Contributions To Modelling, Analysis and Simulation of AC Drive Systems-Application To Larg

“Dunarea de Jos” University - Galati
Department of Electrotechnics, Electrical Machinery and

Drive Systems
Contributions to Modelling, Analysis and

Simulation of AC Drive Systems.
Application to Large Wind Turbines
Ing. FLORIN IOV
Scientific Supervisor:
Prof. Dr. Ing. Calueanu Dumitru
Faculty of Electrical Engineering and Computer Science

“Dunarea de Jos” University - Galati
2003
2
Preface
The present thesis summarizes a part of the work carried out during the research project
“Electrical Design and Control - A Simulation Platform to Model, Optimize and Design Wind
Turbines”. The project is a research collaboration between Aalborg University, Institute of
Energy Technology and RISØ National Laboratory/VEA with the support from the Danish
Energy Agency.
This work concentrates on modelling and analysis of the large wind turbines, with a focus on
the electrical part. A Matlab/Simulink Toolbox which contains models for the electrical
components of a wind turbine as well as the mechanical ones has been developed. However,
an important effort should be done in the future to complete this Toolbox.
I would like to thank Professor Dumitru Calueanu at the Department of Electrotechnics,
Electrical Machines and Drive Systems, Galati University, for being my supervisor.
Very special thanks to Professor Frede Blaabjerg from Department of Energy Conversion,
Aalborg University for his continuous support and valuable guidance during the last years.
I want also to thank Anca-Daniela Hansen and Poul Sørensen from RISØ National
Laboratory/VEA for their assistance with detailed knowledge within control of wind turbines.
Thanks also to the colleagues from the Department of Electrotechnics, Electrical Machines
and Drive Systems, Galati University as well as from the Department of Energy Conversion,
Aalborg University for their good advices and support.
However, the persons who deserve most thanks are my wife and my daughter.
Aalborg, spring 2003
Florin Iov
3
4
Abstract
The present thesis summarizes a part of the work carried out during the research project
“Electrical Design and Control - A Simulation Platform to Model, Optimize and Design
Wind Turbines”. The project is a research collaboration between Aalborg University,
Institute of Energy Technology and RISØ National Laboratory/VEA with the support from
the Danish Energy Agency.
Since the MW-sized wind turbines (up to 3MW) are available on the market and it is
expected that in the near future the size will increase further, as well as the application of
power electronics in wind turbines, the modelling and simulation techniques will play an
important role in the design and analysis of these large wind turbines
This work deals with the modelling and analysis of the electrical generators and power
converters within a large wind turbine as a part of the above mentioned simulation platform
project.
A new Matlab/Simulink Toolbox for wind turbine applications is developed and some
applications of it are shown. Some models from this toolbox have been used and validated in
a particular analysis for the Middelgrunden Windfarm (near Copenhagen), Denmark.
Since the induction machine is currently the most widely used electrical generator in wind
turbine applications, a big effort has been paid in dynamic and steady state modelling of it.
These models are derived in the state space form in the abc/abc reference frame as well as in
the dqo-dqo arbitrary reference frame both with currents and fluxes as state-variables. Some
special features, e.g. deep-bar effect, have been included in models using only parameters
from data-sheets.
As simulation speed is a very important issue in any simulation tool, especially when the
considered system should be analysed in a long time horizon, the implementation methods in
Matlab/Simulink have been investigated in detail. It has been found that using C S-Functions
the simulation speed can be increased with at least a factor of two, while the numerical
stability increases.
One of the main targets has been the development of a new model for induction generator,
both squirrel-cage and doubly-fed machine for the advanced aero elastic simulation tools.
Since here the time scale is in the range 10-1 – 104 sec the reduced order modelling of the
induction machine has been considered. It has been found that using a reduced order model in
the advanced aero elastic simulation tools, some resonances between the mechanical part and
the electrical part of the wind turbine arises due to the dynamics of the induction generator.
This behaviour has been observed also in the field measurements.
Currently, many wind turbines, up to 2.3 MW, are based on the “Danish concept” in which a
squirrel-cage induction generator is directly connected to the grid. An AC-controller is used
during the start-up sequence of the generator in order to limit the inrush currents and the
5
starting torque transients in the drive train. Based on the switching function concept the
mathematical models for these AC-controllers have been developed taking into account the
winding connection type of the generator.
An overview of the control methods for induction generators, both squirrel-cage and doubly-
fed machine is presented. The basic control structures for these generators, both grid-
connected and stand-alone mode are treated. A controllers design method based on the
complete state-space model of the generator including the model of the drive train is
presented. The influence of disturbance in the stator voltage in the rotor current control loops
is shown. Again one of the main targets here was to develop a simple control structure for the
active and reactive power for the doubly-fed induction generator, which will be implemented
in HAWC.
The new Matlab/Simulink Toolbox for wind turbine applications as a part of a new
simulation platform dedicated to model, optimize and design wind turbines has been
presented. The content of the main libraries from this toolbox has been briefly presented.
Then, in order to highlight the capabilities of this Toolbox some simulation demos are shown.
The main contribution of the thesis is in modelling of the components within a wind turbine.
A collection of these mathematical models is available in a new Matlab/Simulink toolbox
dedicated for wind turbine applications. Using the available models from the Toolbox the
main concepts for large wind turbines, namely fixed speed wind turbines with squirrel-cage
induction generator and variable-speed ones using doubly-fed machines, can be analyzed and
simulated under different operation conditions.
6
Table of contents
Preface .......................................................................................................................................3
Abstract......................................................................................................................................5
Table of contents .......................................................................................................................7
Chapter 1 Introduction ...........................................................................................................9

1.1 Background................................................................................................................9
1.2 Scope of work..........................................................................................................11
1.3 Review of related research in modelling .................................................................12
1.4 Contributions ...........................................................................................................13
1.5 Outline .....................................................................................................................14
1.6 References ...............................................................................................................14
Chapter 2 Large Wind Turbines...........................................................................................15

2.1 Proprieties of the wind.............................................................................................15
2.2 Wind turbine performances .....................................................................................17
2.3 Wind turbines technology........................................................................................24
2.4 Summary..................................................................................................................27
2.5 References ...............................................................................................................27
Chapter 3 Induction Machine Modelling .............................................................................29

3.1 The voltage equations formulated in their natural reference frame.........................29
3.2 The voltage equations in arbitrary reference frame variables .................................35
3.3 Mathematical model of a gearbox with torsional torques .......................................41
3.4 Steady-state equations .............................................................................................43
3.5 Modelling deep-bar effect .......................................................................................50
3.6 Modelling saturation................................................................................................53
3.7 Modelling iron losses...............................................................................................55
3.8 Induction machine modelling using Matlab/Simulink ............................................57
3.9 Summary..................................................................................................................61
3.10 References ...............................................................................................................61
7
Chapter 4 Reduced order modelling and eigenvalue analysis of induction machine .......... 63
4.1 Reduced order modelling of induction machine ..................................................... 64
4.2 Linearized equation of induction machine.............................................................. 66
4.3 Eigenvalue analysis................................................................................................. 67
4.4 HAWC implementation .......................................................................................... 77
4.5 Summary ................................................................................................................. 79
4.6 References............................................................................................................... 79
Chapter 5 Power Converters Modelling .............................................................................. 81

5.1 AC Controllers ........................................................................................................ 81
5.2 Voltage Source Converters ..................................................................................... 88
5.3 Summary ................................................................................................................. 93
5.4 References............................................................................................................... 93
Chapter 6 Control of variable-speed wind energy conversion systems............................... 95

6.1 Introduction............................................................................................................. 95
6.2 Control for WECS................................................................................................... 96
6.3 Design case study: Doubly-fed induction generator controllers........................... 102
6.4 Summary ............................................................................................................... 116
6.5 References............................................................................................................. 116
Chapter 7 Wind Turbine Blockset ..................................................................................... 119

7.1 Simulation Platform Concept................................................................................ 119
7.2 Matlab/Simulink “Wind Turbine Blockset” ......................................................... 121
7.3 Wind Turbine Blockset Demos............................................................................. 126
7.4 Summary ............................................................................................................... 130
Chapter 8 A flexible control strategy for small stand alone wind turbines ....................... 131
8.1 Background ........................................................................................................... 131
8.2 System description ................................................................................................ 132
8.3 Grid-connected mode of operation ....................................................................... 134
8.4 Stand-alone mode of operation ............................................................................. 135
8.5 Experimental results.............................................................................................. 136
8.6 Summary ............................................................................................................... 138
Chapter 9 Conclusions and Future work ........................................................................... 139
Appendix A Typical parameters for induction generators and drive train .......................... 143
Appendix B List of publications .......................................................................................... 145
8
Chapter 1
Introduction
1.1 Background
The global energy demand has been in a continuous growth in the last century as the energy
is essential for economic development and rising living standards [1]. There is a strong
relationship between income and energy demand as shown in Fig. 1.1.
Fig. 1.1. Per capita primary energy consumption versus income across countries and time [1].
Per capita primary energy consumption grows with income in a similar pattern across
countries and time and it is clear that in a longer term the energy demand will increase
further.
The energy suppliers are moving from high to low carbon fuel towards electricity as the
dominant energy carrier. This process is driven by some demands as security, cleanliness and
sustainability. The global primary energy shares in the past and as it is expected in the future
is presented in Fig. 1.2. It is evident that since the so-called “oil crisis “ from 1970ths coal and
oil are becoming substituted by renewables and biofuels while the nuclear energy is in a
continuous reduction.
Among the renewables sources, wind energy was the fastest growing energy technology in
the last decade in terms of percentage of yearly growth of installed capacity as shown in Fig.
1.3.
9
Fig. 1.2. Global primary energy share - forecast to 2050 [1]
Fig. 1.3. Actual annual wind power development and forecast for 2005 [3].
However, this growth is not evenly distributed in the world and Europe is by far the leading
region in terms of wind power installed capacity during the year 2001 with around 86% from
the total. The main driver for this development was the initiation of the national programmes
concerning the wind energy in the 1990ths as well as the creation of different financial
support systems, e.g. fixed feed-in tariffs [3].
Today the MW-sized wind turbines are available on the market and it is expected that in the
near future the size will increase further. Some manufacturers have already announced the
new wind turbine class of 4-5 MW. Most of these larger wind turbines are used in offshore
applications due to the fact that the offshore wind energy resources are significantly higher
than the onshore ones as well as due to the environmental issues.
The electric system contained by a large wind turbine as well as within an offshore wind farm
and its connection to the main power system arises new challenges to engineers. Since,
currently both directly grid connected wind turbine and variable speed wind turbines are used
10
it is of great interest to analyse and predict the behaviour of these wind turbine in some
specific operating conditions e.g. steady-state, faults, etc.
1.2 Scope of work

Simulation is the most important technique used today for evaluation of the engineering
solutions, especially in the case of large wind turbines and wind farms. In general different
simulation tools exist to simulate power converters, generators, power systems, mechanical
parts and aero elastic behaviour of wind turbines at different levels and time scale.
Depending on the modelling aspects for each level and therefore each time scale, particular
software should be used as shown in Table 1.1.
Currently the advanced aero elastical tools used by the wind turbine industry, e.g. HAWC,
Flex4/5 have focus mainly on the aerodynamic and mechanical parts of the wind turbines,
while the electrical aspects are neglected. The dedicated power system analyses tools, e.g.
DigSilent, EMTDC, for assessment of power quality and analysis of the wind turbine
interaction with the grid have no focus on the mechanical aspects of the wind turbine.
Moreover, simulation tools as Pspice and Saber, used in circuit and systems power design
including electrical, thermal, magnetic and mechanical components are currently not focused
on wind turbine applications.
Table 1.1. Modelling aspects and modelling tools for wind turbines and wind farms.
Modelling aspect Time scale Tool

Wind turbines mechanical loads 10-1 – 104 sec HAWC, Flex4/5
-1 4
Wind turbines power quality 10 – 10 sec DigSilent, EMTDC, Saber,… Matlab
Wind turbines swirtchings 10-3 – 101 sec DigSilent, EMTDC, Saber,… Matlab
-6 0
Grid faults 10 – 10 sec DigSilent, EMTDC, Saber, … Matlab
Power converters control and switchings 10-6 – 100 sec Saber, Pspice, … Matlab
-3 4
Wind turbine control system 10 – 10 sec Saber,… Matlab
However, Matlab/Simulink, which provides a powerful graphical interface for building new
models as well as new control strategies can be used to cover a wide range of modelling
aspects and time scales. Moreover, using a dSPACE prototyper, who provides a very high
computational power and flexibility degree, new control strategies can easily be implemented
in tested in a Hardware-In–the-Loop structure.
Thus, in order to fill the existing gaps in modelling and analysis of wind turbines, Aalborg
University and RISØ National Laboratory have initiated a project entitled “A simulation
platform to model, optimize and design wind turbines” with the support of the Danish Energy
Agency. A description of this project will be presented, later in Chapter 7. However, the main
goals in this project are:
• Develop an extended simulation platform with special focus on simulating the
electrical parts of the wind turbines.
• Develop new models for the electrical part of the wind turbine, which can be used in
the advanced mechanical/aero elastical design tools (e.g. HAWC and FLEX 4/5):
o reduced order models for the electrical generators;
11
o steady state models for power converters.
• Develop dynamic and steady state models for all components within a wind turbine,
which in a longer term can be used in a complete optimisation of a wind turbine
system.
o Models for mechanical part: wind, drive train, active and passive stall wind
turbine, variable pitch wind turbine;
o Models for generators: squirrel-cage, doubly-fed induction generator,
synchronous generator, permanent magnet synchronous generator;
o Models for power converters: soft-starters, back-to-back voltage source
converter, multi-level converter, matrix converter;
o Models for three-phase transformers: two-winding, three-winding;
o Models for cables and distribution lines;
o Grid models.
Matlab/Simulink is used as a general model-developing tool and also for validation of the
models. After validation of the models they are implemented in the other tools. Finally, all
these models are collected into a dedicated Simulink Toolbox for wind turbine applications.
The overall demands to these models are:
• Models should be open;
• Parameters based on datasheets;
• Models first based on common literature. Next improvement may be added;
• Simulation speed important;
• User friendly: documentation and reliability
• Easy to extend with extra features.
Using these models two main wind turbine configurations should be modelled and analyzed
in the first stage: namely the active stall wind turbine with squirrel-cage induction generator
and variable-speed wind turbine with doubly-fed induction generator.
1.3 Review of related research in modelling

Regarding the induction machine modelling many approaches can be found in literature, but
these papers do not handle the specific problems, which arise into a large machine, both
squirrel-cage and doubly-fed concepts.
Each existing wind turbine design requires a special modelling approach for the generator
and/or power converter. A directly grid-connected wind turbine will demand an abc/abc
model for the induction generator as well as a particular model for each AC-controller (soft-
starter) configuration, e.g. delta or branch-delta. Moreover, in studying of unbalanced or fault
operation conditions, an abc/abc model is much more suitable than a dqo one.
Some eigenvalues analysis based on the commonly accepted fifth order model for doubly-fed
induction machine are reported in literature, but the focus is on slip energy recovery drives,
12
which implies a uni-directional power flow. Although none of these papers do not take into
account the effect of the drive train and the variations of the rotor parameters with the slip
(deep-bar effect), which is significant for a large doubly-fed induction machine.
Considerable attention has been paid to the reduced order modelling of induction machine
and a simplified second-order model has been developed as an alternative to third-order
model. Nevertheless, there are no references regarding the reduced order modelling of large
induction generators and the influence of the drive train in steady-state operation as well as
during the transients.
Regarding the modelling of AC controllers there are few references, and almost all deals with
modelling of the star connected loads.
However a detailed state-of-the art for each major item from this thesis will be presented in
detail in front of the each chapter.
1.4 Contributions
The contributions of this thesis are:
• A unified approach in dynamic modelling of induction machine, both squirrel-cage
and doubly-fed is presented in Chapter 3. The dynamic equations are derived in the
state space form in the abc/abc reference frame as well as in the dqo-dqo arbitrary
reference frame both with currents and fluxes as state-variables. Then, how to include
saturation of the main path, iron losses and the variation of the rotor parameters with
the slip is shown.
• Evaluation of the electrical machine performances using frequency response. Starting
from the state-space model of the machine, the linearized model is obtained and then
the linear system theory can be used in analysis of the machine. The eigenvalue
analysis can be performed and the transfer functions between the desired input-output
pair can be extracted easily. Then the machine performances can be analyzed using
the frequency response.
• A new method based on switching function in modelling of soft-starters for wind
turbine applications is presented in Chapter 5. Then the usual converter topologies,
which are used in wind turbine applications, are modelled based on the switching
function concept. Moreover the connection type of the load is taken into account.
• A synthesis of the control algorithms for wind energy conversion systems using both
squirrel-cage machines and doubly-fed ones.
• A new design method for the control loops which uses transfer functions derived from
the linearized model of the machine.
• Developed models for the main components within a wind turbine, which can be used
in analysis and simulation of wind energy conversion schemes. These models are
collected into a new Matlab/Simulink Toolbox for wind turbine applications.
13
1.5 Outline
• Chapter 2 - A description of the wind proprieties and the performances of the
modern wind turbines, as well as the power control methods are shown in this
chapter. Then, the current status regarding the wind turbine technology is presented
• Chapter 3 - This chapter deals with the induction machine modelling in state space
form, using both fluxes and currents as state variables. How to include saturation, iron
losses and the variation of the rotor parameters with the speed (deep-bar effect) is also
presented.
• Chapter 4 - Reduced order modelling and eigenvalue analysis of the induction
generators are presented. This analysis is performed in two cases: using a simple
equation of motion and the complete model of the drive train. The influence of the
deep-bar effect is also taken into account.
• Chapter 5 - The power converters used currently in wind turbine applications are
modelled based on switching functions. The connection type of the AC controllers fed
induction machine is considered in modelling.
• Chapter 6 - An overview of the control methods for induction generators, both
squirrel-cage and doubly fed machine is presented. The basic control structures for
these generators, both grid-connected and stand-alone mode is shown. A controller
design method based on the complete state-space model of the generator including the
model of the drive train is presented.
• Chapter 7 – A new Wind Turbine Blockset as a part of the Matlab/Simulink
environment is developed. This toolbox contains the electrical models as well as the
mechanical ones within a wind turbine. Some applications of this new toolbox are
shown.
• Chapter 8 – A flexible control strategy for an 11 kW wind energy conversion system
capable of working both in stand-alone and grid-connected configuration is presented
• Chapter 9 – The conclusions of the whole thesis are presented as well as the
proposed future work
• Appendix A – Some typical parameters for induction generators and wind turbine
drive train are presented
• Appendix B – List of publications is presented.
1.6 References
[1]. ***-– Energy needs, choices and possibilities. Scenarios to 2050, Shell International 2001;
[2]. M. Toennes, P. Thoegersen, G.K. Andersen, F. Blaabjerg – Active and Integrated Power – Agile solutions
for low power drives,
[3]. *** - International wind energy development – World market update 2000. Forecast 2001-2005, BTM
Consult ApS, March 2001;
[4]. T. Ackermann – Transmission systems for offshore wind farms, IEEE Power Engineering Review,
December 2002, pp.23-27;
14
Chapter 2
Large Wind Turbines

