, direct driven) in order to get an acceptable switching loss in the power
devices at 10-MW power conversion. The grid-side filter design for MW wind
power converter has been comprehensively investigated in [15]; a filter inductance
is considered in the simulation, and it is designed to limit the maximum current
ripple to 25 % of the rated current amplitude. For simplicity and to focus the
analysis on the device stress, the filter capacitance is not taken into account; the
power grid is considered as three 20 kV/50 Hz ideal AC voltage sources; the
resistance in the generator and the cables is not taken into account; DC bus
capacitance is assumed high, and the transformers are assumed ideal.
For the loss and thermal simulation of power semiconductor devices, the
press-pack IGCT 5SHY 40L4511 (4.5 kV/3.6 kA) and recommended diodes 5SDF
10H4503 are chosen [16, 17]. The thermal model for power devices is similar to
Fig. 3.10, and each of the parameters for thermal impedance can be found from the
manufacturer datasheets and Sect. 7.1. It is noted that the temperature of the heat
sink TH is normally much lower and more stable compared to the junction tem-
perature Tj in a properly designed converter system; therefore for simplicity the heat
sink temperature is considered as an adverse and constant value at 60 °C. However,
in the reality, the heat sink temperature may strongly depend on the operation
environment and the design of the heat sink (air or liquid cooled).
4.2 Stress of Converter Imposed by Wind Speeds
The wind speed decides the amount of active power generated by the wind turbines
and as well the loading level of wind power converter. In this section, the electrical
and thermal stress of converter imposed by the wind speeds are going to be
investigated. The relevant approach for the stress analysis is highlighted in Fig. 4.3.
It is noted that in order to establish a proper operating condition for analysis, all of
the three factors (i.e., mission profile, converter design, and converter control) need
to be involved and they also have impacts to the stresses of converter. However,
only the stresses imposed by the mission profile (in this case wind speed) are going
to be focused in this section.
4.2.1 Thermal Stress Under Steady-State Wind Speeds
A. Thermal Lading of 3L-NPC Converter Under Diffident Wind Speeds

Figure 4.4 shows the junction temperature of the given 3L-NPC converter under
rated wind speed with 10 MW output, where the temperature distributions of power
devices in both the generator side and grid-side converters are presented. Figure 4.5
shows the junction temperature profiles of the 3L-NPC converter under different
Converter design
Topology
Power devices
Pass. components
Mission profile Stress generation Control

Wind profile Mechanical Modulation
Ambient temp. Electrical Basic control
Grid codes Thermal High level control
Electrical & Thermal

stress
Fig. 4.3 Stress analysis approach considering the wind speeds
Fig. 4.4 Junction temperature of 3L-NPC converter under rated wind speed (rated 10 MW
output). a Grid side, grid frequency fo = 50 Hz, switching frequency fs = 750 Hz. b Generator side,
generator frequency fg = 27.5 Hz, switching frequency fs = 412.5 Hz
wind speeds. It can be seen that the thermal stress in the converter is quite unequal
not only among different power devices, but also between generator side and grid
side. On the grid side, the clamping diode Dnpc and outer switch T1 are the hottest
devices with much higher junction temperature level compared to other devices—
that means shorter life time to failure according to Fig. 4.12. On the generator side,
the freewheeling diodes D1 become the most stressed device.
The thermal stress profile under various wind speeds can be used as an insight
indicator for the loading level of power semiconductor devices, and the clues about
how to optimize the device ratings and heat sink are thereby located. The thermal
4.2 Stress of Converter Imposed by Wind Speeds 49
(a) (b)
Junction temperature ( )
Dnpc
D1
T1
T2
Wind speed (m/s) Wind speed (m/s)
Fig. 4.5 Junction temperature of most stressed devices in 3L-NPC converter versus wind speeds.
a Grid side, fo = 50 Hz, fs = 750 Hz. b Generator side, fg = 27.5 Hz, fs = 412.5 Hz
(a) (b)
(c)
Fig. 4.6 The distribution of the three types of power devices in different converter solutions with
direct drive train. The red color components represent the “hot” devices, yellow ones represent the
“warm” devices, and blue ones represent the “cold” devices. a 3L-NPC BTB (direct-drive).
b 3L-HB BTB (direct-drive). c 5L-HB BTB (direct-drive)
profile under different wind speeds in Fig. 4.5 can be also used as a lookup table to
estimate the lifetime of power devices subjected to long-term variations of wind
speeds. The detailed analysis and results can be found in [18].
B. Device Utilization of Different Topologies
With the information of thermal stress under rated wind speed, the loading map of
different converter configurations can be obtained. As shown in Fig. 4.6, several
multi-level topologies mentioned in Chap. 2 are indicated based on the same
10-MW wind turbine. According to the loading level, the power devices in different
topologies can be categorized into three types: the “hot” devices which have the
junction temperature above 105 °C, the “warm” devices which have the junction
temperature between 70 and 105 °C, and “cold” devices which have the junction
Fig. 4.7 Summary of device utilization in different converter topologies
temperature below 70 °C. It can be seen from Fig. 4.6 that the “week point” or “hot
spot” of a certain converter topology can be clearly indicated [7].
The numbers of the three types of power devices in different converters are
summarized in Fig. 4.7, where situation with the direct-drive-train is indicated. The
proportion of the three types of power devices will be a useful tool to indicate the
information of device utilization and reliability for a certain converter solution: The
more “cold” devices a converter has, the less sufficient utilizing of the power
devices; some of the IGCTs or diodes could be barely used. The more “warm”
devices a converter has, the more sufficient use of the power devices; this will be
the ideal operating condition when designing a converter. While the more “hot”
devices a converter has, the more over-loading of the power devices, the heat sink
or device capacity may be under-designed. It can be seen that the given 5L-HB,
3L-HB converter topologies in [7] may reduce the numbers of the “hot” devices,
achieving more equal device utilization compared to the 3L-NPC topology.
4.2.2 Thermal Stress Under Wind Speed Variations
During wind speed variations, the mechanical parts, electric-machine, and power
electronics converter all suffer from loading fluctuations. The excursion of the
thermal stress may quickly trigger many failure mechanisms in power semicon-
ductors and result in significantly reduced life time of the converter. Therefore,
different from the stress analysis under steady-state wind speed, the thermal
excursion in power devices is more focused under the wind speed variations.
By the wind speed models presented in [19] and Sect. 7.1, a wind speed vari-
ation is generated in Fig. 4.8a, which fluctuates between 7.5 and 12 m/s within 20 s.
According to the turbine and machine models presented in Sect. 7.1, this wind
4.2 Stress of Converter Imposed by Wind Speeds 51
Fig. 4.8 Wind gust operation of 3L-NPC inverter without thermal control. a Wind speed
variations and current references. b Thermal distribution of power devices
speed variations can be transferred to the corresponding current references for the
grid-side converter, as also indicated in Fig. 4.8. It is noted that the reactive current
Iq delivered by the grid-side converter is normally set to be zero if the amplitude of
grid voltage is within normal range.
The junction temperature of the power devices in the given 3L-NPC grid-side
converter is shown in Fig. 4.8b, which is the thermal response to the predefined
wind speed variation and corresponding current references in Fig. 4.8a. It can be
seen that the clamping diode Dnpc and outer switch Tout suffer from the highest
thermal stress as well as the excursion amplitude; this is the adverse loading con-
dition from the point view of device lifetime as claimed before.
4.3 Stress of Converter Imposed by Grid Codes
Besides the wind speeds, various grid requirements which have regulations for the
active/reactive power delivery will also modify the loading of wind power con-
verter. In this section, the electrical and thermal behaviors of the given 3L-NPC
wind power converter imposed by the grid codes for normal operation are going to
be investigated. The relevant stress generation process is highlighted in Fig. 4.9. It
can be seen that this analysis also focuses on the estimation of electrical and thermal
stresses considering the impacts by mission profile (in this case the grid codes),
although other factors should be also involved in order to establish a proper
operating condition for the converter.
4.3.1 Converter Efficiency Considering Reactive Power

Demands by Grid Codes
As mentioned in Chap. 1, the reactive power ranges delivered by wind power

converter at different active power outputs have to be regulated by grid codes in a
Converter design
Topology
Power devices
Pass. components


profiles
Fig. 4.9 Stress analysis approach considering grid codes
certain range. The German grid codes are used here as an example, and the cor-
responding ranges of power factors are plotted in Fig. 4.10 according to Fig. 4.1.
The enormously modified power factors under different active power outputs may
change the loss distribution of power devices and lead to unexpected efficiency
performances of the grid-side converter.
The efficiency of the given 3L-NPC grid-side converter under different active
power outputs is shown in Fig. 4.11, in which the conditions of the three extreme
power factors in Fig. 4.10 are indicated. It is noted that the loss in the clamping
circuits and filters are not taken into account. As it can be seen, the converter
efficiency may be much lower than expected when considering the grid
Fig. 4.10 Power factor under

Leading
0.5
different reactive power 0.6 Q-max
boundaries in Fig. 4.1 0.7
Power Factor
0.8
0.9
Q0
1
0.9
0.8
0.7
lagging
0.6
0.5 Q+max
0.4
0.3
0 0.2 0.4 0.6 0.8 1
Active Power (p.u.)
4.3 Stress of Converter Imposed by Grid Codes 53
Fig. 4.11 Efficiency of 99.6%

Q0
grid-side 3L-NPC converter 99.5%
under different reactive power 99.4%
boundaries in Fig. 4.1 99.3% Q-max
Efficiency
99.2% Q+max
99.1%
99.0%
98.9%
98.8%
98.7%
98.6%
0 0.2 0.4 0.6 0.8 1
Active Power (p.u.)
requirements for the reactive power under normal operation, and this should be
taken into account when evaluating and designing the wind turbine system. The
similar efficiency performances of a 5L-MLC2 multi-level converter considering
grid codes can be found in [20].
4.3.2 Thermal Stress Considering Reactive Power Demands

by Grid Codes
In order to investigate the thermal stress imposed by grid codes, the converter’s
electrical behaviors modified by the grid requirements have to be clarified. The
current amplitude and phase angle of converter with relation to the allowable
reactive power range defined by German grid codes are plotted in Fig. 4.12a, b,
respectively, in which the situations when the wind speed is 12 m/s (Po = 1 p.u.),
10 m/s (Po = 0.6 p.u.), and 8 m/s (Po = 0.3 p.u.) are indicated. It can be seen that
when complying with grid codes, the delivered active/reactive power by the wind
power converter is restrained in a narrow range which is much smaller than the
capability of the converter.
The corresponding junction temperature of the 3L-NPC converter is shown in
Fig. 4.13, where the wind turbine is generating 0.63 p.u. active power at 10 m/s wind
speed and three extreme reactive power conditions required by grid codes are
applied: (a) maximum underexcited reactive power when Q−max (10 m/s) = −0.13 p.u.,
(b) no reactive power when PF = 1, and (c) maximum overexcited reactive power
when Q+max (10 m/s) = 0.27 p.u.
It can be seen that when complying with the regulations by grid codes, the
thermal loading in the most stressed devices (Dnpc and T1) has no significant
difference among the three extreme reactive power conditions. More detailed
analysis and function regarding the electrical and loss behaviors modified by
reactive power can be found in the special attached topic 11.
(a) (b)
Load current amplitude Ig (p.u.)
Phase angle , Ig - Uc (degree)

12 m/s
10 m/s
12 m/s
10 m/s
8 m/s
8 m/s
Underexcited Overexcited Underexcited Overexcited
Reactive power (p.u.) Reactive power (p.u.)
Fig. 4.12 Operating range of reactive power when complying with grid codes (dot line represents
underexcited Q−, real line represents overexcited Q+). a Current amplitude Ig versus reactive
power Q. b Phase angle α versus reactive power Q
(a) (b)
)
Junction temperature (
Dnpc1 Dnpc1
T1 T1
T2 D1 T2 D1
D2 D2
Time (s) Time (s)
(c)
Dnpc1
T1
D1
T2
D2
Time (s)
Fig. 4.13 Thermal distribution of 3L-NPC inverter under different reactive power boundaries
defined by the grid codes (vw = 10 m/s, Po = 0.63 p.u.). a Q−max = −0.13 p.u. b Q0 = 0 p.u.
c Q+max = 0.27 p.u
4.4 A Thermal Control Method Utilizing Reactive Power 55
4.4 A Thermal Control Method Utilizing Reactive Power
In this section, a thermal control concept which utilizes the reactive power to relive
the temperature excursion under wind gust is going to be introduced. The relevant
approaches for the analysis are highlighted in Fig. 4.14. It can be seen that this
section mainly investigates the electrical and thermal stress influenced by the
control strategy of wind power converter. It should be noted that a control loop is
created by feedback the thermal stress to the control part of converter.
4.4.1 Control Idea and Diagram
It has been revealed that the thermal stresses in the 3L-NPC converter are quite
fluctuated under the wind gust operation, which is proven to be adverse for the
reliability of power devices. Because the junction temperature of the power devices
is proportional to the consumed power losses, one idea to reduce the junction
temperature fluctuation is to increase the device losses at the period with lower
junction temperature. It is known that the consumed losses of power switching
devices are governed by the delivered active and reactive power of converter. In a
grid connected wind power converter, the delivered active power is set according to
the input mechanical power from wind turbines and it cannot be regulated in order
to maintain the DC bus voltage. As a result, the delivered reactive power is one of
Converter design
Topology
Power devices
Pass. components

Electrical &Thermal
profiles
Fig. 4.14 Stress analysis approach considering control strategy

the left control freedoms to change the loss and thermal behavior in the power
devices.
A junction temperature control method is thereby proposed: The control target is
to maintain the mean junction temperature of the most stressed device (e.g., Dnpc in
the given 3L-NPC converter) at the fixed level. By adjusting the amount of reactive
power delivered by the power converter, it is possible to change the device loss
dissipation and thereby make the real device junction temperature trace the tem-
perature reference. Consequently, the thermal excursion in the power devices can
be more stabilized under wind gust.
One possible control diagram is demonstrated in Fig. 4.15. For simplicity, the
active current reference idref is generated by the wind speed referring to the wind
turbine model. A PI controller is used to control the error between temperature
reference Tjmref and the estimated junction temperature TjmE. Then the output of the
PI controller is used as the reactive current reference iqref for the target converter.
It is noted that in the proposed control system shown in Fig. 4.15, a loss cal-
culation model is important to accurately estimate the junction temperature of the
target power devices in order to close the control loop. The inputs of this loss
estimation block can be the real-time measured voltages, currents, and ambient
temperature from the converter. More detailed explanations for the control models
can be found in the attached special topic Chap. 11 and [21].
4.4.2 Idea to Overcome the Reactive Power Limits
Nevertheless, the controllable reactive power by wind power converter has to be

restrained by grid codes, which normally define much narrower ranges of reactive
power than the converter ability. As a result, the junction temperature controllability
in the case of one converter is severely limited.
Wind turbine idref vdref sa

vw model
PI sb
Modulation Converter
Control
iqref vqref sc
Tjmref +- PI
TjmE -1 Measured iabc, vgabc, etc.

Thermal
Model -iqref
(for other inverters)
Loss
PavgLoss Estimation Measured iabc, vgabc, Vdc, TA, etc.
Fig. 4.15 Junction temperature control method with PI controller for wind power inverter
But the situation is different in the case of a wind park or a multi-cell converter
system where a few converters are connected in parallel. The basic idea to over-
come the range limits of reactive power by grid codes is to circulate the reactive
power among the paralleled power converters, as shown in Fig. 4.16a, b, respec-
tively. It can be seen that for the two cases the reactive power is only trapped among
converters and will not be injected into the power grid. Thereby, the useable
reactive power range is significantly enlarged and only limited by the device
ratings.
4.4.3 Thermal Stress Considering Extended Q Ranges

in Paralleled Converters
When considering the circulated reactive power among paralleled converters, the
current amplitude and phase angle with relation to the delivered reactive power are
plotted in Fig. 4.17a, b, respectively, in which different active power situations
when the wind speed is 12 m/s (P = 1 p.u.), 10 m/s (P = 0.6 p.u.), and 8 m/s
(P = 0.3 p.u.) are indicated. It can be seen that the delivered active/reactive power
by the wind power converter can be dramatically enlarged compared to Fig. 4.12.
The corresponding junction temperature of the 3L-NPC converter is shown in
Fig. 4.18, where the wind turbine is generating 0.63 p.u. active power at 10 m/s
wind speed and three extended reactive power conditions are applied: (a) maximum
underexcited reactive power when Q−max (10 m/s) = −0.82 p.u., (b) no reactive power
when PF = 1, and (c) maximum overexcited reactive power when Q+max (10
m/s) = 0.5 p.u.. It can be seen that the underexcited reactive power can significantly
increase the junction temperature in the most stressed devices of the given 3L-NPC
converter (outer switches T1 and clamping diode Dnpc).
4.4.4 Thermal Control Results
Figure 4.19 shows the thermal stress of the power devices with the same wind speed
variation in Fig. 4.8a. By enabling the proposed thermal control method, in which
the underexcited reactive current is adjusted to keep the average junction temper-
ature in the most stressed device (Dnpc) constant when the wind speed is below
10.5 m/s. It can be seen that junction thermal excursion amplitude in Dnpc reduces
from 32 °C (rising edge) and 25 °C (falling edge) in Fig. 4.8b to 12 °C (rising edge)
and 13 °C (falling edge), respectively in Fig. 4.19b with the proposed thermal
control method.
Figure 4.20 shows the situation for the paralleled converter in Fig. 4.16 which
compensates the reactive power generated by the target converter. The amount of
overexcited reactive current is adjusted to compensate the underexcited reactive
(a) Converter 1 (Underexcited operation)
T1 D1
P1
Dnpc1
D2
T2 Grid
T3
Dnpc2 D3
T4 D4 Underexcited
-Q
Converter 2 (Overexcited operation)

Overexcited
+Q
P2
...
Converter N
(b) Converter 1 (Underexcited operation)
P1
T1 D1
Dnpc1
D2
T2
T3
Dnpc2 D3
T4 D4 -Q
Grid
+Q
P2
...
Converter N
Fig. 4.16 Reactive power circulated in paralleled wind power converters. a Parallel converters in
a wind park. b Parallel converters in a multi-cell converter system
(a) (b)
Phase current amplitude Ig (p.u.)
Phase angle Ig - Uc (degree)

12 m/s
10 m/s 12 m/s
10 m/s
8 m/s
8 m/s
Fig. 4.17 Operating range of reactive power when considering paralleled converters (dot line
represents underexcited Q−, real line represents overexcited Q+). a Current amplitude Ig versus
reactive power Q. b Phase angle α versus reactive power Q
Fig. 4.18 Thermal stress of 3L-NPC inverter under different reactive power boundaries
considering paralleled converters (vw = 10 m/s, Po = 0.63 p.u.). a Q−max = −0.82 p.u. b Q0 = 0
p.u. c Q+max = 0.5 p.u
Fig. 4.19 Wind gust operation of 3L-NPC inverter with thermal control. a Wind speed variations
and current references. b Thermal distribution of each power device
Fig. 4.20 Wind gust operation of 3L-NPC inverter with compensating reactive power. a Wind
speed variations and current references. b Thermal distribution of each power device
power consumed by the converter 1. As a result, the overexcited reactive current

reference in Fig. 4.20a has the same amplitude but opposite direction compared to
Fig. 4.19a. It can be seen from Fig. 4.20b that the maximum junction temperature as
well as thermal excursion of the most stressed devices is not further increased.
4.5 Summary
The thermal profile of power switching devices under steady-state wind speeds is an
important tool which can be used either as a loading indicator for certain converter
topologies or as a lookup table for the lifetime estimation. It was found that wind
speed variations will lead to severe thermal excursion of some power switching
devices in the given 3L-NPC wind power converter. Particularly, the clamping
diode and outer switch suffer from more adverse thermal excursion than the other
4.5 Summary 61
devices. The grid codes even under normal operation may change the delivered
reactive power of wind power converters and thereby have impacts to the converter
efficiency as well as thermal stress distribution of power devices.
By circulating the reactive power among paralleled converters in a wind farm or
multi-cell converter system, it is possible to control the junction temperature and
relive the thermal excursion in most stressed devices under wind gust operation,
leading to higher reliability of the converter, while the increased thermal stresses to
the other devices or paralleled converters are still acceptable.
Relevant special topic
Chapter 11.
References
1. E.ON-Netz—Grid Code (2008) Requirements for offshore grid connections in the E.ON Netz
network, April 2008
2. Tsili M, Papathanassiou S (2009) A review of grid code technical requirements for wind
farms. IET Renew Power Gener 3(3):308–332
3. Kouro S, Malinowski M, Gopakumar K, Pou J, Franquelo LG, Wu B, Rodriguez J, Perez MA,
Leon JI (2010) Recent advances and industrial applications of multilevel converters. IEEE
IEEE Trans Ind Appl 48 (2):708–719
5. Senturk OS, Helle L, Munk-Nielsen S, Rodriguez P, Teodorescu R (2012) Power capability
investigation based on electro-thermal models of press-pack IGBT three-level NPC and ANPC
VSCs for multi-MW wind turbines. IEEE Trans Power Electr 59(3):1462–1476
6. Alvarez R, Filsecker F, Bernet S (2009) Characterization of a new 4.5 kV press pack SPT
+IGBT for medium voltage converters. In: Proceedings of ECCE’09, pp 3954–3962, Sept
2009
wind turbines. In: Proceedings of ISIE’ 2011, pp 340–346, June 2011
8. Maibach P, Faulstich A, Eichler M, Dewar S (2007) Full-scale medium voltage converters for
wind power generators up to 7 MW. ABB technical article, Switzerland www.abb.com.
Accessed Feb 2007
10. Li H, Chen Z, Polinder H (2009) Optimization of multibrid permanent-magnet wind generator
systems. IEEE Trans Energy Conver 24(1):82–92
11. Polinder H, van der Pijl FFA, de Vilder G-J, Tavner PJ (2006) Comparison of direct-drive and
geared generator concepts for wind turbines. IEEE Trans Energy Convers 21(3):725–733
12. Slootweg J, De Haan S, Polinder H, Kling W (2003) General model for representing variable
speed wind turbines in power system dynamics simulations. IEEE Trans Power Syst 18
(1):144–151
13. Holmes DG, Lipo TA (2003) Pulse width modulation for power converters. IEEE
Press/Wiley-Interscience, New York
14. Bruckner T, Holmes DG (2005) Optimal pulse-width modulation for three-level inverters.
IEEE Trans Power Electron 20(1):82–89
15. Rockhill AA, Liserre M, Teodorescu R, Rodriguez P (2011) Grid-filter design for a
multimegawatt medium-voltage voltage-source inverter. IEEE Trans Ind Electr 58(4):1205–
1217
16. Website of ABB Semiconductors (2012) http://www.abb.com/product/us/9AAC910029.aspx.
Accessed Jan 2012
17. ABB Application Note (2007) Applying IGCTs, May 2007
18. Kostandyan EE, Ma K (2012) Reliability estimation with uncertainties consideration for high
power IGBTs in 2.3 MW wind turbine converter system. In:Microelectronics reliability,
Proceedings of ESREF’ 2012
19. Sørensen P, Hansen AD, Rosas PAC (2002) Wind models for simulation of power fluctuations
from wind farms. J Wind Eng 90:1381–1402
20. Ma K, Muñoz-Aguilar RS, Rodríguez P, Blaabjerg F (2013) Thermal and efficiency analysis
of five-level multi-level clamped multilevel converter considering grid codes, IEEE
transactions on industry applications, 2013. In: Proceedings of ECCE’ 2012, pp 1774–178
21. Ma K, Blaabjerg F, Liserre M (2012) Reactive power control methods for improved reliability
of wind power inverters under wind speed variations. In: Proceedings of ECCE’ 2012,
pp 3105–3112
Chapter 5
Stress Analysis of 3L-NPC Wind Power
Converter Under Fault Condition
This chapter investigates the thermal stress of wind power converter when suffering
from grid faults. The comprehensive analysis for the electrical and thermal
behaviors of power devices undergoing various grid faults is conducted on the
3L-NPC wind power converter. Afterwards, a series of thermal-redistributed
modulation methods and a power control strategy which utilizes the zero sequence
current are presented to achieve better performance under this adverse condition.
5.1 Requirements and Conditions Under Fault Operation
Besides the normal operation, the TSOs in different countries have also issued strict
Low Voltage Ride Through (LVRT) codes for the wind turbine system, as shown in
Fig. 1.10. Meanwhile, it is becoming a need that the wind power generation system
should be able to provide reactive power (up to 100 % current capacity) to con-
tribute to the voltage recovery when grid faults are present. Figure 1.11 shows the
required amount of reactive current against to the lowest grid voltage amplitude
regulated by German grid codes. As a case study, the German LVRT codes are
adopted in this chapter for analysis, and the same 3L-NPC wind power converter
model as already illustrated in Sect. 4.1 is again used to demonstrate the converter
behavior under various grid fault conditions. However, in this chapter, only the
grid-side converter is focused.
When a short-circuit fault happens in the power grid, depending on the types and
location of fault, the line impedances, and the connection of transformer windings,
the voltage dips may vary significantly on different locations (buses) in the power
grid system [1, 2]. Therefore, it is important to investigate how the voltage dips
propagate to the bus where the WTS is connected. A typical configuration of a
WTS with the grid system is shown in Fig. 5.1, in which the voltage on Bus 2 is
monitored by the WTS and hence determines the LVRT behavior of the wind
power converter. A delta-star transformer is used to interface the WTS on Bus 2
(e.g., 3.3 kV) and the Point of Common Coupling (PCC) on Bus 1 (e.g., 20 kV). It
is assumed that a short-circuit fault happens somewhere with line impedance ZF to

64 5 Stress Analysis of 3L-NPC Wind Power Converter …
Bus 2 Bus 1 (PCC)

AC DC
DC AC
3.3 kV / 20 kV
Wind Turbine System
ZS
Fault Grid
>=20 kV
ZF
Fig. 5.1 Typical configuration for grid integration of a WTS
the Bus 1 (PCC), and the line impedance from PCC to the grid with a higher voltage
level is Zs.
Define that the voltage dip severity on Bus N is DN, which is related to the
location of grid faults and power line impedance. Providing that the line imped-
ances for the positive and negative sequence components are equal, the dip severity
on the Point of Common Coupling (or Bus 1) D1 can be written as [2]
ZF
D1 ¼ ð5:1Þ
ZF þ ZS
It is obvious that D1 ranges from 0 (if ZF = 0) to 1 (if ZS = 0) and represents the

voltage dipping severities from the most severe case (D1 = 0) to the non-dip case
(D1 = 1). Three typical grid faults, one-phase grounded (1 phase), two-phase
connected (2 phase), and three-phase grounded (3 phase), are assumed to happen,
respectively, at the same location of the given power grid system. The dip type, dip
severity, and voltage amplitude of the three faults seen on Bus 1 and Bus 2 are
summarized in Table 5.1, where the voltage dipping type is defined in Fig. 5.2 [1].
It is noted that the voltage dipping characteristics appeal differently not only among
the three types of grid faults, but also among different buses. Because Bus 2 is
directly connected and monitored by the WTS, only V2min will determine the
amount of reactive current injection into the power grid by the converter when the
three types of grid faults are present.
Figure 5.3 shows the control scheme adopted to deal with unbalanced grid faults:
both positive and negative sequences voltages/currents are detected and controlled.
A sequence decoupling algorithm is used to remove the 100 Hz oscillation com-
ponents in each of the sequence domain [2–5]. It is worth to mention that the
current control schemes of wind power converter under grid faults may have strong
impacts to the loading of power devices, but this is still an open question under
discussion because the definitions for the current and voltage behaviors of each
phase under grid faults have not yet been clearly specified. In this section, it is first
assumed that only positive sequence currents are generated by the converter, and
the negative sequence currents are control to be zero by setting the references for
negative sequence currents to zero when using the control method shown in
5.1 Requirements and Conditions Under Fault Operation 65
Table 5.1 Voltage dip seen on Bus 1 and Bus 2 for various grid faults
Fault type One-phase Two-phasea Three-phase
Bus1 Dip type B C A
(PCC) Dip value Db1 D1 \0
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi
Voltage amplitudec D1 D1
0:5 1 þ ð 3D1 Þ2
V1min (p.u.)
Bus2 Dip type C D A
(WTS) Equivalent dip D1 þ0:5
1:5 \0
D1 \0 D1 \0
value D2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi
Voltage amplitudec D2 D2
0:5 1 þ ð 3D2 Þ2
V2min (p.u.)
Notes
a
Two-phase grounded fault is not included because it seldom happens
b
No phase jump is assumed, i.e., vector D1 has zero phase angle
c
Voltage amplitude means the lowest one among the three phases
VC Type A VC Type B VC Type C VC Type D

VdipC
VdipC VdipA VA VdipA VA VA VA
VdipC
VdipB
VdipB
VdipB
VB VB VB VB
Fig. 5.2 Phasor diagram definitions for the dip types A–D specified in Table 5.1
Fig. 5.3. The current references in positive sequence for converter as a function of
the minimum grid voltage amplitude are specified in Fig. 5.4, where the reactive
current I+q is set according to Fig. 1.11, and the active current I+d is referred to the
generated power by wind turbines when V2min is above 0.5 p.u. It is assumed that
the converter is set to provide as much active power as possible during grid faults,
and the pitch control of the wind turbine does not have enough time to activate [6].
This is the worst LVRT condition for the power converter in terms of device
loading.
The 10 % increased DC bus voltage of converter is taken into account in the
stress analysis under LVRT, and this is due to the short-term input–output power
unbalance and the trigging of DC bus chopper [6]. It is noted that the increased DC
bus voltage may significantly decrease the lifetime of power switching devices due
to the cosmic radiation failure mechanism as reported in [7]. However, this issue
will not be considered in this book.
idref+ ugd+
+
PI + + ud+
- +
id+
abc/ + idq + -ωLg dq0+/
+
dq0+ - iq abc
+
ωLg
idq~
θ + -
PI + + uq+ θ
iqref + + +
ugq
iabc 2θ Sequence + vabc
Decoupling idref - ugd-
+
0
+
PI + + ud-
-θ - -θ
idq~- +
id-
- idq- ωLg
abc/ dq0-/
-
dq0- + iq abc
-ωLg
0
+ -
PI + + uq-
iqref -
-
+
ugq
Fig. 5.3 Current control scheme of the grid-side converter
Fig. 5.4 Current references

Id / Iq reference positive sequence (p.u.)
in positive sequence versus

grid voltage amplitude Iq+ >12 m/s
10 m/s
Id+ 8 m/s
Minimum voltage amplitude on Bus 2 V2min (p.u.)