A description of the wind proprieties and the performances of the modern wind turbines, as
well as the power control methods are shown in this chapter. Then, the current status
regarding the wind turbine technology is presented.
2.1 Proprieties of the wind

The understanding of the characteristics of the wind resource is very important to all aspects
of wind energy exploitation, from the identification of the suitable sites and predictions of the
economic viability of wind turbine/farm projects through to the design of the wind turbines.
Since the wind is highly variable, both geographically and temporally the understanding of
the wind turbines effect on electricity distribution networks became a very important issue
[1].
2.1.1 Wind distribution
The wind sped variations during the year can be well characterized in terms of a probability
distribution. It has been found that the Weibull distribution gives a good representation of the
variation in hourly mean wind speed over a year at many typical sites [2], [3]. This
distribution takes the form:
k
v 
− ∞ 
F ( v∞ ) = e  c 
(2.1.1)
where F ( v ∞ ) is the fraction of time for which the hourly mean wind speed exceeds v∞ , k is
the shape parameter and c is the scale parameter.
The scale parameter c is related to the annual mean wind speed by:
 1
v ∞ = cΓ 1 +  (2.1.2)
 k
where Γ is the complete gamma function. Based on (2.1.2) the probability density function is:
(v )
k
k −1 v 
dF − ∞ 
f ( v∞ ) = − = k ∞k e  c 
(2.1.3)
dv∞ c
since the mean wind speed is given by:
∞
v ∞ = ∫ v∞ f ( v∞ ) dv∞ (2.1.4)
0
A special case of the Weibull distribution is the Rayleigh distribution, with k=2, which is
actually a reasonably value for many sites. In [2] some Weibull distributions for different
values of the shape parameter k. are given as shown in Fig. 2.1.
15
Fig. 2.1. Example of a Weibull distribution [2].
Since for some of the sites the seasonal component of the wind is very important, a double-
peaked “bi-Weibull” distribution, with different scale factors and shape factors in the two
seasons can be used:
k1 k2
v  v 
− ∞  − ∞ 
F ( v ∞ ) = F1e  c1 
+ (1 − F1 ) e  c2 
(2.1.5)
2.1.2 Synoptic and diurnal variations
In addition to seasonal changes in the wind speed there are some variations on the shorter
time scale. These variations are called synoptic variations and they have a peak in frequency
at around 4 days. The synoptic variations are related with large-scale weather patterns such as
areas of high and low pressure and associated weather fronts as they move across the surface
of the earth.
Moreover many locations will show a distinct diurnal peak at a frequency of 24 hours. Local
thermal effects usually drive this diurnal peak. Intense heating in the daytime may cause
large convection cells in the atmosphere, which die down at night.
2.1.3 Turbulence
Besides the seasonal components in the wind speed there is a turbulence component. This
turbulence refers to fluctuations in wind speed on relatively fast time-scale, typically less
than 10 min. In studying the wind energy conversion into a wind turbine it is very useful to
consider the mean wind speed determined by the seasonal, synoptic and diurnal effects,
which varies on a time scale of one to several hours, with turbulent fluctuations
superimposed.
The turbulence can be described based on some indicators as: intensity, frequency spectra,
cross-spectra etc.
The turbulence intensity depends on the surface roughness and it is prescribed in the
standards used for wind turbine design calculations as the Danish Standard (DS 472), the
Germanischer Lyod rules (GL), the IEC standard, EuroCode, etc [2].
16
The turbulence spectra describe the frequency content of wind-speed variations. Commonly,
two alternative formulations for the spectrum are used: the Kaimal and the von Karman
spectra. The von Karman spectrum gives a good description for turbulence in wind tunnels,
although the Kaimal spectrum may give a better fit to empirical observations of atmospheric
turbulence [2]. Some recent work suggested that the von Karman spectrum gives a good
representation of atmospheric turbulence above 150 m, but has some deficiencies at lower
altitudes [2]. Several modifications have been proposed and a modified von Karman
spectrum has resulted [2].
Two new methods for non-stationary turbulence component generation, using adjustable
shaping filters are proposed in [7]. The first method based on the non-integer order filter
leads to a non-parametric model for the wind, but requires considerable computation effort,
while the second one, which use a rational shaping filter is much more faster and therefore
much more attractive for computer simulation.
In addition to the temporal variation there is a space variation of the turbulence in the lateral
and vertical directions due to rotation of the blades. In order to model these effects, the
spectral description of the turbulence must be extended to include information about the
cross-correlations between turbulent fluctuations at points separated laterally and vertically
[2].
2.1.4 Gust
In wind turbine applications it is often to know the maximum gust speed, which can be
expected to occur in any given time interval. This is usually represented by a gust factor G,
which is the ratio of the gust wind speed to the hourly mean wind speed, The gust factor is a
function of the turbulence intensity, and it also depends on the duration of the gust.
2.2 Wind turbine performances
2.2.1 The performance curves
Mainly three indicators, namely power torque and thrust, which are varying with the wind
speed, characterize the performances of a wind turbine. The power determines the amount of
energy captured by the rotor, while the torque developed determines the size of the gearbox
and must be matched the electrical generator which is driven by the rotor. The rotor thrust
has a great influence on the structural design of the tower. In order to characterize the actual
performances of a wind turbine regardless of how it operates e.g. at constant or variable
speed these indicators are expressed in non-dimensional terms as functions of tip speed ratio.
2.2.1.1 Tip speed ratio
The tip speed ratio of a wind turbine is defined as:

ΩR
λ= (2.2.1)
v∞
where: Ω is the mechanical speed at the rotor shaft of the wind turbine, R is the blade radius
and v∞ is the actual wind speed.
17
2.2.1.2 The CP-λ performance curve
The power coefficient CP for a wind turbine is defined based on the extracted power as:
Power
Cp = (2.2.2)
0.5ρv3∞ A
where: ρ is the air density and A is the rotor swept area.
The theoretical maximum value of the power coefficient is CP = 0.593 and it is called the
Betz limit. The CP-λ curve for a modern three blade wind turbine is shown in Fig. 2.2.
Fig. 2.2. CP-λ performance curve for a modern three blade wind turbine.
It can be observed that the maximum value of CP is much less than the Betz limit due to the
imperfect blade design and the drag and the tip losses.
The power developed by the wind turbine is:
P = 0.5ρπR 2 v3∞ CP (2.2.3)
Usually, the CP curve is used in the control issues for a wind turbine in order to maximize the
extracted power from the wind at a given moment.
2.2.1.3 The CQ-λ performance curve
The torque coefficient CQ is derived from the power coefficient simply by dividing it with the
tip speed ratio. The CQ-λ curve is used for torque assessment purposes when the rotor is
connected to a gearbox and generator. The CQ-λ curve for a modern three blade wind turbine
is shown in Fig. 2.3.
The torque developed by the wind turbine is given by:
T = 0.5ρπR 3 v ∞2 CQ (2.2.4)
18
Thus the aerodynamic torque is not computed through power expression T = P/Ω, which
cannot be used at standstill ( Ω = 0 ). Notice that at standstill the aerodynamic power is zero,
while the aerodynamic torque is not.
Fig. 2.3. CQ-λ performance curve for a modern three blade wind turbine.
2.2.1.4 The CT-λ performance curve
The thrust force on the rotor is directly applied to the tower on which the rotor is supported
and so considerably influences the structural design of the tower.
2.2.2 Constant rotational speed operation
At present, the directly grid-connected wind turbine is the most widely used to generate
electricity. These wind turbines both grid-connected or stand alone need to produce electrical
power at constant frequency. So, the most common mode of operation for a wind turbine is
constant rotational speed. A constant speed wind turbine connected to the grid is
automatically controlled whereas a stand-alone machine needs to have speed control and a
means of dumping excess power.
2.2.2.1 The Kp-1/λ curve

An alternative performance curve, so-called the Kp-1/λ curve, can be obtained for a turbine
controlled at constant speed. The CP-λ curve shows how the power would vary with
rotational speed if the wind speed has been held constant, while the Kp-1/λ curve describes
how the power would change with wind speed when constant rotation speed is enforced. The
KP coefficient is defined as:
Cp
KP = (2.2.5)
λ3
The CP-1/λ and Kp-1/λ curves for a typical fixed-pitch wind turbine are shown in Fig. 2.4.
19
Fig. 2.4 Non-dimensional performance curves for constant speed operation for
a modern three-blade wind turbine.
The efficiency of the wind turbine in this case (given by the CP-λ curve) varies greatly with
the speed, which is a drawback of the constant speed operation. However, the wind turbine
should be designed such that the maximum efficiencies are achieved at wind speeds where is
the most energy available.
2.2.2.2 Stall regulation
An important feature of the Kp-1/λ curve is that the power, initially, falls off once stall has
occurred and then gradually increases with the wind speed. This feature provides an element
of passive power output regulation, so that the generator is not overloaded as the wind speed
increases. Currently, the stall regulation is the most commonly adopted control method for
directly connected wind turbines in order to limit the maximum power generated.
The main advantage of the stall control is simplicity, but there are also some drawbacks [2]
as:
• The power versus wind speed curve is fixed by the aerodynamic characteristics of the
blades, in particular the stalling behaviour;
• The post stall power output of a turbine varies unsteadily;
• The stalled blade exhibits low vibration damping and as a result large bending
moments and stresses.
So, a fixed-pitch, stall-regulated turbine will experience more severe blade and tower loads
than a pitch regulated turbine.
2.2.2.3 Effect of rotational speed change

The power output of a turbine, which running at constant speed, is strongly governed by the
operational rotation speed. At low rotation speed the power reaches a maximum at low wind
speed and consequently it is low. To extract energy at wind speeds higher than the stall peak
the turbine must operate in a stalled condition and so is very inefficient. At high rotational
speed the turbine will extract a great deal of power at high wind speeds but at moderate wind
20
speeds it will be operating inefficiently because of the high drag losses. The sensitivity to the
rotational speed of the output power is shown in Fig. 2.5.
Fig. 2.5. Effect on extracted power of rotational speed.
2.2.2.4 Effect of blade pitch angle change
Another parameter, which affects the power output, is the pitch setting angle of the blades β.
Blade design always involves twist but the blade can be set at the root with an overall pitch
angle. The effects of a few degrees of the pitch angle are shown in Fig. 2.6.
Fig. 2.6. Effect on extracted power of blade pitch set angle.
Small changes in pitch setting angle can have a big effect on the output power. A turbine
rotor designed to operate optimally at a given set of wind conditions can be suited to other
conditions by appropriate adjustment of blade pitch angle and rotational speed.
2.2.2.5 Pitch regulation
Some of the drawbacks of fixed-pitch / passive stall regulation can be overcome by providing
an active pitch angle control.
21
The most important application of pitch control is for power regulation. However, the pitch
control has also other advantages as:
• by setting a large positive pitch angle a large starting torque can be generated;
• a 90o pitch angle is used during the shut-down sequence because this minimises the
rotor idling speed.
2.2.2.6 Pitching to stall
At wind speeds below the rated level the blade pitch angle is kept at 0o as shown in Fig. 2.6.
As rated power is reached only a small negative pitch angle is necessary to support stalling
and hence, to limit the power to the rated level. As the wind speed increases small
adjustments in both positive and negative directions will maintain constant power.
2.2.2.7 Pitching to feather
In order the keep the airflow attached to the blade the pitch angle is increased with the wind
speed when the rated power is reached. Moreover a good, positive damping is provided.
Feathered blade parking and assisted parking are also available using this control method.
2.2.3 Variable speed operation
If the speed of the wind turbine rotor is continuously adjusted in order to maintain the tip
speed ratio at the level, which gives the maximum CP, then the efficiency of the turbine will
be considerably increased.
Active pitch control is used to maintain a constant tip speed ratio but only in the process of
adjustment of the rotational speed. However, the pitch angle should always return to the
optimum setting for highest efficiency. Pitch control regulation is also required for wind
speed above the rated value, when the rotational speed is kept constant.
A turbine operating at variable speed (constant tip speed ratio) until maximum power is
reached and then at constant speed, constant power and pitch control would capture the
maximum possible amount of energy in a given time as shown in Fig. 2.7.
Fig. 2.7 Power curve for a modern three-blade variable speed wind turbine.
22
2.2.4 Power control methods
2.2.4.1 Passive stall control
The simplest form of power control is passive stall control, in which the fixed-blade pitch is
chosen so that the turbine reaches its maximum power or rated power at the desired wind
speed.
2.2.4.2 Active stall control
This power control method achieves power limitation above rated wind speed by pitching the
blades initially into stall using a negative pitch angle.
2.2.4.3 Passive pitch control
This method involves a special design for the wind turbine blades and/or its hub mounting, so
that they will twist under the action of loads on its. As a result the desired pitch changes at
higher wind speed will be achieved and so the output power will be limited.
Even if this principle is easy to state, it is difficult to achieve in practice, because the required
variation in the blade twist with the wind speed generally does not match the corresponding
variation in blade load [2]. Since the optimization of the extracted energy is not the key
objective for a stand-alone wind turbine, the passive pitch control is sometimes adopted.
However, this concept has not been utilized yet for grid connected wind turbines.
2.2.4.4 Active pitch control

Active pitch control achieves power limitation above rated wind speed by rotating all or part
of each blade about its axis in order to maximize the extracted power. The extra energy
obtainable with the pitch regulated wind turbine is not that large compared with a stall-
regulated one [2] as shown in Fig. 2.8.
Fig. 2.8. Power curves for stall-regulated and pitch-regulated wind turbine.
Although the active pitch control requires a pitch actuation system in order to modify the
pitch angle. Currently, the hydraulic actuators as well as the electric ones are used, each type
having its own particular advantages and drawbacks.
23
2.3 Wind turbines technology
2.3.1 Present status