5.2 Stress Analysis of Converter Under LVRT 67
5.2 Stress Analysis of Converter Under LVRT
The ride-through operation of converter will significantly change the electrical

behaviors compared to the normal conditions, and thereby result in quite different
loading conditions of power devices. In this section, the electrical and thermal
behaviors of converter under LVRT are going to be investigated. The relevant
approaches for analysis are highlighted in Fig. 5.5. It can be seen that this section
mainly focuses on the electrical and thermal stresses influenced by the mission
profile, although the other stress-determining factors are needed to be involved in
order to establish a proper operating condition for the converter.
5.2.1 Electrical Behaviors
After the converter and grid conditions under grid faults are defined, the electrical
behaviors of the given 3L-NPC converter can be simulated and specified. Figure 5.6
gives the profiles for the amount of active and reactive powers delivered by con-
verter under various voltage dipping severities on Bus 1. The situations with
one-phase and two-phase grid faults are shown in Fig. 5.6a, b, respectively. It can
be seen that the delivered active/reactive powers by converter under the two types
of grid faults are significantly different, especially when dip severity D1 is below
0.5 p.u.
Converter design
• Topology
• Power devices
• Pass. components

• Wind profile • Mechanical • Modulation
• Ambient temp. • Electrical • Basic control
• Grid codes • Thermal • High level control
Electrical & thermal

profiles
Fig. 5.5 Stress analysis approach considering LVRT grid codes

(a) 0.58 0.66 0.75 0.83 0.92 1 (b) 0 0.2 0.4 0.6 0.8 1
Minimum voltage on Bus 2 V2min (p.u.) Minimum voltage on Bus 2 V2min (p.u.)
>12 m/s >12 m/s
Average P +/Q + (p.u.)
Average P +/Q + (p.u.)

+
P
+
Q
10 m/s 10 m/s
8 m/s 8 m/s
+
P
Q+
Dip severity on PCC D1 (p.u.) Dip severity on PCC D1 (p.u.)
Fig. 5.6 Active and reactive power versus dip severity on PCC. a Single-phase grounded fault.
b Two-phase connected fault
Figure 5.7 shows the phase angle displacement (between the load current of
converter and grid voltage on Bus 2) with relation to the dipping severity on PCC. It
is noted that the phase angle difference among the three phases of the converter
becomes larger at smaller D1, and the maximum difference achieves 60° when
D1 = 0 p.u. for both of the grid fault conditions. It can also be observed from
Fig. 5.7 that the wind speeds have strong impacts to the phase angle of the converter
under LVRT operation, especially when D1 is larger than 0.5 p.u..
Figure 5.8 shows the modulation index of the given converter at various grid
faults and dip severities on PCC. It can be seen that the two-phase grid fault leads to
very large difference for modulation index among the three phases of converter (up
to 0.7 difference when D1 = 0 p.u.). It is interesting to see that the wind speeds do
not have strong impacts on the modulation index of converter under various LVRT
conditions.
(a) 0.58 0.66 0.75 0.83 0.92 1 (b) 0 0.2 0.4 0.6 0.8 1
Minimum voltage V2min (p.u.) Minimum voltage V2min (p.u.)
Phase angle I-Vg2 (°)
Phase angle I-Vg2 (°)
Phase B
Phase C
Phase A
Phase A
Phase C >12 m/s >12 m/s
8 m/s 8 m/s
Phase B
Fig. 5.7 Phase displacement between load current and Bus 2 voltage versus dip severity on PCC.
a Single-phase grounded fault. b Two-phase connected fault
5.2 Stress Analysis of Converter Under LVRT 69
(a) 0.58 0.66 0.75 0.83 0.92 1 (b) 0 0.2 0.4 0.6 0.8 1
Minimum voltage on Bus 2 V2min (p.u.) Minimum voltage on Bus 2 V2min (p.u.)
Phase B
Modulation index m
Modulation index m
Phase A
Phase C
Phase C
Phase B Phase A
>12 m/s >12 m/s

8 m/s 8 m/s
Fig. 5.8 Modulation index of grid converter versus dip severity on PCC. a Single-phase grounded
fault. b Two-phase connected fault
The active and reactive power oscillations under unbalanced grid voltage are
inevitable if only positive sequence current is injected by the converter, and this
problem can be explained by instantaneous power theory as detailed in [8].
Figure 5.9 shows the oscillation amplitude of the power delivered by the converter
with relation to the dipping severity on PCC, where various grid faults and wind
speed conditions are indicated. It can be seen that there is no power oscillation of
the converter under the three-phase balanced grid fault, while the two-phase fault
condition introduces larger power oscillation than the condition of one-phase fault
at the same dipping severity on PCC. The maximum power oscillation amplitude
achieves 0.5 p.u. at D1 = 0 when the two-phase grid fault is present.
The analytical functions and simulation examples for Figs. 5.6, 5.7, 5.8, and 5.9
are shown in Chap. 14. It can be expected that the dramatically different behaviors
compared to the normal operation in respect to the delivered power, phase angles,
Fig. 5.9 Active/reactive 0.5

power oscillation versus dip
2 phase
P/Q oscillation amplitude (p.u.)
severity on PCC 0.4 >12 m/s

8 m/s
0.3
0.2
1 phase
0.1
3 phase
0
0 0.2 0.4 0.6 0.8 1

Dip severity on PCC D1 (p.u.)
and modulation index under LVRT operations may lead to significantly different
thermal loadings of power semiconductor devices.
5.2.2 Thermal Behaviors
As an example, the junction temperatures for the three phases of 3L-NPC converter
undergoing the extreme two-phase connected grid fault (Type D voltage dip on Bus
2 with D1 = 0 p.u., vw = 12 m/s) are shown in Fig. 5.10. It can be seen that the
thermal loading behaviors in the three phases of the converter are totally different
from each other.
The mean junction temperature Tm of the switches and diodes under various dip
severities and grid faults are summarized in Figs. 5.11 and 5.12, respectively. For
the one-phase grid fault, Tout and Dnpc are the most stressed devices under the whole
range of dipping severity, while for the two-phase and three-phase grid fault con-
ditions, Tin and Dnpc are more stressed, when D1 is below 0.5 p.u..
Special attention should be given to the power device Tin, Dout, Din, and Dnpc
under the LVRT operation of 3L-NPC. In fact those devices may have even higher
junction temperature than in the cases of normal operation. (Up to 40 °C higher for
Dout, 20 °C higher for Tin, 15 °C higher for Din, and 10 °C higher for Dnpc). This
overloading should be taken into account when designing the power devices and
heat sink system for the wind power converter.
In Chap. 12, the loading of power device in several promising multi-level
converter topologies will be investigated based on the LVRT conditions.
Fig. 5.10 Thermal distribution under two-phase grid fault (Type D voltage dip on Bus 2 with
D1 = 0 p.u., vw = 12 m/s, horizontal axis means time with unit of second)
5.3 Thermal Redistributed Modulations Under LVRT 71
(a) (b)
120.0 120.0
Junction temperature (°C)

Tout Tout
110.0 110.0
100.0 100.0
90.0 90.0
Tin Tin
80.0 80.0
Phase A Phase A
70.0 Phase B 70.0 Phase B
Phase C Phase C
60.0 60.0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(c)
120.0
Tout
110.0
100.0
90.0
Tin
80.0
Phase A
70.0 Phase B
Phase C
60.0
0 0.2 0.4 0.6 0.8 1
Fig. 5.11 Junction temperature profiles of switches under various dipping severities.
a Single-phase grounded fault. b Two-phase connected fault. c Three-phase grounded fault
5.3 Thermal Redistributed Modulations Under LVRT
Based on the thermal stress analysis under LVRT, it is found that the thermal
distribution of 3L-NPC inverter is quite unequal and inefficient: as the “hottest”
power devices, the NPC diodes Dnpc and inner switch Tin have up to 40 K higher
junction temperature than the most “cold” device Din. In this section, a series of
thermal redistributed modulations for 3L-NPC converter are going to be presented,
which aims to achieve a more even junction temperature among different power
devices under extreme grid faults. The relevant determining factors for analysis are
highlighted in Fig. 5.13, and it can be seen that this section mainly focuses on the
electrical and thermal stresses influenced by the modulation strategy of the wind
power converter.
5.3.1 Basic Idea
The loss distribution of the most stressed clamping diode Dnpc and inner switch Tin
in 3L-NPC inverter under extreme grid voltage dips is shown in Fig. 5.14a, b,
(a) (b)
120.0 120.0

Dnpc Dnpc
110.0 110.0
100.0 100.0
90.0 Phase A 90.0 Phase A

Phase B Phase B
80.0 Phase C 80.0 Phase C
70.0 Dout 70.0 Din Dout

Din
60.0 60.0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(c)
120.0
Dnpc
110.0
100.0
90.0 Phase A
Phase B
80.0 Phase C
70.0 Din Dout
60.0
0 0.2 0.4 0.6 0.8 1
Fig. 5.12 Junction temperature profiles of diodes under various dipping severities. a Single-phase
grounded fault. b Two-phase connected fault. c Three-phase grounded fault
Converter design
Topology
Power devices
Pass. components


profiles
Fig. 5.13 Stress analysis approach considering modulation of converter

(a) 4.0 (b) 4.0

Lcon Lcon Lsw
3.5 3.5
Lsw
3.0 3.0
Loss (kW)
Loss (kW)
2.5 2.5
2.0 2.0
1.5 1.5
1.0 1.0
0.5 0.5
0.0 0.0
0.05 0.2 0.35 0.5 0.05 0.2 0.35 0.5
Grid voltage (p.u.) Grid voltage (p.u.)
Fig. 5.14 Loss distribution of the 3L-NPC wind power inverter under LVRT (Lcon conduction
loss, Lsw switching loss). a Clamping diode Dnpc. b Inner switch Tin
respectively, where only three-phase balanced grid fault condition is assumed for
simplicity. Referring to the switching process of the 3L-NPC converter [9], the
losses in Dnpc and Tin will be effectively relieved by reducing the dwelling time of
zero voltage level (reduced conduction loss) or reducing the commutations
involving zero voltage level (reduced switching loss).
When implementing the space vector modulation for 3L-NPC inverter, the
reference vector under extreme LVRT is mainly located in the inner hexagon of
space vector diagram. In this special area, there are one or two switching redun-
dancies for each state vector, and it is possible that some of the switching redun-
dancies could modify the loss distribution and achieve better thermal distribution
for the 3L-NPC wind power inverter under LVRT operation.
For simplicity, the sequence generation method is demonstrated only based on
sector I (0°–60°) in the space vector diagram for 3L-NPC converter. The detailed
information of sector I is shown in Fig. 5.15. As mentioned before, the voltage
reference vector Vref in this sector mainly locates in region A under extreme LVRT
condition. The reference can be synthesized by the three nearest state vectors: one
Fig. 5.15 Sector I of the 220

space vector diagram for
3L-NPC converter
D
221 (p)
110 (n) 210
A2
B
A C
Vref
222 A1
111 200
211 (p)
000
100 (n)
000 100 110 111 211 221 222 221 211 111 110 100 000
Fig. 5.16 The “complete” vector sequence in region A
from the redundant short vector group 211/100 (red), one from the other redundant
short vector group 221/110 (blue), and one from the redundant zero vector group
000/111/222 (black). The numbering “2,” “1,” and “0” of the state vectors represent
that a certain phase is connected by the converter to the positive DC bus, the neutral
point, and the negative DC bus, respectively.
It is interesting that when using all of the state vectors including the redundant
ones in region A to synthesize the reference vector Vref, the sequence arrangement
is unique if the least switching process and symmetric pulses for each phase are
realized, as shown in Fig. 5.16. All of the state vectors in region A are visited
without unnecessary switching processes and abrupt voltage level changes.
Therefore, other applicable modulation sequences can be generated by elimi-
nating some redundant state vectors in the “complete” vector sequence. However,
the sequence generation method must follow two criteria:
i. Atleast one state vector from each of the “three nearest” redundant vector group
has to be used in order to synthesize the desired reference voltage.
ii. The state vectors have to be arranged in reverse fashion over one carrier cycle
in order to smoothly be connected with the succeeded vector sequence without
extra switching processes.
5.3.2 A Group of Modulation Methods
In order to reduce the output time for the zero voltage level of 3L-NPC inverter
during LVRT, one effective way is to reduce the activating time for the zero state
vector “111” when generating the modulation sequence (because vector 111 outputs
zero voltage level for all of the three phases). Figure 5.17a indicates a special
modulation sequence, and it is called Optimized sequence 1 (O1) for convenience.
(a) (b)
Vo
Current or voltage (A or V)
A
Iload
Vg
C
000 100 110 111 211 221 222 221 211 111 110 100 000
Ts/2 Ts/2 Ts/2 Ts/2

Time (s)
(c) (d)
Vo
A
Iload
Vg
C
000 100 110 111 211 221 222 221 211 111 110 100 000
Ts/2 Ts/2 Ts/2 Ts/2

Time (s)
(e) (f)
Vo
A
Iload
Vg
C
000 100 110 111 211 221 222 221 211 111 110 100 000
Ts/2 Ts/2 Ts/2 Ts/2

Time (s)
Fig. 5.17 The loss and thermal-redistributed modulation schemes. (Vo-output voltage pulses, Vg-
grid voltage, Iload-phase current. Vg = 0.05 p.u., 100 % rated reactive current). a O1 sequence
arrangement. b Output of O1 sequence. c O2 sequence arrangement. d Output of O2 sequence.
e O3 sequence arrangement. f Output of O3 sequence
The output time for zero vector 111 in O1 sequence is reduced and partly replaced
by the equivalent zero state vectors 000 and 222.
Another direct and basic idea is trying to avoid using the state vector “111”
based on the “complete” sequence, as shown in Fig. 5.17b which is called
Optimized sequence 2 (O2) for convenience. It can be seen that the zero state
vectors 111 are totally replaced by the equivalent state vectors 000 and 222. The
output waveforms of the O2 sequence are shown in Fig. 5.17d, in which the widths
of both positive and negative output voltage pulses are expanded, compared to the
O1 sequence—which means further reduced dwelling time for zero voltage level.
In order to reduce the commutations involving zero voltage level, the third
modulation sequence is generated in Fig. 5.17e, which eliminates the zero state
vectors 000 and 222, and is called Optimized sequence 3 (O3) for convenience. The
Fig. 5.18 SVM sequences in the form of carried-based modulation references (duty ratio for the
inner switch S1 and outer switch S2, M = 0.3). a Normal sequence. b O1 sequence. c O2 sequence.
d O3 sequence
output waveforms of the O3 sequence are shown in Fig. 5.17f, in which there are
less output voltage pulses compared to the O1 and O2 sequences—which means
reduced commutations and less switching loss in Dnpc and Tin.
It is noticed that because there is redundancy for each of the state vector, all the
used vectors in the proposed sequences are arranged within two switching cycles
rather than one, as indicated at the bottom of Fig. 5.17a, c, e, respectively. As a
result, compared to the normal sequence, the equivalent switching frequency in the
proposed sequences is not further increased especially for O1 and O2.
For better understanding the proposed space vector modulation sequences in the
form of carried-based modulation, they are transferred to the voltage
references/duty ratio for the inner and outer switches of 3L-NPC converter, as
shown by DTout and DTin in Fig. 5.18. It can be seen that compared to the normal
sequence, the O1 and O2 sequences generally move the level of voltage references,
introducing larger duty ratio for the outer switch and smaller duty ratio for the inner
switch—which will lead to the longer dwelling time for the positive/negative
voltage output and reduce the dwelling time of the zero voltage output. In respect to
the O3 sequence, the duty ratio is very similar to the normal sequence but
(a) (b)
Vll Vll
Vo Vo
Iload Iload
(c) (d)
Vll Vll
Vo Vo
Iload Iload
Fig. 5.19 Experimental outputs of different modulation sequences, Vll-line-to-line voltage pulses
(100 V/div), Vo-phase voltage pulses (100 V/div), Iload-load current (5 A/div), modulation index
M = 0.5 p.u., phase angle θ = 80°. a Normal sequence. b O1 sequence. c O2 sequence. d O3
sequence
deviations can be found around the time for zero crossing of voltage reference—
which means that the dwelling time for zero voltage output is not significantly
reduced. The O3 sequence is very similar to the modulation sequence proposed by
[10, 11].
The proposed modulation sequences are also validated on a downscaled
experimental setup, as shown in Fig. 5.19, whose detailed parameters are listed in
Appendix II, where a special passive RL load is used to emulate the LVRT
operation of converter. Based on the same condition of the experimental setup, a
series of simulations are also carried out and the results are shown in Fig. 5.20. It
can be seen that the experimental results of line-to-line voltage pulses, phase
voltage pulses, and load current have good agreement with the simulations.
5.3.3 Loss and Thermal Improvements
The loss distributions of the 3L-NPC wind power converter under LVRT when
utilizing the normal and optimized vector sequences are compared in Fig. 5.21a.
The comparison of steady-state mean junction temperature distribution is shown in
Fig. 5.21b. It can be seen that the optimized sequences O1, O2, and O3 all achieve
the loss and thermal redistribution among the power devices, and the stress of Dnpc
and Tin can be relieved under LVRT operation. The mean junction temperature
(a) (b)
Vll (V)
Vll (V)
Vo (V)
Vo (V)
Iload (A)
Iload (A)
(c) (d)
Vll (V)
Vll (V)
Vo (V)
Vo (V)
Iload (A)
Iload (A)
Fig. 5.20 Simulated outputs of different modulation sequences in Fig. 5.19, Vll-line-to-line
voltage pulses, Vo-phase voltage pulses, Iload-load current, modulation index M = 0.5 p.u., phase
angle θ = 80°. a Normal sequence. b O1 sequence. c O2 sequence. d O3 sequence
(a) 6 (b)110.0 Norm

Dcon 105.0 O1
Dsw
5 100.0 O2
Junction Temp. (°C)
Tcon
O2
Tsw 95.0
4
Loss (kW)
90.0
3 85.0
80.0
2
75.0
1 70.0
65.0
0
Norm O1 O2 O3 Norm O1 O2 O3 Norm O1 O2 O3 60.0
Tout & Dout Tin & Din Dnpc Tout Dout Tin Din Dnpc
Fig. 5.21 Thermal and loss comparison for the 3L-NPC wind power inverter under LVRT for
different modulation sequences, Vg = 0.05 p.u., 100 % reactive power. a Loss distribution. b Mean
junction temperature comparison
reductions in the most stressed devices Dnpc and Tin are estimated to be up to 12 and
8 K, respectively, when the O2 modulation sequence is applied, while the O3
sequence shows another advantage in reducing the stress in Tout and Dout by 3 and
5 K, respectively.
The dynamic thermal performance of 3L-NPC wind power inverter which goes
from normal operation to extreme LVRT and then back to normal operation can
also be simulated. For simplicity, only the results when applying O2 sequence are
indicated and more results can be found in [12]. It can be seen from Fig. 5.22 that
the maximum junction temperatures in Dnpc and Tin are both reduced, and more
(a) (b)
Tjmax=104
Tjmax =92
Dnpc Dnpc
Tout Tout
Dout Tin Tin

Dout
Din Din
Normal Voltage Normal Normal Voltage Normal
operation dips operation operation dips operation
Time (s) Time (s)
Fig. 5.22 Junction temperature dynamic response with a voltage dip time of 500 ms of different
modulation sequences (from normal operation with wind speed 8 m/s to 0.05 p.u. LVRT, and then
back to normal operation). a Normal modulation during LVRT. b O2 modulation during LVRT
equal thermal distribution can be achieved compared to the normal modulation.

This improvement may contribute to a lifetime extension of the converter according
to the important Coffin-Masson lifetime model.
5.3.4 Neutral Point Potential Control and Total Harmonic

Distortion
In order to analyze the control ability of Neutral Point (NP) potential by the pro-
posed modulation sequence, the neutral point current iNP which is the main reason
for the DC bus unbalance will be focused.
For the proposed three sequences O1, O2, and O3, both of the used short vectors
have redundancies, which mean full control freedom for the NP current, and it is
possible to achieve zero iNP within one switching cycle under all power factors and
applicable modulation indexes. Consequently, the NP potential control ability of all
the proposed modulation sequences should be better than the normal SVM
sequence. Figure 5.23 indicates the experimental NP current of the 3L-NPC con-
verter when different modulation methods are applied. It is noted that the active NP
current control is not active. For clarity, both the instantaneous neutral point current
iNP and its switching-cycle-averaged value iNPave are indicated. It can be seen that
the iNPave in the proposed modulation sequences O1–O3 are almost eliminated,
while the normal modulation sequence has iNPavg which oscillates at 0.2 p.u.
amplitude and three times of the fundamental frequency.
Nevertheless, the proposed modulation sequences especially for O2 and O3 have
no Total Harmonic Distortion (THD) improvements compared to the normal
sequence. This can be also observed from the load current waveforms in Fig. 5.17d,
f, where the current distortions in the O2 and O3 sequences are actually more than
that in the normal sequence of Fig. 5.17a. However, it is noted that under the LVRT
operation, which is an abnormal condition and normally last up to few seconds, the
Fig. 5.23 Instantaneous and mean value of experimental MP current in 3L-NPC converter (mean
value averaged within one switching cycle). a Normal modulation sequence. b O2 Sequence. c O2
sequence. d O3 sequence
first control target is to make the grid converter survive when withstanding the grid
voltage dips and providing large amount of reactive power support. Therefore, THD
is relatively less important performance for the wind power converter under the
LVRT.
Based on the similar idea, there are many groups of modulation sequences that
can be generated, which are able to modify the thermal distribution of 3L-NPC
converter under LVRT operation. More details can be found in Chap. 13.
5.4 New Power Control Methods Under Unbalanced AC

Source
This section targets to improve the power control limits of typical three-phase wind
power converter system under unbalanced AC source (e.g., grid or generator with
faults). A new series of control strategies which utilizes the zero sequence com-
ponents are proposed to enhance the power control ability under this adverse
condition. The relevant determining factors for analysis are highlighted in Fig. 5.24,
5.4 New Power Control Methods Under Unbalanced AC Source 81
Converter design
• Topology
• Power devices
• Pass. components

• Wind profile • Mechanical • Modulation
• Ambient temp. • Electrical • Basic control
• Grid codes • Thermal • High level control
Electrical profile
Fig. 5.24 Electrical stress analysis mainly considering the basic control of converter
and it can be seen that the electrical behavior influenced by the basic control of
wind power converter is mainly focused, although the rest factors may need to be
involved in order to establish a proper converter operating condition.
5.4.1 Applicable Conditions and Control Structure
When the voltages become distorted and unbalanced under faults or disturbances,
special control methods of the converter which can regulate both the positive and
negative sequence currents have been introduced to handle these problems [3–5,
12–15]. However, the resulting performances by these control methods are not
satisfactory: either distorted load currents or power oscillations will be introduced,
and thereby not only the grid/generator but also the power converter will be further
stressed.
The main reason for the unsatisfactory control performances is detailed in
Chap. 14. In the typical three-phase three-wire converter structure, there are only
four current control freedoms, which are not enough under unbalanced AC source
condition because there are six power parameters which need to be controlled [3–5].
Therefore, more control freedoms would be needed in order to improve the control
performance of the converter under adverse AC source condition.
Another type of converter structures which have been presented before for the
next generation wind turbines is shown in Fig. 5.25a, b, respectively. Compared to
the traditional three-wire structure, this series of converters introduces six wires
(a) (b)
Open winding Open winding
Transformer/ Transformer/
Generator Generator
Filter
Filter
Fig. 5.25 Converter system with zero-sequence current path. a Six-wire three-level system.
b Six-wire five-level system
(open winding) and thus zero-sequence current path, which may enable extra two
current control freedoms to achieve better control performances. The control dia-
gram for the positive, negative, and zero sequence current used for this six-wire
converter structure is shown in Fig. 5.26. It is noted that if the converter is used at
the grid side in the wind turbine system, the zero sequence current is not injected
into the grid but trapped in the typically used d-Y transformer.
5.4.2 Control Ideas and Methods
With the zero sequence current, the three-phase currents generated by the converter
can be written as [12–17]
IC ¼ Iþ þ I þ I0 ð5:2Þ
By operating the voltage of AC source and current controlled by power con-

verter (5.2), the instantaneous generated real power p, imaginary power q in the αβ
coordinate, and the real power p0 in the zero coordinate can be calculated as
2 3 2 3 2 3
p va ia þ vb ib P þ Pc2 cosð2xtÞ þ Ps2 sinð2xtÞ
4 q 5 ¼ 4 va ib va ib 5 ¼ 4 Q þ Qc2 cosð2xtÞ þ Qs2 sinð2xtÞ 5 ð5:3Þ
p0 v0 i 0 P0 þ P0c2 cosð2xtÞ þ P0s2 sinð2xtÞ
Then the instantaneous three-phase real power p3Φ and imaginary power q3Φ of
the converter can be written as

p3/ p þ p0
¼
q3/ q
P þ P0 Pc2 þ P0c2 Ps2 þ P0s2
¼ þ cosð2xtÞ þ sinð2xtÞ ð5:4Þ
Q Qc2 Qs2
idref+ ugd+
+ + u d+
PI
-
id+
abc/ + idq + -ω Lg dq+/
+
dq+ - iq abc
ω Lg
- +
θ idq~+ +
PI u q+ θ
+
iqref ugq+
2θ Sequence
Decoupling idref- ugd-
idq~- + u d-
-θ PI -θ
- +
id-
iabc - ω Lg vabc
abc/ dq-/ +
- -
dq- + idq iq abc
-ω Lg
+ - +
PI u q-
-
iqref
ugq-
iRe Pole/ iref0 Zero Sequence Current Control
iIm 0
θ u g0
+ 0
abc/ - PR + u 0/
0 i 0 abc
HC0
Fig. 5.26 Control structure for converter system with zero sequence current
It is noted that the voltage and current in zero sequence only contribute to the
real power p3Φ of the converter. Each part of (5.4) can be calculated as
3
P ¼ ðvþ i þ þ v
d id Þ
2 d d
3
Pc2 ¼ ðv i þ þ vþ
d id Þ
ð5:5Þ
2 d d
3
Ps2 ¼ ðv þ þ
d i q þ vd i q Þ
2
3
Q ¼ ðvþ þ
d i q vd i q Þ
2
3
Qc2 ¼ ðv þ þ
d i q vd i q Þ
ð5:6Þ
2
3
Qs2 ¼ ðv þ þ
d i d þ vd i d Þ
2
3
P0 ¼ ðv0Re i0Re Þ
2
3
P0c2 ¼ ðv0Re i0Re Þ ð5:7Þ
2
3
P0s2 ¼ ðv0Re i0Im Þ
2
where the zero sequence voltage and current are more like a single-phase AC
component at the fundamental frequency. They can be represented by the real part
and imaginary part as
v0Re ¼ V 0 cosðu0 Þ ¼ V 0
v0Im ¼ V 0 sinðu0 Þ ¼ 0
ð5:8Þ
i0Re ¼ I 0 cosðd0 Þ
i0Im ¼ I 0 sinðd0 Þ
It can be seen from (5.4) to (5.8) that if the three-phase AC source voltage is
decided, and then the converter has six controllable freedoms (i+d , i+q , i−d , i−q , iRe0, and
iIm0) to regulate the current flowing in AC source, which means six control
targets/functions can be established by the converter using the zero-sequence cur-
rent path. Normally, the three-phase average active and reactive powers delivered
by the converter are two basic requirements for a given application, and then two
control functions need to be settled first as
P3/ ¼ P þ P0 ¼ Pref
ð5:9Þ
Q3/ ¼ Q ¼ Qref
Then for the converter system in Fig. 5.25, there are four control freedoms left to
achieve two extra control targets than the traditional three-wire system. This feature
also means extended controllability and better performance under the unbalanced
AC source. Some of the new control strategies which can relive the stress of the DC
bus voltage or load current to the power switching devices under unbalanced AC
source are shown as follows.
A. Elimination of Active and Reactive Power Oscillations