Currently, all installed wind turbines are based on one of the four main wind turbines types as
shown in Fig. 2.9:
Fig. 2.9. Basic topologies for wind turbines:

a) fixed speed with directly grid connected squirrel-cage induction generator,
b) variable speed with squirrel-cage induction generator,
c) variable speed with doubly-fed induction generator,
d) variable speed with direct driven synchronous generator.
Apart from these design types, some manufacturers have developed other technologies during
the time. An overview regarding the current design and power ranges for different
manufacturers is presented in Table 2.1 ([12] March 2003).
Currently, the “Danish Concept” of the directly grid-connected wind turbine is widely used
for power ratings up to 2.3 MW. The scheme consists of a squirrel-cage induction generator
(SCIG), connected via a transformer to the grid as shown in Fig. 2.9a. These turbines operate
at approximately fixed speed (1-2% speed variation) and both passive and active stall control
concepts are used. Since a squirrel-cage induction generator always draws reactive power
from the grid, a capacitor bank is used in order to compensate the output power factor. The
main drawbacks of this solution are: the wind turbine has to operate at constant speed, it
requires a stiff power grid to enable stable operation and implies an more expensive
mechanical construction in order to be capable to absorb high mechanical stress [6].
However, this is a very robust and cheap solution.
Some manufacturers have developed variable speed wind turbines with a SCIG and a power
converter as shown in Fig. 2.9b. Usually, a back-to-back voltage source converter is used in
order to achieve the full control for the active and reactive power. Since the power converter
is designed to carry full load this solution is preferred for low power (up to 1MW), especially
in stand-alone and/or hybrid systems.
The most attractive topology seems to be the variable speed doubly-fed induction generator
(DFIG) as shown in Fig. 2.9c. Some manufacturers have developed the semi-variable speed
wind turbine concept, in which the rotor resistance of a DFIG is varied using power
electronics. In this way the speed range can be extended up to 10% of the rated value.
However, using a power converter in the rotor circuit a full control of the active and reactive
power can be achieved. The main advantage is that the converter power rating is around 25%
of the total generator power with a speed range of ± 30%.
24
Finally, a wind turbine topology based on synchronous generator is shown in Fig. 2.9d. In
general, this topology use a multi-pole machine, both wound rotor and permanent magnet
rotor in order to eliminate the gearbox.
It should be noticed that all the turbines presented in Table 2.1 are 3-bladed upwind with
tubular tower.
Table 2.1 Wind turbines manufacturers – current designs and power ranges
Power rating Generator
Manufacturer Design Power Converter
[MW] type
0.6 FS/PS SCIG Soft-starter
Bonus (Denmark)
1, 1.3, 2, 2.3 FS/AS SCIG Soft-starter
0.66, 1.8 FS/AS DFIG optislip
Vestas (Denmark)
0.85, 1.75, 2 VS/AP DFIG optispeed
0.75, 0.9, 1.5, 2 FS/AS SCIG Soft-starter
Neg Micon (Denmark)
2.75 VS/AP DFIG VSC IGBT
DeWind (UK/Germany) 0.6, 1.25, 2, 3.5 VS/AP DFIG VSC IGBT
Enercon (Germany) 0.3, 0.6, 1 1.8 VS/AP DD-SG -
0.6, 0.8, 1.3 FS/PS SCIG Soft-starter
Nordex (Germany) 1.5, 2.3, 2.5, FS/AS DFIG -
5 VS/AP DFIG -
0.6, 1.3 FS/PS SCIG Soft-starter
MADE (Spain)
0.8, 2 VS/AP SG R+B+VSI-IGCT
GE Wind Energy
0.9, 1.5, 3.2, 3.6 VS/AP DFIG VSC-IGBT
(USA/Germany)
REpower Systems 0.6, 0.75 FS/PS SCIG Soft-starter
(Germany) 0.75, 5 VS/AP DFIG -
Jeumont Industrie (France) 0.75 VS/AP PMSG VSC-IGBT
0.75 VS/AP DD-SG -
Lagerwey (Netherlands)
1.5, 2 VS/AP DD-PMSG -
Legend:
• FS/PS – fixed speed/passive stall
• FS/AS – fixed speed/active stall
• VS/AP – variable speed/active pitch
• SCIG – squirrel-cage induction generator
• DFIG – doubly-fed induction generator
• SG – synchronous generator
• PMSG – permanent magnet synchronous generator
• DD – direct driven (gearless)
• VSC – back-to-back voltage source converter
• R – diode bridge rectifier
• B – DC-boost converter
Each of these topologies has benefits and drawbacks. A fixed speed wind turbine is relatively
simple, so the price tends to be slightly lower [9]. Since the rotor speed cannot be varied,
these turbines must be more robust than the other designs due to the higher structural loads
involved.
A variable speed wind turbine will generate more energy for a given wind speed profile.
Moreover, the active and reactive power can be easily controlled and there is less mechanical
stress. Unfortunately, the power electronics are sensitive to voltage dips caused by faults
and/or switching and so, they are much more expensive.
The major drawback of the direct-driven topologies is the large and relatively heavy
generator. Moreover, for high power ratings the power converter should be designed to
handle the full generated power.
25
The basic elements for these converters are diodes and electronic switches as: gate turn-off
thyristor (GTO), integrated gate commutated tyristor (IGCT), bipolar junction transistors
(BJT), metal oxide semiconductor field effect transistor (MOSFET) and insulated gate
bipolar transistor (IGBT). A comparison of characteristics and ratings for some of these
switches is presented in Table 2.2 (based on [4], [5] and [8]).
Table 2.2 Maximum ratings and characteristics for electronic switches used in power converters
Switch type GTO IGCT BJT MOSFET IGBT

Voltage [V] 6000 4500 1200 1000 1700/3300
Current [A] 4000 2000 800 28 3600/2000
Drive requirements high low medium low low
It is obvious that due to the rapid development in power electronics the power handling
capability has increased in last decade while the price/kW has decreased. In these conditions
the application of power electronics in wind turbines will increase further
2.3.2 Future trends in wind turbines technology
Based on Table 2.1 the present ”state-of-the-art” in large wind turbine technology can be
summarized as follows:
• 3-bladed upwind turbine with tubular tower
• fixed-speed active stall with two speed squirrel-cage induction generator
• variable-speed pitch-control with doubly-fed induction generator
Only two manufacturers offer direct driven design as shown in Table 2.1.
Currently the back-to-back voltage source converter seems to be the most used power
converter topology. However, others power converters topologies are possible candidates for
the next generation of variable-speed large wind turbines e.g.: multilevel converter, tandem,
matrix converter, and resonant converter.
A comparison of these power converters topologies is carried out with advantages and
drawbacks for each one in [5]. As the back-to back voltage source converter is state-of-the-
art today in wind turbine technology it can be used as a reference in a benchmark of the other
converter topologies regarding the number of the components and their ratings, the auxiliary
components, the efficiency, the harmonic performances and implementation. However, the
comparison presented in [5] concludes and recommends further studies on back-to-back
converter, the matrix converter and the multilevel converter.
It is expected that in the near future the size of the wind turbines will increase further. Some
manufacturers have already announced the new wind turbine class of 4-5 MW. Most of these
larger wind turbines are used and will be used in offshore applications due to the fact that the
offshore wind energy resources are significantly higher than the onshore ones as well as due
to the environmental issues.
In the last three years the first MW-sized offshore wind farms have been materialized in
Denmark, Netherlands and Sweden and further offshore projects are planned as shown in
Table 2.3 (based on [1] and [5]).
26
Table 2.3 Offshore energy projects in Europe.
Name Year Capacity [MW]
Utgrunden, Baltic Sea, Sweden 2000 7 * 1.425
Blyth, North Sea, UK 2000 2*2
Exiting
Middelgrunden, Baltic Sea, Denmark 2001 20 * 2

Yttre Stengrund, Baltic Sea, Sweden 2001 5*2
Horns Rev, Denmark 2002 80 * 2
Rødsand, Denmark 2003 72 * 2.3
Læsø, Baltic Sea, Denmark 2003 150
Lillgrund, Sweden 2003 72
Gros. Vogels, Germany 2003 80
NOVEM, Netherlands 2003 100
Planned
Lubeck, Germany 2004 100

Dublin Bay, Ireland 2003-2004 240
Electrabel, Belgium 2004 100
Scroby Sands, UK 2002-2003 76
E-Connection, Netherlands 2003 240
PNE, Germany 2004 60
Eirtricity, Ireland 2005 500
Helgoland, Germany 2005 500
Since the major investment of wind turbines will in the future be done in wind farms with
hundreds of MW power capacity the electric system contained by a large wind turbine as
well as within an offshore wind farm and its connection to the main power system became
more and more important. Moreover, it is important to identify and to improve the power
system stability limit in these future power systems with a high penetration of wind energy.
2.4 Summary
A description of the wind proprieties and the performances of the modern wind turbines, as
well as the power control methods are shown in this chapter. Then, the current status
regarding the large wind turbine technology is presented.
2.5 References
[1]. T. Ackermann – Transmission systems for offshore wind farms, IEEE Power Engineering Review,
December 2002, pp.23-27;
[2]. T. Burton, D. Shape, N. Jenkins, E. Bossany – Wind Energy Handbook, J. Wiley & Sons, Ltd. 2001, ISBN
0-471-48997-2;
[3]. H. Dumitrache, A.Georgescu – Calculus for Propellers, Ed. Academiei Romane, 1990, ISBN 973-27-
0053-X;
[4]. T. Gilmore, R.G. Sladky – Ratings of semiconductors for AC drives, IEEE Trans. on Industry
Applications, Vol. 37, No. 2, March/April 2001, pp. 434-441;
[5]. H. Hansen, L. Helle, F. Blaabjerg et all – Conceptual survey of generators and power electronics for wind
turbines, RISØ-R-1205EN, 2002, ISBN 87-550-2743-1;
[6]. S. Müller, M. Deicke, R.W. De Doncker – Doubly-fed Induction Generator Systems, IEEE Industry
Applications Magazine, May/June 2002, pp. 26-33;
[7]. C. Nichita, D. Luca, B. Dakyo, E. Ceanga – Large band simulation of the wind speed for real time wind
turbine simulators, IEEE, Trans on Energy Conversion, Vol. 17, No. 4, December 2002, pp. 523-529;
[8]. M. Pfaffenlehner, T. Laska, R. Mallwitz, A. Mauder, F. Pfirsch. C. Schaeffer – 1700V-IGBT3: Field Stopp
technology with optimized trench structure – Trend setting for the high power applications in industry and
traction,
27
[9]. H Slootweg, E. DeVries –Inside wind turbines. Fixed speed vs. variable speed, Journal of Renewable
Energy World, Vol. 6, No. 1, January/February 2003;
[10]. *** - International wind energy development – World market update 2000. Forecast 2001-2005, BTM
Consult ApS, March 2001;
[11]. ***-– Energy needs, choices and possibilities. Scenarios to 2050, Shell International 2001
[12]. Wind Turbines Manufacturers: www.bonus.dk, www.vestas.dk, www.negmicon.dk, www.made.es
28
Chapter 3
Induction Machine Modelling

First dynamic model of the induction machine for transient studies has been elaborated in
1920 by R.H. Park and it was based on the two-axis theory. Later on, using the geometrical
symmetry of these machines, K.P. Kovacs and I. Racz have proposed a dynamic model based
on the complex phasors theory. Then, this model has been extended to the steady-state
analysis. Currently both approaches are used by different authors; B.K. Bose, T.A. Lipo,
D.W. Novotny use the two axis representation, while W. Leonhard, H. Buhler, M.P.
Kazmierkowski, H. Tunia prefers the complex representation.
It is not the aim of this chapter to make an extended presentation of the basic principles for
induction machines in terms of magnetically coupled circuits, machine windings and air-gap
mmf. All these issues can be found in many references [1]-[9]. It is much more important to
have the dynamic equations in such a manner which makes easier the implementations in any
commercial software (e.g. Matlab, Spice, Saber, etc.) used for dynamic and/or steady-state
analysis of the induction machine.
In this chapter the general dynamic equations of induction machine both in the natural
reference frame (ABC/abc) and the dqo reference frame using matrix form will be presented.
Based on the dqo model the complex form can easily be derivate and then the steady-state
equations obtained. In each situation the dynamic equations will be treated with fluxes and
currents as state variables.
Moreover, some specific features of the large induction machines will be presented e.g. deep-
bar effect, saturation, as well as the methods, which can be used in order to implement these
effects.
Finally, some implementation methods for the induction machine model in Matlab/Simulink
as well as a comparison regarding the simulation time are shown.
3.1 The voltage equations formulated in their natural

reference frame
The dynamic equations of the induction machine written in the natural reference frame
(ABC/abc) is suitable for studying the complex unbalance operation modes both in stator and
rotor circuit. Moreover some power converters associated with the induction machine e.g.
AC controllers (soft-starters) require such a model because of their operation principles.
3.1.1 Fluxes as state variables
All machine models for the transient analysis of the induction machine are based on the so-
called ABC/abc model, which is the model of the machine in the natural reference frame
[10],[11]. The equations of this model are derived under assumptions that the machine has
geometrical symmetry, the mmf in the airgap of the machine is sinusoidal; there is negligible
saturation and negligible losses in the core of the machine [1]-[9]. Then, the simplified
29
schematic of the stator and rotor windings as shown in Fig. 3.1, can be used to derive the
dynamic equations.
Fig. 3.1. Equivalent diagram of the induction machine windings in their natural reference frame.
The voltage equations for the induction machine can be written as:
 vA  R s  i A  Ψ A 
v      Ψ 
 B  Rs 0  iB   B
 vC   Rs  iC  d  Ψ C 
 =  +   (3.1.1)
 va   Rr   i a  dt  Ψ a 
 vb   0 Rr   ib   Ψb 
      
 v c   R r   i c   Ψ c 
or in compact form:
d
[ V ] = [ R ][ I] + [ Ψ ] (3.1.2)
dt
where:
vA, …, vc are the instantaneous phase voltages of the stator and the rotor
windings;
iA, …, ic are the instantaneous currents through the stator and the rotor
windings;
ψA, …, ψc are the instantaneous values of the stator and the rotor flux linkage.
The indices s and r related with the stator and the rotor, as well as the prime indices for the
rotor variables have been neglected for the sake of simplicity.
The relationship between the linkage fluxes and the currents through windings are given by:
30
 Ψ A   LσA + M AA M AB M AC M Aa M Ab M Ac  i A 
Ψ   M LσB + M BB M BC M Ba M Bb M Bc   i B 
 B  BA
 Ψ C   M CA M CB LσC + M CC M Ca M Cb M Cc   i C 
 =    (3.1.3)
 Ψ a   M aA M aB M aC Lσa + M aa M ab M ac   i a 
 Ψ b   M bA M bB M bC M ba Lσb + M bb M bc   i b 
    
 Ψ c   M cA M cB M cC M ca M cb Lσc + M cc   i c 
or in the compact form:

[ Ψ ] = [ L][ I] (3.1.4)
where
MAA, …. , Mcc are mutual inductances between each phase of the stator and
rotor respectively and each other stator or rotor phases;
LσA, …., Lσc are the leakage inductances for each stator and rotor phase.
The mutual inductances between stator and rotor phases are a function of the electrical angle
θr (between the axis of stator phase A and axis of rotor phase a). The mutual inductances
between each stator phase and also between each rotor phase are a function of geometrical
angle 120o.
Assuming a geometrical symmetry for the magnetic and electrical circuit and the inductance
matrix can be simplified as follows:
 Ls −0.5M s −0.5M s M sr f1 M sr f 2 M sr f 3 
 −0.5M Ls −0.5M s M sr f 3 M sr f1 M sr f 2 
 s
 −0.5M −0.5M L M sr f 2 M sr f 3 M sr f1 
[ L] =  M f s M f s M sf Lr −0.5M r −0.5M r 
 (3.1.5)
 sr 1 sr 3 sr 2
 M sr f 2 M sr f1 M sr f 3 −0.5M r Lr −0.5M r 
 
 M sr f 3 M sr f 2 M sr f1 −0.5M r −0.5M r L r 
where:
3
Lss = Lσs + M s - stator self-inductance;
2
3
L rr = Lσr + M r - rotor self-inductance;
2
L σs , Lσr - stator and rotor leakage inductance;
Ms , M r - mutual inductances between two stator phases, and
two rotor phases respectively;
M sr - mutual inductance between stator phase A and
rotor phase a;
The coefficients f1,…, f3 are defined as:
31
 2π   2π 
f1 = cos θr , f 2 = cos  θr +  , f 3 = cos  θr −  (3.1.6)
 3   3 
If the rotor parameters are referred to stator, then the mutual inductances are
equal: M s = M r = M sr . If it is taken into account that:
i A + i B + i C = 0 and i a + i b + i c = 0
the inductance matrix [ L ] can be rewritten as:
 Ls 0 0 M sr f1 M sr f 2 M sr f 3 
 0 Ls 0 M sr f 3 M sr f1 M sr f 2 

 0 0 L M f M f M f 
[ L] =  M f M f M sf Lsr 2 0sr 3 0sr 1  (3.1.7)
 sr 1 sr 3 sr 2 r 
 M sr f 2 M sr f1 M sr f 3 0 Lr 0 
 