Because of more current control freedoms, the power converter with zero-sequence
current path can not only eliminate the oscillation in the active power, but also
cancel the oscillation in the reactive power, and this control targets can be written as
Pc2 þ P0c2 ¼ 0 Qc2 ¼ 0

ð5:10Þ
Ps2 þ P0s2 ¼ 0 Qs2 ¼ 0
It can be seen that the power oscillations caused by zero sequence current P0c2
and P0s2 are used to compensate the power oscillations Pc2 and Ps2 caused by the
positive and negative sequence currents. Translating the control targets into (5.9)
and (5.10), all the controllable current components of the converter with
zero-sequence current path can be calculated as
2 Pref
iþ
d ¼ þ þ
3 ðvd vd Þ ð1 v

d =vd Þ
ð5:11Þ
2 Qref
iþ
q ¼
3 vþ 2 þ
d þ ðvd Þ =vd
v
i
d ¼
d
iþ
vþ d
d
ð5:12Þ
vd þ
i
q ¼ þ iq
vd
2 Pref P
i0Re ¼
3 v0Re
ð5:13Þ
vþ þ
d iq vd iq
i0Im ¼
v0Re
When applying the control targets in (5.9) and (5.10), the corresponding source
voltage, load current, sequence current amplitudes, and the instantaneous power
delivered by converter are shown in Fig. 5.27, where the grid voltage on phase A is
dipping to 0 p.u. It can be seen that by this control strategy, the twice fundamental
frequency oscillation of both active and reactive power can be totally eliminated.
Moreover, compared to all the control strategies for three-wire system, the ampli-
tude of load current in each phase is not further increased, and the current in the
faulty phase is smaller than the other two normal phases.
The current amplitude in different sequences and the delivered active/reactive
power with relation to the voltage amplitude on the dipping phase are shown in
Fig. 5.28a, b, respectively. It is noted that the converter has to deliver positive,
negative, and zero sequence currents to achieve this control strategy, and there is no
power oscillations for both the active and reactive powers under various voltage
dipping severities.
Fig. 5.27 Simulation of converter control with no active and reactive power oscillation
(three-phase converter with zero-sequence path, Pref = 1 p.u., Qref = 0 p.u., Ps2 = 0 p.u., Pc2 = 0 p.
u., Qs2 = 0 p.u., Qc2 = 0 p.u., VA = 0 p.u.)
(a) (b)
Current Amplitude (p.u.)
Izero
Q
Ipositive
Inegative
Voltage of the dip phase (p.u.) Voltage of the dip phase (p.u.)
Fig. 5.28 Profile of converter control with no active and reactive power oscillation (three-phase
converter with zero-sequence path, Pref = 1 p.u., Qref = 0 p.u., Ps2 = 0 p.u., Pc2 = 0 p.u., Qs2 = 0 p.
u., Qc2 = 0 p.u.). a Sequence current amplitude versus VA. b P and Q range versus VA
B. Elimination of Active Power Oscillation and Negative Sequence Current

Another promising control strategy for the converter with zero sequence current is
to eliminate the active power oscillation and negative sequence current. The extra
four control targets besides (5.9) can be written as
Pc2 þ P0c2 ¼ 0 i
d ¼0
ð5:14Þ
Ps2 þ P0s2 ¼ 0 iq ¼ 0
Translating the control targets into (5.9) and (5.14), all the controllable current
components of the converter with zero-sequence current path can be calculated as
2 Pref
iþ
d ¼
3 ðvþ
d vd Þ
ð5:15Þ
2 Qref
iþ
q ¼
3 vþd
i
d ¼0
ð5:16Þ
i
q ¼0
þ
v
d id
i0Re ¼
v0Re ð5:17Þ
i0Im ¼0
When applying the control targets in (5.9) and (5.14), the corresponding source
voltage, load current, sequence current amplitude, and the instantaneous power by
converter are shown in Fig. 5.29. It can be seen that by this control strategy, the
twice fundamental frequency oscillation of active power can be eliminated, and the
load current in the faulty phase is reduced to zero.
power with relation to the dipping phase voltage are shown in Fig. 5.30a, b,
respectively. It is noted that the converter has to deliver constant positive and zero
sequence currents to achieve this control strategy under different source voltage
dips. The fluctuation of reactive power is maintained in a smaller range (up to
±0.3 p.u.) compared to that in the three-wire system (up to ±1.3 p.u.).
Fig. 5.29 Simulation of

converter control with no
active power oscillation and
no negative sequence
(three-phase converter with
zero-sequence current path,
Pref = 1 p.u., Qref = 0 p.u.,
Ps2 = 0 p.u., Pc2 = 0 p.u.,
id- = 0 p.u., iq- = 0 p.u.,
VA = 0 p.u.)
(a) (b)
P

Qmax
Izero Ipositive
Qmin
Inegative
Fig. 5.30 Profile of converter control with no active power oscillation and no negative sequence
(three-phase converter with zero-sequence current path, Pref = 1 p.u., Qref = 0 p.u., Ps2 = 0 p.u.,
Pc2 = 0 p.u., i−d = 0 p.u., i−q = 0 p.u.). a Sequence current amplitude versus VA. b P and Q range
versus VA
C. Other Control Strategies

Another group of control strategies for the converter structure in Fig. 5.25 is to
eliminate the zero sequence current, as given in (5.18). Meanwhile, another two
arbitrary control targets can be added. This series control strategies will result in the
same control performances as the typical three-wire converter system, and it could
also be used under the normal AC source operation to limit the zero sequence
current.
i0Re ¼ 0
ð5:18Þ
i0Im ¼ 0
The converter stresses for the active/reactive power oscillations and the current
amplitude in the faulty/normal phases are compared in Table 5.2, where different
control strategies and converter structures are indicated, respectively. It can be seen
that by introducing the converter structures and controls with zero-sequence current
path, the power oscillations under unbalanced AC source are significantly reduced;
meanwhile, the current amplitude in the normal phases are not further stressed, and
the normal phases are not further stressed, and the current stress in the faulty phases
is significantly relieved.
D. Experimental Results
The control results by different converter structures and control strategies are val-
idated on a downscale DC-AC converter. As shown in Fig. 5.31, the circuit con-
figurations and setup photo are both illustrated. A three-phase two-level converter
with corresponding LCL filter is used to interconnect two DC voltage sources and a
programmable three-phase AC voltage source. The detailed parameters of the
experimental setup are shown in Table 5.3. It is noted that the converter is
Table 5.2 Converter stress comparison by different control strategies (values are represented
in p.u., Pref = 1 p.u., Qref = 0 p.u., VA = 0 p.u.)
Converter stress Typical 3-wire system 4 or 6-wire system
Control A Control B Control A Control B
Active power osc. Posc 0.5 0 0 0
Reactive power osc. Qosc 0.5 1.3 0 0.3
Current in faulty phase Ifault 1.5 3 1 0
Current in normal phase Inorm 1.5 2 2 2
Typical three-wire converter
Control A no negative sequence current
Control B no active power oscillation
Converter with zero-sequence current path
Control A no active and reactive power oscillations
Control B no active power oscillation and no negative seq. current
(a) (b)
Vdc /2
C1
DC+
Lf Lg ic
ib
N
ia
DC-
Rd
C2 va vb vc
Vdc/2
Cf N
Dip
Switch
Fig. 5.31 Configurations of experimental setup. a Circuit topology. b Setup photo
Table 5.3 Detail parameters DC bus voltage Vdc 700 V DC

of experimental setup
DC capacitance C1/C2 3300 µF
Nominal power Pnorm 5.5 kW
Nominal AC voltage Vnorm 311 V
Nominal load current Inorm 11.8 A
Fundamental frequency fo 50 Hz
Switching frequency fs 20 kHz
Filter inductance Lf 11 mH
AC source side inductance Lg 7.3 mH
Filter capacitance Cf 2.2 µF
Damping resistance Rd 3.5 Ω
Fig. 5.32 Experimental control performance of converter with only three wires (units are
normalized by parameters in Table 5.3, reference given: Pref = 0.5 p.u., Qref = 0 p.u., AC source
condition: amplitude of phase voltage VA = 0.1 p.u., VB = VC = 1 p.u.). a No negative sequence
current control. b No P oscillation control
controlled to operate at inverter mode, where the active power is flowing from DC
source to AC source. By opening and closing a switch shown in Fig. 5.31a, the
converter can be shifted between typical three-wire system and four-wire system
with zero-sequence current path. The amplitude of phase A voltage in the pro-
grammable AC source is adjusted to 0.1 p.u. (22 Vrms) in order to establish an
adverse unbalanced condition.
The control performance of the converter with three-wire structure is shown in
Fig. 5.32, where the given conditions and the very typical two control strategies are
applied, respectively. It can be seen that the experimental results agree well with the
analysis and simulation results, where either significant power oscillations or
over-loaded currents in the faulty phase are presented.
After enabling the zero-current path and proposed controls, the performances of
the given converter are shown again in Fig. 5.33, where the same conditions and
two control strategies mentioned in Figs. 5.27 and 5.29 are applied, respectively. It
can be seen that the experimental results also agree well with the simulation results,
where the power oscillations are much more reduced or even totally canceled;
meanwhile, the current stress in the faulty phase is significantly relived. These
critical performances are hard to be achieved by a single three-wire converter
structure using existing control strategies.
5.5 Summary 91
Fig. 5.33 Experimental control performance of converter after enabling zero-sequence current
path (units are normalized by parameters in Table 5.3, reference given: Pref = 0.5 p.u., Qref = 0 p.u.,
amplitude of phase voltage VA = 0.1 p.u., VB = VC = 1 p.u.). a No P and No Q oscillation control.
b No negative sequence current and No P oscillation control
5.5 Summary
Depending on the types and severity of grid faults as well as the corresponding
LVRT control behaviors, the operation conditions of grid connected power con-
verter like delivered power, phase angles, and modulation index are significantly
different compared to the normal operation. It should be noted that the power device
Tin, Dout, Din, and Dnpc under the LVRT operation of 3L-NPC converter may have
even higher junction temperature than the most stressed normal operation condition.
This should be taken into account when choosing the power devices and heat sink
system for the wind power converter.
According to the investigations, the thermal optimization target for 3L-NPC
wind power inverter under extreme LVRT is to reduce the junction temperature in
the NPC diode and inner switch. Compared to the normal modulation, the proposed
thermal-redistributed modulation sequences, which all enable full neutral point
potential control ability, can achieve more equal thermal distribution and reliving
the hottest power devices under extreme LVRT operation of 3L-NPC inverter. The
proposed thermally optimized modulation methods are especially feasible during
the LVRT operation, where the modulation index is relatively low, and more
redundant switching states can be utilized.
In the typical three-phase three-wire converter structure, there are four current
control freedoms, and it may not be enough to achieve satisfactory performances
under unbalanced AC source condition. In the three-phase converter structure with
zero-sequence current path, there are six current control freedoms. The extra two
control freedoms coming from the zero sequence current can be utilized to extend
the controllability of converter and improve the performance under unbalanced AC
source condition. By the proposed control strategies, it is possible to totally cancel
the oscillation in both the active and reactive powers, or reduced the oscillation
amplitude in the reactive power. Meanwhile, the current stress in the faulty phase is
also relived compared to the typical three-wire system.
Relevant attached special topics
Chapter 12.
Chapter 13.
Chapter 14.
References
1. Saccomando G, Svensson J, Sannino A (2002) Improving voltage disturbance rejection for

variable-speed wind turbines. IEEE Trans Energy Convers 17(3):422–428
2. Teodorescu R, Liserre M, Rodriguez P (2011) Grid converters for photovoltaic and wind
power systems. Wiley-IEEE Press, New York
3. Rodríguez P, Luna A, Muñoz-Aguilar R, Etxeberria-Otadui I, Teodorescu R, Blaabjerg F
(2012) A stationary reference frame grid synchronization system for three-phase
grid-connected power converters under adverse grid conditions. IEEE Trans Power Electron
27(1):99–112
4. Rodriguez P, Timbus AV, Teodorescu R, Liserre M, Blaabjerg F (2007) Flexible active power
control of distributed power generation systems during grid faults. IEEE Trans Ind Electron 54
(5):2583–2592
5. Song Hong-Seok, Nam Kwanghee (1999) Dual current control scheme for PWM converter
under unbalanced input voltage conditions. IEEE Trans Ind Electron 46(5):953–959
6. Muyeen SM, Takahashi R, Murata T, Tamura J (2010) A variable speed wind turbine control
strategy to meet wind farm grid code requirements. IEEE Trans Power Syst 25(1):331–340
7. Kaminski N, Kopta A (2011) Failure rates of HiPak Modules due to cosmic rays. ABB
application note 5SYA 2042-04, Mar 2011
8. Akagi H, Kanazawa Y, Nabae A (1984) Instantaneous reactive power compensators
comprising switching devices without energy storage components. IEEE Trans Ind Appl
IA-20(3):625–630
9. Semikron Application Note: 3L NPC & TNPC Topology, AN-11001, 2011
10. Busquets-Monge S, Bordonau J, Boroyevich D, Somavilla S (2004) The nearest three virtual
space vector PWM—a modulation for the comprehensive neutral-point balancing in the
three-level NPC inverter. IEEE Power Electron Lett 2(1):11–15
11. Busquets-Monge S, Bordonau J, Beristain JA (2006) Comparison of losses and thermal
performance of a three-level three-phase neutral-point-clamped DC-AC converter under a
conventional NTV and the NTV2 modulation strategies. In: Proceedings of IECON’ 2006,
pp 4819–4824
12. Ma K, Blaabjerg F (2012) Loss and thermal redistributed modulation methods for three-level
neutral-point-clamped wind power inverter undergoing low voltage ride through. IEEE Trans
Ind Electron (Also in Proceedings of ISIE’ 2012, pp 1880–1887, 2012)
13. Ng CH, Ran L, Bumby J (2008) Unbalanced-grid-fault ride-through control for a wind turbine
inverter. IEEE Trans Ind Appl 44(3):845–856
References 93
14. Miret J, Castilla M, Camacho A, Vicuña L, Matas J (2012) Control scheme for photovoltaic
three-phase inverters to minimize peak currents during unbalanced grid-voltage sags. IEEE
15. González-Espín F, Garcerá G, Patrao I, Figueres E (2012) An adaptive control system for
three-phase photovoltaic inverters working in a polluted and variable frequency electric grid.
IEEE Trans Power Electron 27(10):4248–4261
16. Ma K, Blaabjerg F, Liserre M (2013) Power controllability of three-phase converter with
unbalance AC source. In: Proceedings of APEC’ 2013
17. Ma K, Blaabjerg F, Liserre M (2011) Thermal analysis of multilevel grid side converters for
10 MW wind turbines under low voltage ride through. IEEE Trans Ind Appl (Also in
Proceedings of ECCE’ 2011, pp 2117–2124, 2011)
Chapter 6
Conclusions and Future Works
6.1 Conclusions
From this book, it can be concluded that power electronics will play more important
role for the next generation of wind turbine systems. In this case, the thermal stress
in the power semiconductors is a critical performance parameter because the con-
verters need to carry all the generated power from wind turbines with very high
power density. It is found that the thermal behaviors in wind turbine system could
be rather adverse under either normal or faults conditions. On the other hand, it is
also possible to improve the thermal behaviors of wind power converter by multiple
methods like the control, modulation, modeling, and converter designs. The
detailed conclusions are given as follows in several aspects:
Topologies and device for wind power converter
It is becoming more and more difficult for the traditional two-level converter
topologies to achieve acceptable performances in the wind power application.
Thereby, some multi-level and multi-cell converters with higher power handling
ability are becoming promising for the next generation wind turbines. Regarding to
the power semiconductor devices, press-packing IGCT and IGBT show significant
improvement in respect to the thermal resistance and power density compared to the
module packaging devices. Due to the rating limits on the market, paralleled
connection of devices may be needed in order to achieve the required amount of
power for the next generation wind turbines—this will modify the loss/thermal
behaviors of device as well as the overall power density of converter.
Criteria and tools for the analysis of wind power converter
Due to the growing power and limited space in the wind power application, the
thermal loading of the wind power converter is becoming significant at multi-MW
level of power conversion. It has been calculated in this work that the thermal stress
of power devices has close relationship with the reliability and cost of the converter.

96 6 Conclusions and Future Works
Therefore, thermal stress analysis is crucial important for the next generation wind
power converter system.
The loading of wind power converter is influenced by many factors which may
involve multidisciplinary approaches of analysis under various time constants.
Thereby, a framework with multi-domain models of the wind turbine system is
established for the stress analysis. In this framework, the factors of mission profile,
converter design, and converter controls are taken into account and they can be
translated into the corresponding stress profile in the power semiconductor devices.
Electrical and thermal stresses under normal operation
The thermal profile of the power switching devices under steady-state wind speeds
is an important tool which can be used either as a loading indicator for certain
converter topologies or as a lookup table for the lifetime estimation. It was found
that the wind speed variations will lead to severe thermal excursion of some power
switching devices in the given 3L-NPC wind power converter. Meanwhile, the grid
codes even under normal operation may change the delivered reactive power of the
wind power converters and thereby also have impacts to the converter efficiency as
well as thermal stress of the power devices.
By circulating the reactive power among paralleled converters in a wind farm or
multi-cell converter system, it is possible to control the junction temperature and
relieve the thermal excursion under wind gust operation, leading to higher reli-
ability of the converter, while the increased thermal stresses to the other devices or
paralleled converters are still acceptable.
Electrical and thermal stresses under fault condition
Depending on the types and severity of grid faults as well as corresponding LVRT
control behaviors, the operating condition of the grid connected power converter is
significantly different compared to the normal operating condition. It should be
noted that some power devices under the LVRT operation of 3L-NPC converter
may have even higher junction temperature than the most stressed normal
operation.
According to the investigations, the thermal optimization target for 3L-NPC
wind power inverter under extreme LVRT is to reduce the junction temperature in
the NPC diode and inner switch. The proposed thermal redistributed modulation
sequences, which all enable full neutral point potential control ability, can achieve
more equal thermal distribution and relieving the hottest power devices under
extreme LVRT operation of 3L-NPC inverter. The proposed thermal optimized
modulation methods are especially feasible during the LVRT operation, where the
modulation index is relative low and more redundant switching states can be
utilized.
In the typical three-phase three-wire converter structure, the control freedom
may be not enough to achieve satisfactory performances under unbalanced AC
source condition. However, in the converter structure with zero sequence current
path (4 or 6 wire system), two extra control freedoms coming from the zero
sequence current can be utilized to extend the controllability of converter and to
improve the performance under unbalanced AC source. By the proposed control
6.1 Conclusions 97
strategies, it is possible to totally cancel the oscillation of both the active and
reactive power, or to reduce the oscillation amplitude in the reactive power.
Meanwhile, the current stress in the faulty phase is also relived compared to the
typical three-wire system.
6.2 Proposals for Future Research Topics
1. Impacts of mechanical parts to the thermal loading of wind power converter

• Realization and simplification of pitch and speed control of wind turbines in
order to analyze the generator side converter.
• Thermal stress and lifetime with long-term wind profiles and different
roughness classes of wind condition.
• Improved design and control methods for thermal stress improvement and
lifetime extension of wind power converter.
2. Advanced analyzing tools for wind power converter
• Lifetime evaluation tools considering long-term mission profile.
• Thermal-oriented design of converters.
• Thermal analysis considering device and loading variations/distributions.
• Stress benchmarking of different converter configurations.
• Grid code influences to the reliability of converter.
3. Zero-sequence current-control method of wind power converter.
• Thermal stress analysis of power control strategies with zero sequence
current.
• Experimental validation in respect to feasibility.
• Other improved control strategies under grid faults.
4. Experimental validations of thermal stress.
• Evaluation of different temperature measurement methods.
• Validation of proposed thermal model for IGBT module.
• LVRT modulation methods in respect to thermal performances.
• Zero-sequence current-control methods in respect to thermal performances.
• Thermal distribution analysis inside one IGBT module.
Chapter 7
Appendix
7.1 Used Models for Analysis
7.1.1 Wind Speed Generator
The used wind speed model in this monograph has been developed at RISØ
National Laboratory based on the Kaimal spectra. The rotor speed and the desired
average wind speed are the two input parameters for the model. Then the wind
speed is calculated as an average value over the whole rotor, and the tower shadow
and the rotational turbulences are taken into account. A generated wind speed
variation is shown in Fig. 7.1. This speed profile corresponds to a Class A turbu-
lence and three wind speed variations within 0.5 h with an average value of 7, 11
and 23 m/s can be generated respectively.
7.1.2 Wind Turbine Model
The wind turbine model transfers the wind speed into the mechanical power for the
generator. One easy and accurate method to model the wind turbines is to look up
the power curve provided by the turbine manufacturer. In these power curves the
output mechanical power of wind turbines in relation to the wind speeds are nor-
mally defined. Figure 7.2 is an example for the designed power curve for 10 MW
wind turbine provided by American Superconductor Corporation (AMSC).
Another way to model the wind turbines is to use the general and analytical
functions:
1
Po ¼ qpR2 Cp ðb; xr ; vw Þ ð7:1Þ
2

100 7 Appendix
Fig. 7.1 Generated medium

term wind speed variations
(0.5 h)
Fig. 7.2 Generated

medium-term wind speed
variations (0.5 h)
where the air density ρ = 1.225 kg/m3, turbine radius R = 85 m. The power
coefficient Cp which is related to the pitch angle β, rotational speed ωr, and wind
speed vw can be written as

116 21
Cp ðb; xr ; vw Þ ¼ 0:5176 0:4b 5 e X þ 0:0035kðxr ; vw Þ ð7:2Þ
X
where

1 0:035
X ¼ 1= and k ¼ xr R=vw ð7:3Þ
k þ 0:08b b3 þ 1
7.1 Used Models for Analysis 101
If the wind turbine contains a gear box for the generator, then the drive train model
which converts the rotational speed and mechanical torque has also to be integrated.
7.1.3 Generator Model
The generator model transfers the mechanical torque and rotational speed from
wind turbines to the corresponding electrical power flowing into the power elec-
tronic converter. The generator model can be acquired from the machine manu-
facturer or by the general design laws. As an example, the parameters of a 10 MW
permanent magnet synchronous generator with one-stage gear box (PMSG_1G) and
direct (PMSG_DD) drive are shown in Table 7.1. It is noted that the gear box ratio
R for PMSG_1G is designed at 9.7 (resulting in 120 Hz rated electrical frequency)
in order to achieve optimal tradeoff between the generator and the gear box cost.
7.1.4 Parameter for Thermal Impedance of Used IGCT
The thermal models of used IGCT for both switch and freewheeling diode are
indicated in Fig. 7.3, in which the thermal impedance from junction to case Z(j–c) is
modeled as a multi-layer Foster RC network.
Table 7.1 Parameters for Generator type PMSG_DD PMSG_1G

10 MW PMSG
Rated wind speed vw (m/s) 11.7
Rated rotor speed nr (rpm) 10
Air gap diameter D (m) 10 4.2
Stator length l (m) 1.8 0.9
Number of pole pairs Np 165 74
Gear box ratio R – 9.7
Rated shaft speed ns (rpm) 10 97
Rated electrical frequency fe 27.5 120
(Hz)
No load-induced voltage Ep (V 3748/1874 3600/1800
rms)
Synchronous inductance Ls 11/2.76 3.16/0.79
(mH)
102 7 Appendix
Table 7.2 Parameters of internal thermal impedance (5SHY 40L4511 and 5SDF 10H4503)
Thermal ZT/D(j–c) ZT/D(c–h)
impedance Sector 1a Sector 2 Sector 3 Sector 4
RiIGCT (K/kW) 5.562 1.527 0.868 0.545 3
τiIGCT (s) 0.5119 0.896 0.0091 0.0024 –
RiDiode (K/kW) 7.705 2.748 1.009 0.539 3
τiDiode (s) 0.5244 0.0633 0.0065 0.0015 –
a
Sector N means different layers of thermal RC lump in Fig. 7.3
Sector 1 Sector 2 Sector 3 Sector 4
Tj Rth1 Rth2 Rth3 Rth4
Cth1 Cth2 Cth3 Cth4
TA ZT/D(j-c)
TC
Fig. 7.3 Thermal model of the impedance ZT(j − c) or ZD(j − c) from junction to case
Fig. 7.4 Schemes of experimental setup for 3L-NPC converter
Each of the thermal parameters can be found from the manufacturer datasheets
and they are summarized in Table 7.2, where the thermal resistance Rth will decide
the steady-state level of junction temperature, and the time constant τ (decided by
Rth and thermal capacitance Cth) will decide the dynamic performance of the
junction temperature.
7.2 Experimental Setup 103
Fig. 7.5 Photo of

experimental setup for
3L-NPC converter
Table 7.3 Parameters for the DC bus voltage Vdc 200 V

experimental setup of
3L-NPC converter Modulation index m 0.5
Switching frequency fs 2 kHz
Fundamental frequency fo 50 Hz
Load inductance Lload 28 mH
Load resistance Rload 1.1 Ω
Maximum load current iloadM 3A
Phase angle θ 80º
7.2 Experimental Setup
The proposed modulation sequences are validated on a downscale 3L-NPC con-

verter controlled by dSPACE system, as shown in Figs. 7.4 and 7.5, whose
parameters are indicated in Table 7.3. It can be seen that, a special RL passive load
is used to simulate the LVRT operating condition for the grid-connected converter
with large amount of reactive power (θ = 80º) and low modulation index (m = 0.5).
Part II
Specially Selected Topics
Chapter 8
The Impacts of Power Switching Devices
to the Thermal Performances of 10 MW
Wind Power NPC Converter
8.1 Wind Power Converter for Case Study
It is expected that 10 MW wind turbines with full-scale power converter will be the
next long-term target to be conquered according to the technology trends. In most
cases, the multi-level converter topologies will be demanded to handle such a high
power with medium voltage ratings. As the most commercialized multilevel con-
verter, three-level neutral-point-clamped (3L-NPC) topology seems to be a prom-
ising candidate, as shown in Fig. 8.1, where Tout represents the outer switch, Dout is
the outer freewheeling diode, Tin is the inner switch, Din is the inner freewheeling
diode, and Dnpc is the clamping diode. This configuration is selected for analysis in
this special topic.
A 10 MW medium-voltage 3L-NPC wind power inverter is first designed for a
case study. As summarized in Table 8.1, all of the power devices have the com-
mutated voltage of 2.8 kV in order to utilize the dominant 4.5 kV high-power
semiconductors available on the market, and the rated DC bus voltage can be
determined at 5.6 kV. Normal phase-disposition sinusoidal-pulse-width-modulation
method for 3L-NPC converter is applied, and the carrier frequency is typically
designed to be 800 Hz in order to reach an acceptable switching loss in the power
switching devices. The output filter inductance is designed to limit the maximum
current ripple to 25 % of the rated current amplitude, and the filter capacitance is not
taken into account. For simplicity of analysis and to keep the analytical focus on the
power loss and thermal behavior of the power semiconductors, the power grid is
considered as three ideal AC voltage sources and the DC bus capacitance is assumed
high. The used generator and wind turbine models can be found from [1, 2].
Three power switching device solutions by using IGCT press-pack, IGBT
press-pack, and IGBT module for the given 10 MW 3L-NPC wind power converter
This chapter is co-authored by F. Blaabjerg.

108 8 The Impacts of Power Switching Devices …
Generator Tout Dout Transformer

Dnpc
Tin Din
Power
Iload
Filter Filter Grid
3L-NPC rectifier 3L-NPC inverter
Fig. 8.1 A back-to-back three level neutral-point-clamped wind power converter system
Table 8.1 Rated parameters Rated output active power Po 10 MW

of 10 MW 3L-NPC wind
DC bus voltage Vdc 5.6 kV DC
power inverter for case study a
Rated primary side voltage Vp 3.3 kV rms
Rated load current Iphase 1750 A rms
a
are listed in Table 8.2, where the device ratings, configurations, total device
numbers, and used device models are indicated respectively [3, 4]. It is noted that in
order to justify and facilitate the comparison, each device solution is designed to
handle the same load current up to 3.6 kA, thereby two 1.8 kA IGBT press-packs
and three 1.2 kA IGBT modules have to be paralleled, due to the limits of maxi-
mum available current rating for the corresponding power switching devices.
As it can be seen, in order to achieve 10 MW power conversion, the power
semiconductors needed for the three device solutions are quite different: IGBT
module solution consumes up to 80 % more device numbers than the other two
press-pack solutions, this is mainly because of the maximum current rating limits
for a single IGBT module. When taking into account the extra auxiliary parts like
drives, fans, heat sink, wires, etc., the IGBT module solution may result in much
lower power density and higher component counts, which may be unpreferable in
the wind power application.
8.2 Thermal-Related Characteristics of Different Power

Switching Devices
The junction temperature of power switching devices is decided by the losses

during switching and conducting, as well as the thermal impedance from junction to
the ambient [5–7]. Therefore, the characteristics related to the power loss and
8.2 Thermal-Related Characteristics of Different Power Switching Devices 109
Table 8.2 Three power switching device solutions for 10 MW 3L-NPC converter
Solutions Main switch (each) Clamping diode (each) Device numbers
IGCT 1 × IGCT (3.6 kA/4.5 kV) 1 × (3.6 kA/4.5 kV) Diode 30
press-pack 1 × Diode (3.6 kA/4.5 kV)
IGBT 2 × IGBT (1.8 kA/4.5 kV) 2 × (1.8 kA/4.5 kV) Diode 36
press-pack Integrated Diode
IGBT module 3 × IGBT (1.2 kA/4.5 kV) 3 × (1.2 kA/4.5 kV) Diode 54
Integrated Diode
IGCT press-pack ABB 40L4511 with recommended freewheeling diode 10H4503
IGBT press-pack Westcode T1800GB45A with integrated freewheeling diode
IGBT module ABB 1200G450350 with integrated freewheeling diode
thermal impedance of power switching devices in Table 8.2 are going to be eval-
uated first.
8.2.1 Switching Loss
The switching loss of the switch (i.e., IGBT or IGCT) is generated when it is
turning on and turning off (Eon and Eoff), while the majority switching loss of the
diode is generated when it is turning off (Err). As an indicator for the switching loss
characteristic, the consumed energy during the switching process (in the unit of
Joule) in relation to the current flowing in the switch (Ic) or diode (IF) can be
derived from the datasheets of manufacturers.
The switching loss profiles of IGCT/IGBT (Eon + Eoff) and diode (Err) for the
designed power device solutions are compared in Fig. 8.2a, b respectively. It is
noted that, in order to unify and compare solutions with different numbers of
parallel devices, the vertical axis in Fig. 8.2 represents the switching loss for only
one switch/diode in the parallel device solutions, and the horizontal axis is
(a) (b)
Switching loss on Diodes Err (J)
Switching loss on switches
20 20
18 18
16 IGBT Presspack 16
14 14
Eon+Eoff (J)
IGCT
12 12
Diode for IGCT
10 10
8 8
Diode for IGBT
6 IGBT Module
6 Diode for IGBT Presspack
4 4 Module
2 2
0 0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
Load current Iload (A) Load current Iload (A)
Fig. 8.2 Normalized switching loss profiles for different power devices (recommended test
conditions, 125 °C). a Switching loss on IGCT/IGBT Eon + Eoff. b Switching loss on diode Err
(a) 12 (b) 12
IGCT
Turn off loss Eoff (J)

Turn on loss Eon (J)
10 IGBT
10
Presspack IGBT
8 8 Presspack
6 6
IGBT Module
4 4 IGBT
Module
2 2
IGCT
0 0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
Fig. 8.3 Normalized switching loss profiles for switches (recommended test conditions, 125 °C).
a Turn on switching loss Eon. b Turn off switching loss Eoff
normalized by the load current Iload of the 3L-NPC converter rather than current
flowing in the devices (Ic or IF).
It can be seen that for the switches (i.e., IGCT or IGBT), IGBT press-pack has
larger total switching loss Eon + Eoff compared to the IGCT press-pack and IGBT
module. As for the freewheeling diodes, the IGCT solution shows significantly
larger Err compared to the other two IGBT solutions based on press-pack and
module packaging technology. This is maybe because the normally used clamping
circuit for the IGCT will significantly slow down the rising time of load current
during the turning on process, and thereby result in very large reverse recovery
charge Qrr on the freewheeling diodes. As for the IGBT module, it shows the lowest
overall switching loss either in the switches or in the freewheeling diodes.
When investigating the Eon and Eoff on switch separately, as shown in Fig. 8.3, it
can be found that the large switching loss of IGBT press-pack comes from the
larger turn on loss Eon. Due to the use of clamping circuit, IGCT can achieve
smaller turn on loss Eon, but with the cost of larger switching loss Err in the
freewheeling diode, as indicated in Fig. 8.2b, and as can be seen from Fig. 8.3b
IGCT shows a large turn off loss Eoff in the switch.
8.2.2 Conduction Voltage and Loss
The conduction loss of the power switching devices is generated when the switch or
diode is conducting load current. As an important indicator for the conduction loss
characteristic, the conduction voltage of IGCT/IGBT vce or diode vF in relation to
the current flowing in switch (Ic) or diode (IF) can be also derived from the data-
sheets of manufacturers.
The profiles of vce and vF for the chosen power switching devices with relation to
the load current of the 3L-NPC inverter Iload are compared in Fig. 8.4a, b respec-
tively. It can be seen that the IGCT and its freewheeling diode show lower con-
duction voltage compared to the other two IGBT solutions.
8.2 Thermal-Related Characteristics of Different Power Switching Devices 111
(a) 4.0 (b)

4.0
Conduction voltage Vce (V)
Conduction voltage VF (V)

3.5 3.5
3.0 3.0
2.5 2.5
2.0 2.0
IGCT IGCT
1.5 1.5
IGBT P IGBT P
1.0 1.0
IGBT M IGBT M
0.5 0.5
0.0 0.0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
Fig. 8.4 Normalized conduction voltage for different power devices (recommended test
conditions, 125 °C). a Switches. b Diode
However, different from the switching loss characteristic in Fig. 8.5, the con-
duction voltage does not directly reflect the conduction loss characteristic of the
power switching devices because the current flowing in switch Ic and diode IF have
to be also taken into account. In order to better compare the conduction loss
performance among different device solutions, a series new profiles are plotted in
Fig. 8.5, where the vertical axis is changed to Vce*Ic for switches and VF*IF for
diodes. These profiles contain the information of both conduction voltage and
switching current, and thereby can directly reflect the conduction loss level of
different device solutions. It is noted that the conduction loss characteristics is only
for one single device in parallel device solutions.
As it can be seen, although the IGCT solution has the lowest conduction voltage
level, it shows the largest conduction loss profile for each device because of much
larger current flowing in each device (the current is 3 times of that in each IGBT
(a) 10 (b) 10 IGCT

9 IGCT 9
IGBT P IGBT P
8 8
IGBT M IGBT M
7 7
Vce × Ic (kW)
VF × IF (kW)
6 6
5 5
4 4
3 3
2 2
1 1
0 0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
Fig. 8.5 Normalized conduction loss profiles for different power devices (recommended test
conditions, 125 °C). a Switches. b Diode
module and 2 times of that in each IGBT press-pack as indicated in Table 8.2).
Again, the IGBT module solution shows overall lower conduction loss level in each
switching device either in the switch or in the diode.
8.2.3 Thermal Resistance
The thermal resistance is another important characteristic which can determine the
thermal performances of power switching devices. The thermal resistance from
junction to heat sink of each device solution is shown in Fig. 8.6, where the
switches and diodes are indicated respectively.
It can be seen that, the thermal resistance of power switching device is closely
related to the packaging technology: the press-pack devices IGCT and IGBT have
significantly smaller thermal resistance both in the switches and in the freewheeling
diodes than the module packaging device “IGBT M”. The freewheeling diodes in
all of the three device solutions have significant larger thermal resistance than the
corresponding IGBT or IGCT. It is noted that, the major thermal resistance dif-
ference between the press-pack and module packaging devices comes from the
thermal resistance outside the devices, i.e., from case to heat sink Rth(CH), which is
much larger for the IGBT module.
8.3 Thermal Analysis of Different Device Solutions
The used thermal models of a single switch and clamping diode are indicated in
Fig. 8.7 in which the thermal impedance from junction to case Z(j−c) is modeled as a
multi-layer Foster RC network. Each of the thermal parameters can be found from
40 Rth(CH)
Thermal resistance (K/kW)
Rth(JC)
35
30
18
25
20
Rth(JH)
15 3 9
10 3 Rth(JH) 19
14.4
12
5 8.5 7.3 9.5
0
Switch Diode Switch Diode Switch Diode
IGCT IGBT P IGBT M
Fig. 8.6 Thermal resistance from junction to heat sink for different power devices. Rth(CH) and
Rth(JC) represent the thermal resistance from junction to case and case to heat sink respectively
8.3 Thermal Analysis of Different Device Solutions 113
Switch Clamped Diode

Tj Tj Tj
IGCT Diode Diode Rth1 RthN
Tj
ZT(j-c) ZD(j-c) ZD(j-c)
TA TA TA τ1 τN
TC TC TC ZT/D(j-c)
TA TC
RT(c-h) RD(c-h) RD(c-h)
TH
Thermal model of the impedance ZT(j-c)
TH
or ZD(j-c) from junction to case.
Note:
Tj: junction temperature, TC: case temperature, TH: heat sink temperature, TA: ambient temperature
Z(j-c): thermal impedance from junction to case, Z(c-h): thermal impedance from case to heat sink
Fig. 8.7 Thermal model for the analysis and simulation of 3L-NPC converter
the manufacturer datasheets, where the thermal resistance Rth will decide the
steady-state mean value of junction temperature, and the thermal capacitance (with
time constant τ) will decide the dynamic change or fluctuation of the junction
temperature.
It is noted that, normally the IGBT manufacturer will only provide thermal
parameters inside IGBT modules with Foster RC network, in order to establish the
complete thermal models from junction to the ambient, the thermal impedance of
ZT/D(j−c) has to be transferred to the equivalent Cauer RC network to facilitate the
thermal impedance extension outside IGBT modules. Because the temperature of
the heat sink TH is normally much lower and more stable compared to the junction
temperature Tj in a properly designed converter system, so the heat sink temperature
is considered as a constant value at 60 °C in this special topic. However, the heat
sink temperature may strongly depend on the operation site and the design of the
heat sink system.
After the parameters for the converter as well as the loss and thermal models for
the power switching devices are settled, some important operation modes of the
3L-NPC converter can be simulated. The simulations are carried out based on
PLECS blockset in Simulink, and the simulation parameters are consistent with the
ones in Tables 8.1 and 8.2, the main focus will be on the grid connected inverter.
8.3.1 Normal Operation
The thermal cycling performance of the 3L-NPC wind power inverter when the
wind turbine is running at rated wind speed 12 m/s and rated grid voltage will first
be analyzed. Figure 8.8 shows the converter output voltage pulses, phase current,
and grid voltage under rated and normal condition of the wind turbines. It can be
seen that the load current is in phase with the grid voltage, i.e., power factor PF = 1.
Fig. 8.8 Simulation outputs of 3L-NPC inverter under rated normal operation (output voltage
pulses-green, grid voltage-blue, phase current-red, Vg = 1 p.u., PF = 1, vw = 12 m/s, DC bus
voltage VDC = 1 p.u.)
8.3.1.1 Loss Distribution
The loss calculation method is a commonly accepted method for the loss evaluation
of power semiconductor devices. It is noted that the switching loss profile of the
chosen switching devices only have the test condition at 125 °C on datasheet,
therefore, the loss models in this special topic are considered temperature inde-
pendent during the simulation. However, if the device characteristics under other
temperatures are provided, the loss model with junction temperature dependence
can be established and simulated by iteration in the simulation software.
The loss distribution of the 3L-NPC converter with different device solutions
under normal rated condition is shown in Fig. 8.9, where only one single device is
indicated if parallel power device solutions are applied.
As it can be seen from Fig. 8.9, the major losses for all of the device solutions
are consumed by the outer switches Tout, inner switches Tin and the clamping diodes
4.5
Lcon
4.0
Lsw
Loss pr. device (kW)
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
IGCT IGBT IGBT IGCT IGBT IGBT IGCT IGBT IGBT IGCT IGBT IGBT IGCT IGBT IGBT
P M P M P M P M P M
Tout Dout Tin Din Dnpc
Fig. 8.9 Loss distribution for different power switching devices under normal operation of
Fig. 8.8 (Lcon means conduction loss-red, Lsw means switching loss-blue)
Dnpc under normal operation of the wind power converter. The IGCT solution
shows significantly larger loss in Dnpc due to larger switching loss, on the other
hand, the IGBT module solution shows significantly lower overall loss level
especially in Tout.
8.3.1.2 Thermal Performances
With the thermal model in Fig. 8.7 and loss information in Fig. 8.9, the junction
temperature for the switching devices in the given 3L-NPC inverter can be also
simulated, as shown in Fig. 8.10, where only one single device is indicated if the
parallel power device solutions are applied. It can be seen that the thermal distri-
bution with different device solutions are quite different.
The junction temperature mean value Tjm and temperature fluctuation ΔTj for
each of the switching device by different device solutions in Fig. 8.11 are sum-
marized in Fig. 8.12. Obviously, the IGCT solution has larger Tjm and ΔTj in almost
all of the switching devices, especially the clamping diode Dnpc. For the IGBT
press-pack solution, it shows the lowest Tjm level but relatively higher ΔTj level.
The IGBT module solution does not achieve the best thermal performance with the
lowest power losses because of much larger thermal resistance.
8.3.2 Low-Voltage-Ride-Through Operation
The simulation output of the 3L-NPC inverter undergoing low voltage rid through
(LVRT) operation is shown in Fig. 8.13, in which the grid voltage is set to 0.05 p.u.
of the rated value as a severe voltage dip, and the converter has to provide 100 %
rated reactive current according to the grid codes. It can be seen that the phase
current lags 90° to the grid phase voltage, and the inverter achieves 100 % rated
reactive current injection (1.75 kA rms) into the power grid.
8.3.2.1 Loss Distribution
The loss distribution of the 3L-NPC inverter undergoing the given LVRT operation
is shown in Fig. 8.14, where different device solutions are indicated and compared.
It can be seen that, the loss distribution is quite different from the one undergoing
normal operation in Fig. 8.10, Tin has the highest power loss for all of the device
solutions. IGCT solution again shows significantly larger loss in Dnpc, while IGBT
module solution has the lowest power loss level.
Fig. 8.10 Thermal cycling of

the 3L-NPC inverter with
different switching device
solutions (normal operation
condition in Fig. 8.9). a IGCT
presspack solution. b IGBT
press-pack solution. c IGBT
module solution
(a) (b)
Temperature fluctuation ΔTj (K)

)
120 18
IGCT IGBT P IGBT M IGCT IGBT P IGBT M
Mean temperature Tjm (
16
110
14
100 12
10
90
8
80 6
4
70
2
60 0
Tout Dout Tin Din Dnpc Tout Dout Tin Din Dnpc
Fig. 8.11 Thermal cycling profile of the 3L-NPC inverter with different switching device
solutions under normal operation (Vg = 1 p.u., PF = 1, vw = 12 m/s). a Mean junction temperature
Tjm. b Junction temperature fluctuation ΔTj
Fig. 8.12 Simulation outputs of 3L-NPC inverter under LVRT operation (output voltage pulses-
green, grid voltage-blue, phase current-red, Vg = 0.05 p.u., 100 % rated reactive current, DC bus
voltage VDC = 1.1 p.u.)
4.5
Lcon
4.0
Lsw
Loss pr. device (kW)
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
IGCT IGBT IGBT IGCT IGBT IGBT IGCT IGBT IGBT IGCT IGBT IGBT IGCT IGBT IGBT
P M P M P M P M P M
Fig. 8.13 Loss distribution for different power switching device solutions under LVRT operation
of Fig. 8.13 (Lcon means conduction loss-red, Lsw means switching loss-blue)
Fig. 8.14 Thermal cycling of

the 3L-NPC inverter with
different switching device
solutions during
Low-Voltage-Ride-Through
operation condition in
Fig. 8.13. a IGCT press-pack
solution. b IGBT press-pack
solution. c IGBT module
solution
(a) (b)
Temperature fluctuation ΔTj (K)

)
120 14
Mean temperature Tjm (
IGCT IGBT P IGBT M IGCT IGBT P IGBT M

110 12
10
100
8
90
6
80
4
70 2
60 0
Tout Dout Tin Din Dnpc Tout Dout Tin Din Dnpc
Fig. 8.15 Thermal cycling profile of the 3L-NPC inverter with different switching device
solutions under low-voltage-ride-through operation (Vg = 0.05 p.u., 100 % rated reactive current).
a Mean junction temperature Tjm. b Junction temperature fluctuation ΔTj
The junction temperature for each of the power switching device in the given
3L-NPC inverter can be also simulated under LVRT operation, as shown in
Fig. 8.15, where only one single device is indicated if parallel power device
solutions are applied. It can be seen that the thermal distribution with different
device solutions are still quite different from each other under LVRT operation of
wind turbines.
The junction temperature mean value Tjm and temperature fluctuation ΔTj for
each of the switching devices by different device solutions in Fig. 8.15 are sum-
marized in Fig. 8.16. It can be seen that, again the IGCT solution has larger Tjm and
ΔTj in almost all of the switching devices, especially for the clamping diode Dnpc.
The IGBT press-pack solution shows the lower Tjm but higher ΔTj compared to the
other two device solutions.
Wind speed (m/s)
Fig. 8.16 Wind speed and

current references of grid side
3L-NPC inverter in a defined
wind gust
Current references (A)
Active current IP
Reactive current IQ
Time (s)
(a) (b)
(c) (d)
35
Max fluctuation ΔTj (K)

IGCT IGBT P IGBT M
30
25
20
15
10
5
0
Fig. 8.17 Thermal cycling of the 3L-NPC inverter with different switching device solutions (wind
gust operation in Fig. 8.13). a IGCT press-pack solution. b IGBT press-pack solution. c IGBT
module solution. d Junction temperature fluctuation ΔTj
8.3.3 Wind Gust Operation
The wind gust operation of wind turbines may impose power switching devices
with large thermal cycling amplitude and longer time disturbances compared to the
LVRT and normal operation modes. According to the one-year return period wind
gust definition by IEC standards, a wind gust condition for a case study is indicated
in Fig. 8.17, where the wind speed is set from 10 m/s dropping to 8 m/s and rising
to 16 m/s, then the reverse fashion is continued.
The junction temperature distributions of the target 3L-NPC inverter during the
given wind gust operation condition are indicated in Fig. 8.17. It is obvious that,
three device solutions show significant differences in the most stressed devices,
temperature fluctuation amplitude, and thermal distribution.
The maximum junction temperature fluctuation amplitude with different device
solutions are summarized in Fig. 8.17d, it can be seen that the temperature in Tout is
Table 8.3 Most stressed Solutions Normal LVRT Wind Gust

devices of different power
IGCT press-pack Tout & Dnpc Dnpc & Tin Dnpc & Tout
device solutions
IGBT press-pack Tout Tin Tout
IGBT module Tout Dnpc & Tin Tout
Table 8.4 Least stressed Solutions Normal LVRT Wind Gust

devices of different power
device solutions IGCT press-pack Dout & Din Din Dout & Din
IGBT press-pack Dout & Din Din Dout & Din
IGBT module Dout & Din Din Dout & Din
the most fluctuated device in all of the proposed solutions, and IGCT solution
shows more temperature fluctuation amplitude in Tin and Dnpc under the given wind
gust defined by the IEC standards.
8.3.4 Summary of Thermal Performances Under Different

Operation Modes
Finally, both the most stressed and the least stressed devices of 3L-NPC inverter
with different device solutions are summarized in Tables 8.3 and 8.4 respectively,
where the defined three operation modes of wind turbines are included.
It is interesting to see that for all of the power switching device solutions, the
outer switches Tout, clamping diodes Dnpc, and inner switch Tin are likely to become
the most stressed devices under various operations of wind turbines; while the outer
diode Dout and inner diode Din are barely used with all of the device solutions under
various operation modes. This information may be used to guide the design and
selection of power device for 3L-NPC wind power converter.
8.4 Conclusions
The thermal performance of power switching devices is important for the modern
wind power converter system. It is found that, the thermal-related characteristics of
the three dominant power switching devices in wind power applications are quite
different.
For all of the power switching device solutions in the 3L-NPC grid side inverter,
the outer switches Tout, clamping diodes Dnpc and inner switch Tin are likely to
become the most stressed devices under various important operation modes of the
wind turbines, therefore, they are critical components for the 3L-NPC wind power
converter; while the performance of outer freewheeling diode Dout and inner
freewheeling diode Din is less important because they are barely used.
Regarding the most stressed devices Tout, Tin and Dnpc, three device solutions
show quite different loading behaviors because of different power loss and thermal
impedance characteristics: for the IGCT solution, the switching loss in the diode is
much larger because of the normally used clamping circuit, this disadvantage may
lead to much higher junction temperature in Dnpc in comparison with the other two
IGBT solutions. The IGBT press-pack solution tends to have larger switching loss
in the switch, however, because of smaller thermal resistance, it shows similar
junction temperature level in the switches as the IGCT solution, but with much
better thermal performance in the clamping diode. IGBT module solution shows the
best loss performances among the three device solutions, but due to much large
thermal resistance, especially from case to heat sink, the junction temperature level
is generally high. It is noted that the IGBT module solution may result in a large
component counts in 10 MW power conversion system, which may be unpreferable
in the wind power application.
Finally it is worth to mention that, the paralleling of power switching devices
may change the loading profile significantly and have strong impacts on the loss,
thermal, cost, and power density performances of the converter, thereby, the parallel
numbers of power switching devices should be carefully evaluated in the design
process.
References
1. Li H, Chen Z, Polinder H (2009) Optimization of multibrid permanent-magnet wind generator

systems. IEEE Trans Energy Convers 24(1):82–92
2. Polinder H, van der Pijl FFA, de Vilder G-J, Tavner PJ (2006) Comparison of direct-drive and
geared generator concepts for wind turbines. IEEE Trans Energy Convers 21(3):725–733
3. Website of ABB Semiconductors. http://www.abb.com/semiconductors
4. Website of Westcode. http://www.westcode.com/igbt.htm
5. Infineon Application Note: Thermal Resistance Theory and Practice, Jan 2000
6. ABB Application Note: Applying IGBTs, May 2007
7. Semikron Application Manual Section 5: Application Notes for IGBt and MOSFET Modules
(2011)
Chapter 9
Reliability-Cost Models for the Power
Switching Devices of Wind Power
Converters
In a traditional wind power converter design process, the power switching device
ratings are normally decided based on the potential current/voltage stresses, and
some rating margins may be reserved to ensure certain reliability requirements.
However, it is found that the loading distribution of power devices may be quite
unequal under various converter topologies as well as operation conditions [1–4].
With this traditional “rating-oriented” design process, it may easily lead to capacity
waste of some less loaded devices. Therefore, more advanced models for wind
power converter design are needed in order to satisfy the growing reliability
requirements together with the most cost-effective solutions.
If the relationship between the reliability and corresponding cost of power
semiconductor for a wind power converter is related, as indicated in the curve of
Fig. 9.1, the “just-right” device cost/ratings can be determined according to the
specific reliability requirements by the mission profile. With this reliability-cost
model and the “reliability-oriented” design process, it is possible to enable more
accurate and cost-effective design for the wind power converter achieving the target
reliability requirements.
With the reliability-cost performances profiles, it is also possible to unify and
compare different converter solutions in a more sensible way, as demonstrated in
Fig. 9.2 (i.e., converter solutions with different topologies, voltage ratings,
switching frequencies, etc. for a given mission profile). Therefore, a new
reliability-cost evaluation benchmark for wind power converters can be established
and help to guide the selection of various converter solutions.
As a result, this special topic will focus on the development of such models
which can relate the reliability and cost of power switching devices for wind power
converters, and it is different from the system-level reliability-cost analysis in [5]. In
order to easier quantify the reliability and cost performances, as shown in Fig. 9.3,
the reliability is represented by the power device junction temperature mean value
Tm and fluctuation ΔTj according to the manufacturer’s life-time models for power
semiconductors [6–10]. The cost is represented by the used chip or cell numbers

124 9 Reliability-Cost Models for the Power Switching Devices …
Fig. 9.1 Reliability-cost Reliability

profile of a converter
Life time
requirements
Design
point
Cost needed Cost
Fig. 9.2 Reliability-cost Reliability

profile comparison of
different converter solutions Converter 3
Converter 2
Converter 1
Cost
N in IGBT/Diode modules according to the dominant packaging structure by IGBT

manufacturers [11, 12]. Consequently, the relationship between the reliability and
cost of converters is transferred to the relationship between junction temperature
and chip numbers of power switching devices.
According to the widely used loss and thermal models for power switching
devices in [7, 13– 15], the junction temperature is generated by applying the losses
to the thermal impedance. Therefore, the key method to relate the junction tem-
perature and chip numbers of power devices is to include the chip number infor-
mation into the loss model as well as the thermal impedance model, respectively.
9.1 Loss Model with Chip Number Information
In most of the cases, a series of IGBT modules at certain voltage rating is composed
of different amounts of IGBT and diode dies/chips/cells, which are paralleled
together in order to achieve various current ratings. The datasheets of the IGBT
modules as well as the used chips are available on the website of semiconductor
manufacturers [16], from which the 4.5 kV series IGBT modules which consist of
different numbers of IGBT chips (4.5 kV/55 A) and freewheeling diode chips
9.1 Loss Model with Chip Number Information 125
Reliability Cost
Thermal Performance Installed Switch Power

vs.
Junction Temp. Tj, Tm Chip Numbers N
Fig. 9.3 Quantification of reliability and cost performances
Generator T1 D1 Transformer
Dnpc
T2 D2
Power
Filter Grid
3L-NPC inverter
Fig. 9.4 Three-level neutral-point-clamped converter used in a wind turbine
Table 9.1 Parameters of a Rated output active power Po 10 MW