 M sr f 3 M sr f 2 M sr f1 0 0 L r 
Based on (3.1.1) and (3.1.4) the general dynamic equations of the induction machine using
fluxes as state variables can be expressed in compact form as:
d
[ Ψ ] = − [ R ][ L] [ Ψ ] + [ V ]
−1
(3.1.8)
dt
3.1.2 Inductance matrix analytical inversion
The main problem in implementing (3.1.8) is to calculate the inverse of the inductance matrix
(3.1.7) at each simulation step [11]. A solution is to use modern software tools, which have
these capabilities (e.g. Matlab). However, it is useful an analytical expression for this matrix
since the coefficients depend on the electrical angle θr. One approach is to apply Cholesky
decomposition since the inductance matrix is “symmetric positive definite” [10].
Starting from the general form (3.1.7) the inverse of the inductance matrix can be written as:
 b11 b12 b13 b14 b15 b16 
b b 22 b 23 b 24 b 25 b 26 
 21
b b32 b33 b34 b35 b36 
[ L] = [ B] =  b31
−1
 (3.1.9)
 41 b 42 b 43 b 44 b 45 b 46 
 b51 b52 b53 b54 b55 b56 
 
 b 61 b 62 b 63 b64 b 65 b 66 
Using a symbolic mathematical tool e.g. Matlab or Mathcad the coefficients from (3.1.9) can
be determined easily. First, some coefficients are calculated as:
3 3 1
K1 − − −
Ls L r 4 4 M sr2
K1 = K2 = K3 = K4 = (3.1.10)
M sr2 9 9 9
K1 − K1 − K1 −
4 4 4
32
and then all other elements of the matrix B:
K2
b11 = b 22 = b33 =
Ls
K2
b 44 = b55 = b66 =
Lr
K3
b12 = b13 = b 21 = b 23 = b31 = b32 =
Ls
K3
b 45 = b 46 = b54 = b56 = b64 = b65 = (3.1.11)
Lr
b14 = b 25 = b36 = b 41 = b52 = b63 = K 4 f1
b15 = b 26 = b34 = b 43 = b51 = b 62 = K 4 f 2
b16 = b 24 = b35 = b 42 = b53 = b61 = K 4 f 3
Based on (3.1.10) and (3.1.11) the dynamic model of the induction machine can be easily
implemented in any commercial simulation tool, e.g. Matlab/Simulink. Moreover, some extra
features for the model can be taken into account e.g. deep-bar effect, saturation, influence of
temperature, etc.
3.1.3 Currents as state variables
Based on (3.1.2) and (3.1.4) the voltage equations using currents as state variables can be
written as follow:
d d [ L] d [ I]
[ V ] = [ R ][ I] + ([ L][ I]) = [ R ][ I] + [ I] + [ L] (3.1.12)
dt dt dt
Or in the state-space form:
d [ I] −1  d [ L] 
= [ L ] ⋅ − [ R ] −  ⋅ [ I] + [ L] ⋅ [ V ]
−1
(3.1.13)
dt  dt 
Taking into account that the coefficients of the inductance matrix are functions of rotor
position θr results:
d [ L ] d [ L] dθr d [ L ]
= ⋅ = ⋅ ωr (3.1.14)
dt dθr dt dθr
where ωr is the electrical speed of the rotor.
Based on (3.1.14) performing some simple mathematical calculations will result:
33
0 0 0 g1 g 2 g 3 
0 0 0 g 3 g1 g 2 

d [ L] 0 0 0 g 2 g 3 g1 
= − M sr  s  (3.1.15)
dθ r  g1 g3 g 2 0 0 0 
g 2 g1 g 3 0 0 0 
 
 g 3 g 2 g1 0 0 0 
where the coefficients g1,…, g3 are defined as follows:

 2π   2π 
g1 = sin θr , g 2 = sin  θr +  , g 3 = sin  θr − 
 3   3 
Grouping (3.1.13) and (3.1.14) results the state-space form using currents as state variables:
d [ I] −1  d [ L] 
= [ L ] ⋅ − [ R ] − ωr ⋅  ⋅ [ I] + [ L] ⋅ [ V ]
−1
(3.1.16)
dt  dθr 
3.1.4 Expression of the electromagnetic torque
The electromagnetic torque can be obtained starting from (3.1.12) and multiplying it with the
transpose of the currents matrix:
d [ L] d [ I]
[ I] [ V ] = [ I] [ R ][ I] + [ I]  [ Ψ ]  = [ I] [ R ][ I] + [ I]
d
[ I] + [ I] [ L]
t t t t t t
(3.1.17)
 dt  dt dt
or using (3.1.14)
d [ L] d [ I]
[ I] [ V ] = [ I] [ R ][ I] + ωr [ I] [ I] + [ I] [ L]
t t t t
(3.1.18)
dθr dt
The terms from (3.1.18) have the following meaning:
Pi = [ I] [ V ] - instantaneous power
t
Pcopper = [ I ] [ R ][ I ] - copper losses in machine windings

t
d [ I]
Pmag = [ I ] [ L ]
t
- magnetic power stored in machine (due to the variation in
dt
time of the magnetic energy)
d [ L]
Pm = ωr [ I ] [ I] - mechanical power.
t
dθr
The electromagnetic torque is then:
Pm P t d [ L]
Te = = p m = p [ I] [ I] (3.1.19)
Ωr ωr dθr
where: p is the number of pole pairs and Ωr is the mechanical speed of the rotor.
34
Substituting (3.1.15) in (3.1.19) the electromagnetic torque as a function of currents is:
Te = − pM sr {( i A i a + i Bi b + i Ci c ) sin θ + ...
+ ( i A i b + i Bi c + i Ci a ) sin ( θ + 2π / 3) + ... (3.1.20)
+ ( i A i c + i Bi a + i Ci b ) sin ( θ − 2π / 3)}
3.2 The voltage equations in arbitrary reference frame

variables
Some of the machine inductances are functions of the rotor speed, whereupon the coefficients
of the state-space equations (voltage equations), which describe the behaviour of the
induction machine, are time-varying except when the rotor is stalled. A change of variables is
often used to reduce the complexity of these state-space equations. There are several changes
of variables, which are used but there is just one general transformation [7]. This general
transformation refers machine variable to a frame of reference, which rotates at an arbitrary
angular velocity.
In this reference frame the machine windings are replaced with some equivalent windings as
shown in Fig. 3.2.
Fig. 3.2. Induction machine windings in the dqo-arbitrary reference frame.
Based on (3.1.1) and using the general transformation, the voltage equations in the arbitrary
reference frame can be written as follows:
 vsd   R s  isd   Ψ sd   0 −ωg 0 0 0 0 Ψ 

v    i  
 Ψ  ωg 0 0 0 0 0   Ψ 
sd
 sq   Rs 0   sq   sq    sq 
 vso   Rs  iso  d  Ψ so   0 0 0 0 0 0  Ψ 
= + +   so (3.2.1)
       0 0 0 0 − ( ωg − ωr ) 0   Ψ rd 
 v rd   Rr  i rd  dt  Ψ rd    
 
 v rq   0 Rr  i rq   Ψ rq  0 0 0 ( ωg − ωr ) 0 0  rq Ψ
         
 ro  
v R r
  ro 
i Ψ 
 ro   0 0 0   Ψ ro 
0 0 0 
35
where: ωr is the electrical speed of the machine and ωg is the speed of the general reference
frame.
Or (3.2.1) in compact form is:
d
[ V ] = [ R ][ I] + [ Ψ ] + [Ω][ Ψ ] (3.2.2)
dt
The relation between the linkage fluxes and the currents is given by:
 Ψ sd   L σs + L m 0 0 Lm 0 0  isd 
Ψ   0 Lσs + L m 0 0 Lm 0  isq 
 sq  
 Ψ so   0 0 Lσs 0 0 0  iso 
 =  
 Ψ rd   L m 0 0 Lσr + L m 0 0  i rd 
 Ψ rq   0 Lm 0 0 Lσr + L m 0  i rq 
    
 Ψ ro   0 0 0 0 0 Lσr   i ro 
(3.2.3)
 Ls 0 0 Lm 0 0  isd 
0 Ls 0 0 Lm 0  isq 

0 0 Lσs 0 0 0  iso 
=  
Lm 0 0 Lr 0 0  i rd 
0 Lm 0 0 Lr 0  i rq 
  
 0 0 0 0 0 Lσr  i ro 
or in compact form:
[ψ ] = [ L][ I] (3.2.4)
where:
3
Lm = M sr is the mutual inductance in the general reference frame;
2
Ls is the stator self inductance;
Lr is the rotor self inductance.
The voltage equations are written again in terms of currents and flux linkages. Clearly, these
variables are related based on the matrix inductance [ L ] and both cannot be independent or
state variables.
3.2.1 Fluxes as state variables
If the fluxes are selected as state variables the currents can be expressed in compact form by:
[ I] = [ L] [ Ψ ]
−1
(3.2.5)
The inverse of the inductance matrix is:
36
 Lr 0 0 −Lm 0 0 
 0 Lr 0 0 − L m 0 

 D 
 0 0
Lσr
0 0 0 
1
[ L] =  
−1
(3.2.6)
D −L m 0 0 Ls 0 0 
 
 0 −Lm 0 0 Ls 0 
 D 
 0 0 0 0 0 
 Lσs 
where D = Ls L r − L2m .
After some mathematical manipulation, the state-space form of the dynamic equations (3.2.1)
is:
d
[ Ψ ] = − [ R ][ L] [ Ψ ] + [ V ]
−1
(3.2.7)
dt
or in expanded form:
 R s Lr R s Lm 
 D −ω g 0 − 0 0 
D
 
 ω R s Lr RL
0 0 − s m 0  Ψ
 Ψ sd   g
D D   sd   vsd 
Ψ    Ψ   v 
 sq  Rs
 0 0 0 0 0   sq   sq 
d  Ψ so   Lσs   Ψ so   vso 
  = −   +   (3.2.8)
dt  Ψ rd  − R r Lm R r Ls Ψ   v rd 
0 0 − ( ωg − ωr ) 0   rd
 Ψ rq   D D   Ψ rq   v rq 
   RL R r Ls    
 Ψ ro   0 − r m 0 ( ωg − ωr ) 0   Ψ ro   v ro 
 D D 
 Rr 
 0 0 0 0 0 
 Lσr 
3.2.2 Currents as state variables
If the currents are selected as state variables the fluxes will be replaced based on (3.2.3)
Assuming that all parameters are constant (saturation is not taken into account) (3.2.1)
become:
d [ I]
dt
( −1
)
= − [ L ] [ R ] + [ L ] [ Ω ][ L ] [ I ] + [ L ] [ V ]
−1 −1
(3.2.9)
So the state-space form of the dynamic equations using currents as state variables is:
d [ I]
= [ A ][ I ] + [ B][ V ] (3.2.10)
dt
where:
37
 RL LL L2 R r Lm Lr Lm 
 − s r + s r ωg − m ( ωg − ωr ) 0 ωr 0 
 D D D D D 
 Ls L r L2m R s Lr Lr Lm R r Lm 
 D g D ( g
− ω + ω − ωr) −
D
0 −
D
ω r
D
0 
 
 RD
0 0 − s 0 0 0 
 Lσs 
[ A] =  
 R s Lm Ls L m R r Ls 2
Lm Ls L r 
 D
−
D
ωg 0 −
D
−
D
ωg +
D
( g r)
ω − ω 0 
 2

 Ls L m R s Lm Lm Ls L r R r Ls 
 D
ωg
D
0
D
ωg −
D
( ωg − ωr ) −
D
0 
 
 RD
0 0 0 0 0 − r 
 Lσr 
 Lr 0 0 −Lm 0 0 
 0 Lr 0 0 −L m 0 

 D 
 0 0 0 0 0 
1 L σr
[ B] = [ L] =  
−1
D −L m 0 0 Ls 0 0 
 
 0 −Lm 0 0 Ls 0 
 D 
 0 0 0 0 0 
 Lσs 
3.2.3 Electromagnetic torque
The electromagnetic torque can be obtained starting from (3.2.2) and multiplying it from the
left with the transpose of the currents vector:
d
[ I] [ V ] = [ I] [ R ][ I] + [ I] [ Ψ ] + [ I] [Ω][ Ψ ]
t t t t
(3.2.11)
dt
where
Pi = [ I] [ V ] is the instantaneous power;
t
Pcopper = [ I ] [ R ][ I ] are the copper losses in machine windings;

t
d [ψ]
Pmag = [ I ]
t
is the magnetic power stored in machine (due to the variation in
dt
time of the magnetic energy);
Pm = [ I ] [ Ω ][ Ψ ] is the mechanical power.
t
The electromagnetic torque is then:

Pm P 3
Te = = p m = p ( Ψ sd isq − Ψ sq isd ) (3.2.12)
Ωr ωr 2
where: p is the number of pole pairs and Ωr is the mechanical speed of the rotor.
Other equivalent expressions for the electromagnetic torque of an induction machine are:
38
3
• Te = p ( Ψ rq i rd − Ψ rd i rq )
2
3
• Te = pL m ( isq i rd − isd i rq )
2
3 Lm
• Te = p
2 D
( Ψ sq Ψ rd − Ψ sd Ψ rq )
Above equations may be somewhat misleading since they seem to imply that the leakage
inductances are involved in the energy conversion process. This, however, is not the case.
Even though the flux linkages contain the leakage inductances, they are eliminated by the
algebra [7].
3.2.4 Commonly used reference frames

Although the behaviour of a symmetrical induction machine may be described in any
reference frame, there are three, which are commonly employed. Namely, the stationary
reference frame first employed by H.C. Stanley [7], the rotor reference frame, which is Park’s
transformation applied to induction machines, and the synchronously rotating reference
frame. The voltage equations for each of these reference frames may be obtained from the
voltage equations in the arbitrary reference frame by assigning the appropriate speed to ωg.
That is, ωg = 0 for the stationary, ωg = ωr, for the rotor, and ωg = ωs for the synchronously
rotating reference frame.
Generally, the conditions of operation will determine the most convenient reference frame for
analysis and/or simulation purposes [7]. If the stator voltages are unbalanced or
discontinuous and the rotor applied voltages are balanced or zero (squirrel-cage machines);
the stationary reference frame should be used to simulate the performance of the induction
machine. If, on the other hand, the external rotor circuits are unbalanced but the applied stator
voltages are balanced, then the reference frame fixed in the rotor is most convenient. Either
the stationary or synchronously rotating reference frame is generally used to analyze
balanced or symmetrical conditions. Linearized machine equations, which are used to
determine eigenvalues and to express linearized transfer functions for use in control system
analysis, are obtained from the voltage equations expressed in the synchronously rotating
reference frame. The synchronously rotating reference frame is also particularly convenient
when incorporating the dynamic characteristics of an induction machine into a computer
simulation tool used to study the transient and dynamic stability of large power systems.
Moreover, this reference frame may also be useful in variable frequency applications if it is
permissible to assume that the stator voltages are balanced sinusoidal set.
Regardless of the reference frame being used the stator and voltages and currents must be
properly transformed to and from this reference frame.
3.2.5 Per unit system
It is often convenient to express machine parameters and variables as per unit quantities.
Base power and base voltage are selected and all parameters and variables are normalized
using these base quantities. When the machine is a part of a power system and if it is
desirable to convert the entire system to per unit quantities, then only one power base (VA
base) is selected which would most likely be different from the rating of any machine in the
39
system. In the following the machine will be considered separately with the rating of the
machine taken as base power.
The RMS value of the rated phase voltage is generally selected as base voltage for the abc
variables while the peak voltage is generally selected as base voltage for the dqo variables.
That is, VB(abc) is the RMS voltage selected as base voltage for the abc variables then
VB(dqo) = 2VB(abc) . The base power may be expressed as:
PB = 3VB(abc) I B(abc) (3.2.13)
or
3
PB = VB(dqo) I B(dqo) (3.2.14)
2
Therefore, since the base power and the base voltage are selected, base current can be
calculated from (3.2.13) or (3.2.14). It follows that the base impedance can be written as:
2
VB(abc) VB(abc)
ZB = =3 (3.2.15)
I B(abc) PB
or
2
VB(dqo) 3 VB(dqo)
ZB = = (3.2.16)
I B(dqo) 2 PB
The dqo equations written both in terms of reactances can be readily converted to per unit by
dividing the voltages by VB(dqo) , the currents by I B(dqo) , and the resistances and reactances by
ZB. Since a flux linkage per second is a volt, it is per-unitized by dividing by base voltage.
Although the voltage and flux linkage per second equations do not change form when per
unitized, the torque equation is modified by the per unitizing process. For this purpose the
base torque may be expressed:
PB
TB = p (3.2.17)
ωb
where ωb corresponds to rated or base frequency of the machine. If in (3.2.17) PB is the rated
output of the machine then the base torque TB will not be rated torque. In the case of an
induction machine, rated power output generally occurs at a speed slightly less than
synchronous speed. Hence, TB will be less than rated torque by the ration of rated speed to
synchronous speed.
So, the per unit torque will be:
Te = ( Ψ sd isq − Ψ sq isd ) (3.2.18)
If the electrical variables are expressed in volts, amperes and watts then the inertia of the
rotor is expressed in mks units. If, however, the per unit system is used the inertia is
expressed in seconds. By definition, the inertia constant expressed in seconds is:
40
1 Jωb 1 1 Jω2b
H= p = (3.2.19)
2 TB 2 p 2 PB
where J is the inertia of the rotor and connected mechanical load in kg ⋅ m 2 .