10 MW 3L-NPC wind power
converter shown in Fig. 9.4 a
Rated load current Iphase 1750 A rms
a
(4.5 kV/110A) are chosen as an example in this special topic [14]. The current
ratings of these series IGBT modules range from 150 to 1200 A. A grid-side
10 MW three-level Neutral-Point-Clamped (3L-NPC) wind power converter is
chosen and designed as a case study in this special topic, as shown in Fig. 9.4, and
the detailed converter parameters are shown in Table 9.1 [1–3].
A. Chip Number Extraction from IGBT Modules
Unlike the switching behavior, the conduction voltage of both IGBT and diode is a
relative “clean” characteristic which only depends on the junction temperature at
the suggested gate drive voltage, while the switching characteristic depends not
only on the junction temperature, but also on drive resistance, commutated voltage,
di/dt, line inductance, etc. [11]. Normally, the used chip numbers of an IGBT
module is not provided by the manufacturer in the datasheet; however, it is possible
to calculate the chip numbers N according to the module’s conduction voltage at a

certain load current Iload.
If the load current is equally distributed in each IGBT/diode chip, the total
conduction voltage with N paralleled chips at the load current of Iload should be
equal to the conduction voltage of a single IGBT/Diode chip at the load current of
Iload/N, and this relationship can be written as

Iload
vce=N ðIload Þ ¼ vce=1 ð9:1Þ
N
where the vce/N is the conduction voltage of N parallel IGBT chips and vce/1 is the
conduction voltage of a single IGBT chip, which can be written as

iload ðtÞ iload ðtÞ Ace=1
vce=1 ¼ Vce0=1 þ rce=1 ð9:2Þ
N N
The Vce0/1, rce/1, and Ace/1 are the fitting parameters of Vce − Ic relationship curve
for a single IGBT chip [12], which can be found in the datasheets of manufacturer.
Putting (9.2) into (9.1), the chip numbers inside an IGBT module can be cal-
culated as
,
vce=N ðIload Þ Vce0=1 1=Ace=1
N ¼ Iload ð9:3Þ
rce=1
Figure 9.5 shows the rated conduction voltage for 4.5 kV series IGBT modules
at various current ratings, where the conduction voltage of both IGBT vce and
freewheeling diode vf at junction temperature of 25 and 125 °C are indicated,
respectively. When using this information to Eq. (9.3), the used chip numbers of
these modules can be calculated, and the results are indicated in Table 9.2.
Fig. 9.5 Rated conduction 3.8

voltage for a series of 4.5 kV
IGBT modules of a 3.6
Conduction Voltage (V)
manufacturer
3.4
3.2
2.8
2.6
2.4
Vce 25 Vce 125 Vf 25 Vf 125
2.2
0 200 400 600 800 1000 1200
Current ratings of modules (A)
Table 9.2 Calculated chip Module ratings (A) 150 650 800 1200
numbers for different current
ratings of IGBT modules IGBT chip numbers 3 11 15 22
(4.5 kV series) Diode chip numbers 1 6 7 11
B. Conduction Loss Model with Chip Number Information

With the chip numbers in Table 9.1, it is possible to acquire the conduction voltage
characteristic for a single chip inside the IGBT modules with different current
ratings. Figures 9.6 and 9.7 show the conduction voltage–load current of a single
IGBT and freewheeling diode chip in different rating modules listed in Table 9.1. It
can be seen that the conduction voltage characteristic of a single IGBT or Diode
chip inside different current rating modules are quite consistent with each other.
Therefore, the instantaneous conduction loss of N chips paralleled IGBT module
Pcond/N can simply be expressed as the sum of conduction loss for each single chip
Pcond/1:
Fig. 9.6 Conduction voltage 6.0

Vce versus load current Ic for a 5.5
single IGBT chip in different 5.0
modules (125 °C) 4.5
Vce pr. chip (V)
4.0
3.5
3.0
55A
2.5
150A
2.0
650A
1.5
800A
1.0
1200A
0.5
0.0
0 20 40 60 80 100 120 140
Current pr. chip (A)
Fig. 9.7 Conduction voltage 5.0

VF versus load current IF for a 4.5
single freewheeling diode
4.0
chip in different modules
(125 °C) 3.5
Vf pr. chip (V)
3.0
2.5
2.0
110A
1.5
650A
1.0
800A
0.5
1200A
0.0
0 30 60 90 120 150 180 210 240

iload ðtÞ iload ðtÞ
Pcond=N ðN; tÞ ¼ N Pcond=1 ðN; tÞ ¼ N vce=1 dðtÞ
N N
iload ðtÞ
¼ vce=1 iload ðtÞ dðtÞ ð9:4Þ
N
where iload(t) is the load current and d(t) is the duty ratio of the IGBT module.
The average mean conduction loss of N chips paralleled in an IGBT Module
PCond avg/N can be expressed as
Z1=fo
Pcondavg=N ðNÞ ¼ fo Pcond=N ðN; tÞ dt ð9:5Þ
0
C. Switching Loss Model with Chip Number Information

As mentioned before, the switching loss characteristic of power devices depends on
many factors, such as the drive resistance, di/dt, line inductance, etc. These factors
are sensitive to the operating conditions and can be easily deviated when different
numbers of chips are packaged. Therefore, it is inaccurate to calculate the switching
loss of N chips paralleled IGBT module by simply summing up the switching loss
of each single chip.
The solution is to investigate the switching loss characteristic pr. chip inside
various current ratings of IGBT modules, and then an average Loss-Current curve is
chosen to be summed up to get the total switching loss of N paralleled chips. With
the information of chip numbers in Table 9.1, the switching loss characteristic pr.
chip inside various current rating modules is plotted in Fig. 9.8. It can be seen that
the switching loss characteristic pr. chip of different rating modules slightly deviates
from each other. An average curve is chosen and fitted with two order functions as
indicated in Fig. 9.8.
With the fitting function for the average switching loss characteristic pr. chip, the
instantaneous switching loss of N chips paralleled IGBT module can be calculated
as
Psw=N ðN; tÞ ¼ N Psw=avg1 ðN; tÞ

" #
jiload ðtÞj 2 jiload ðtÞj ; ð9:6Þ
¼ N S1=avg1 þ S2=avg1 þ S3=avg1
N N
where S1/avg1, S2/avg1, and S3/avg1 are the fitting parameters for average switching
loss characteristic pr. chip, as indicated in Fig. 9.8. The average mean switching
loss of N chips paralleled IGBT module PSWavg/N with chip number information can
be calculated as
(a) 700 (b) 600

600 y = 4.4105x + 26.866
Eon pr. chip (mJ)
y = 0.0262x 2 + 1.7853x + 21.198 500
Eoff pr. chip (mJ)

500
400
400
300 300
200 200
100 100
55A 150A 650A 800A 1200A 55A 150A 650A 800A 1200A
0 0
0 30 60 90 120 150 0 30 60 90 120 150
Current pr. chip (A) Current pr. chip (A)
(c) 400
350
Err pr. chip (mJ)
300
250
200
y = 0- .0057x 2 + 2.6009x + 34.364
150
100
50
110A 650A 800A 1200A
0
0 30 60 90 120 150 180 210 240
Fig. 9.8 Switching loss pr. chip versus load current Ic for different current ratings of IGBT
modules (4.5 kV series). a Eon versus Ic pr. chip (125 °C). b Eoff versus Ic pr. chip (125 °C). c Err
versus Ic pr. chip (125 °C)
Z1=fo
PSWavg=N ðNÞ ¼ fo Psw=N ðN; tÞ dt ð9:7Þ
0
D. Loss Calculation Results

The instantaneous loss of the most stressed switching devices for the 3L-NPC
grid-side inverter under rated condition is shown in Fig. 9.9, where the chip number
N is 22 and switching frequency is 800 Hz. The average losses of the most stressed
switching devices with relation to the chip numbers are shown in Fig. 9.10.
9.2 Thermal Impedance Model with Chip Number

Information
According to the datasheets of IGBT modules, the thermal impedance is modeled as

three layers Foster RC network inside the modules (from junction to case ZthJ−C),
and a thermal resistance outside the modules (from case to heat sink RthC−H), as
(a)
Instantaneous Loss
(b)
Instantaneous Loss
Total loss
Switching loss
(W)
(W)
Total loss
Conduction loss Conduction loss

Switching loss
Time (s) Time (s)

(c)
Instantaneous Loss
(W)
Total loss Switching loss

Conduction loss
Time (s)
Fig. 9.9 Instantaneous loss of switching devices for three-level neutral-point-clamped grid
inverter (normal operation, N = 22, fs = 800 Hz. PF = 1, Po = 10 MW, red line total loss, blue line
conduction loss, green line switching loss). a T1. b T2. c Dnpc
(a) (b)
Total loss
Average Loss (W)
Average Loss (W)
Switching loss
Total loss
Conduction loss Conduction loss
Switching loss
Chip numbers Chip numbers
(c)
Average Loss (W)
Total loss
Switching loss
Conduction loss
Chip numbers
Fig. 9.10 Average loss versus chip numbers of switching device for three-level neutral-point-
clamped grid inverter (normal operation, fs = 800 Hz. PF = 1, Po = 10 MW, red line total loss, blue
line conduction loss, green line switching loss). a T1. b T2. c Dnpc
9.2 Thermal Impedance Model with Chip Number Information 131
Rth1 Rth2 Rth3

Tj TC
Rc-h
1/Rth1 2/Rth2 3/Rth3
TH
Fig. 9.11 Thermal impedance model for ABB 4.5 kV series IGBT modules, different chip
numbers have different parameters in Rth and τ
(a) (b)
Thermal resistance (K/W) 120
Thermal resistance (K/W)
50
Rth1 Rth1
y = 132.96x -1 100 y = 129.57x -1
40 Rth2
Rth2 80
30 Rth3
Rth3 60
20 y = 43.843x -1 y= 45.489x -1
40
10 20
y = 21.64x -1 y = 22.441x -1
0 0
0 5 10 15 20 25 0 2 4 6 8 10 12
Fig. 9.12 RthJ-C inside the modules (junction to case) versus numbers of chips. a IGBT. b Diode
indicated in Fig. 9.11. The thermal impedance model with chip number information
can be acquired by fitting function of the thermal parameters provided by datasheets
of various current ratings of IGBT modules.
The thermal resistances Rth1, Rth2, and Rth3 in Fig. 9.11 of various ratings of
IGBT modules with relation to the calculated chip numbers can be plotted in
Fig. 9.12 [16], where the fitting function is also indicated. It can be seen that the
thermal resistances Rth1 − Rth3 of both IGBT and diode inside the IGBT modules
are inverse proportional to the chip number N, which is consistent with the thermal
resistance physical model [15]:
d
Rth ðNÞ ¼ ¼ MRth N 1 ð9:8Þ
kAN
where N is the number of chips, A is the physical area pr. chip, d is the thickness of
material, λ is the thermal conductivity constant (W/m*K), and MRth is a fitting
parameter.
(a) (b)

50 120
RcsT RcsD
y = 115.61x - 0.839 100
40
80
30 y = 109.73x -0.751
60
20
40
10 20
0 0
0 5 10 15 20 25 0 2 4 6 8 10 12
Fig. 9.13 RthC−H outside the modules (case to heat sink) versus numbers of chips. a IGBT.
b Diode
(a) (b)
250 250
Thermal time constant
Thermal time constant
t1 t2 t3 t1 t2 t3
200 200
150 150
(ms)
(ms)
100 100
50 50
0 0
0 5 10 15 20 25 0 5 10 15
Fig. 9.14 Time constants t1 − t3 in Fig. 9.11 (junction to case) versus numbers of chips
(tx = Rthx*Cthx). a IGBT. b Diode
In respect to the thermal resistance from case to heat sink RthC−H (outside the
module), it is more related to the base plate size of module case as well as pack-
aging technology; therefore, RthC−H is no longer inverse proportional to the chip
numbers N, as plotted and fitted in Fig. 9.13.
The time constant of the thermal impedance in Fig. 9.11 (τ1 − τ3) with relation to
the chip numbers of IGBT modules is plotted in Fig. 9.14. It can be seen that the
time constants of the thermal impedance inside the IGBT modules are kept constant
with different chip numbers, and this characteristic is consistent with the time
constant physical model, which is represented as [15]
cq 2
s ¼ Rth Cth ¼ d ð9:9Þ
k
9.2 Thermal Impedance Model with Chip Number Information 133
where the thermal capacitance Cth can be represented as
Cth ðNÞ ¼ c q A d N ð9:10Þ
The c is the factor proportional to heat in (Ws/g*K), and ρ is the density of

materials (g/cm3). In the time constant function (9.9), it can be seen that there is no
item for the chip numbers N, which means that the time constant of thermal
impedance is only related to the thickness of the chips, which should not be
deviated in various current ratings of IGBT modules.
In summary, the thermal impedance model with the chip number information for
both IGBT and diode of IGBT modules can be written as
X
3
Zth ðN; tÞ ¼ N 1 MRthx ð1 et=sx Þ ð9:11Þ
x¼1
9.3 Analytical Solution of Junction Temperature

with Chip Number Information
With the instantaneous loss model (9.4), (9.6) and thermal impedance model (9.11)
(in which the chip number information is both included), it is possible to calculate
the instantaneous junction temperature of switching devices by convoluting the loss
and thermal impedance models as
Zt
d
Tj ðN; tÞ ¼ Tc þ Ploss=N ðN; zÞ Zth ðN; t zÞdz ð9:12Þ
dz
0
However, this calculation is complicated and time consuming. Actually, only the
temperature average mean value Tm and fluctuation amplitude ΔTj are related to the
life-time of power devices by most of the reliability models for power semicon-
ductors [6–10]. As a result, simplified junction temperature solutions, which can
directly extract the information of Tm and ΔTj, have to be developed, and it is done
in this special topic.
A. Simplified Solution for Tm
According to [15], the steady-state junction temperature mean value Tm is only
related to the total thermal resistance as well as the average loss dissipation in the
power devices. The Tm can be written as follows, where the chip number infor-
mation is included:
!
X
3
Tm ðNÞ ¼ Tref þ Pavg=N ðNÞ Rthx þ RthCH ð9:13Þ
x¼1
Fig. 9.15 Mean junction 250

temperature Tm of each power
Mean Junction temperature Tm ( )

device in 3L-NPC grid
inverter versus chip numbers T1
(normal operation, 200
Po = 10 MW, fs = 800 Hz, Dnpc
Vll = 3.3 kVrms)
150
T2
100
D1
D2
50
20 30 40 50 60 70
Chip numbers N
The calculated mean junction temperature Tm with relation to the chip numbers
N is shown in Fig. 9.15, in which each switching device in a switching arm of
10 MW 3L-NPC wind power inverter is indicated as an example.
B. Simplified Solution for ΔTj
According to the instantaneous power loss dissipation, as shown in Fig. 9.8, the
time domain of losses in the most stressed devices is more or less sinusoidal
distributed within a half fundamental cycle. On this case, it is not easy to acquire the
fluctuation amplitude of junction temperature ΔTj, because the exact time when the
junction temperature achieves its maximum/minimum value is hard to be derived
by (9.12).
One possible simplification is to use square wave loss pulses which share the
same loss-time area as the original sinusoidal loss distribution within a half fun-
damental cycle. In this case, the loss is more constant and the time when the
junction temperature achieves its maximum value can be determined.
In Fig. 9.16, three kind of loss pulses (one step, two steps, and three steps) which
share the same loss-time area as the original sinusoidal-like losses are generated and
applied to the same thermal impedance of IGBT module, and the corresponding
junction temperatures are also indicated. It can be seen that the two-step loss pulses
can achieve an acceptable consistency of junction temperature fluctuation with the
original loss distribution.
The more detailed two-step loss pulses and its resulting junction temperature are
indicated in Fig. 9.17. It is relatively easy to calculate the temperature fluctuation
amplitude ΔTj with the information of loss pulse amplitude and step time. The
approximate function is shown as follows [17]:
9.3 Analytical Solution of Junction Temperature with Chip Number Information 135
(a)
Loss (kW)
Time (s)
(b)
Loss (kW)
Time (s)
(c)
Loss (kW)
Time (s)
Fig. 9.16 Loss pulses approximation to get analytical solution of junction temperature. (green
original loss and corresponding junction temperature, red approximate loss pulses and
corresponding junction temperature) a One-step loss pulses approximation. b Two-step loss
pulses approximation. c Three-step loss pulses approximation
Ploss
3Pavg
Tj
Tj
2Pavg
Pavg
t1 t2 t3 1/(2fo) time
Fig. 9.17 The calculation of junction temperature fluctuation by two-step loss pulses (t1 = 1/(8 fo),
t2 = 3/(8 fo), t3 = 1/(2 fo), fo is the fundamental frequency)
Fig. 9.18 Junction 40

temperature fluctuation ΔTj of
Junction temperature fluctuation Tj (K)
each power device in 3L-NPC T1

grid inverter versus chip
numbers (normal operation, 30
Po = 10 MW, fs = 800 Hz, Dnpc
Vll = 3.3 kVrms)
20 T2
10
D1
D2
0
20 30 40 50 60 70
Chip numbers N
DTj ¼ Pavg Zth ðt2 Þ þ ð3Pavg Pavg Þ Zth ðt2 t1 Þ

3 1 ð9:14Þ
¼ Pavg Zth þ 2Pavg Zth
8fo 4fo
The calculated junction temperature fluctuation ΔTj with relation to the chip
numbers N is shown in Fig. 9.18, in which each switching device in a switching arm
of 10 MW 3L-NPC inverter is indicated.
C. Temperature-Cost Profile of 3L-NPC Inverter
With the relationship between junction temperature and corresponding chip num-
bers for each of the power switching devices in Figs. 9.15 and 9.18, the same
junction temperature (maximum or fluctuation) for each of the switching device can
9.3 Analytical Solution of Junction Temperature with Chip Number Information 137
Fig. 9.19 Junction (a) 130

temperature versus chip
numbers for one leg of 120
)
3L-NPC grid-side converters.
110
Maximum Tj (
(All of the devices are set to
have the same junction
100
temperature by adjusting the
chip numbers, respectively, 90
Po = 10 MW, fs = 800 Hz,
Vll = 3.3 kVrms) a Maximum 80
junction temperature.
b Fluctuation of junction 70
80 110 140 170 200 230 260
temperature
Chip numbers for one leg
(b) 18
16
14
12
Tj ( )
10
8
6
4
2
0
80 110 140 170 200 230 260
Chip numbers for one leg
be set by adjusting the chip numbers, respectively. The characteristics of the

junction temperature for all of the switching devices with relation to the needed
total chip numbers for one phase of the given 10 MW three-level Neutral-Point-
Clamped wind power inverter can be plotted in Fig. 9.19.
It can be seen that Fig. 9.19 shows a unique performance profile of a given
converter solution (topology, voltage rating, etc.); therefore, it is possible to unify
and compare different converter solutions with this profile in a more sensible way.
Because the junction temperature is close related to the life-time of power switching
devices according to [6], and the chip numbers will decide the current rating and
packaging of IGBT modules, the connection between reliability and cost of the
power devices for a certain converter solution is thereby established.
9.4 Conclusions
The relationship between the reliability and cost of converters is quantified by the
junction temperature and chip numbers of power switching devices, respectively, in
this special topic. It is proved that the conduction loss, switching loss, and thermal
impedance models of power switching devices (IGBT module) can be included

with chip number information. A simplified analytical solution which directly
extracts the junction temperature mean value Tm and fluctuation ΔTj is proposed.
The proposed reliability-cost model is demonstrated on a 10 MW 3L-NPC wind
power inverter as case study.
With the developed reliability-cost profile for the given converter solution, it is
possible to enable more accurate and cost-effective design for wind power converter
achieving the target reliability performance, and it is also possible to unify and
compare different converter solutions in a more sensible way.
References

wind turbines. In: Proceedings of ISIE 2011, pp 340–346, June 2011
2. Ma K, Blaabjerg F (2011) Multilevel converters for 10 MW wind turbines. In: Proceedings of
EPE 2011, pp 1–10
3. Ma K, Blaabjerg F, Liserre M (2011) Thermal analysis of multilevel grid side converters for
10 MW wind turbines under low voltage ride through. In: Proceedings of ECCE 2011,
pp 2117–2124
IEEE Trans Ind Appl 48(2):708–719
5. Yu X, Khambadkone AM (2012) Reliability analysis and cost optimization of parallel-inverter
system. IEEE Trans Ind Electron 59(10):3881–3889
6. Kaminski N (2004) Load-cycle capability of HiPaks, ABB Application Note 5SYA 2043-01,
Sept 2004
power applications. Microelectr Reliab 51(9–11):1903–1907
8. Wintrich A, Nicolai U, Reimann T (2011) Semikron application manual, pp 128, 2011
9. Kovacevic IF, Drofenik U, Kolar JW (2010) New physical model for lifetime estimation of
power modules. In: Proceedings IPEC’10, pp 2106–2114
10. Due J, Munk-Nielsen S, Nielsen R (2011) Lifetime investigation of high power IGBT
modules. In: Proceedings of EPE 2011, pp 1–10
11. ABB Application Note: Applying IGBTs, May 2007
12. ABB Application Note: Applying IGBT and diode dies, March 2010
13. Blaabjerg F, Jaeger U, Munk-Nielsen S, Pedersen J (1995) Power losses in PWM-VSI inverter
using NPT or PT IGBT devices. IEEE Trans Power Electron 10(3):358–367
14. Lixiang W, McGuire J, Lukaszewski RA (2011) Analysis of PWM frequency control to
improve the lifetime of PWM inverter. IEEE Trans Ind Appl 47(2):922–929
15. Infineon Application Note: Thermal Resistance Theory and Practice, Jan 2000
16. Website of ABB semiconductors. http://www.abb.com/product/us/9AAC910029.aspx)
17. Semikron Application Manual section 5: Application notes for IGBt and MOSFET Modules,
2011
Chapter 10
Electro-Thermal Model of Power
Semiconductors Dedicated for Both Case
and Junction Temperature Estimation
The thermal model for power semiconductors devices from junction to ambient can
be modeled by a series thermal resistance Rth and capacitance Cth networks [1–3].
The power loss generated on the semiconductor junction or chips first flows through
the internal thermal networks of power device to its case or base plate, and then the
left power loss continues going through the attached thermal grease and heat sink
until the ambient. When the loss passes though the thermal network outside power
devices, the case temperature is then established; when the loss passes though the
thermal network inside power devices, the junction temperature will be established
based on the case temperature level.
As a study case, the power semiconductor device of IGBT module is applied in a
typical three-phase DC-AC two-level voltage source converter. The rated power of
converter is set as 1 MW with unit power factor, the input DC voltage is 1100 V
and output line to line voltage is 690 Vrms, switching frequency of IGBT module is
designed at 1950 Hz. IGBT module 5SNA1600N170100 (1600 A/1700 V) from
ABB [4, 5] is chosen with enough datasheet information for thermal analysis.
Normally the multilayer Foster thermal RC network inside the IGBT module
(i.e. from junction to case) is widely provided by manufacturers on their device
datasheets. This type of thermal model is just a mathematical fitting of the measured
external thermal behavior of power device and don’t represent any physical
meaning for each RC layer [1, 3, 6]. As an example to demonstrate the limits of the
Foster thermal network, a complete thermal impedance train including thermal
networks inside and outside the given IGBT module is shown in Fig. 10.1.
Based on the thermal impedance in Fig. 10.1. and the given converter condition,
(Fig. 10.2) the simulated power loss generated inside individual IGBT Pin, power
loss output from IGBT base plate Pout, as well as the IGBT junction and case
temperature Tj and TC are shown in Fig. 10.3a respectively. It is noted that the Pin
which is periodically changed at 50 Hz with many switching loss pulses trans-
parently passes through the internal Foster thermal network of power device (i.e.
Pin = Pout). When the un-filtered power loss Pout passes through the thermal
This chapter is co-authored by F. Blaabjerg and M. Liserre.