Thus the equation of motion using a one-mass model become:
d ωr
Te m Td = 2H (3.2.20)
dt ωb
where the minus sign is used in motor operation and plus for generator.
3.3 Mathematical model of a gearbox with torsional torques

The equivalent model of a wind turbine drive train is presented in Fig. 3.3. It is essentially a
two mass model connected by a flexible shaft. The masses correspond to a large mass of the
wind turbine rotor and a small mass of the generator respectively.
Fig. 3.3. Equivalent diagram of a gearbox with torsional torques.
The moment of inertia for the shaft and the gearbox wheels can be neglected because they are
small compared with the moment of inertia of the wind turbine or generator. Taking into
account the shafts stiffness and the damping factors the mechanical equations can be written
as:
dΩ wtr dθ
Twtr = J wtr + D wtr wtr + k swtr (θ wtr − θ1 )
dt dt
dΩ1 dθ
T1 = J1 + D wtr wtr + k swt (θ1 − θwtr )
dt dt
(3.3.1)
dΩ 2 dθ
T2 = J 2 + Dgen gen + k sgen (θ2 − θgen )
dt dt
dΩgen dθ
−Tgen = J gen + Dgen gen + k sgen (θgen − θ2 )
dt dt
where: Twtr – wind turbine torque; Jwtr – wind turbine moment of inertia; Ωwtr – wind turbine
mechanical speed; kswtr – spring constant indicating the torsional stiffness of the shaft on
wind turbine part;Tgen – generator torque; Jgen – generator moment of inertia; Ωgen – generator
mechanical speed; ksgen – spring constant indicating the torsional stiffness of the shaft on
41
generator part; T1 – torque that goes in the gearbox; T2 – torque out from the gearbox;
1
T2 = ⋅ T1 , Kgear – the gearbox ratio; Ω 2 = k gear ⋅ Ω1
k gear
The system complexity can be reduced by considering an equivalent system with an
equivalent stiffness and damping factor. Only the gearbox ratio has influence on the new
equivalent system.
The equations for equivalent system can be written in two points: on the wind turbine side
with the influence of generator component through the gearbox and on the generator side
respectively. The equivalent system on the generator side is shown in Fig. 3.4.
Fig. 3.4. Equivalent diagram of the wind turbine drive train on the generator side.
The equivalent dynamic equations of the gearbox written on the generator side are:
dΩ' wtr
T ' wtr = J ' wtr + D'e (Ω' wtr − Ω gen ) + k 'se (θ' wrt − θgen )
dt
dθ' wtr
= Ω ' wtr
dt
(3.3.2)
dΩ gen
−Tgen = J gen + D'e (Ω gen − Ω ' wtr ) + k 'se (θgen − θ' wtr )
dt
dθgen
= Ωgen
dt
where: the equivalent stiffness is given by:
1 1 1
= + (3.3.3)
'
k se k wtr k gen
2
k gear
and the equivalent moment of inertia for the rotor is:
1
J ' wtr = 2
⋅ J wtr (3.3.4)
k gear
Some typical values for a 2MW wind turbine are:

• moment of inertia for the rotor Jwtr= 50*105 Kgm2;
• stiffness of the rotor shaft Kswtr=1.5*108 Nm/rad;
• moment of inertia for the generator is Jgen=100 Kgm2;
42
• stiffness for the generator shaft Ksgen=105 Nm/rad;
• equivalent damping factor of the entire drive-train is Dwtr=8*105 Nmsec/rad.
3.4 Steady-state equations

The voltage equations, which describe the balanced steady-state operation of an induction
machine, can be obtained in several ways. For balanced steady state conditions the d and q
variables are sinusoidal in all reference frames except the synchronously rotating reference
frame wherein they are constants [7]. In the following paragraphs the prime index will be
used for the rotor equations in order to highlight that these are related to the stator circuit of
the machine.
Starting from (3.2.1) written in the synchronously rotating reference frame ωg = ωs and
neglecting the time rate changing of all flux linkages and then employ the relationships:
~
2 Fas = Fsd − jFsq
~
(3.4.1)
2 F = F − jF
'
ar
'
rd
'
rq
the steady-state voltage equations for induction machine can be written as follows:
V s = ( R s + jX σs ) Is + jX m Is + I r ( '
) (3.4.2)
V = ( R r + jsX
'
r
'
σr )I
'
r (
+ jsX m Is + I
'
r )
ωe − ωr 2πf s − pΩ r pΩ r
where: slip s = = = 1− ; f s is the frequency of stator voltage; Ω r is the
ωe 2πf s 2πf s
mechanical rotor speed; p is the number of poles pair.
Equation (3.4.2) suggests the per-phase equivalent circuit shown in Fig. 3.5.
Fig. 3.5. Equivalent circuit for steady state operation of a symmetrical induction machine.
The steady-state voltage equations (3.4.2) can be arranged as follows:
V s = ( R s + jX s ) Is + jX m I r
'
(3.4.3)
V r = ( R 'r + jsX 'r ) I r + jsX m I r
' ' '
where:
43
X s = X σs + X m is the self reactance of the stator windings;
X 'r = X 'σr + X m is the self reactance of the rotor windings.

Or in matrix form:
 V s   R s + jX ss jX m   Is 
 '= ⋅  (3.4.4)
 V r   jsX m R 'r + jsX 'rr   I'r 
 
The compact form of (3.4.4) is:
 V  =  Z ⋅  I  (3.4.5)
     
Based on (3.4.5) the stator and rotor currents are obtained from:
−1
 I  =  Z ⋅  V  (3.4.6)
     
So, the induction machine can be treated as a complex quadripole as shown in Fig. 3.6 where
the complex impedance is a function of slip.
Fig. 3.6. Equivalent quadripole per-phase for the induction machine used in steady-state analysis.
Recalling (3.2.11) the electromagnetic torque is given by:

3
pX m Re  jIs I r 
* '
Te = (3.4.7)
2  
*
where Is is the conjugate of Is
The active power is given by:
3 *
Px = ⋅ Re al(V x ⋅ I x ) (3.4.8)
2
The reactive power is given by:
3 *
Q x = ⋅ Im ag(V x ⋅ I x ) (3.4.9)
2
Equations can be used both for stator and rotor windings by replacing the indices x with s and
r respectively.
Using this approach the induction machine can be completely characterized in terms of
steady-state values of the stator and the rotor currents, phase angle for both currents,
electromagnetic torque, active and reactive power for both stator and rotor circuit. Moreover
an analysis both in motor and generator mode of operation can be performed.
44
3.4.1 Analysis of steady-state operation for the squirrel-cage induction
generator
Starting from (3.4.4) and assuming that the rotor voltage is zero and neglecting the iron
losses, the equivalent diagram per-phase for a squirrel cage induction machine is shown in
Fig. 3.7.
Fig. 3.7. Equivalent diagram per-phase for a squirrel-cage induction machine in motoring mode.
The differences between motor and generator operation can be seen by reference to Fig. 3.8,
where phasor diagram are shown with the arrangement of rotor quantities expressed in
equivalent stator values [13],[14].
Fig. 3.8. Phasor diagram for squirrel-cage induction machine: a) motor and b) generator.
Fig. 3.8a is the normal phasor diagram for an induction motor operating below synchronous
speed and with the stator current lagging the stator voltage.
If the shaft speed is raised by the application of power to it, the slip, induced e.m.f. and rotor
current will all begin to decrease and finally become zero at synchronous speed, after the
rotor current has moved along the dotted circle. At this point the supply voltage feeds the
stator copper losses and the magnetization, while the friction and windage losses are supplied
by the mechanical power source, e.g. a wind turbine. The rotor copper losses are zero.
Further increase of speed above synchronous will cause the rotor e.m.f. and current to
reappear but in phase opposition to the previous position. Likewise the stator current phasor
moves to a new position. The shaft power input is converted into electrical energy and
45
supplies the machine losses and its power output. The magnetization current increases
somewhat but remains constant in direction, being still supplied by the external system.
Fig. 3.8a is based on an internal view of the machine, usually adopted for the motor phasor
diagram. The normal way to draw the generator phasor diagram is to change the viewpoint to
an external one, and the result is the redrawn phasor diagram of Fig. 3.8b in which the stator
voltage and current occupy positions leading the system voltages.
Based on (3.4.4) to (3.4.9) and taking into account the generator convention for negative slip
the steady state characteristics for a 2MW squirrel-cage induction generator with constant
parameters are shown in Fig. 3.9 to Fig. 3.11.
Fig. 3.9. Electromagnetic torque for a 2 MW squirrel-cage induction machine

in motor and generator operation as a function of slip.
Fig. 3.10. Stator current and phase angle for a 2 MW squirrel-cage induction machine
46
The electromagnetic torque will be negative in generator operation, while the stator current
will lead the stator voltage.
Fig. 3.11. Active and reactive power for a 2 MW squirrel-cage induction machine
The active power will be positive for negative slip while the reactive power is negative, this
means that the machine will deliver active power to the grid and will draw reactive power.
It is obviously that the squirrel cage induction machine will operate in generator mode only
for over synchronous speeds. However using power converters it is possible an operation in a
wide range of speed.
3.4.2 Analysis of steady-state operation for the doubly-fed induction

generator
Starting from (3.4.4) and neglecting the iron losses, the equivalent diagram per-phase for a
doubly-fed induction machine is shown in Fig. 3.12.
Fig. 3.12. Equivalent diagram per phase for a doubly-fed induction machine.
Supplying the stator windings with a constant frequency and magnitude voltage and varying
the amplitude and the phase for the rotor voltage will result a four quadrant operation for the
double fed induction machine as shown in Fig. 3.13.
47
Fig. 3.13. Operating regions for a doubly-fed induction machine with a rotor side control.
Based on (3.4.4) - (3.4.9) a detailed analysis of these operating regions for the doubly-fed
induction machine can be performed. Assuming a positive rated slip (0.01) for a 2MW
induction machine and varying the amplitude and the phase of the rotor voltage the
electromagnetic torque will result as shown in Fig. 3.14.
Fig. 3.14. Electromagnetic torque for different magnitudes and phases of the rotor voltage
at rated sub-synchronous speed.
Both the stator and rotor active and reactive power when the magnitude and phase of the
rotor voltage is varied are presented in Fig. 3.15.
a) b)
Fig. 3.15. Active and reactive power for different magnitude and phases of the rotor voltage
at rated sub-synchronous speed: a) stator power, b) rotor power.
48
If the machine will run at rated over synchronous speed (s = -0.01) the steady state
characteristics for the doubly-fed induction machine for different magnitudes and phases of
the rotor voltage are shown in Fig. 3.16 and Fig. 3.17.
Fig. 3.16. Electromagnetic torque for different magnitudes and phases of the rotor voltage
at rated over-synchronous speed.
a) b)
Fig. 3.17. Active and reactive power for different magnitude and phases of the rotor voltage
at rated over-synchronous speed: a) stator power, b) rotor power.
The possible operation modes of a doubly-fed induction machine are related with the
magnitude and phase of the rotor voltage and the slip. When the rotor voltage is in phase
opposition with the e.m.f. the sub-synchronous operation will result. If the magnitude of the
V'
rotor voltage is greater than the e.m.f r > E s the machine will run in motoring mode and if it
s
'
V
is lower r < E s the generator mode will result. However, when the rotor voltage is in phase
s
with the e.m.f. the over-synchronous operation will be obtained. When the magnitude of the
V'
rotor voltage is greater than the e.m.f. r > E s , the generator operation is achieved otherwise
s
the machine will run in motoring mode.
Modifying the phase of the rotor voltage in respect with the e.m.f. the reactive power on the
stator side can be adjusted in both situations: over- and sub-synchronous operation.
49
Independent the phase of the rotor voltage a special mode of operation can be achieved when
the magnitude of the rotor voltage is equal with the e.m.f. This operation mode is called
asynchronous compensator.
Fig. 3.18 summarizes the possible operation modes for a doubly-fed induction machine as
well as the power flow on stator and rotor side in motor and generator.
Fig. 3.18. Possible operation modes for a doubly-fed machine.
Moreover some special operation modes (e.g. electromagnetic break, asynchronous

compensator) can be obtained.
One problem related to the doubly-fed induction machine is that the frequency of the rotor
voltage is given by the rotor speed. Although, using a fraction from the stator rated power (up
to 30%) a fully 4-quadrant operation in a wide range of speed can be obtained by varying the
magnitude and phase of the rotor voltage.
3.5 Modelling deep-bar effect

In larger machines rotor deep-bar effect causes the equivalent rotor circuit resistance and
leakage reactance to vary significantly with the slip [8]. Fig. 3.19 illustrates the leakage flux
paths for a squirrel-cage induction machine with deep-bar design. The parts of each bar
extending deeply into the rotor iron have higher leakage inductances than those parts of the
bar cross-section near the air gap, because more of the leakage flux links the deeper parts of
the bar. One may think of each bar as being composed of several layers of equal cross
section, and thus of equal resistance, but with inductances increasing with depth.
Fig. 3.19. Deep-bar effect for a squirrel cage induction machine.
50
Under running conditions, the slip is quite small and the frequency of the rotor currents is
only 1-2 Hz. As a result, the leakage reactance is neglected, and the current distribution is
essentially uniform throughout each rotor bar. The effective resistance of the rotor is that of
all layers in parallel. This low resistance makes for low slip and high efficiency at full load.
As the motor is overloaded, however, slip and frequency increase. Thus, the value of rotor
leakage reactance in the circuit model becomes smaller and the value of rotor resistance
becomes greater.
Another important need is to include the leakage path saturation effect into reduced model.
The main saturation leakage path is at the tooth tips over the closed or nearly closed slots [8].
However, large machines usually have large slots openings, and hence the leakage path
saturation effect is not significant for large machines [15].
Numerous approaches regarding the modelling of deep-bar effect have been presented in
literature; therefore, all these models are complicated and require complex mathematical
models [15], [16].
In [15] is presented a simplified approach for the modelling of deep bar effect, which gives
accurate results. The equivalent rotor winding resistance and leakage reactance are modelled
as follows:
R 'r = R 'r _ DC + (R 'r _ 50Hz − R 'r _ DC ) ⋅ s
(3.5.1)
X 'σr = X 'σr _ DC + (X 'σr _ 50Hz − X 'σr _ DC ) ⋅ s
where:
R 'r _ DC is the equivalent rotor winding resistance at zero frequency (DC value or at
synchronous speed) ;
R 'r _ 50Hz is the equivalent rotor winding resistance at 50Hz (at standstill - s=1);
X 'σr _ DC is the equivalent stator leakage reactance at zero frequency (DC value or
at synchronous speed);
X 'σr _ 50Hz is the equivalent stator leakage reactance at 50Hz (at standstill - s=1).
Therefore, this approach assumes a linear dependency of rotor resistance and rotor leakage
reactance against entire range of slip. Moreover, requires extra parameters determined from
tests at synchronous speed.
Usually, the parameters from data sheets are given for rated operating point, near the
synchronous speed and cannot be applied.
Based only on parameters from datasheet, the equivalent rotor winding resistance and
leakage inductance at any operating point can be modelled as:
51
R 'r , s ≤ s m

R 'r _ dbe = '  1 − K rsm 
R r ( K r − 1) s + 1 − s  , s > s m
  m 
(3.5.2)
X 'σr , s ≤ s m

X 'σr _ dbe = '  1 − K xsm 
X σr ( K x − 1) s + 1 − s  , s > s m
  m 
where:
R 'r and X 'σr are the values the from data-sheet given for rated operating point
K r is the resistance deep-bar factor;
K x is the leakage reactance deep-bar factor;
R s2 + X s2
s m = R 'r is the breakdown slip in motoring mode.
( X 2m − Xs X'r ) + R s2 X'2r
This approach assumes that the parameters of induction machine are constant for the values
of the slip less or equal with the breakdown slip.
Resistance and leakage reactance deep-bar factors can be determined based on rated voltage
and current, starting current and phase angle (from short-circuit test at rated current). If the
phase angle is not available, a value around 80o can be used with a good approximation in the
first step of iteration, than it can be corrected so that the calculated starting torque will fit the
value from datasheet or test reports.
The deep-bar factors for rotor resistance and rotor leakage reactance are calculated as
follows:
1  U ph _ rated 1 
Kr =  − Rs 

R 'r  β I ph _ rated 1 + tan ϕsc
2
 (3.5.3)
1
K x = ' ( R s + K r R 'r ) ⋅ tan ϕsc − X σs 
X σr
where:
U ph _ rated is the rated voltage per phase;
I ph _ rated is the rated current per phase;
β is the relative starting current (from data sheet);

R s is the stator resistance per-phase (from data sheet);
X σs is the stator leakage reactance per-phase (from data sheet);
R 'r is the rotor resistance per-phase for the rated operating point (from data
sheet);
52
X 'σr is the rotor leakage reactance per-phase for the rated operating point (from
data sheet);
ϕsc is the phase angle at standstill (s=1): from short-circuit test at rated current or
around 800 for large induction machines.
Fig. 3.20a shows the rotor resistance and the rotor leakage reactance against slip based on
(3.5.2) and (3.5.3), while Fig. 3.20b shows the electromagnetic torque with and without deep-
bar effect.
a) b)
Fig. 3.20. Deep-bar effect for a 2 MW squirrel-cage induction machine:
a) rotor resistance rotor leakage reactance as a function of slip and b) electromagnetic torque.
The same approach can be used for doubly-fed induction machine since the manufacturers
performs the standard tests as for the squirrel-cage induction machines.
3.6 Modelling saturation

The flux equations highlight the terms, which are depended on magnetic saturation and
therefore introduce non-linearity. It will be assumed that the saturation of the leakage fluxes
into a large induction machine can be neglected [6], [8], thus the stator and rotor leakage
inductances are constant.
The magnetizing inductance is a function of magnetization current as shown in Fig. 3.21.
Fig. 3.21. Magnetizing flux versus magnetizing current.