140 10 Electro-Thermal Model of Power Semiconductors Dedicated …
Loss Tj Tc Th Ta
Pin Pout Thermal

Grease
IGBT module Heat sink
multi-layers Foster Z j-c Zc-h+Zh-a
Fig. 10.1 Thermal train based on Foster thermal network
Loss Tj IGBT module Tc Th Ta
Pin Pout Thermal

Grease
Heat sink
multi-layers Cauer Z j-c Z c-h+Zh-a
Fig. 10.2 Thermal train based on equivalent Cauer thermal network
Fig. 10.3 Loss and temperature responses of the thermal models, a Foster-based model.
b Cauer-based model
resistance of the thermal grease, abrupt changed IGBT case temperature TC as well
as the junction temperature Tj are observed. Therefore, the Foster thermal RC
network inside IGBT module should be only connected to a temperature reference
and cannot be extended with any other thermal RC networks like the thermal grease
or heat sink, this limit is also introduced and described in [1, 3].
10 Electro-Thermal Model of Power Semiconductors Dedicated … 141
In order to estimate the correct case and junction temperature by the given power
semiconductor devices and the corresponding heat sink design, the Foster thermal
network inside IGBT module has to be mathematically transferred to the equivalent
Cauer type RC network. As shown in Fig. 10.2, in which the thermal impedance
inside IGBT module are replaced by the equivalent multilayer Cauer RC network.
The corresponding power loss generated inside individual IGBT Pin, power loss
output from IGBT base plate Pout, as well as the IGBT junction and case temper-
ature Tj and TC are shown in Fig. 10.3b respectively. It is noted that the unregulated
loss pulses of Pin are significantly filtered into smooth Pout by the parallel multilayer
thermal capacitances inside IGBT module, a much more stable case temperature TC
is thereby established compared to Fig. 10.3a and a correct device junction tem-
perature Tj profile at steady state can be observed.
However, the mathematical transformation from Foster to Cauer thermal net-
work does not gain any physical meaning for each RC layer either, this means the
Pin in Fig. 10.3b may not be correctly filtered by the equivalent Cauer thermal
model. Normally over-filtering will be introduced when doing the mathematically
transformation from Foster to Cauer form (it can be verified from Fig. 10.3b that
there is almost no fluctuation in Pout) [6], and this over-filtering in the equivalent
multilayer Cauer network will lead to slower dynamic performance for the Pout and
eventually for the case temperature TC and junction temperature Tj estimation of
power devices.
A new thermal model which targets to overcome the shortages of this two
thermal models are given in this topic. The proposed thermal model is shown in
Fig. 10.4, it can be seen that it looks like a combined solution and contains two
paths for the thermal flow:
The first thermal path is used for the junction temperature estimation. In this path
the original datasheet-based multilayer Foster thermal network inside power devi-
ces are used. Different from Fig. 10.1, only a temperature potential, whose value is
Loss Tj TA
Pin
IGBT module
multi-layers Foster Zj-c
IGBT module TC TH TA
Pout Thermal
Grease
Heat sink
equivalent 1 layer Cauer Zj-c Z c-h +Z h-a
Fig. 10.4 Proposed thermal model

142 10 Electro-Thermal Model of Power Semiconductors Dedicated …
determined by the case temperature TC from the other thermal path, is connected to
the Foster network. As a result the abrupt change of case temperature and junction
temperature in the thermal model of Fig. 10.1 can be avoided.
The second thermal path is used for the case and heat sink temperature esti-
mation. In this path the thermal network inside IGBT module is just used for the Pin
loss filtering rather than the junction temperature estimation. While the complete
thermal network outside IGBT module (i.e. thermal grease and heat sink) have to be
included. It is noted that the multilayer Foster thermal network inside IGBT module
is mathematically transformed to a Single-layer Cauer RC unit. This transformation
will lose some accuracy for the dynamic performance of junction temperature Tj,
but the physical meaning is somehow regained because any object can be repre-
sented as a Cauer RC unit from the thermal point of view, consequently, the
single-layer equivalent Cauer RC unit inside the IGBT module can achieve more
correct filtering of Pin, and better dynamic performance for the Pout and TC com-
pared to the multilayer equivalent Cauer network in Fig. 10.2.
When applying the proposed thermal model, the corresponding power loss
generated inside individual IGBT Pin, power loss output from IGBT base plate Pout,
as well as the IGBT junction and case temperature Tj and TC are shown in
Fig. 10.5a respectively, in which the behaviors of the equivalent Cauer thermal
model in Fig. 10.2 are also indicated as a comparison. It can be seen that under the
same generated power loss Pin, the proposed new thermal model can achieve almost
the same junction and case temperatures as the Cauer thermal model in Fig. 10.2 at
steady state, the Pout and TC are less filtered (or more fluctuated) in the new thermal
model.
Fig. 10.5 Loss and temperature responses of the new thermal model, a Steady state. b Dynamic
loss change
10 Electro-Thermal Model of Power Semiconductors Dedicated … 143
Figure 10.5b shows a dynamic change of the converter loading from rated 1 MW
power output to 0.2 MW output at the 0.6 s. It can be seen that the new thermal
model can achieve faster response in the Pout, TC and Tj.
10.1 Conclusion
The Foster thermal model and the equivalent Cauer form both have their limits to
correctly estimate the case and junction temperature of power semiconductor
devices. By the proposed thermal model, it is possible to extend the Foster thermal
network with other thermal impedances, and thereby acquire more accurate junction
and case temperature of power semiconductor devices not only under steady state
but also during dynamic thermal changes.
References
1. Marz M, Nance P Thermal modeling of power electronic system, Infineon application note
2. Infineon application note: thermal resistance theory and practice, Jan 2000
3. Infineon application note AN2008-03: thermal equivalent circuit models, Jun 2008
4. Website of ABB semiconductors
5. ABB application note: applying IGBTs, May 2007
6. ABB application note 5SYA 2093-00: thermal design and temperature ratings of IGBT
modules, 2012
Chapter 11
Reactive Power Influence on the Thermal
Cycling of Multi-MW Wind Power
Inverter
The growing requirements for reliability and grid integration push the solutions of
wind power generation system from doubly fed induction generator (DFIG) with
partial-rated power converter to synchronous or asynchronous generator with
full-scale power converter. In the full-scale power converter solutions, there are
more flexibilities to control the reactive power, and it is easier to satisfy the grid
voltage support requirements by most of the grid standards under both normal and
faults condition of the power grid.
However, the added reactive power, especially in the application of multi-MW
wind power conversion, may increase the converter output voltage (modulation
index) and change the amplitude as well as phase angle of the load current in the
grid side inverter, as shown in Fig. 11.1. The voltage and current phasor diagram
from the point view of power grid by introducing the overexcited and underexcited
reactive current IQ in grid connected converter are indicated respectively. Referring
to the commonly used and accepted loss model, the modified output voltage and
load current will change the loss consumption and distribution, and thereby lead to
the change of thermal cycling (or junction temperature) of power switching devices.
According to most of the reliability models for power semiconductor, the lifetime of
the wind power inverter is closely related to the thermal cycling performance––both
average mean junction temperature and temperature fluctuation amplitude, therefore
it is interesting to investigate the relationship between the reactive power and the
resulting modified thermal cycling for multi-MW wind power inverters.
In this special topic, the allowable reactive power ranges as well as their impacts
on thermal cycling of wind power inverters are presented at different wind speeds.
The situations when considering single converter and paralleled converter systems
in a wind park are analyzed respectively. A new concept is also proposed to
stabilize the device thermal fluctuation during wind gust by controlling the reactive
power circulated among parallel converter systems in a wind park.
This chapter is co-authored by M. Liserre and F. Blaabjerg.

146 11 Reactive Power Influence on the Thermal Cycling …
(a) (b)
Uc
UL IQ- Ig
IP Uc
Ug UL
IQ+ IP
Ig Ug
Fig. 11.1 The phasor diagram from the power grid point of view by introducing reactive power in
grid converter (Uc represents the converter output voltage, Ug grid voltage, UL voltage drop on
filter inductor, Ig load current, IP active current, IQ reactive current, cos θ power factor, α the phase
angle). a Receive overexcited reactive power from converter (IQ+). b Generate underexcited
reactive power to converter (IQ−)
11.1 Effect of Reactive Power in Case of Single Converter
As the most commercialized multilevel topology which is widely used in the

high-power medium-voltage drives for industry, mining, and traction applications,
the three-level neutral-point-clamped (3L-NPC) converter seems to be a promising
candidate for the Multi-MW full-scale wind power conversion system, as shown in
Fig. 11.2. This converter is chosen and basically designed based on a 10 MW wind
turbine as a case study, where the major design parameters are summarized in
Table 11.1.
For a single wind power generation system, the maximum reactive power
achieved by the grid side inverter has to be restrained in a certain range according to
the grid standards [1–3], as shown in Fig. 11.3, in which the allowable boundaries
of reactive power with relation to the generating active power is defined by German
grid code when the grid voltage is within ±5 % range around nominal value. It can
be seen that in the definition of Variant 1 the overexcited reactive power Q+ should
be less than 48 % of the rated active power Prated, and the underexcited reactive
power Q− should be less than 23 % of Prated.
Referring to the phasor diagram of the grid side converter in Fig. 11.1, the
amplitude of the load current Ig in relation to the reactive power Q and active power
P can be calculated as
Generator T1 D1
Filter Transformer
Dnpc1
D2
T2
T3 Grid
Filter Dnpc2 D3 Lg
T4 D4
Focused part
Fig. 11.2 Three-level neutral point clamped converter used in a wind turbine
11.1 Effect of Reactive Power in Case of Single Converter 147
Table 11.1 Parameters of Rated active power (P) 10 (MW)

three-level neural point
clamped inverter for case DC bus voltage Vdc 5.6 kV DC
a
study Rated primary side voltage Vll 3.3 kVmis
b
Rated phase current Iphase 1.94 kA rms
a
Line-to-line voltage in the primary windings of the transformer
b
Phase current when power factor is 0.9
Variant Variant Variant P/Prated (p.u.)

3 2 1
1.0
0.8
0.6
0.4
0.2
-0.41 -0.33 -0.23 0.33 0.41 0.48

Q/Prated (p.u.)
Fig. 11.3 The P, Q range of wind power converter during normal operation defined by the
German grid codes [1]
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P2 þ Q2
Ig ¼ ð11:1Þ
3 Ug
where Ug is the rms value of grid phase voltage. The phase angle α between load
current Ig and converter output voltage Uc in Fig. 11.1 can be calculated as

Ug sin h þ 2pfo Lf Ig
a1 ¼ arctan ð11:2Þ
Ug PF
when introducing overexcited reactive power, and

Ug sin h 2pfo Lf Ig
a2 ¼ arctan ð11:3Þ
Ug PF
when introducing underexcited reactive power.

In (11.2) and (11.3), the power factor angle θ is
P
h ¼ arccos pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð11:4Þ
P2 þ Q2
Referring to the grid code defined in Fig. 11.3, the restrained condition has to be
added as
0:23P Q 0:48P ð11:5Þ
given the condition that the power factor are maintained and generating active
power P is larger than 20 % of its rated value.
Then the amplitude and phase angle of the load current Ig in relation to the
reactive power p.u. can be plotted in Fig. 11.4a, b respectively, in which the
situations when the wind speed is 12 m/s (P = 10 MW, 1 p.u.), 10 m/s
(P = 6.3 MW, 0.63 p.u.), and 8 m/s (P = 3.2 MW, 0.32 p.u.) are indicated.
In order to investigate the maximum impact of the reactive power to the losses
and find the thermal distribution of the power devices, three extreme conditions
when complying with grid codes are chosen based on 10 m/s wind speed from
Fig. 11.4 (10 m/s is the typical average annual offshore wind speed defined by IEC I
wind class standard [4]): the maximum underexcited reactive power boundary when
Q−max (10 m/s) = −0.13 p.u., no reactive power when PF = 1, and maximum
overexcited reactive power boundary when Q+max (10 m/s) = 0.27 p.u.
The loss model is a commonly accepted method for loss evaluation of power
semiconductor devices; the conduction loss and switching loss are accumulated by
switching cycles according to the information of conduction voltage as well as
switching energy in relation to the load current provided by the datasheets of
manufacturers. The simulation is carried out based on PLECS Blockset in Simulink.
Press-pack IGCT 5SHY35L4512 (commutated voltage 2.8 kV/maximum current
3.3 kA rms) and diodes 5SDF16L4503 (2.8 kV/2.6 kA rms) from ABB are chosen
as the power switching devices. For simplicity, losses dissipated in the devices are
considered as temperature independent.
(a) (b)
Load current amplitude Ig (p.u.)
12 m/s
Ig - Uc (degree)
10 m/s
12 m/s
10 m/s
8 m/s
Phase angle
8 m/s
Fig. 11.4 Load current versus reactive power when complying with grid codes (dot line
The thermal models of a single switch and clamped diode are indicated in
Fig. 11.5, in which the thermal impedance from junction to case Z(j−c) is modeled as
a four-layer Foster RC network, as shown in Fig. 11.6. Each of the thermal
parameters can be found from the manufacturer datasheets and they are summarized
in Table 11.2, where the thermal resistance Rth will decide the steady-state mean
value of junction temperature, and the thermal capacitance (with time constant τ)
will decide the dynamic change or fluctuation of the junction temperature. The
ambient temperature is set to 50 °C and considered constant during the operation of
converter, however, it may be changed depending on the operation site.
It is noted that the separately packaged IGCT and diodes are chosen because of
the limitation for available products which can be found on the market. However, in
a practical converter design, the IGCT/IGBT and its freewheeling diode are usually
integrated and packaged together, the chip size for diode is about half of that for the
IGCT/IGBT, accordingly, the thermal resistance of the diode from junction to heat
sink is not consistent with its datasheet values but set to twice of the IGCT.
It is also noted that the thermal impedance between heat sink and ambient Z(h−a)
in Fig. 11.5 is closely related to the design of the heat sink and to the layout for the
Switch Clamped Diode Note:

Tj Tj Tj Tj: junction temperature
IGCT Diode Diode TC: case temperature
ZT(j-c) ZD(j-c) ZD(j-c) TH: heat sink temperature
TA TA TA TA: ambient temperature
TC TC TC Z(j-c): thermal impedance
from junction to case
ZT(c-h) ZD(c-h) ZD(c-h) Z(c-h): thermal impedance
from case to heat sink
TH TH Z(h-a): thermal impedance
from heat sink to ambient
Z(h-a) Z(h-a)
TA TA
Fig. 11.5 Thermal models of the power devices
Fig. 11.6 Thermal model of the impedance ZT(j−c) or ZD(j−c) from junction to case in Fig. 11.5
Table 11.2 Parameters of thermal impedance for IGCT/diode

Thermal impedance ZT/D(j−c) ZT/D(c−h)
Sector 1* Sector 2 Sector 3 Sector 4
RiIGCT (K/kW) 5.562 1.527 0.868 0.545 3
τiIGCT (s) 0.5119 0.896 0.0091 0.0024 –
RiDiode (K/kW) 11.124 3.054 1.736 1.09 6
τiDiode (s) 0.5119 0.896 0.0091 0.0024 –
*Sector N means different layers of thermal RC lump in Fig. 11.6
power devices, which are application-dependent and they are out of the scope of
this special topic, therefore, the Z(h−a) is not included in the thermal analysis.
However, it is worth to mention that the cost and structure of heat sink systems do
have impacts on the thermal impedance and temperature profile of power devices. If
heat sinks with small thermal resistance and large thermal capacitance are used,
normally the temperature rise on the heat sink has smaller amplitude and longer
time constant compared to the temperature rise inside the power devices. After the
heat sink system is chosen and optimized, the temperature rise on the heat sink will
be known and it is possible to be included in the analysis.
The simulation results regarding inverter outputs and current distribution are
shown in Figs. 11.7 and 11.8 respectively, where the three extreme reactive power
Fig. 11.7 Output waveforms of 3L-NPC inverter under different reactive power (boundaries
defined by grid codes, 10 m/s wind speed), output voltage pulses (green), grid voltage (blue),
phase current (red). a Q−max = −0.13 p.u. b Q = 0 p.u. c Q+max = 0.27 p.u
(a) (b)
T1 T1
D1 D1
T2 T2
D2 D2
T3 T3
D3 D3
T4 T4
D4 D4
Dnpc1 Dnpc1
Dnpc2 Dnpc2
Time (s) Time (s)
(c) T1
D1
T2
D2
T3
D3
T4
D4
Dnpc1
Dnpc2
Time (s)
Fig. 11.8 Current distribution of 3L-NPC inverter under different reactive powers (boundaries
defined by grid codes, 10 m/s wind speed). a Q−max = −0.13 p.u. b Q = 0 p.u. c Q+max = 0.27 p.u
conditions are introduced to the inverter when generating 0.63 p.u. active power at
10 m/s wind speed. It can be seen that when complying with the regulations by grid
codes, the differences of converter outputs with the given three extreme reactive
powers are insignificant.
The loss simulation results of the 3L-NPC wind power inverter are presented in
Fig. 11.10, in which three given reactive power conditions are compared. It can be
seen that there is no significant loss difference between the three conditions where
different amounts of reactive power are introduced. The situation at the maximum
overexcited reactive power boundary (Q+max = 0.27 p.u.) consumes slightly more
loss in the outer switches and diode T1, D1 and inner switch T2 of the 3L-NPC
inverter.
According to the junction temperature simulation results in Fig. 11.9, the thermal
cycling in the most stressed devices of the clamping diode Dnpc as well as the outer
switches T1 has no significant difference among the three reactive power operating
points.
(a) (b)
Dnpc1 Dnpc1
T1 T1
T2 D1 T2 D1
D2 D2
Time (s) Time (s)
(c)
Dnpc1
T1
D1
T2
D2
Time (s)
Fig. 11.9 Thermal cycling of 3L-NPC inverter under different reactive powers (boundaries
defined by grid codes, 10 m/s wind speed). a Q−max = −0.13 p.u. b Q = 0 p.u. c Q+max = 0.27 p.u
5000
4000
Loss (W)
3000 Q- max
No Q
2000 Q+ max
1000
0
T1 D1 T2 D2 Dnpc
Fig. 11.10 Loss distribution of 3L-NPC inverter with different extreme reactive powers
(Complying with grid codes)
11.2 Effect of Reactive Power in Case of Paralleled

Converters
For wind power converter systems in a wind farm, the grid side inverters are
connected as a local grid, therefore, the reactive power can basically be circulated
among the inverters locally and not necessarily be seen in the power grid, as shown
11.2 Effect of Reactive Power in Case of Paralleled Converters 153
Converter 1 (Underexcited operation)
T1 D1
P1
Dnpc1
D2
T2 Grid
T3
Dnpc2 D3
T4 D4 Underexcited
-Q
Overexcited
+Q
P2
...
Converter N
Fig. 11.11 Reactive power circulated in paralleled wind power converters
in Fig. 11.11. It can be seen that the underexcited reactive power –Q absorbed by
converter 1 is generated/compensated by converter 2 which is under overexcited
operation with +Q, and there is no (or less) reactive power injected into the power
grid. On this condition the operation range of reactive power in converter 1 and
converter 2 is not restrained by the grid codes but by the amplitude of load current
as well as the converter output voltage, which is due to the limitations of device
current rating and maximum modulation index of the power converter.
The new ranges of amplitude and phase angle of the load current with relation to
the reactive power are shown in Fig. 11.12, which is restrained by
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
S2 P2 Q S2 P2 ð11:6Þ
M1 ð11:7Þ
where the modulation index M can be calculated as

pffiffiffi
6 Uc
M¼
Vdc
pffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi ð11:8Þ
6 ðUg PFÞ2 þ ½Ug sin h 2pfo Lf Ig 2
¼
Vdc
(a) (b)
12 m/s
Phase current amplitude Ig (p.u.)
Phase angle Ig - Uc (degree)

10 m/s 12 m/s
10 m/s
8 m/s
8 m/s
Fig. 11.12 Operation range of reactive power when complying with grid codes (dot line
It can be seen that the operating range of reactive power is significantly extended
compared to Fig. 11.4, and the ranges for overexcited and underexcited reactive
current are unsymmetrical: The increase of Q+ will cause an increase of converter
output voltage Uc, therefore the Q+ is first limited by the maximum modulation
Fig. 11.13 Output waveforms of 3L-NPC inverter under different reactive power (Considering
interaction of inverters, 10 m/s wind speed), output voltage pulses (green), grid voltage (blue),
phase current (red). a Q−max = −0.82 p.u. b Q = 0 p.u. c Q+max = 0.55 p.u
(a) (b)
T1 T1
D1 D1
T2 T2
D2 D2
T3 T3
D3 D3
T4 T4
D4 D4
Dnpc1 Dnpc1
Dnpc2 Dnpc2
Time (s) Time (s)
(c) T1
D1
T2
D2
T3
D3
T4
D4
Dnpc1
Dnpc2
Time (s)
Fig. 11.14 Current distribution of 3L-NPC inverter under different reactive powers (Considering
interaction of inverters, 10 m/s wind speed). a Q−max = −0.82 p.u. b Q = 0 p.u. c Q+max = 0.55 p.u
index before it gets to the limitation of maximum device current rating. While the
increase of Q− will cause a decrease of Uc and the modulation index, therefore Q− is
only limited by the device current rating.
Another three extreme reactive power conditions when considering the inter-
action between paralleled converters are investigated based on 10 m/s wind speed:
the maximum underexcited reactive power boundary when Q−max (10 m/s) = −0.82 p.
u., no reactive power when PF = 1, and maximum overexcited reactive current
boundary when Q+max (10 m/s) = 0.5 p.u.
The simulation results regarding inverter outputs and current distribution are
shown in Figs. 11.13 and 11.14 respectively, where the three extreme reactive
power conditions are introduced when generating 0.63 p.u. active power at 10 m/s
wind speed. It can be seen that when considering the interaction between paralleled
converters, the converter output pulse width, amplitude, and phase angle of load
current with the given three extreme reactive powers significantly differ from each
other (Fig. 11.15).
The loss simulation results are presented in Fig. 11.16, in which three given
reactive power conditions are compared when the interaction between parallel
converters is considered. It can be seen that the situation at maximum overexcited
(a) (b)
Dnpc1
Dnpc1
T1
D1
T1
T2
T2 D1
D2 D2
Time (s) Time (s)
(c)
Dnpc1
T1
D1
T2
D2
Time (s)
Fig. 11.15 Thermal cycling of 3L-NPC inverter under different reactive powers (Considering
interaction of inverters, 10 m/s wind speed). a Q−max = −0.82 p.u. b Q = 0 p.u. c Q+max = 0.55 p.u
5000
4000
Loss (W)
3000 Q- max
No Q
2000 Q+ max
1000
0
T1 D1 T2 D2 Dnpc
Fig. 11.16 Loss distribution of 3L-NPC inverter with different extreme reactive powers
(Considering parallel converters)
reactive power boundary (Q+max = 0.55 p.u.) consumes significantly more loss than
the situation without reactive power in the outer switch T1, outer diode D1, inner
switch T2 and clamping diode Dnpc of the 3L-NPC inverter. The major difference
between overexcited and underexcited reactive power boundary is at the stress of
the clamping diode, the latter has much less loss than the former one in Dnpc.
According to the junction temperature simulation results in Fig. 11.15, the

thermal cycling of the most stressed device of the clamping diode Dnpc as well as
outer switch T1 is significantly higher than the other two reactive power operating
points.
In this special topic the operation conditions of converter with various reactive
powers are analyzed based on p.u. system; with other converter ratings and designs
the basic operation status, e.g., current loading and voltage outputs will be similar
except for the switching frequency. Due to different parameters for loss and thermal
models, the loss and junction temperature profile could deviate from the results
obtained in this special topic if other power devices are applied. However,
according to the device current loadings in Fig. 11.14, the extreme underexcited
operation of converter will further increase the stress of Dnpc no matter which
device is adopted, because the extreme underexcited operation leads to much higher
current amplitude with more conduction time of Dnpc. The extreme overexcited
operation of converter will further increase the stress of T1 for the same reasons.
11.3 Conclusions
The introduction of reactive current may change the thermal distribution of power
devices. The operation ranges of reactive current not only depend on the grid codes,
but also relate to the interaction between paralleled inverters—the latter situation
has much wider allowable reactive current range than the former one.
By introducing reactive current during the lower wind speed of a wind gust, the
junction temperature fluctuation in the most stressed devices of a 3L-NPC wind
power inverter can be significantly stabilized, and the reliability could thereby be
improved according to, e.g., the Coffin-Masson lifetime models, while the increased
stresses to the other devices or paralleled converters are still acceptable.
References
1. Altin M, Goksu O, Teodorescu R, Rodriguez P, Bak-Jensen B, Helle L (2010) Overview of

recent grid codes for wind power integration. In: Proceedings of OPTIM’2010, pp 1152–1160
2. Tsili M, Papathanassiou S (2009) A review of grid code technical requirements for wind farms.
IET Renew Power Gener 3(3):308–332
3. E.ON-Netz (2006) Grid code high and extra high voltage
4. International Electrotechnical Commission (2005) Wind turbines, part 1: design requirements,
IEC 61400-1, 3rd edn
Chapter 12
Thermal Loading of Several Multilevel
Converter Topologies for 10 MW Wind
Turbines Under Low Voltage Ride
Through
In this special topic, three promising grid-side multilevel converters for 10 MW

wind turbines are proposed and basically designed. The evaluation criteria will
mainly aim at the utilization and thermal performances of power switching devices
during various LVRT conditions. Studies regarding the converter output, as well as
loss and thermal distributions under different grid voltage dips/wind speeds are
presented and compared.
12.1 Promising Topologies and Basic Design
The concept and major parts of a variable speed wind turbine with full-scale power
converters are shown in Fig. 12.1.
As mentioned before, the power capacity of a single wind turbine keeps
climbing up even to 7 MW, and medium voltage (1–10 kV) would be interesting
and needed to reduce the current rating in the wirings and switching devices under
such a high power level. It is more difficult for the traditional two-level voltage
source converter to achieve acceptable performance with the available switching
devices. With the abilities of more output voltage levels, higher voltage amplitude
and larger output power, multilevel converter topologies are becoming the most
promising candidates in the application of full-scale power/medium-voltage wind
power conversion.
Because the grid side converter in Fig. 12.1 is directly interfaced with the power
grid, and plays a key role to comply with the stricter standards during grid faults, the
discussions will mainly focus on this part of the generation system. Among various
multilevel topologies, three of them are of interest (Figs. 12.2, 12.3, and 12.4):
The basic design of each converter topology for a case study is as follows: All of
the power switching devices have the commutated voltage of 2.8 kV in order to
utilize the available and dominant 4.5 kV high-power IGCT/IGBT on the market,

160 12 Thermal Loading of Several Multilevel Converter Topologies …
Transformer
AC DC Grid
Filter DC AC Filter
Gear
Generator Focused part
Fig. 12.1 Wind power generation system with full-scale converter
Fig. 12.2 Three-level

neutral-point-clamped
converter topology (3L-NPC)
T1 D1
Dnpc D2
T2
T3
Dnpc D3
D4
Fig. 12.3 Three-level

H-bridge converter topology
(3L-HB)
T1 D1 T3 D3
T2 D 2 T4 D4
Fig. 12.4 Five-level

H-bridge converter topology
(5L-HB)
T1 D1
Dnpc D2
T2
T3
Dnpc D3
T4 D4
12.1 Promising Topologies and Basic Design 161
Table 12.1 Parameters of different converter topologies for case study

Configurations 3L-NPC 3L-HB 5L-HB
Rated active power 10 MW
Equivalent sw. freq. 800 Hz
Modulation method PD-PWM Unipolar-PWM POD-PWM
DC bus voltage 5–6 kV 2.8 kV 5.6 kV
Primary side voltagea 1.9 kV rms 1.9 kV rms 3.8 kV rms
Rated phase current 1.75 kA rms 1.75 kA rms 972 A rms
Filter inductance 1.13 mH (0.2 p.u.) 1.13 mH (0.2 p.u.) 2.89 mH (0.12 p.u.)
a
Rated phase voltage in the primary windings of transformer
then the DC bus and maximum output voltage of each configuration can be
determined. The most commonly used carrier-based PWM method for each con-
verter topology is applied and the equivalent switching frequency is typically
designed to be 800 Hz in order to get an acceptable switching loss in the power
devices. The output filter inductance is designed to limit the maximum current
ripple to 25 % of the rated current amplitude, and the filter capacitance is not taken
into account. The references of active and reactive current delivered by the inverter
under different voltage dips are selected according to the German grid codes. The
design parameters are summarized in Table 12.1. For simplicity of analysis, the
power grid is considered as three ideal AC voltage sources, and the transformers are
assumed ideal. The DC bus voltage during LVRT is assumed to be controlled at
110 % of the rated value by a DC bus chopper, which is a typical industrial solution
used to absorb the active power from the generator.
12.2 Operation Status Under Balanced LVRT
After the parameters for each converter topology are settled, the operation status
with the information of output voltage, load current and delivered power under
various LVRT conditions can be derived and simulated. In order to facilitate the
investigation and demonstration of the converter operation characteristics under
LVRT, three-phase balanced grid faults are first taken into account.
Figure 12.5 summarizes the active/reactive power delivered by the grid side
converter under various balanced grid voltage dips of three-phases. Because the
injected reactive current under LVRT is defined by German grid codes, the reactive
power Q delivered by the converter is only decided by the grid voltage. However,
there is still some flexibility to adjust the active current when the grid voltage is
above 0.5 p.u. In order to reduce the stress of braking chopper on DC bus and
maintain the DC bus voltage, the active power P delivered by the converter under
grid voltage above 0.5 p.u. should refer to the generated power by wind turbines.
The worst condition is assumed when the generation system is set to provide as
much active power as possible, and the pitch control of wind turbine does not have
12
Active / reactive power (MW / MVar)

P Q
>12 m/s
10
8
10 m/s
6
4 8 m/s
0
0 0.2 0.4 0.6 0.8 1
Grid Voltage Vg (p.u.)
Fig. 12.5 Active and reactive power delivered by converter during balanced LVRT (based on
German grid codes)
enough time to activate. The situations of 12 m/s wind speed (10 MW generated
power), 10 m/s (6.3 MW generated power), and 8 m/s (3.2 MW generated power)
are indicated, respectively, in Fig. 12.5.
The current amplitude and phase angle (between load current and grid voltage)
under three-phase balanced LVRT are shown in Fig. 12.6, in which the situation of
different wind speeds at 12, 10, and 8 m/s are indicated respectively. It can be seen
that when the grid voltage is below 0.5 p.u., both the current amplitude and phase
angle are kept constant because of 100 % rated reactive current injection; however,
when the grid voltage is above 0.5 p.u., the current amplitude and phase angle
dramatically change with the variation of grid voltage and wind speed.
The simulations are carried out based on PLECS Blockset in Simulink, and the
simulation parameters are consistent with the ones in Table 12.1. A normal
3000 100
Phase angle I - Vg (degree)
Current amplitude (3L-HB, 3L-NPC) 80

2500
Current amplitude (A)
60
40
2000
20
1500 >12 m/s 0
10 m/s -20
1000 8 m/s
-40
-60
500
Phase angle -80
0 -100
0 0.2 0.4 0.6 0.8 1
Grid Voltage Vg (p.u.)
Fig. 12.6 Amplitude and phase angle of load current during balanced LVRT (based on German
grid codes in Fig. 12.2, the 5L-HB converter has half current amplitude)
12.2 Operation Status Under Balanced LVRT 163
operation status of each converter is first assumed at wind speed of 10 m/s, which is
the typical average annual offshore wind speed defined by IEC I wind class stan-
dard. As an extreme example, the converters are subjected to 0.05 p.u. balanced
grid voltage dips for 150 ms during the defined normal operation status. Studies of
output pulses (green), load current (red), and grid voltage (blue) of each converter
topology are presented in Fig. 12.7. It is obvious that, the current amplitude during
LVRT with 0.05 p.u. grid voltage increases significantly in all of the three converter
topologies compared to the normal operation with 10 m/s wind speed.
When zooming in the indicated areas of Fig. 12.7, the detailed output waveforms
before and after LVRT are shown in Fig. 12.8. Compared to the normal operation with
10 m/s wind speed, the converter outputs when undergoing LVRT with 0.05 p.u. grid
voltage have significant changes on the current amplitude, phase angle, and
voltage-pulse width. It can be seen that the load current lags the grid voltage by 90 °
because of 100 % reactive current injection, and the output voltage-pulse width is largely
reduced because of lower modulation index. It is noted that the output voltage level
reduces from five to three in the 5L-HB topology when LVRT operation is presented.
The current distribution in the power switching devices of Fig. 12.8 are shown in
Fig. 12.9. It is found that the current load moves from transistor to diodes (both
freewheeling and clamped diodes) in the entire three converter topologies during
(a) (b)
Current (A) / Voltage (V) Current (A) / Voltage (V)
Zoom in Zoom in
Normal LVRT Normal Normal LVRT Normal

operation 0.05 p.u. operation operation 0.05 p.u. operation
Time (s) Time (s)

Current (A) / Voltage (V)
(c)
Zoom in
Normal LVRT Normal

operation 0.05 p.u. operation
Time (s)
Fig. 12.7 Simulation outputs when LVRT is presented (normal operation: vw = 10 m/s,
PG = 6.3 MW/LVRT: Vg = 0.05 p.u., Ireactive = 100 % Irated), output voltage pulses (green),
grid voltage (blue), phase current (red). a 3L-NPC. b 3L-HB. c 5L-HB
(a) (b)
Current (A) / Voltage (V) Current (A) / Voltage (V)
Normal operation LVRT 0.05 p.u. Normal operation LVRT 0.05 p.u.
Time (s) Time (s)
(c) Current (A) / Voltage (V)
Normal operation LVRT 0.05 p.u.
Time (s)
Fig. 12.8 Zoom in area of Fig. 12.7, output voltage pulses (green), grid voltage (blue), phase
current (red). a 3L-NPC. b 3L-HB. c 5L-HB
balanced LVRT, and the increased current amplitude will increase the stress of the
power devices compared to the normal operational condition.
12.3 Loss Distribution Under Balanced LVRT
Press-pack IGCT were introduced into the medium voltage converters in 1990s and
are already becoming state of the art in high power electric drives (e.g,. for oil and
gas applications), but not yet widely adopted in the wind turbine industry also
because of cost issues. In this special topic, the press-pack IGCT 5SHY35L4512
(commutated voltage 2.8 kV/maximum current 3.3 kA rms) and diodes
5SDF16L4503 (2.8 kV/2.6 kAr ms) from ABB are chosen as the switching power
devices for 3L-NPC and 3L-HB topologies, IGCT 5SHY35L4510 (2.8 kV/2.7 kA
rms) and diodes 5SDF10H4503 (2.8 kV/1.8 kA rms) are chosen for the 5L-HB
topology. The used loss model is a commonly accepted method for loss evaluation
of power semiconductor devices, and the loss simulation is carried out based on
PLECS Blockset in Simulink. Losses dissipated in the power devices are consid-
ered temperature independent during the simulation.
12.3 Loss Distribution Under Balanced LVRT 165
(a) (b)
T1 T1
D1 D1
T2 T2
D2 D2
T3 T3
D3 D3
T4 T4
D4 D4
Dnpc1
Dnpc2
Normal operation LVRT 0.05 p.u. Normal operation LVRT 0.05 p.u.
Time(s) Time(s)
(c) T1
D1
T2
D2
T3
D3
T4
D4
Dnpc1
Dnpc2
Normal operation LVRT 0.05 p.u.
Time(s)
Fig. 12.9 Current distribution in the power devices. a 3L-NPC. b 3L-HB. c 5L-HB
It is noted that during the LVRT operation, the DC bus of power converter may
probably increase because of the short-term mismatch in the input and output active
power through the converter. Normally, the increased DC bus voltage should be
limited (e.g., maximum 110 % rated) for hundreds of milliseconds by triggering the
braking chopper. According to the loss model in [22], the DC bus voltage has
important impacts on both the switching loss and the conduction loss in power
switching devices. As a result, the increased DC bus voltage should be taken into
account in the loss analysis during LVRT. Moreover, the increased DC bus voltage
may significantly decrease the lifetime of power switching devices due to the
cosmic radiation failure mechanism; however, this issue will not be discussed in
this special topic.
The loss distribution in the power switching devices under normal operation
with wind speed of 8, 10, 12 m/s, as well as balanced LVRT condition with 0.05 p.
u. grid voltage are compared in Fig. 12.10, in which 10 % higher DC bus voltage
(3.1 or 6.2 kV) are applied for LVRT condition. It can be seen that, the LVRT
operation may impose the diodes (D1, D2, Dnpc) and inner switches (T2 in
3L-NPC and 5L-HB) with significantly larger losses than the most stressed normal
operation condition (12 m/s wind speed, 10 MW rated output power). The loss
simulation results are consistent with the current distributions in Fig. 12.9, in which
the diodes and inner switches are heavily loaded because of increased current
amplitude and moved phase angle between load current and grid voltage.
(a) (b)
5000 5000
Normal 8m/s Normal 10m/s Normal 8m/s Normal 10m/s
Normal 12m/s LVRT 0.05 p.u. Normal 12m/s LVRT 0.05 p.u.
4000 4000
Loss (W)
Loss (W)
3000 3000
2000 2000
1000 1000
0 0
T1 D1 T2 D2 Dnpc T1 D1 T2 D2 Dnpc
(c)
5000
Normal 8m/s Normal 10m/s
Normal 12m/s LVRT 0.05 p.u.
4000
Loss (W)
3000
2000
1000
0
T1 D1 T2 D2 Dnpc
Fig. 12.10 Loss distribution of converters under normal and balanced LVRT. a 3L-NPC.
b 3L-HB. c 5L-HB
12.4 Thermal Distribution Under Balanced LVRT
The thermal performance of power devices are closely related to the reliability of
the converter, current rating of power devices, and the cost of the cooling system.
Therefore, it is an important indicator for large-scale wind power converters. In
order to conduct thermal performance evaluation, an appropriate thermal model
should first be acquired.
Thermal impedance from junction to case Z(j−c) is modeled as a four-layer
Foster RC network. Each of the thermal parameters can be found from the man-
ufacturer’s datasheet and they are summarized in Table 12.2, where the thermal
resistance Rth will decide the steady-state mean value of the junction temperature,
and the thermal capacitance (with time constant τ) will decide the dynamic change
or fluctuation of the junction temperature. The ambient temperature is set to 50 °C
and considered constant during the operation of converter. However, it may be
changed depending on the operation site.
It is noted that the separately packaged IGCT and diodes are chosen because of
the limitation for available products, which can be found on the market. However, in
a practical converter design, the IGCT/IGBT and its freewheeling diode are usually
integrated and packaged together, the chip size for diode is usually about half of that
for the IGCT/IGBT; accordingly, the thermal resistance of the diode from junction to
heat sink is not consistent with its datasheet but set to twice the IGCT.
12.4 Thermal Distribution Under Balanced LVRT 167
Table 12.2 Parameters of thermal impedance for IGCT/diode

Thermal ZT/D(j−c) ZT/D(c−h)
impedance Sector 1 Sector 2 Sector 3 Sector 4
RiIGCT (K/kW) 5.562 1.527 0.868 0.545 3
τiIGCT (s) 0.5119 0.896 0.0091 0.0024 –
RiDiode (K/kW) 11.124 3.054 1.736 1.09 6
τiDiode (s) 0.5119 0.896 0.0091 0.0024 –
Normally, the thermal capacitance outside a power device from case to ambient
are much larger compared to that inside a power device from junction to case in a
properly designed cooling system. The larger thermal capacitance, which has longer
time constant ranging from hundreds of milliseconds to hundreds of seconds,
mostly decide the time to achieve steady-state junction temperature, and have no
significant impact on the dynamic junction temperature fluctuation within a fun-
damental cycle of the converter output (dozens of milliseconds). Therefore, it is
efficient to make a simplification which ignores the relatively larger thermal
capacitances in Z(c−h) and Z(h−a) to realize a faster thermal simulation. In megawatts
power converter systems, separated heat sink is typically used and a good thermal
decoupling among the power devices can be achieved, so the thermal resistance
between the heat sink and ambient is considered small.
Based on the previous loss simulation results and thermal model, the junction
temperature of the power devices in each of the converter solution can be inves-
tigated by the PLECS blockset in Simulink.
A. 3L-NPC
The simulated junction temperature in 3L-NPC converter under normal operation

with 10 m/s wind speed and three- phase balanced LVRT condition with 0.05 p.u.
grid voltage are shown in Fig. 12.11. It can be seen that the thermal distribution is
quite unequal under both operation modes. The LVRT operation has higher junc-
tion temperate in all of the switching devices except the outer switch T1, and the
maximum temperature which is located in the clamped diode Dnpc increases about
20 K compared to the normal operation at 10 m/s wind speed.
According to the important Coffin-Masson lifetime model, the junction tem-
perature mean value Tm and the fluctuation amplitude ΔTj are two of the most
important information for the reliability of power semiconductor devices; the
simulated Tm and ΔTj of each switching device in 3L-NPC converter in relation to
the grid voltage are shown in Fig. 12.12a, b respectively. The change in junction
temperature keeps relatively smooth when the grid voltage is below 0.5 p.u., and
becomes dramatic when the grid voltage is above 0.5 p.u. It is noted that there is a
temperature rise in Dnpc and T1 when the grid voltage is around 0.7 p.u., which is
due to the change in current phase angle and the fast growing switching loss in T1
and Dnpc.
Dnpc
Dnpc
T2
T1
T1 D1
T2 D1
D2
D2
Normal operation (10 m/s) LVRT condition (0.05 p.u.)
Fig. 12.11 Junction temperature in normal operation versus LVRT in 3L-NPC converter (normal
operation: vw = 10 m/s, PG = 6.3 MW/LVRT: Vg = 0.05 p.u.)
Fig. 12.12 Junction (a) 140 T1 D1 T2 D2 Dnpc

temperature distribution under 130
balanced LVRT in 3L-NPC
Mean value Tm ( )
120
converter (vw = 10 m/s). 110
a Junction temperature mean
100
value Tm versus grid voltage.
90
b Junction temperature
80
fluctuation ΔTj versus grid
voltage 70
60
50
40
0 0.2 0.4 0.6 0.8 1
Grid voltage (p.u.)
(b) 22 T1 D1 T2 D2 Dnpc
20
18
Fluctuation Tj (K)
16
14
12
10
8
6
4
2
0
0 0.2 0.4 0.6 0.8 1
Grid voltage (p.u.)
B. 3L-HB
The simulated junction temperature in 3L-HB converter, under normal operation

with 10 m/s wind speed and LVRT condition with 0.05 p.u. grid voltage is com-
pared in Fig. 12.13. It can be seen that the junction temperature is equally dis-
tributed among all the switching devices under normal operation, and significantly
increases, especially in the diodes under LVRT. The maximum temperature which
is located in the freewheeling diodes D1/D2 increases about 35 K compared to the
normal operation with 10 m/s wind speed.
The simulated temperatures Tm and ΔTj in 3L-HB converter in relation to the
grid voltage are shown in Fig. 12.14. It is interesting to see that the thermal
distribution in 3L-HB topology under LVRT is much more equal than the 3L-NPC
in Fig. 12.12, and both the junction temperature mean value and amplitude keep
reducing when the grid voltage is above 0.5 p.u.
C. 5L-HB
The junction temperature in the 5L-HB converter under normal operation with
10 m/s wind speed and LVRT condition with 0.05 p.u. grid voltage are compared in
Fig. 12.15, and the simulated Tm and ΔTj in 5L-HB converter in relation to the grid
voltage are shown in Fig. 12.16.
It can be seen that the trends of thermal performance in the 5L-HB topology
under LVRT is quite similar to that of the 3L-NPC topology, but the junction
temperature keeps at a much lower level. This is because of the half current rating
compared to the 3L-NPC topology.
D1 D2
T1 T2
T2 T1
D1 D2
Fig. 12.13 Junction temperature in normal operation versus LVRT in 3L-HB converter
(simulation results, normal operation: vw = 10 m/s, PG = 6.3 MW/LVRT: Vg = 0.05 p.u.,
Ireactive = 100 % Irated)
(a) 140 T1 D1 T2 D2
(b) 22 T1 D1 T2 D2
130 20
Mean value Tm ( )
Fluctuation Tj (K)
120 18
110 16
100 14
12
90
10
80 8
70 6
60 4
50 2
40 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fig. 12.14 Junction temperature distribution under balanced LVRT in 3L-HB converter
(vw = 10 m/s). a Junction temperature mean value Tm versus grid voltage p.u. b Junction
temperature fluctuation ΔTj versus grid voltage p.u
Dnpc
Dnpc T1
T2
T2
D2 D1 D2 T1 D1
Fig. 12.15 Junction temperature in normal operation versus LVRT in 5L-HB converter
(simulation results, normal operation: vw = 10 m/s, PG = 6.3 MW/LVRT: Vg = 0.05 p.u.,
Ireactive = 100 % Irated)
D. Comparison of Topologies
The junction temperature comparison of the most stressed power device in each of
the converter topologies is shown in Fig. 12.17, in which the fluctuation range of
the junction temperature is indicated. It can be seen that the 3L-NPC converter has
the most stressed switching devices Dnpc among the three converter topologies,
and the most extreme condition happens when the grid voltage is close to 0 p.u.
(due to the large conduction losses) and 0.7 p.u. (due to the large switching losses).
For the 3L-HB topology, it has better thermal performance compared to the
(a) 140 T1 D1 T2 D2 Dnpc

(b) 20 T1 D1 T2 D2 Dnpc
130 18
)
Fluctuation Tj (K)
120 16
Mean value Tm (
110 14
100 12
90 10
80 8
70 6
60 4
50 2
40 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fig. 12.16 Junction temperature distribution under balanced LVRT in 5L-HB converter
(vw = 10 m/s). a Junction temperature mean value Tm versus grid voltage p.u. b Junction
temperature fluctuation ΔTj versus grid voltage p.u
Fig. 12.17 Junction 140

temperature comparison 130 max
Junction temperature Tj ( )
between converters under Dnpc

120 min
balanced LVRT (10 m/s, most
max
stressed device) 110
D1/D2
min
100
90
80 max
70 min
60 Dnpc
50
3L-NPC 3L-HB 5L-HB
40
0 0.2 0.4 0.6 0.8 1
Grid voltage (p.u.)
3L-NPC topology, especially when the grid voltage is above 0.5 p.u. The 5L-HB
shows the best temperature performance among the three topologies, and shares the
similar junction temperature changing trends as 3L-NPC converter. This also means
that it has more potential to handle higher power or reduce the cost for cooling
system and power semiconductors.
12.5 Unbalanced LVRT
In reality, the unbalanced grid faults (e.g., one-phase grounded or two-phase con-
nected) are more likely to happen. In these cases, the LVRT operation of the grid
side converter is more complicated compared to the balanced fault condition
(three-phase grounded). During the unbalanced LVRT, the types and location of
grid faults, the connection of transformer windings, and the power control strategies
of the converter dramatically change the loading of power switching devices.
Therefore, the operation conditions for unbalanced grid fault have to be carefully
specified.
A. Propagation of Voltage Dips
A typical configuration for grid integration of wind turbine is shown in Fig. 12.18,
in which a Dy transformer is used to interface the power converter output on Bus 2
(e.g. 3.3 kV) and the distribution line of wind farm on Bus 1 (e.g. 20 kV).
A short-circuit fault in the grid integration system will cause voltage dips on Bus 1
and Bus 2.
Defining the voltage dip value DN which represents the lowest phase voltage
amplitude in p.u. on the Bus N. Three typical grid faults: one-phase grounded (1
phase), two-phase connected (2 phase), and three-phase grounded (3 phase) with
the same dip values are assumed to happen respectively on Bus 1 (D1 = 0.5, no
phase jump). Due to the Dy connection of the transformer windings, the voltage
dips propagated from Bus 1 may look different on Bus 2. As summarized in
Table 12.3, in which the voltage dip type and dip value on Bus 1 and the corre-
sponding voltage characteristics on Bus 2 are included. The voltage dip types A-D
are defined as phasor diagrams in Fig. 12.19 respectively.
It can be seen that the three-phase balanced grid fault (3 phase) on Bus 1
propagates the same dip type and dip value on Bus 2. While the unbalanced grid
faults (1 phase and 2 phase) with the same dip value on Bus 1 cause different dip
types and dip values on Bus 2, which is monitored by the grid side converter and
result in different amount of delivered reactive currents to help the grid to recover
from faults.
The whole range of voltage dip value on Bus 1 in relation to the corresponding
dip value on Bus 2 is shown in Fig. 12.20, in which balanced (3 phase) and
unbalanced (1 phase and 2 phase) grid faults are indicated respectively. It is clear
that the balanced grid fault on Bus 1 will propagate the same dip value on Bus 2.
While the single-phase unbalanced grid fault happened on Bus 1 has a higher dip
value when propagating on Bus 2, and the two-phase unbalanced grid fault on Bus
1 has lower dip value when propagating on Bus 2.
Bus 2 Bus 1
AC DC
DC AC
3.3 kV / 20 kV
ZS
Fault Grid
>=20 kV
ZF
Fig. 12.18 Typical configuration for grid integration of wind power generation system
12.5 Unbalanced LVRT 173
Table 12.3 Voltage dip type/value on Bus 1 and Bus 2 for different grid faults
Fault type 1 phase 2 phase 3 phase
Bus 1 Dip type B C A
Dip value D1 (p.u.) 0.5 0.5 0.5
Bus 2 Dip type C D A
Dip value D2 (p.u.) 0.76 0 0.5
Positive sequence V+ (p.u.) 0.83 0.5 0.5
Negative sequence V− (p.u.) 0.17 −0.5 0
Note Two-phase grounded fault is not included because it seldom happens
VC Type A VC Type B VC Type C VC Type D

VdipC
VdipC VdipA VA VdipA VA VA VA
VdipC
VdipB
VdipB
VdipB
VB VB VB VB
Fig. 12.19 Phasor diagram definitions for the dip types A-D
Fig. 12.20 Dip value on Bus 1

2 versus dip value on Bus 1
Dip Value on Bus 2 (p.u.)
0.8
0.6
0.4
1 phase
2 phase
0.2 3 phase
0
0 0.2 0.4 0.6 0.8 1
Dip Value on Bus 1 (p.u.)
For simplicity of the analysis, only the single-phase grounded fault on Bus 1 is
chosen in this special topic as an example of the unbalanced LVRT. The injected
reactive current to the grid by wind power converter is set according to the lowest
phase voltage amplitude on Bus 2 referring to the grid demands. It is required that
the active/reactive current generated by the converters only contain positive
sequence component, and the negative sequence currents are controlled to be zero.
The behavior of converters under various LVRT is still a continuous discussion for
the future grid standards.
B. Operation Status Under Single-Phase Unbalanced Grid Fault
The active/reactive power delivered by the grid side converter under single phase
unbalanced grid voltage dip is shown in Fig. 12.21, where the horizontal axis
represents the voltage dip values on Bus 2. It is noted that the single-phase voltage
dip on Bus 1 (type B) propagates two-phase voltage dip on Bus 2 (type C), whose
dip value D2 cannot be lower than 0.577 due to the characteristic of Dy trans-
former. The situations of 12 m/s wind speed (10 MW generated power), 10 m/s
(6.3 MW generated power) and 8 m/s (3.2 MW generated power) are indicated
respectively.
The current amplitude as well as the phase angle under single-phase unbalanced
grid fault is shown in Fig. 12.22, where the phase angle represents the angle
between the load current and grid voltage of A phase in the type C grid fault
definition diagram (Fig. 12.19). It can be seen that the current amplitude and phase
angle dramatically change with the variation of grid voltage and wind speed.
As an example, Fig. 12.23 shows the grid voltage, load current as well as
instantaneous active/reactive power of the wind power converter undergoing
single-phase unbalanced grid fault on Bus 1. The wind speed is at 10 m/s (6.3 MW),
the dip value D1 is at 0 p.u. (0.577 p.u. when propagated to Bus 2), the positive
sequence active current is at 0.533 p.u. and the reactive current is at 0.864 p.u.
according to the German grid codes.
It can be seen that the grid voltage on Bus 2 is consistent with the type C fault
definition in Fig. 12.19, and the currents in the converters are symmetrical among
three phases, which means only positive sequence currents are generated. Due to
the existence of negative sequence voltage, there is a 100 Hz fluctuation in the
delivered active- and reactive power of converter, which is assumed to be absorbed
by the DC bus chopper.
Fig. 12.21 Active and 0 0.5 1

reactive power delivered by 12
Active / reactive power (MW / MVar)
converter during single phase Dip value of Bus 1 (D1 in p.u.)

grid fault (based on German >12 m/s
10
grid codes)
8
10 m/s
6
4 8 m/s
2
P Q
0
0 0.2 0.4 0.6 0.8 1
Dip value of Bus 2 (D2 in p.u.)
0 0.5 1
3000 100
Dip value of Bus 1 (D1 in p.u.)
Current amplitude (3L-HB, 3L-NPC) 80

2500
Current amplitude (A) 60
40
2000
20
1500 >12 m/s 0
10 m/s -20
1000 8 m/s
-40
-60
500 Phase angle of
none dip phase -80
0 -100
0 0.2 0.4 0.6 0.8 1
Dip value on Bus 2 D2 (p.u.)
Fig. 12.22 Positive sequence amplitude and phase angle of the load current during single phase
grid fault (based on German grid codes, 5L-HB has half the current amplitude)
Grid voltage on
Non dip phase A Dip phase B

Bus 2(p.u.)
Dip phase C
Load current
(p.u.)
P
Output Power
Q
(p.u.)
Time (s)
Fig. 12.23 Output waveforms during unbalanced grid fault. (Type C fault on Bus 2 with
D2 = 0.577 p.u., vw = 10 m/s)
By looking at the phasor diagram definition for grid voltage dip type C in
Fig. 12.19, it is interesting to see that when the single-phase grid fault is presented
on Bus 1, there is a phase shift in the voltage of phase B and C on Bus 2, and
thereby the angle between each phase voltage is no longer 120°. On the other hand
it is required that only positive sequence current should be delivered by the wind
power converter, therefore the angle between each phase current is still kept at
120°, as shown in Fig. 12.23.
Figure 12.24 shows the phase angles (between phase voltage and current) for the
three phases of wind power converter in relation to the voltage dips when
single-phase unbalanced grid fault is presented. The conditions with wind speeds of
12 and 8 m/s are indicated respectively.
As mentioned before, the phase angles as well as the voltage/current amplitude
are closely related to the loading of power switching devices. From Figs. 12.23 and
12.24, it can be seen that the voltage amplitude and phase angle of the three phases
of converter are quite different from each other, therefore, the device loss and
thermal distribution should also be different for each phase of the converter when
undergoing unbalanced LVRT.
C. Loss Analysis Under Single-Phase Unbalanced Grid Fault
The loss distribution of the power switching devices under normal operation (with
wind speed of 12 m/s), and the loss distribution for the three phases of converters
under unbalanced LVRT, are compared in Fig. 12.25. The 10 % higher DC bus
voltage is applied for LVRT condition. It can be seen that, the LVRT operation still
imposes the diodes and inner switches with significantly larger losses than the most
stressed normal operation condition, and the loss distribution among the three
phases is asymmetrical for each topology.
D. Thermal Distribution Under Single-Phase Unbalanced Grid Fault
The simulated junction temperatures for the three phases of converters are shown in
Fig. 12.26, in which the converters are undergoing unbalanced LVRT (Type C fault
on Bus 2 with D2 = 0.577 p.u., vw = 10 m/s). It can be seen that for the 3L-NPC
topology, the thermal distribution is unequal not only among the devices but also
among the three phases. Phase B has a more stressed Dnpc and T1, while phase A
Fig. 12.24 Phase angle (load 0.5 0.6 0.7 0.8 0.9 1
current to grid voltage) for 0
Dip value on Bus 2 (p.u.)
three phases of the power -10
converter (Type C fault on Phase B

-20
Bus 2)
-30
-40
Phase A -50
-60
>12 m/s
8 m/s -70
-80
Phase C -90
-100
0 0.25 0.5 0.75 1
(a)
5000
4000
Loss (W)
Normal 12 m/s
3000
Dip phase A
Dip phase B
2000 Dip phase C
1000
0
T1 D1 T2 D2 Dnpc
(b)
5000
4000
Loss (W)
Normal 12 m/s
3000
Dip phase A
Dip phase B
2000 Dip phase C
1000
0
T1 D1 T2 D2 Dnpc
(c)
5000
4000
Loss (W)
Normal 12 m/s
3000
Dip phase A
Dip phase B
2000 Dip phase C
1000
0
T1 D1 T2 D2 Dnpc
Fig. 12.25 Loss distribution of converters under normal and unbalanced LVRT. (Type C fault on
Bus 2 with D2 = 0.577 p.u., vw = 10 m/s). a 3L-NPC. b 3L-HB. c 5L-HB
and phase C have more stressed T2, D1, and D2. It is found that Dnpc in phase B is
the most stressed device of the converter under the given condition.
For the 3L-HB topology, the thermal distribution is more equal both among the
devices and the three phases, D1/D2 in phase C is the most stressed devices of the
converter.
For the 5L-HB converter, thermal distribution tendency is similar to the 3L-NPC
topology, but the inequity among the devices and the three phases are significantly
improved.
(a) Dnpc
Dnpc
Dnpc
)
Junction Temperature (
T1 D1
T1
T1
T2
T2 D1
D1
D2
D2
D2
Phase A Phase B Phase C

(b)
)
D1 D2
D2
T1 T2
T1 T2 D1 D2 D1
T1 T2

(c)
)
Dnpc
Dnpc T1 Dnpc D1
T1
D2 D1 T2 D2 T2 D1 T1 D2
Fig. 12.26 Thermal distribution in three phases of converters under unbalanced LVRT. (Type C
fault on Bus 2 with D2 = 0.577 p.u., vw = 10 m/s). a 3L-NPC. b 3L-HB. c 5L-HB

0 0.5 1
150
max
) 140
Dnpc (phase B)
130
Junction temperature Tj (
min
120 max
110
100 min 3L-NPC
D1/D2 (phase C) 3L-HB
90
80 max 5L-HB
70 min
60 Dnpc (phase B)
50
40
0.5 0.6 0.7 0.8 0.9 1
Fig. 12.27 Junction temperature comparison between converters under unbalanced LVRT (most
stressed device in three phases, Type C fault on Bus 2, vw = 10 m/s)
The junction temperature comparison of the most stressed power device among
the three phases in each of the converter topology is shown in Fig. 12.27, in which
the fluctuation range of the junction temperature is indicated. Again, 3L-HB and
5L-HB topologies show advantages under unbalanced LVRT condition (Type C
fault on Bus 2, vw = 10 m/s).
12.6 Conclusion
The reactive current injection requirements for full-scale wind power converter
during LVRT will impose some power switching devices (especially the diodes)
with even larger stress than the rated normal operation condition.
The device loading of the grid side converter under balanced LVRT changes
dramatically under different grid voltage dips. When the grid voltage is below 0.5 p.u.,
100 % rated reactive current is needed, the amplitude and position of the current is
kept fixed and only the conduction loss is changed in the power devices. While when
the grid voltage is above 0.5 p.u., both switching loss and conduction loss are changed
dramatically in the power devices because the grid code allows some room and
flexibility for the active current, which is related to the wind speed as well as to the
pitch angle/rotation speed control strategies for wind turbines during LVRT.
When undergoing single phase unbalanced grid fault, it is found that the device
loading among the three phases of converter is asymmetrical for all of the interested
topologies. And it is also found that both the three-level and five-level H-bridge
topologies show more potential to reduce and more equally distribute the stress in
the power switching devices, compared to the well-known three-level neutral point
clamped topology under various LVRT conditions.
Chapter 13
Another Groups of Thermal Optimized
Modulation Methods of Three-Level
Neutral-Point-Clamped Inverter Under
Low Voltage Ride Through
13.1 Basic Principles
In order to reduce the output time for zero voltage level of 3L-NPC inverter during
LVRT, a direct and basic idea is trying to generate a special vector sequence which
avoids using the state vector “111,” (because state vector 111 outputs zero voltage
level for all of the three phases). Following the two criteria for generating the
modulation sequences and avoiding use of the state vector 111, one way to syn-
thesize the desired voltage reference is to choose the state vectors 000, 110, 211,
222 based on the “unique” vector sequence is shown in Fig. 13.1. The special
modulation sequence is generated in Fig. 13.2a, which is called Optimized
sequence 1 (O1) for convenience. The gray blocks indicate the eliminated redun-
dant state vectors.
The simulation results of this optimized sequence are shown in Fig. 13.2b. It can
be seen that the output voltage of this modulation sequence alternates between
positive, zero, negative voltage levels within neighboring switching cycles, which is
quite different from the traditional three-level output pattern of 3L-NPC converter.
Also the amplitude and phase angle of the load current are kept unchanged. After
modification, the average voltage reference will be kept unchanged, but the duration
time of the zero voltage level is significantly reduced, and the widths of both
positive and negative output voltage pulses are expanded.
It is noticed that the equivalent switching frequency is doubled by the optimized
modulation sequence O1, the carrier frequency thereby needs to be adjusted to half
in order to acquire the same switching frequency as the normal SVM sequence.