53
Assuming that the machine has magnetic, electric and geometric cylindrical symmetry, then
the magnetic saturation caused by the main magnetic field is independent by the direction of
this field and depends only on the absolute value of it.
The magnetizing flux can be written as:
d d d ( L m ⋅ i m ) di m
( Ψ m ) = ( Lm ⋅ im ) = ⋅ (3.6.1)
dt dt di m dt
The magnetization current is independent on the reference frame. It can be defined two
inductances as follows:
Ψm
Lm = = tgα1 = f1 (i m ) (3.6.2)
im
d ( L m ⋅ i m ) dΨ m
Ld = = = tgα 2 = f 2 (i m ) (3.6.3)
di m di m
where: Lm is the magnetizing inductances and Ld is the dynamic inductance.
Fig. 3.22 shows some typical dependencies of these two inductances with the magnetizing
current.
Fig. 3.22. Magnetizing and dynamic inductance versus magnetizing current.
Based on the no-load curve for an induction machine these two inductances can be
determined as a function of no-load current. Then, using a mathematical approximation e.g. a
function or a look-up table, the desired values for these inductances for a given current can be
found.
Taking into account the above mentioned, the phasor voltage equations of the induction
machine in general reference frame including the saturation of the main path can be written
as follow:
vs = R s is + Lσs
d
dt
( )
is + L d
di m
dt
(
+ jωg Lσs is + L m i m )
v r = R r i r + Lσr
d
dt
( )
i r + Ld
di m
dt
(
+ j ( ωg − ωr ) Lσr i r + L m i m )
i m = is + i r (3.6.4)
( )
L m = f1 i m
Ld =f (i )
2 m
54
3.7 Modelling iron losses
Considerable attention has been paid during the last years in modelling of iron losses into an
induction machine. A simple search in IEEE database will produce around 120 results. All
these paper dealing with the iron loss modelling for an inverter-fed induction motor and take
into account the hysteresis losses as well as the eddy-current losses in a wide range of speed
and frequency. In wind turbine applications usually a large squirrel-cage induction generator
is directly connected to grid and the rotor speed is almost fixed. When a doubly-fed machine
is used, again the stator windings are directly connected to the grid and the variable speed
operation is achieved using a power converter in the rotor side. Usually, the speed varies in a
range of ± 30% from the synchronous speed. In data sheets the no-load power, current and
power factor are provided for the machine rated voltage as well as the electrical parameters
for the rated operating point. So, it is convenient to use only this information in the case of a
large induction machine, both squirrel-cage and wound rotor.
A classical approach in iron losses modelling for an induction machine involves a resistance
inserted in parallel with the magnetizing reactance in the equivalent per-phase diagram (Fig.
3.5) as shown in Fig. 3.23
Fig. 3.23. Equivalent diagram per-phase for induction machine including iron losses resistance.
One major drawback of this approach is that the magnetizing current is divided in two
components: an active component and a reactive component. So, it is difficult to use this
equivalent diagram for a dynamic model as well as for a steady state analysis. A modified
equivalent scheme, called Γ has been developed. Unfortunately, the currents in this
equivalent scheme are multiplied with some factors and the dynamic modelling became
difficult.
A simple approach assumes the iron losses resistance in series with the magnetizing
reactance as shown in Fig. 3.24.
Fig. 3.24. Equivalent T diagram of the induction machine with iron losses.
55
Since the machines manufactures gives in data sheet the electrical parameters for both
schemes presented in Fig. 3.23 and Fig. 3.24, or at least for the first one it is easier to use this
approach. The electrical parameters for the parallel mode can be transformed through some
simple mathematical calculations to the series mode equivalent circuit.
Recalling (3.1.2) the iron losses can be introduced into the dynamic equations written in the
ABC/abc reference frame as follows:
d
[ V ] = [ R ][ I] + [ R o ][ I] + [ Ψ ] (3.7.1)
dt
where the iron losses resistance matrix is defined in a similar manner as (3.1.5) by:
1 0 0 1 0 0
0 1 0 0 1 0 

0 0 1 0 0 1
[R 0 ] = R o ⋅   = R o ⋅ [ To ] (3.7.2)
1 0 0 1 0 0
0 1 0 0 1 0
 
 0 0 1 0 0 1 
In this conditions the inductance matrix (3.1.7) will be rewritten as:
 Ls 0 0 M o f1 M o f 2 M o f 3 
 0 Ls 0 M o f3 M o f1 M o f 2 

 0 0 L Mf Mf Mf 
[ L] =  M f M f M sf Lo 2 0o 3 0o 1  (3.7.3)
 o1 o 3 o 2 r 
 M o f 2 M o f1 M o f 3 0 Lr 0 
 
 M o f 3 M o f 2 M o f1 0 0 L r 
where:
3
Ls = Lσs + M o - stator self-inductance;
2
3
L r = Lσr + M o - rotor self-inductance;
2
Mo - is the magnetizing inductance for the series circuit.
Finally the state space equations with fluxes as state variables are:
d
[ Ψ ] = − ([ R ] + [ R 0 ] ) [ L ] [ Ψ ] + [ V ]
−1
(3.7.4)
dt
or using currents as state variables:
d [ I] −1  d [ L] 
= [ L ] ⋅  − ([ R ] + [ R 0 ]) − ωr ⋅  ⋅ [ I] + [ L] ⋅ [ V ]
−1
(3.7.5)
dt  dθ r 
Similar approach can be used to introduce the iron loss resistance into the dynamic equations
written in arbitrary reference frame.
56
3.8 Induction machine modelling using Matlab/Simulink
Using the available blocks from Matlab/Simulink three methods have been considered which
implements the dynamic equations of the induction machine in different ways. In Fig. 3.25 is
shown the general structure for all considered models. These models implement the induction
machine model in state-space form given by (3.2.9). The equation of motion is implemented
in a separate block in order to gain flexibility in modelling. So, in this way a simple equation
of motion or the complete model of a wind turbine drive train can be used.
Fig. 3.25. General structure of induction machine model in Simulink.
The first two models, Model A and Model B, have the same structure as shown in Fig. 3.26
Fig. 3.26. Implementation of induction machine model in dq rotor reference frame used
in Model A and Model B.
The only difference consists in the implementation of the state-space equations with currents
as state-variables as shown in Fig. 3.27.
a) `b)
Fig. 3.27. Simulink implementation of dynamic equations for induction machine:
a) Model A and b) Model B.
57
While Model A is based on Simulink blocks Sum, Product, Gain, GoTo, From and
Integrator, Model B use only Function and Integrator blocks.
The third model - Model C - is based on an S-function written in “C” language according
with the Simulink format for this type of function (Fig. 3.28).
Fig. 3.28. Simulink implementation of Model C for induction machine using a C S-Function
The coordinate transformations abc/dq and dq/abc are also included in this function as well
as the calculation of the reference frame angle. So, the S-function called “scim” implements
five differential equations
Using Simulink Performance Tools Toolbox® the considered models have been compared
regarding the simulation speed. The Simulink profiler collects performance data while
simulating a model and generates a report, called a simulation report. The profile is useful to
find out how much time Simulink spends on executing each block from a model and hence
where to focus the model optimization efforts.
Fig. 3.29 shows the pseudo-code, which summarizes the execution of a dynamic system in
Simulink.
Fig. 3.29. Pseudo-code for execution of a dynamic system model in Simulink
According to this conceptual model, Simulink executes a Simulink model by invoking the
functions presented in Table I zero, one, or more times, depending on the function and the
model [17].
58
Table 3.1 Function invoked by Simulink during simulation.
Function Purpose Level
Simulate the model. This top-level function invokes the other
sim functions required to simulate the model. The time spent in this System
function is the total time required to simulate the model.
ModelInitialize Set up the model for simulation System
Execute the model by invoking the output, update, integrate, etc.,
ModelExecute functions for each block at each time step from the start to the end of System
simulation.
Output Compute the outputs of a block at the current time step. Block
Update Update a block’s state at the current time step. Block
Compute a block’s continuous states by integrating the state
Integrate Block
derivatives at the current time step.
MinorOutput Compute a block’s output at a minor time step. Block
MinorDeriv Compute a block’s state derivatives at a minor time step. Block
MinorZeroCrossings Compute a block’s zero crossing values at a minor time step. Block
ModelTerminate Free memory and perform any other end-of-simulation cleanup. System
The profiler measures the time required to execute each invocation of these functions and
generates a report at the end of the model that details how much time was spent in each
function.
The considered models have been tested using Simulink Profiler for a simulation time of 0.3
s with a simulation step of 1 ms. Ode5 fixed step solver has been used.
Table 3.2. Simulink Profiler Summary Report for the considered models.
Parameter Model A Model B Model C

Total recorded time 13.35 sec 6.98 sec 3.02 sec
Number of block methods 71 40 15
Number of internal methods 9 9 9
The parameters recorded in Table 2 are:

• Total recorded time – total time required to simulate the model using Simulink
Profiler;
• Number of block methods – total number of invocations of block-level functions
(e.g. Output());
• Number of internal methods – total number of invocations of system-level functions
(e.g. ModelExecute).
The number of blocks used in a model has a big influence on the total simulation time as
shown in Table 2. Model A, which used much more blocks than Model B, has the biggest
recorded time even if the number of Integrator blocks is equal.
Fig. 3.30 shows the simulation time in percent for functions invoked to simulate the entire
model. It has been omitted the functions for the constitutive blocks from the models.
Function Integrate is the most time consuming procedure during the simulation around 70-80
% from the total simulation time. So, the total simulation time for a Simulink model is
affected first by the number of blocks used in the model and then by the number of Integrator
blocks.
59
Fig. 3.30. Total time spent executing all invocations of specific functions for the considered models as a
percentage of the total simulation time for Model A.
Fig. 3.31 shows a comparative study regarding execution time for a C S-function block and
one Integrator block (compute the rotor speed) from the Model C.
Fig. 3.31. Average time for execution of a C S-Function and one Integrator block.
It has been considered the average time (self-time) required to execute Update, Output and
MinorDeriv functions for these blocks as a percent from the sum of the self-time for these
functions for the Integrator block. It is obvious that the total time required for all invocations
of functions related to the S-function block is smaller than the corresponding time for one
Integrator block even if this S-function implements five differential equations.
Therefore, using a S-function which implements all the differential equations from the
considered dynamic system, the simulation speed can be increased with at least a factor of
two. Moreover, the numerical stability increases using an S-Function.
60
So, in modelling of large dynamic systems, which involves a big number of Integrator blocks
it is desirable to use C S-Function concept in order to increase the simulation speed and the
numerical stability.
3.9 Summary
In this chapter the general dynamic equations of induction machine both in the natural
reference frame (ABC/abc) and the dqo reference frame using matrix form are presented.
Based on the dqo model the complex form can easily be derivate and then the steady-state
equations obtained. In each situation, the dynamic equations are treated with fluxes and
currents as state variables. Moreover, some specific features of the large induction machines
are presented e.g. deep-bar effect, saturation, as well as the methods, which can be used in
order to implement these effects.
Finally, some implementation methods for the induction machine model in Matlab/Simulink
as well as a comparison are shown.
3.10 References
[1]. I. Bala – Electrical Machines, Ed. Didactica si Pedagogica, Bucharest 1982 (in romanian);
[2]. I. Boldea, S.A. Nasar – Electric Drives, CRC Press, 1999, ISBN 0-8493-2521-8;
[3]. I. Boldea - Electrical Machines and Transformers, Ed. Didactica si Pedagogica, Bucharest 1994, ISBN
973-30-2341-8 (in romanian);
[4]. I. Boldea IM handbook
[5]. A. Fransua, R. Magureanu – Electrical Machines and Drives Systems, Technical Press, Oxford, 1984,
ISBN 0-291-39503-1;
[6]. P. Kovacs– Transient analysis of electric machines, Ed. Tehnica, Bucuresti 1980;
[7]. P.C. Krause, O. Wasynczuk, S.D. Sudhoff – Analysis of Electric Machinery, IEEE Press 1995, ISBN 0-
7803-1101-9;
[8]. G, McPherson, R.D. Laramore – Electrical Machines and Transformers, J. Wiley & Sons, 1990, ISBN 0-
471-63529-4;
[9]. P. Vas – Electrical Machines and Drives. A Space-Vector Theory Approach, Clarendon Press, Oxford
1992, ISBN 0-19-859378-3;
[10]. C. Goldemberg, A. de Arruda Penteado – Improvements on the inductance matrix inversion simplifying
the use of the ABC/abc induction machine model, Proceed. Of IEMD ’99, pp. 422-424;
[11]. P. Pillay, V. Levin – Mathematical models for induction machines;
[12]. C.L.C Allan – Water-turbine-driven induction generators, Trans on IEE, Dec 1959, pp. 529 – 550;
[13]. J.E. Barkle, R.W. Ferguson – Induction generator theory and application, Trans. on AIEE, February 1954,
pp. 12-19;
[14]. C.L.C Allan – Water-turbine-driven induction generators, Trans on IEE, Dec 1959, pp. 529 – 550;
[15]. M. Akbaba, S.Q. Fakhro – New model for single-unit representation of induction motor loads, including
skin-effect, for power system transient stability studies, IEE Proceedings-B, Vol. 139, No.6, November
1992;
[16]. K. Macek-Kaminska – Estimation of the induction machine parameters, Electric machines and power
systems, 1995, pp. 329-344;
[17]. K. Macek-Kaminska, P. Wach – Estimation of the parameters of mathematical models of squirrel-cage
induction motors, IEEE 1996, pp. 337-342;
[18]. *** - Simulink: Dynamic System Simulation for Matlab. Using Simulink. The Mathworks Inc.;
61
62
Chapter 4
Reduced order modelling and eigenvalue

analysis of induction machine
Over the years considerable attention has been paid to the reduced order modelling of
induction machine primarily for the purpose of predicting the dynamic behaviour of it during
large excursions in some or in all of the machine variables. First, the dynamic behaviour of
the induction machines was predicted with the steady-state voltage equations and the
dynamic relationship between rotor speed and torque. Then, the reduced order model, which
neglects the stator transients, has been developed in the power systems analysis.
However the reduced order models are very useful in the advanced aero elastical simulation
tools (e.g. HAWC), where the focus is on the frequency scale between 0 –20 Hz, which is the
most relevant for the dynamics of the wind turbine.
Regarding the “classical” steady-state model it is reported in [1], [6] and [7] that the dynamic
performances of the relatively large induction machines are not predicted accurately.
In the last years some new approaches in the reduced order modelling of the induction
machine are reported in the literature. A simplified second-order model has been developed
as an alternative to third-order model in [2], [3]. In [4] and [5] based on integral manifolds
theory the reduced order models for synchronous machine are obtained. Then this theory is
extended to induction machines in [6] and [7].
All this papers deals with the eigenvalue analysis of the reduced order models based on the
commonly accepted fifth order model for the induction machine, where 4 differential
equations describes the electrical transients both in stator and rotor windings and one
differential equation which describes the mechanical transients. Moreover, the analysis is
performed for the squirrel-cage induction motor and do not take into account the variations of
the rotor parameters with the slip (deep-bar effect), which is significant for a large machine.
Some eigenvalue analysis for doubly-fed induction machine are reported in literature [8]-
[11], but the focus is on slip energy recovery drives, which implies a uni-directional power
flow. Although none of these papers take into account the deep-bar effect.
Unfortunately, these papers do not consider the effect of the drive train as in the case of a
wind turbine. The dynamic behaviour of the large machines should be investigated
considering the mechanical and aero elastic aspects as in the case of a wind turbine.
This chapter focuses on the analysis of the eigenvalues for the complete and reduced order
models of large induction generators, which are used in wind turbine applications. The non-
linear dynamic equations of the induction machine presented in Chapter 3 are linearized
using a Taylor expansion about an operating point. The resulting set of linear differential
equations describes the dynamic behaviour of small displacement about this operating point.
Then the machine can be treated as a linear system with regard of small disturbances,
whereupon basic linear theory can be used to calculate eigenvalues and to establish transfer
functions for use in the design of the control associated with the induction machine, both
63
squirrel-cage and doubly-fed with and without the deep-bar effect. In each case the one mass
model and the two-mass model of the drive-train model are taken into consideration in the
analysis.
Using Control System Toolbox from Matlab the reduced order models are extracted. The
main interest for a aero elastic simulation tools is in the transfer function, which describe the
output of the system, namely electromagnetic torque, as a function of the input, which is the
wind turbine torque, when the others inputs are constant. Using this transfer function, the
frequency response of the complete and reduced order models in both cases are treated and
compared using Bode plots.
Then, the considered models are compared under different operating modes e.g. start-up,
fault on the stator side and normal continuous operation.
4.1 Reduced order modelling of induction machine

There are two reduced order models commonly employed in the case of the induction
machine. The simplest of these is the one, which neglects both stator and rotor transients or
so called the steady-state model presented in Chapter 3. In the following the reduced order
model, which neglects only the stator transients will be briefly presented.
Starting from (3.3.1) the voltage equations written in dq-synchronous reference frame (or xy)
are:
dψ sd
vsd = R s ⋅ isd + − ωe ψ sq
dt
dψ
vsq = R s ⋅ isq + sq + ωe ψ sd
dt
(4.1.1)
dψ
v rd = R r ⋅ i rd + rd − (ωe − ωr )ψ rq
dt
dψ
v rq = R r ⋅ i rq + rq + (ωe − ωr )ψ rd
dt
The relation between fluxes and currents is given by:
 ψ sd   Lss 0 L m 0  isd 
ψ    
 sq  =  0 Lss 0 L m  isq 
⋅ (4.1.2)
 ψ rd   L m 0 L rr 0  i rd 
     
 ψ rq   0 L m 0 L rr  i rq 
Substituting the fluxes in (4.1.1) will result:
64
 R s L rr RL 
 −ωe − s m 0 
• D D
0   ψ sd   
 vsd  1 0 0
 R s L rr R s L m  ψ sd 
     ωe −  
0   ψ sq  
0
 vsq  =  0 1 0 D D   ψ sq 
⋅ +  ⋅ ψ  (4.1.3)
 v rd   0 0 1 0   ψ rd   R r L m R r Lss
      − 0 − ( ωe − ωr )   rd 
 v rq   0 0 0 1   ψ rq  D D   ψ rq 
 R r Lm R r Lss 
 0

− ( ωe − ωr ) 

D D
where: D = Lss L rr − L m 2
The theory of neglecting electric transients is presented in [1]. For balanced steady-state
operation, the variables in synchronous reference frame are constants. Hence, the electric
transients of the stator can be neglecting if the derivative of the stator fluxes in the d and q
axis are set to zero in (4.1.3) as:
 R s L rr RL 
 −ωe − s m 0 
• D D
0   ψ sd   
 vsd   0 0 0
 R s L rr R s L m  ψ sd 
     ωe −  
0   ψ sq  
0
 vsq  =  0 0 0 D D   ψ sq 
⋅ +  ⋅ ψ  (4.1.4)
 v rd   0 0 1 0   ψ rd   R r L m R r Lss
      − 0 − ( ωe − ωr )   rd 
 v rq   0 0 0 1   ψ rq  D D   ψ rq 
 RL R r Lss 
 0 − r m ( ωe − ωr ) 
 D D 
In order to avoid the computational problems due to algebraic loops (4.1.4) should be written
in state space form only in terms of the rotor fluxes as:
  τrm    ατrm    ατrm ατrm  v 
•   ατrm σ r −    L m + 1 ωe − ωr  
 1 0   sd 
ψ rd    σs    τsr     ψ rd   τsr τ sr
2
v
 =   ψ  +  ατ ατ   sq  (4.1.5)
ψ rq     ατrm   τ   rq   − 2 rm  v 
− L m + 1 ωe − ωr  ατrm σ r − rm
rm
0 1   rd 
   τ σs   τ sr τsr   v rq 
   sr   
R s L rr RL L L 1
where: τsr = , τrm = r m , σ r = m , σs = m , α = 2
.
D D L rr Lss ω 
1+  e 
 τsr 
Based on rotor fluxes and input voltages the stator fluxes can be obtained using:
 α   α α 
 ασ r L m ωe   ωe 
 Ψ sd  τsr  ψ rd  τsr τ sr
2
 vsd 
Ψ  =     +     (4.1.6)
 sq   − α L ω   ψ rq   α α   vsq 
 τ m e ασ r   − τ2 ωe τ 
 sr   sr sr 
Then the electromagnetic torque is calculated based on:
65
3 Lm
Te =
p⋅ (ψ sq ⋅ ψ rd − ψ rq ⋅ ψ sd ) (4.1.7)
2 D
However, the reduced order model, which neglects the stator transients, can be extracted
from the full order model given by (4.1.3) using Control System Toolbox from Matlab. This
toolbox is a very powerful tool for reduced order modeling. Based on some specific functions
as modred, ssbal, minreal, and balreal the desired state variables can be eliminated from the
state space equations and the reduced order models are obtained.
4.2 Linearized equation of induction machine

In Chapter 3 the voltage equations for induction machine have been written in dqo arbitrary
reference frame. In the followings these voltage equations will be treated in dq-synchronous
reference frame.
The dynamic equations of the induction machine model given by (4.1.3) has appropriate use
because the Control System Toolbox from Matlab® Environment is used to analyze the
system behaviour. The Control System Toolbox needs a state-space representation of the
system like:
•
x = Ax + Bu (4.2.1)
y = Cx + Du
where: A, B, C and D are matrices, x is the states vector, u is the input vector and y is the
output.
The voltage and the equation of motion determine the fifth order system for the induction
machine. This model can be linearized using Taylor’s expansion about a fixed value [1].
4.2.1 One-mass model
If we take the one-mass model of the wind turbine drive train into account, the fifth order
linearized system for the induction machine is:
 R s L rr RL 
 0 − s m 0 0 
1 0 0 0 0  • D D
∆ψ    ∆ψ 
 ∆u sd     
 R s L rr R s Lm
0 
sd sd
 ∆u   0 1 0 0 0   0 0 − 
 ∆ψ sq   D D ∆ψ
  sq 
 sq   0 0 1 0 0     ∆ψ  (4.2.2)
 ∆u rd  =    ∆ψ rd  +  − R r L m 0
R r Lss
ωr 
ψ rq0 ⋅ p  rd 
  0 0 0 1 0  
∆ψ  D D   ∆ψ rq 
 ∆u rq   J
 rq 
 
 ∆Twtr   0 0 0 0 −   ∆Ω r   0 R r Lm R r Lss 
 p  − ω r −ψ rd0 ⋅ p   ∆Ω r 
 D D 
 − kψ rq 0 kψ sd0 kψ sq0 −kψ rd0 0 
 
3 Lm
where: k = p . Or in compact matrix form:
2 D
[ u ] = [ E ] ⋅  x  + [ F][ x ]
•
(4.2.3)
 
66
[ u ] =  ∆u sd
'
where: ∆u sq ∆u rd ∆u rq ∆T ' wtr  - input vector
[ x ] =  ∆ψsd
'
∆ψ sq ∆ψ rd ∆ψ rq ∆Ω gen  - state vector
The state space form of (4.2.3) is:

•
 x  = [ E ] ⋅ [ F] ⋅ [ x ] + [ E ] [ u ] = [ A ] ⋅ [ x ] + [ B] ⋅ [ u ]
−1 −1
(4.2.4)
4.2.2 Two-mass model
The voltage equations of induction machine (4.1.3) and the two-mass model of the drive train
with torsional torques define the complete system under analysis.
 R s L rr RL 
 0 − s m 0 0 0 0 0 
• D D
'
 ∆u sd   1   ∆ψ sd      ∆ψ 
R s L rr RL
0 
sd
 ∆u   1   ∆ψ  0 0 − s m 0 0 0 
 sq     sq   D D   ∆ψ sq 
 ∆u rd   1   ∆ψ rd   R r L m R r Lss
  ∆ψ 
 − ωr ψ rq0 p 0 0  rd
     0 0 
 ∆u rq  =  1  ⋅  ∆ψ rq  +  D D   ∆ψ rq  (4.2.5)
 0   − J gen   ∆Ω gen  RL R r Lss  
       0 − r m ωr −ψ rd0 p 0 0 0   ∆Ω gen 
D D
 0   −1   ∆θgen     ∆θgen 
 ∆T '   J '   ∆Ω '   − kψ rq0 kψ sd0 kψ sq0 −kψ rd0 D e − k se − D e k 'se   ' 
' ' '
  '   0   ∆Ω wtr
   0  ' 
wtr wtr wtr
0 0 0 −1 0 0
 0   −1   ∆Ω wtr    ∆θ wtr 
0 0 0 0 D e − k se − D'e
' '
k 'se 
 
 0 0 0 0 0 0 −1 0 
or in compact matrix form:
[ u ] = [ E ] ⋅  x  + [ F][ x ]
•
(4.2.6)
 
[ u ] =  ∆u sd
'
where: ∆u sq ∆u rd ∆u rq 0 0 ∆T ' wtr 0  - input vector
[ x ] =  ∆ψsd
'
∆ψ sq ∆ψ rd ∆ψ rq ∆Ω gen ∆θgen ∆Ω ' wtr ∆Ω ' wtr  - state vector
The state space form of (4.2.6) is:

•
 x  = [ E ] ⋅ [ F] ⋅ [ x ] + [ E ] [ u ] = [ A ] ⋅ [ x ] + [ B] ⋅ [ u ]
−1 −1
(4.2.7)
4.3 Eigenvalue analysis

The roots of the characteristic matrix [A] from (4.2.4) and (4.2.7) gives the eigenvalues of the
considered system.
The eigenvalues provide a simple means of predicting the behaviour of an electrical machine
at any balanced operating condition. Eigenvalues may either be real or complex and when
complex they occur as conjugate pairs signifying a mode of oscillation of the state variables.
67
Negative real parts correspond to state variables or oscillations of state variables, which
decrease exponentially with time. Positive real pars indicate an exponential increase with
time, an unstable condition.
These eigenvalues can be obtained when one of the input parameters is varied in the
characteristic equation e.g. rotor speed, rotor resistance and rotor leakage reactance, etc.
For the eigenvalue analysis a 2 MW wind turbine has been considered. First a squirrel-cage
induction generator (SCIG) and then a doubly-fed induction generator (DFIG) has been used.
The electrical parameters for these machines are given in Table 4.1.
Table 4.1. Typical electrical parameters for 2 MW induction machines in wind turbine applications.
Parameters at rated speed SCIG DFIG
Rated power [MW] 2 2
Stator resistance [Ω] 0.005 0.001
Stator leakage inductance [mH] 0.4 0.07
Rotor resistance [Ω] 0.01 0.0013
Rotor leakage inductance [mH] 0.3 0.08
Magnetizing inductance [mH] 16 3
Number of pole pairs 2 2
Moment of inertia [kgm2] 90 65
Rated speed [rpm] 1512 1512
The parameters of the drive train in the case of a 2MW wind turbine are given in Table 4.2
Table 4.2. Typical parameters of a drive train for 2 MW wind turbine.
Parameters
Moment of inertia of the wind turbine rotor [kgm2] ≅ 50 ⋅105
Equivalent spring constant indicating the torsional stiffness of the shaft on
≅ 1.7 ⋅104
the generator side
Equivalent damping coefficient of the shaft on the generator side ≅100
Gear box ratio ≅90
The eigenvalue analysis has been performed in two cases: using the one-mass model and with
the two-mass model of the drive train both for SCIG and DFIG. Moreover, the influence of
the deep-bar effect has been taken into account for a large DFIG.
The main interest in these analysis is in the transfer function, which describe the output of the
system, namely electromagnetic torque, as a function of the input, which is the wind turbine
torque, when the others inputs are constant.
4.3.1 Squirrel-cage induction machine with a one-mass model for the

wind turbine drive train
When the one-mass model is use in conjunction with (4.1.3) results the 5th order model,
neglecting stator transients and then rotor transients the 3rd and respectively the 1st are
obtained. The eigenvalues for the full 5th and reduced order models in the case of a 2MW
wind turbine are shown in Table 4.3.
68
Table 4.3 Eigenvalues for complete and reduced order models of SCIG at rated speed using
a one-mass model for the wind turbine drive train
Complete model Reduced model Steady-state model
5th order 3rd order 1st order
Frequency Frequency Frequency
Eigenvalues Eigenvalues Eigenvalues
[rad/sec] [rad/sec] [rad/sec]
Torque
-12.2 12.2 -12.2 12.2 -19.6 19.6
eigenvalue
Rotor
-6.69 ± j 12.1 13.8 -6.69 ± j 12.1 13.8 - -
eigenvalue
Stator
-7.13 ± j 314 314 - - - -
eigenvalue
Based on Table 4.3 it is possible to identify an association between eigenvalues and the
machine variables. The complex conjugate pair, which remains at a frequency close to
synchronous frequency, is primarily associated with the transient offset currents in the stator
windings that reflect into the synchronously rotating reference frame as a decaying 50 Hz
variation. This complex conjugate pair is denoted as the stator eigenvalue and is not present
when the electric transients are neglected in the stator voltage equations. The complex
conjugate pair, which changes in frequency as the rotor speed varies is associated primarily
with the electric transients in the rotor circuits and is denoted as the rotor eigenvalue and is
not present when the rotor electric transients are neglected. The negative real eigenvalue
signifies an exponential response and will be denoted as torque eigenvalue. It will
characterize the behaviour of the machine equations if all electric transients are
mathematically neglected.
The Bode plots for the considered transfer function Te = f ( Twt ) are shown in Fig. 4.1.
Fig. 4.1. Frequency response for complete and reduced order models of the SCIG using
a one-mass model for the wind turbine drive train.
It can be observed from Fig. 4.1 that the response in frequency for the full and 3rd order
model are basically the same in the frequency range of interest, while the response of the
steady-state model is not accurate both in magnitude and phase.
69
4.3.2 Squirrel-cage induction machine with a two-mass model for the
The eigenvalues of the complete and reduced order models when the two-mass model of the
drive train is used are shown in Table 4.4.
In this case a new complex pair of eigenvalues is presented in the system formed by the
voltage equations of the induction machine and the drive train model. This complex
conjugate pair is related to the shaft angle and is denoted as the twist eigenvalue.
Table 4.4 Eigenvalues for complete and reduced order models of SCIG at rated speed using
a two-mass model for the wind turbine drive train
8th order 6th order 4th order
removed 0 0 - - - -
Twist
-0.732 ± j 4.90 4.95 -0.732 ± j 4.90 4.95 -0.732 ± j 4.90 4.95
eigenvalue
Torque
-12.4 12.4 -12.4 12.4 -227 227
eigenvalue
Rotor
-5.09 ± j 42.7 43 -5.09 ± j 42.7 43 - -
eigenvalue
Stator
-7.18 ± j 314 314 - - - -
eigenvalue
Fig. 4.2. Frequency response for complete and reduced order models of the SCIG using
a two-mass model for the wind turbine drive train.
When a two-mass model for the wind turbine drive train is used for analysis all the
considered models have the same frequency response. A change in the sign of the magnitude
around 6 rad/sec can be observed as well as in the phase.
70
4.3.3 Doubly-fed induction machine with a one-mass model for the
When a one-mass model for the wind turbine drive train is used in conjunction with (4.1.3)
results the 5th order model, neglecting stator transients and then rotor transients the 3rd and
respectively the 1st are obtained. The eigenvalues for the complete and reduced order models
in the case of a 2MW wind turbine are shown in Table 4.5.
Table 4.5 Eigenvalues for complete and reduced order models of DFIG at rated speed using
5th order 3rd order 1st order
Torque
-8.66 8.66 -8.66 8.66 -205 205
eigenvalue
Rotor
-4.69 ± j 32.9 33.2 -4.69 ± j 32.9 33.2 - -
eigenvalue
Stator
-7.72 ± j 314 314 - - - -
eigenvalue
The frequency response of the considered models in this case is shown in Fig. 4.3.
Fig. 4.3. Frequency response for complete and reduced order models of the DFIG using
Using a one-mass model only the complete and the reduced order model presents the same
frequency response. Again, the electromagnetic torque predicted with the steady-state model
will not have a good frequency response.
71
4.3.4 Doubly-fed induction machine with a two-mass model for the
When a two-mass model for the wind turbine drive train is used the 8th order model will
result, neglecting stator transients and then rotor transients the 6th and respectively the 4th
order models are obtained. The eigenvalues for the complete and reduced order models in
this case are shown in Table 4.6.
Table 4.6 Eigenvalues for complete and reduced order models of DFIG at rated speed using
a two-mass model for the wind turbine drive train.
8th order 6th order 4th order
removed 0 0 - - - -
Twist
-0.0326 ± j 5.23 5.23 -0.0326 ± j 5.23 5.23 -0.0326 ± j 5.23 5.23
eigenvalue
Torque
-8.68 8.68 -8.68 8.68 -2160 2160
eigenvalue
Rotor
-3.4 ± j 107 107 -3.4 ± j 107 107 - -
eigenvalue
Stator
-8.15 ± j 314 314 - - - -
eigenvalue
Fig. 4.4. Frequency response for complete and reduced order models of the DFIG using
the complete model of the drive train.
Changes in sign both for magnitude and phase around 6 rad/sec (≅ 0.9 Hz) can be observed as
shown in Fig. 4.4.
72
4.3.5 Start-up and fault analysis on the stator side for a squirrel-cage
induction machine
In fault analysis on the stator side the interest is in the transfer function, which describes the
q-axis component of the stator voltage, when the other inputs are constants.
The frequency response of the transfer function Te=f(vqs) for the complete and the reduced
order models using a one-mass model as well as a two-mass model for the wind turbine drive
train are shown in Fig. 4.5 and Fig. 4.6 respectively.
Fig. 4.5. Frequency response of Te=f(vqs) for complete and reduced order models using
the one-mass model for the wind turbine drive train.
Fig. 4.6. Frequency response of Te=f(vqs) for complete and reduced order models using
the two-mass model for the wind turbine drive train.
In this case the response of the reduced order model, which neglects the stator transients, is
more accurate than the steady-state one.
73
Then, the considered models have been analyzed under different transient conditions: the
direct start-up sequence (without soft-starter) and a three-phase fault at the stator terminals.
For the start-up sequence the generator is connected to the grid at synchronous speed with a
50 % driven torque, then a three-phase short-circuit at the stator terminals occurs for 180
msec. The electromagnetic torque and the rotor speed in both cases are presented in
a) b)
Fig. 4.7. Electromagnetic torque and rotor speed for different operating modes:
a) start-up and b) three-phase short-circuit at the stator terminals.
It can be observed that during the direct start-up sequence the reduced order models cannot
predict accurately the electromagnetic torque and the rotor speed. When a fault occurs on the
stator side the reduced order model, which neglects the stator transients, gives relatively good
results in terms of torque and rotor speed, while the steady-state model cannot predict the
fault transients.
4.3.6 Analysis of normal continuous operation for a squirrel-cage