182 13 Another Groups of Thermal Optimized Modulation Methods …
000 100 110 111 211 221 222 221 211 111 110 100 000
Fig. 13.1 The “complete” vector sequence in region A of space vector diagram of 3L-NPC
converter
(a)
000 100 110 111 211 221 222 221 211 111 110 100 000
Ts/2 Ts/2 Ts/2 Ts/2
(b)
Fig. 13.2 Thermal optimized modulation sequence 1 (O1) for 3L-NPC inverter during LVRT.
a Sequence generating method. b Simulation results: output voltage pulses (green), grid voltage
(blue), phase current (red). Vg = 0.05 p.u
13.2 Neutral Point Potential Control Method 183
13.2 Neutral Point Potential Control Method
Regarding the Neutral Point (NP) potential, sequence O1 utilizes the short vectors
which have zero accumulated NP current iNP over a fundamental cycle, as sum-
marized in Table 13.1. Therefore, the positive and negative DC bus can naturally be
balanced without any active methods. However, under some nonideal conditions
(e.g., unbalance loads or power devices), the active NP potential control method for
the normal SVM sequence cannot be applied, because only one type of short
vectors with negative NP current are used and no redundant short vectors are
included in the optimized sequence O1.
A solution to this problem is to generate another compensating vector sequences
with the same zero vectors 000, 222, but the other pair of redundant short vectors
with positive NP current. The new compensating sequence, which is called O2 here,
has the opposite impact on the direction and amplitude of iNP compared to the O1
sequence, as compared in Table 13.2, in which only sector I are indicated for the
simplicity. The generation of the compensating sequence O2 is shown in Fig. 13.3a.
The simulation results of the compensating sequence O2 is shown in Fig. 13.3b,
and it can be seen that when the phase current achieves its maximum value, the
3L-NPC inverter output only two voltage levels between positive and negative
voltage level for 60° (in Fig. 13.3a zero voltage level for B phase is eliminated).
This feature will contribute to the loss reduction in the NPC diodes, because there is
no current conducting or switching in the NPC diodes during these periods when
the 3L-NPC inverter eliminate zero output voltage level. Therefore, sequence O2
can be considered as a thermal optimized sequence for the 3L-NPC inverter during
Table 13.1 The used vectors and their neutral point current of O1 sequence over an output
fundamental cycle
Sector Short vectors 1 iNP Short vectors 2 iNP Zero vectors iNP
I 211 −ia 110 −ic 000 0
II 121 −ib 110 −ic
III 121 −ib 011 −ia
IV 112 −ic 011 −ia 222
V 112 −ic 101 −ib
VI 211 −ia 101 −ib
Table 13.2 Comparisons of State vectors Short Short Zero

the used vectors and their type vectors 1 vectors 2
neutral point current in sector
I between O1 and O2 Sequences O1 O2 O1 O2 O1 O2
sequences Used vectors 211 100 110 221 000/222
Neutral point −ia ia −ic ic 0
current iNP
(a)
A
000 100 110 111 211 221 222 221 211 111 110 100 000
Ts/2 Ts/2 Ts/2 Ts/2
(b)
Fig. 13.3 Thermal optimized modulation sequence 2 (O2) for 3L-NPC inverter during LVRT.
a Sequence generating method. b Simulation results: output voltage pulses (green), grid voltage
(blue), phase current (red). Vg = 0.05 p.u
LVRT, and also it can compensate the neutral point current of optimized sequence
O1.
It can be seen that both of the sequences O1 and O2 start and end with zero
vector 000, and thereby they can be alternated in the neighboring carrier cycles to
enable the NP potential control ability without any extra switching processes.
Because both of the short vectors which disturb the NP potential in sequence O1 or
O2 will be compensated in the next carrier cycle, it is expected that the NP potential
control ability by alternating the O1 and O2 sequence will be better than the normal
sequence, in which only one short vector can be compensated in a switching cycle.
It is noticed that, the equivalent switching frequency is also doubled by the
optimized modulation sequence O2, the carrier frequency thereby needs to be
adjusted to half in order to acquire the same switching frequency as the normal
SVM sequence.
13.3 Loss and Thermal Performances 185
13.3 Loss and Thermal Performances
The loss distributions of the 3L-NPC wind power converter under LVRT when
utilizing the optimized vector sequences O1, O2, and O1 + O2 (enabling the NP
potential control ability) are compared in Fig. 13.4a, where Dcon and Tcon are the
conduction loss in diode and IGCT, respectively, Dsw and Tsw are the switching
loss in diode and IGCT, respectively. The comparison of thermal distribution is
shown in Fig. 13.4b. It can be seen that, both of the optimized sequences O1 and
O2 achieve the loss and thermal transferring from NPC diodes to other power
devices, and sequence O2 shows the best thermal distribution performance because
of the eliminated zero voltage level at maximum phase current. When alternating
the sequences O1 and O2 in neighboring switching cycles to enable the NP
(a) 6
5
4
Loss (KW)
3
Dcon
Dsw
2
Tcon
1 Tsw
0
O1+O2
O1+O2
O1+O2
Normal
Normal
Normal
O1
O2
O1
O2
O1
O2
Tout & Dout Tin & Din Dnpc
(b) 130.0
120.0
)
110.0
Junction Temp. (
100.0
Normal
90.0 O1
O2
80.0 O1+O2
70.0
60.0
50.0
Fig. 13.4 Thermal and loss among normal and optimized modulation sequences for the 3L-NPC
wind power inverter under LVRT. Vg = 0.05 p.u. (note: thermal resistances for diodes Dnpc, Din,
Dout are twice of that for the IGCT Tin, Tout). a Loss distribution comparison. b Mean junction
temperature comparison
potential control ability, the loss and thermal performances are a tradeoff between
the sequences O1 and O2.
The dynamic thermal performance of 3L-NPC wind power inverter which goes
from normal operation to extreme LVRT can be also simulated. The process is
conducted as follows: the inverter is first operating under the normal grid voltage
with normal modulation method, and the wind speed is set as 8 m/s with 3.2 MW
output active power. Then the grid voltage dips to 0.05 p.u. for 500 ms, during this
period both normal and optimized modulation methods are applied, respectively.
(a) Tjmax=116
Dnpc
Tout
Tin Dout
Din
Normal Voltage Normal

operation dips operation
Time (s)
(b)
Tjmax=94
Dnpc
Tout
Tin Dout
Din
Normal Voltage Normal
operation dips operation
Time (s)
Fig. 13.5 Junction temperature dynamic response with a voltage dip time 500 ms (from normal
operation with wind speed 8 m/s to 0.05 p.u. LVRT, and then back to normal operation). a Normal
modulation method is applied when LVRT. b Thermal optimized modulation method (O1 + O2) is
applied when LVRT
13.3 Loss and Thermal Performances 187
Afterwards the grid voltage recovers back to the normal operation status before the
LVRT. The junction temperature of each power device when applying different
modulation methods are shown in Fig. 13.4b, due to the effect of thermal capaci-
tances, the junction temperature of each power device fluctuates with fundamental
frequency.
As it can be seen, before the LVRT (simulation time 2 s), Dnpc is the most
stressed device with 90 °C maximum junction temperature and 9 K temperature
fluctuation amplitude. When LVRT is presented and normal modulation method is
still applied, as shown in Fig. 13.5a, the maximum junction temperature in Dnpc
significantly rise to 108 °C with 18 K fluctuation at the end of LVRT (simulation
time 2.5 s). If the O1 + O2 optimized modulation method is applied, as shown in
Fig. 13.5b, the maximum junction temperature in Dnpc keeps nearly unchanged and
it is reduced by 22 with 8 K less fluctuation compared to the normal modulation.
The junction temperature reduction may contribute to a lifetime extension of the
converter according to, e.g., the well-known Coffin-Masson lifetime model.
13.4 Conclusions
According to the investigations in this paper, the thermal optimization target for
3L-NPC wind power inverter under extreme LVRT is to reduce the junction
temperature in the NPC diode, which is the hottest power device of the whole
inverter system, and the conduction loss is dominant on this condition.
By the proposed insight generation method for modulation sequence of 3L-NPC
inverter, it is possible to develop a desired thermal optimized modulation as well as
a neutral point potential control method. The proposed thermal optimized modu-
lation sequence, which enables neutral point potential controllability for 3L-NPC
wind power inverter, can effectively reduce the conduction time of NPC diodes, and
achieve 32 K less steady-state junction temperature in NPC diodes under 0.05 p.u.
LVRT compared to the normal modulation method. However, due to the thermal
time constants as well as the short duration period of extreme LVRT, the tem-
perature improvement by the optimized modulation method is limited in the
dynamic LVRT process. The proposed thermal optimized modulation methods are
special feasible during the LVRT operation, where the modulation index is rela-
tively low and more redundant switching states can be utilized.
Nevertheless, the optimized modulation still achieves significant reduction of the
maximum junction temperature and temperature fluctuation by 22 and 8 K,
respectively, within the 500 ms LVRT duration, and more equal thermal distribu-
tion is realized between the different switching devices. It is expected that the
proposed optimized modulation can extend the lifetime of 3L-NPC wind power
inverter under low voltage ride through and satisfy stricter grid codes in the future
which may require longer LVRT time and higher reliability of the converter.
Chapter 14
Limits of the Power Controllability
of Three-Phase Converter
with Unbalanced AC Source
In order to analyze the controllability and performance of power electronics con-

verter under adverse AC source, a severe unbalanced AC voltage is first defined as a
case study in this special topic. As shown in Fig. 14.1, the phasor diagram of the
three-phase distorted AC voltage is indicated, it is assumed that the type B fault
happens with significant voltage dip on phase A of the AC source. Also there are
many other types of voltage faults which have been defined as type A–F.
Any distorted three-phase voltage can be expressed by the sum of components in
positive sequence, negative sequence, and zero sequence. For simplicity of analysis,
only the components with fundamental frequency are considered in this special
topic; however it is also possible to extend the analysis to higher order harmonics.
The distorted three-phase AC source voltage in Fig. 14.1 can be represented by
VS ¼ Vþ þ V þ V0
2 3 2 3
va sinðxt þ uþ Þ
6 7 6 7
¼ 4 vb 5 ¼ V þ 4 sinðxt 120 þ uþ Þ 5
vc sinðxt þ 120 þ uþ Þ ð14:1Þ
2 3 2 3
sinðxt þ u Þ sinðxt þ u0 Þ
6 7 6 7
þ V 4 sinðxt þ 120 þ u Þ 5 þ V 0 4 sinðxt þ u0 Þ 5
sinðxt 120 þ u Þ sinðxt þ u0 Þ
where V+, V−, and V0 are the voltage amplitude in positive, negative, and zero
sequence, respectively. And φ+, φ−, and φ0 represent the initial phase angles in
positive sequence, negative sequence, and zero sequence, respectively. The pre-
defined voltage dip as indicated in Fig. 14.1 should contain voltage components in
all the three sequences.
A typical used three-phase three-wire two-level voltage source DC-AC converter
is chosen and basically designed, as shown in Fig. 14.2 and Table 14.1, where the
converter configuration and the parameters are indicated, respectively. It is noted

190 14 Limits of the Power Controllability …
Type B
VC
VdipA
VB
Fig. 14.1 Phasor diagram definitions for the voltage dips in the AC source. VA,VB,VC means the
voltage of three-phases in the AC source
Fig. 14.2 Typical

three-phase three-wire
2L-voltage source converter
VA,VB,VC
Filter
Transformer
/Generator
Table 14.1 Converter Rated output active power Po 10 MW

parameters for case study
a
Rated load current Iload 1.75 kA rms
a
that the three-phase AC source is represented here by three windings with a

common neutral point, which can be the windings of electric machine or
transformer.
Because there are only three wires and a common neutral point in the windings
of AC source, the currents flowing in the three-phases do not contain zero sequence
component. As a result, the three-phase load current controlled by the converter can
be written as
14 Limits of the Power Controllability … 191
IC ¼ Iþ þ I ð14:2Þ
With the voltage of AC source in (14.1) and current controlled by converter in

(14.2), the instantaneous real power p and imaginary power q in αβ coordinate, as
well as the real power p0 in the zero coordinate can be calculated as
2 3 2 3
p va ia þ vb ib
6 7 6 7
4 q 5 ¼ 4 va ib va ib 5
p0 v0 0
2 3 ð14:3Þ
P þ Pc2 cosð2xtÞ þ Ps2 sinð2xtÞ
6 7
¼ 4 Q þ Qc2 cosð2xtÞ þ Qs2 sinð2xtÞ 5
0
Then the instantaneous three-phase real power p3Φ and imaginary power q3Φ of
the AC source/converter can be written as

p3/ p þ p0
¼
q3/ q
" # ð14:4Þ
P Pc2 Ps2
¼ þ cosð2xtÞ þ sinð2xtÞ
Q Qc2 Qs2
where P and Q is the average part of the real and imaginary power, Pc2, Ps2 and Qc2,
Qs2 are the oscillation parts, which can be calculated as
3
P ¼ ðvþ i þ þ vþ þ
q i q þ vd i d þ v q i q Þ
2 d d
3
Pc2 ¼ ðv i þ þ v þ þ þ
q i q þ vd i d þ v q i q Þ
ð14:5Þ
2 d d
3
Ps2 ¼ ðv i þ v þ þ þ
d i q vq i d þ v d i q Þ
2 q d
3
Q ¼ ðvþ iþ vþ þ
d iq þ vq id vd iq Þ
2 q d
3
Qc2 ¼ ðv iþ v þ þ þ
d iq þ vq id vd iq Þ
ð14:6Þ
2 q d
3
Qs2 ¼ ðv þ þ þ þ
d i d vq i q þ vd i d þ v q i q Þ
2
where a positive dq synchronous reference frame and a negative dq synchronous

reference frame are applied, respectively, to the positive and negative sequence
voltage/current. Each of the component on corresponding positive and negative dq
axis can be written as
vþ þ þ
d ¼ V cosðu Þ
vþ þ þ
q ¼ V sinðu Þ
ð14:7Þ
v
d ¼ V cosðu Þ
v
q ¼ V sinðu Þ
þ
iþ þ
d ¼ I cosðd Þ
þ
iþ þ
q ¼ I sinðd Þ

ð14:8Þ
i
d ¼ I cosðd Þ

i
q ¼ I sinðd Þ
Then the Eqs. (14.5) and (14.6) can be formulated as a matrix relation as
2 3 2 þ 32 3
P vd vþ
q v
d v
q iþ
6 Q 7 36 76 þ 7
d
6 6 v
7 ¼ 6 q
þ
vþ v vd 76 iq 7
4 Ps2 5 2 4 v
d q
7 4 i 5 ð14:9Þ
q v d vþq vþ
d 5 d
Pc2 vd v
q vþ
d vþ
q
i
q
It can be seen from (14.9) that if the AC source voltage is decided, then the
converter has four controllable freedoms (id+, iq+, id− and iq−) to regulate the
current flowing in the AC source. That also means: four control targets/functions
can be established. Normally, the three-phase average active- and reactive powers
delivered by the converter are the two basic requirements for a given application,
then two control targets have to first be settled as
P3/ ¼ P ¼ Pref
ð14:10Þ
Q3/ ¼ Q ¼ Qref
It is noted that different applications may have different requirements for the control
of average power,. E.g., in the power production application, the active power
reference Pref injected to the grid is normally set as positive, meanwhile large
amount of reactive power Qref may be needed in order to help and to support the
grid voltage. As for the electric machine application, the Pref is set as negative for
generator mode and positive for motor mode, there may be no or just a few reactive
power Qref requirements for magnetizing of electric machine. While in most power
quality applications, e.g., STACOM, Pref is normally set to be very small to provide
converter loss, and a large amount of Qref is normally required.
Consequently, for the three-phase three-wire converter system, there are only
two more current control freedoms left to achieve another two control targets
besides (14.10). These two additional control targets may be utilized to further
improve the performances of the converter under unbalanced AC source. However,
this special topic more focuses the evaluation of control limits and the control
possibilities under the whole voltage dipping range. In the following two of the
most mentioned control methods achieved by three-wire converter structure are

investigated under unbalanced AC source.
A. Elimination of Negative Sequence Current
In most of the grid integration applications, there are strict grid codes to regulate the
behavior of the grid connected converters. The negative sequence current which
always results in unbalanced load current may be unacceptable from the point view
of Transmission System Operator (TSO). Therefore, extra two control targets which
aim to eliminate the negative sequence current can be added as
i
d ¼0
ð14:11Þ
i
q ¼0
Translating the control targets in (14.10) and (14.11), all the controllable current
components can be calculated as
þ
2 vd Pref þ vþ
q Qref
iþ
d ¼ þ 2
3 ðvd Þ ðv
dÞ
2
ð14:12Þ
2 Pref vþ
iþ
q ¼ dþ iþ
3 vþ
d vq d
i
d ¼0
ð14:13Þ
i
q ¼0
When applying the current references in (14.12) and (14.13), the AC source
voltage, load current, sequence current amplitude, and the instantaneous power
Fig. 14.3 Simulation of the

converter with no negative
sequence current control
(three-phase three-wire
converter, Pref = 1 p.u.,
Qref = 0 p.u., Id− = 0 p.u.,
Iq− = 0 p.u., VA = 0 p.u., I+,
I−, I0 means the amplitude of
the current in the positive,
negative, and zero sequences,
respectively)
delivered by the converter are shown in Fig. 14.3. The simulation is based on the
parameters predefined in Fig. 14.2 and Table 14.1. The AC source voltage is set
with VA dipping to zero. The average active power reference Pref for the converter is
set as 1 p.u. and reactive power reference Qref is set as 0.
It can be seen from Fig. 14.3 that with the extra control targets in (14.11), there is
no zero sequence nor negative sequence component in the load current, i.e., the
currents among the three-phases of converter are symmetrical under the given
unbalanced AC source condition.
power with relation to the voltage amplitude of the dipping phase VA are shown in
Fig. 14.4a, b, respectively. It is noted that only positive sequence current is generated
by the converter, and there is up to ±0.5 p.u. oscillations both in the active- and reactive
power when VA dips to zero. The significant fluctuation of active power would result in
the voltage fluctuation of the DC bus, compromising not only the THD, but also the
reliability performances of the converter.
B. Elimination of Active Power Oscillation
In order to overcome the disadvantage of the active power oscillation under

unbalanced AC source, another two extra control targets, which aim to cancel the
oscillation items in the instantaneous active power, can be used to replace (14.11) as:
(a) (b) Pmax

I+ Pmin
Qmax
Qmin
I0 I-
Fig. 14.4 Profile of converter control with no negative sequence current (three phase three-wire
converter, Pref = 1 p.u., Qref = 0 p.u., Id− = 0 p.u., Iq− = 0 p.u.). a Sequence current amplitude
versus VA (I+, I−, I0 means the amplitude of the current in the positive, negative and zero
sequences, respectively). b P and Q oscillation range versus VA
P3/c2 ¼ Pc2 ¼ 0
ð14:14Þ
P3/s2 ¼ Ps2 ¼ 0
Then according to (14.9) each of the current components can be calculated as
2 3 2 þ 31 2 3 2 þ 3
iþ vd vþ
q v
d v
q Pref vd
d
6 iþ 7 6 vþ vþ v v 7 6 7 2Pref 6 vþ 7
6 q 7 ¼ 2 6 q d q d 7 6 0 7
¼ 6 q 7 ð14:15Þ
4 id 5 3 6 4 v v vþ þ 7 4
vd 5 0 5 3M vd 5
4
q d q
i
q v
d v
q vþ
d
þ
vq 0 vq
where
M ¼ ðvþ 2 þ 2 2 2
d Þ þ ðvq Þ ðvd Þ ðvq Þ ð14:16Þ
When applying the current references in (14.15), the corresponding source voltage,
load current, sequence current, and the instantaneous power delivered by the
converter are shown in Fig. 14.5. It can be seen that the active power oscillation at
twice of the fundamental frequency can be eliminated.
However, the disadvantage of this control strategy is also significant: First the
converter has to deliver up to 3 p.u. load current in the faulty phase, which is much
larger than the currents in other two normal phases—this large current may cause
over loading of the system and may result in failures. Moreover, significant fluc-
tuation of reactive power will be presented compared to the control strategy in
Fig. 14.5. In case of grid connected application, this significant reactive power
oscillation may cause grid voltage fluctuation—which is unpreferred especially
with weak grid and grid faults.
Fig. 14.5 Simulation of the

converter control with no
active power oscillation
(three-phase three-wire
converter, Pref = 1 p.u.,
Qref = 0 p.u., Ps2 = 0 p.u.,
Pc2 = 0 p.u., VA = 0 p.u. I+,
I−, I0 means the amplitude of
the current in the positive,
negative and zero sequences,
respectively
(a) (b)
P
I+

Qmax
I-
Qmin
I0
Fig. 14.6 Profile of converter control with no active power oscillation (three-phase three-wire
converter, Pref = 1 p.u., Qref = 0 p.u., Ps2 = 0 p.u., Pc2 = 0 p.u.). a Sequence current amplitude
versus VA. (I+, I−, I0 means the amplitude of the current in the positive, negative, and zero
sequences, respectively). b P and Q range versus VA
The current amplitude in the different sequences, as well as the delivered

active/reactive power with relation to the voltage amplitude on dipping phase is
shown in Fig. 14.6a, b, respectively. It is noted that the converter has to deliver both
positive and negative sequence current to achieve this control strategy, and up to
±1.3 p.u. oscillation in the reactive power is generated when VA dips to zero.
Another three possible control strategies which can eliminate the oscillation of
reactive power as shown in (14.17), or reduce the oscillations of both active-and
reactive power as shown in (14.18) and (14.19), are also possible for the
three-phase three-wire converter under unbalanced AC source
Qc2 ¼ 0
ð14:17Þ
Qs2 ¼ 0
Pc2 ¼ 0
ð14:18Þ
Qs2 ¼ 0
Ps2 ¼ 0
ð14:19Þ
Qc2 ¼ 0
14.1 Conclusion
In a typical three-phase three-wire converter structure, there are four current control
freedoms, and it may be not enough to achieve satisfactory performances under
unbalanced AC source, because either significantly oscillated power or over-loaded
current will be presented.

(Research Topics in Wind Energy 5) Ke Ma (Auth.) - Power Electronics For The Next Generation Wind Turbine System-Springer International Publishing (2015) PDF

Enviado por

Dados do documento

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

(Research Topics in Wind Energy 5) Ke Ma (Auth.) - Power Electronics For The Next Generation Wind Turbine System-Springer International Publishing (2015) PDF

Enviado por

Direitos autorais:

Formatos disponíveis

RESEARCH

More information about this series at http://www.springer.com/series/11859

ISSN 2196-7806 ISSN 2196-7814 (electronic)

Library of Congress Control Number: 2015944177

Springer Cham Heidelberg New York Dordrecht London

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media

2 Promising Topologies and Power Devices for Wind Power

3 Criteria and Tools for Evaluating Wind Power Converter . . . . . . 31

3.2 Classification and Approach for the Thermal Stress

4 Thermal Stress of 10-MW Wind Power Converter

5 Stress Analysis of 3L-NPC Wind Power Converter Under

5.4 New Power Control Methods Under Unbalanced

6 Conclusions and Future Works . . . . . . . . . . . . . . . . . . . . . . . . . . 95

Part II Specially Selected Topics

8 The Impacts of Power Switching Devices to the Thermal

9 Reliability-Cost Models for the Power Switching Devices

9.3 Analytical Solution of Junction Temperature with Chip

10 Electro-Thermal Model of Power Semiconductors Dedicated

11 Reactive Power Influence on the Thermal Cycling of Multi-MW

12 Thermal Loading of Several Multilevel Converter Topologies

13 Another Groups of Thermal Optimized Modulation Methods

14 Limits of the Power Controllability of Three-Phase

to many determining factors in wind power applications such as reliability, cost,

1.1 State-of-the-Art for Wind Power Generation

© Springer International Publishing Switzerland 2015 3

Worldwide wind power capacity

Fig. 1.2 Distribution of wind

1980 1985 1990 1995 2000 2005 2011 2018 (E)

1.2 Development of Wind Power Technologies

1.2.1 Evolution of Wind Turbine Concepts

1.2.2 Evolution of Power Electronics for Wind Turbines

1.3 Emerging Challenges for Wind Power Converter

Accompanying with the advancements of technology and power capacity, the

1.3.1 More Grid Supports

Fig. 1.8 Frequency control Available power

Fig. 1.9 Reactive power P/Prated (p.u.)

Fig. 1.10 Voltage proﬁle for

Fig. 1.11 Reactive current Dead band

1.3.2 Higher Reliability

Turbines Gearbox Generator Electrical Control

Stress Analysis Strength Modeling

Mission profile to component stress. Design tools for Lifetime model

Monitoring & Control Probability & Statistics

Temperature measurement Reliability engineering

Fig. 1.13 Multi-discipline approaches for more reliable power electronics

1.3.3 Special Cost Considerations

Estimated Cost of Energy in 2018

1.3.4 Formulation of Overall Requirements

Generator side Wind Power Grid side

1. Controllable I 1. Energy balance/storage 1. Fast/long P response

1.4 Scopes of the Book

1. Liserre M, Cardenas R, Molinas M, Rodriguez J (2011) Overview of Multi-MW wind turbines

2.1 Promising Converter Topologies

2.1.1 Traditional Two-Level Converters

© Springer International Publishing Switzerland 2015 19

in the case of long cable connections. Consequently, it is becoming more difﬁcult

2.1.2 Multilevel Converters

B. Three-Level H-Bridge Back-to-Back Topology (3L-HB BTB)

2.1.3 Multi-cell Converters

Besides the solution of multilevel converters, some conﬁgurations with multi-cell

A. Multi-cell Converter with Paralleled Converter Cells (MC-PCC)

Fig. 2.7 Cascade H-bridge