induction generator under a given wind profile
Based on the developed models from the “Wind Turbine Blockset”, which will be presented
later in Chapter 7, the considered models for the induction generator have been tested under a
given wind profile The model used for simulation comprises a wind model, an aerodynamic
model of the wind turbine rotor, the two-mass model of the wind turbine drive train, the 2
MW squirrel-cage induction generator models and the grid model.
The wind speed is calculated as an average value of the fixed-point wind speed over the
whole rotor, and takes the shadow tower and the rotational turbulences into account. The
generated wind profile used in simulation is shown in Fig. 4.8 a.
The aerodynamic model of the wind turbine is based on the torque coefficient CQ look-up
table and the aerodynamic torque is determined directly using (2.2.4).
For clearance of plots, the electromagnetic torque and the stator active power are shown in
Fig. 4.8 b only for 10 sec interval.
74
a) b)
Fig. 4.8. Analysis of the normal continuous operation of the wind turbine:
a) considered wind profile and b) electromagnetic torque and produced stator active power.
In steady state operation there is a difference around 10 – 15 % in amplitude and a phase lag
between the steady state model and the others. This difference results also from the Bode
plots around 6 Hz (see Fig. 4.2) and is mainly due to the rotor eigenvalues.
4.3.7 Influence of the deep-bar effect for doubly-fed induction

generator
In the case of a variable speed wind turbine, which uses a DFIG and a power converter in the
rotor side of the machine, the steady state characteristics of the system will be modified by
the deep-bar effect. As shown in Chapter 3 due to this effect, the electromagnetic torque is
changed for a given slip. Since the variable speed operation of the wind turbine typically
requires a speed range of ±30% the eigenvalues analysis should take into account the deep-
bar effect.
The variation of the rotor parameters with the rotor speed will not affect only the behaviour
of the system but will require also an adequate control of the power converter during variable
speed operation of the wind turbine.
A slip range of -0.1 – 0.5 has been considered in the eigenvalues analysis. The eigenvalues of
the system, when a two-mass model for the wind turbine drive train is taken into account, are
calculated at 0.5 slip. Table 4.7 summarize these eigenvalues for both situations: without and
with deep-bar effect.
Table 4.7. Eigenvalues for the complete model of DFIG without and with deep-bar effect at 0.5 slip.
Without deep-bar effect With deep-bar effect
Frequency Frequency
Eigenvalues Eigenvalues
[rad/sec] [rad/sec]
removed 0 0 - -
Twist eigenvalue -4.64 ± i 12.8 13.6 -6.84 ± i 10.9 12.9
Torque eigenvalue -1.83 1.83 -2.96 2.96
Rotor eigenvalue -1.69 ± i 190 190 -16.8 ± i 190 191
Stator eigenvalue -8.68 ± i 314 314 -8.68 ± i 314 314
The stator and the rotor eigenvalues for the entire slip range are shown in Fig. 4.9.
75
Fig. 4.9. Loci of the stator and rotor eigenvalues without and with deep-bar effect
for a slip range of –0.1 … 0.5.
Due to the deep-bar effect only the real part of the rotor eigenvalue varies, while the
imaginary part is almost fixed for the entire considered range. The stator eigenvalues are not
modified due to the deep-bar effect.
The twist eigenvalues and the torque eigenvalue in both considered situations are shown in
Fig. 4.10.
Fig. 4.10. Loci of the torque and twist eigenvalues without and with deep-bar effect
for a slip range of –0.1 … 0.5.
Both the torque and the twist eigenvalues are modified due to the deep-bar effect as shown in
Fig. 4.10. Finally, the frequency response of the complete and reduced order models for
doubly-fed induction generator with deep-bar effect is presented in Fig. 4.11 for 0.5 value of
slip.
76
Fig. 4.11. Frequency response of complete and reduced order models for DFIG with the deep-bar effect using a
two-mass model for the wind turbine drive train at 0.5 slip.
In this case the Bode plots are different then those for the rated speed, which indicates
another behaviour of the system under analysis.
4.3.8 Some conclusions regarding analysis of complete and reduced

order models for induction generators
Based on the analysis performed in § 4.4.1-4.4.4 it can be concluded that the two-mass model
normally used in aero elastic simulations should not be reduced to a one-mass model. The
frequency scale between 0 – 20 Hz is the most relevant for the dynamics of the wind turbine.
Since , the frequency of the rotor eigenvalues for the reduced order model, which neglects the
stator transients, is in the interest range of 0 – 20 Hz, this reduced order model is much more
suitable in fault analysis within a wind turbine.
Special care should be taken in evaluation of the eigenvalue frequencies for stator and rotor
because the analysis has been performed in terms of electrical speed. So, using the number of
poles (4 in this case) the eigenfrequencies for the electrical part of the generators can be
evaluated in the mechanical part of the wind turbine.
The “classical” steady state model for induction generator can be used with good results in
analysis of the wind turbine power performances if a two-mass model of the wind turbine
drive train is taken into account.
In the case of a variable-speed large wind turbine in both configurations, SCIG or DFIG, the
deep-bar effect should be taken into consideration in order to predict the transients as well as
to tune the parameters for the controllers.
4.4 HAWC implementation

The reduced order models for the induction generator neglecting stator transients and a
linearized steady-state model have been implemented and tested in the advanced aero elastic
code HAWC for a 2 MW stall regulated wind turbine.
77
HAWC, an aero elastic tool developed at RISØ National Laboratory, is a computer program
with the purpose of predicting load response for a horizontal axes 2 and 3 bladed wind
turbines in time domain [13].
It has been considered a 2 MW induction generator with short-circuited rotor windings
without any control on the electrical part (stator and rotor). In order to illustrate the behaviour
of the considered models a time simulation at low wind speed (6 m/sec) and high wind speed
(20 m/sec) with a turbulence intensity of 10% has been performed. The results of 100 sec
time simulation are shown in Fig. 4.12 in terms of power spectra of the rotor speed, active
power and electromagnetic torque of the generator.
a) b)
Fig. 4.12. Simulation results from HAWC using reduced order models for the induction generator: a) at 6 m/s
wind speed and b) at 20 m/s wind speed with a turbulence intensity of 10%.
In “dynamic model” means the reduced order model neglecting stator transients, while the
“static slip model” is the linearized steady-state model.
Fixed speed wind turbines exhibits periodic power pulsations in addition to the stochastic
power pulsations caused by the wind as shown in Chapter 2. Usually, the rotor frequency is
referred as 1p: 0.90 Hz in this case, as shown in Fig. 4.12.
In Fig. 4.4 changes in sign both for magnitude and phase for the electromagnetic torque
around 0.9 Hz appears due to this fundamental pulsation.
The high order components 3p, 6p, etc. are related with the frequency with which the blades
pass in front of the tower. In Fig. 4.12 these peaks appear around 2.5 Hz and 5 Hz and their
amplitude is related with the wind speed.
78
In Table 4.6 it has been found that the frequency of the rotor eigenvalues is around 107
rad/sec. Taking into account that the machine has 4 poles the rotor eigenvalues frequency
related to the mechanical part of the wind turbine is around 5 Hz. This peak is damped in the
case of the linearized steady-state model. The same peak has been observed also in the field
measurements.
It can be concluded that due to the dynamics of the induction generator some resonances
between the mechanical part and the electrical part of the wind turbine arises.
Thus, using a reduced order model, which neglects the stator transients, the interaction
between the mechanical structure of the wind turbine and the electrical part can be taken into
account in the case of a wind turbine.
4.5 Summary
This chapter focuses on the analysis of the eigenvalues for the complete and reduced order
models of large induction generators, which are used in wind turbine applications. The non-
linear dynamic equations of the induction machine presented in Chapter 3 are linearized
using a Taylor expansion about an operating point. The resulting set of linear differential
equations describes the dynamic behaviour of small displacement about this operating point.
Then the machine can be treated as a linear system with regard of small disturbances,
whereupon basic linear theory can be used to calculate eigenvalues and to establish transfer
functions for use in the design of the control associated with the induction machine, both
squirrel-cage and doubly-fed with and without the deep-bar effect. In each case a one-mass
model and a two-mass model of the wind turbine drive train are taken into consideration in
the analysis.
Using Control System Toolbox from Matlab the reduced order models are extracted. The
main interest for a aero elastic simulation tools is in the transfer function, which describe the
wind turbine torque, when the others inputs are constant. Using this transfer function, the
frequency response of the complete and reduced order models in both cases are treated and
compared using Bode plots.
Then the considered models have been analyzed under different operating conditions e.g.
start-up, fault on the stator side and normal continuous operation.
Finally, a comparison for the time response of the reduced order models implemented in
HAWC is shown.
4.6 References
[1]. P.C. Krause, O. Wasynczuk, S.D. Sudhoff – Analysis of Electric Machinery, IEEE Press 1995, ISBN 0-
7803-1101-9;
[2]. G.G. Richards, O.T. Tan – Simplified models for induction machines transients under balanced and
unbalanced conditions, IEEE Trans. on Industry Applications, Vol. 17, No. 1, January/February 1981, pp.
15-21;
[3]. G.G. Richards – Reduced order model for single and double cage induction motors during start-up, IEEE
Trans. on Energy Conversion, Vol. 3,No. 2, June 1988, pp. 335-341;
79
[4]. P.W. Sauer, S. Ahmed-Zaid, P.V. Kokotovic – An integral manifold approach to reduced order dynamic
modeling of synchronous machines, IEEE Trans. on Power Systems, Vol. 3, No. 1, February 1988, pp. 17-
23;
[5]. P.V. Kokotovic, P.W. Sauer – Integral manifold as a tool for reduced-order modelling of non-linear
systems: a synchronous machine case study, IEEE Trans. on Circuits and Systems, Vol. 36, No. 3, March
1989, pp. 403-410;
[6]. E. Drenan, S. Ahmed-Zaid, P.W. Sauer – Invariant manifolds and start-up dynamics of induction
machines, Power Symposium, 1989, Proceedings of the Twenty-First Annual North-American, 1989, pp.
129 –138;
[7]. S. Ahmed-Zaid, M. Taleb – Structural modeling of small and large induction machines using integral
manifolds, IEEE Trans. on Energy Conversion, Vol. 6, No. 3, September 1991, pp. 529-535;
[8]. V. Akhamtov - Modelling of Variable-Speed Wind Turbines with Double-fed Induction Generators in
Short Term Stability Investigations, Proceed. of 3rd Int. Workshop on Transmission Networks for Offshore
Wind Farms, April 11-12, 2002, Stockholm, Sweden, pp. 1-23;
[9]. S.A. Papathanasiou, M.P. Papadopoulos – State-Space Modeling and Eigenvalue Analysis of the Slip
Energy Recovery Drive, IEE Proceed. on Electr. Power Appl., Vol 144, No. 1, January 1997, pp. 27-36;
[10]. A.M. Stankovic, B.C. Lesieutre – Parametric Variations in Dynamic Models of Induction Machine
Cluster, IEEE Trans on Power Systems, Vol. 12, No.4, November 1997, pp. 1549-1554;
[11]. L. Congwei, W. Haiqing, S. Xudong, Li Fahai – Research of Stability of Double-fed Induction Motor
Vector Control System, Proceed. of ICEMS 2001, Vol. 2, August 2001, pp. 1203 -1206;
[13]. T.J. Larsen, M.H. Hansen, F. Iov – Generator dynamics simulated in the aeroelastic code HAWC Risø
report, Risø-R-1395E;
[14]. T. Thiringer, J-A Dahlberg – Periodic pulsations from a three-bladed wind turbine, IEEE Trans. on
Energy Conversion, Vol. 16, No. 2, June 2001, pp.128-133;
80
Chapter 5
Power Converters Modelling

Since the directly-grid connected wind turbine and the variable-speed one, which use a
doubly-fed induction generator, are currently the state-of the-art in wind turbine technology,
as shown in Chapter 2, in the following paragraphs some aspects regarding the modelling of
the power converter topologies, which are used for this two concepts will be presented.
The focus will be on the modelling of different soft-starters-fed induction machine topologies
as well as on the voltage source converters modelling using the switching function concept.
The influence of the load connection is taken into account in both cases.
5.1 AC Controllers
5.1.1 Background
Many authors deal with the theory of operation for three-phase AC-controllers (soft-starters)
[1]-[5], however, only few of them present in details the theory of operation with a 3-phase
resistive-inductive load [1], [2]. Especially the influence of the load connection type and the
steady-state analysis are not presented in detail.
In [6] a comparison for full and half-wave controlled load both for star and delta connection
is carried out, unfortunately, the branch delta connection is not treated. An analysis of
variable-voltage thyristor controlled induction motors in terms of specific operation modes,
harmonic content and performances for a star connected machine is presented in [7]. In [8] is
presented a hybrid abc-dqo model for the induction machine, unfortunately this model has a
complicated mathematical model and it can be used only for a star connection of the stator
windings. A generalized approach in modelling of the power converters-fed induction
machine, which involves a time-domain static network, is presented in [9]. This model use
some “dummy” shunt resistances in order to eliminate the computation errors, however the
proposed method can be used for a star or delta-connected machine. Moreover these
approaches do not take into account the deep-bar effect, which is present into a large
induction machine.
Therefore in analysis of a soft-starter-fed induction machine is very convenient to use the
abc/abc model presented in Chapter 3, since it is based on per-phase quantities.
5.1.2 Topologies of AC Controllers-fed induction machine in wind

turbine applications
At present many wind turbines, up to 2.3 MW, are based on the “Danish concept” in which a
squirrel-cage induction generator is directly connected to the grid as shown in Fig. 5.1.
81
Fig. 5.1. Block diagram of directly grid-connected wind turbine.
The scheme comprises the wind turbine rotor, linked via a gearbox to a generator, which
through an electrical interface is connected to the grid. A control system is necessary to
assure a proper operation of the wind turbine under all conditions. The electrical interface
consists in: an soft-starter, a capacitor bank and a transformer, which makes the connection
with the medium voltage grid (typically 30 kV). The capacitor bank is used to control the
power factor of the generator output.
Soft-starters are used only during the start-up sequence of the generator in order to limit the
inrush currents and the starting torque transients in the drive train.
There are many configurations of soft-starters, which fed an induction machine [1]-[5],
although of interest in wind turbine applications are those presented in Fig. 5.2.
Fig. 5.2. Possible configurations of soft-starter-fed induction machine in wind turbine applications:
a) star connection, b) delta connection and c) branch-delta connection.
The star and delta configurations have basically the same layout for the semiconductors
(SCRs), the difference consists in the machine winding connection. There are two anti-
parallel thyristors on each phase, each one conducting on the positive cycle of the applied
voltage. For a star connection the applied per phase voltages depend on the on-state of the
SCRs on each phase. So, the soft-starter can operated only when two or three SCRs are
conducting in the same moment. The similar considerations can be done for a delta
connection (Fig. 2b).
However a soft-starter with a branch delta connection as shown in Fig. 2c, will operate only
with one SCR conducting at a given moment.
In wind turbine applications mainly the delta connection for the induction machine are used
because the current rating of the stator windings can be reduced, and the third harmonic in
the line currents is eliminated in this case.
82
5.1.3 Operation modes for a soft-starter with resistive-inductive load
Depending on the firing angle α, three different modes of operation of the soft-starter can be
distinguished when a star or delta connected resistive load is used [1]:
• Mode 1 - 0 ≤ α < 600 - two or three SCRs are conducting (in either direction);
• Mode 2 - 600 ≤ α < 900 - two SCRs are conducting;
• Mode 3 - 900 ≤ α < 1500 - none or two SCRs are conducting.
where α is the firing angle for the soft-starter.
When a resistive-inductive load is used the analysis of the controller is difficult, since the
operation modes depend on the extinction angle ξ and the limit angle αlim, both dependent on
the phase angle ϕ.
Mode 2 of operation, characterized by rapid changes of the output current, is not possible due
to the load inductance. The ranges of the two remaining operation modes are:
• Mode 1: ϕ ≤ α < α lim - two or three SCRs are conducting;
• Mode 3: α lim ≤ α < 150o - none or two SCRs are conducting.
The limit angle can be determined numerically from:

 4π  π
sin  α lim − ϕ −  −
 3 = 2e 3tan ϕ
−1
(5.1.1)
sin ( α lim − ϕ ) −
π
2 − e 3tan ϕ
5.1.4 Switching function concept
Switching functions SwA, SwB, and SwC in two levels can be introduced in modelling of the
SCRs and defined as equal to one when a given thyristor is conducting and equal to zero
otherwise as shown in Fig. 5.3.
Fig. 5.3. Modelling two anti parallel SCRs using a switching function.
These switching functions can be determined based on phase voltage and phase (or line)
current for each particular topology.
83
5.1.5 Star connected 3-phase load
A soft-starter-fed 3-phase resistive-inductive load in star connection is shown in Fig. 5.4.
Fig. 5.4. Soft-starter-fed 3-phase R-L load in star connection.
Analyzing this topology the applied per-phase voltages at the machine terminals can be
written as:
 v UX   SwA −SwB −SwC   v AN 
1
 v  =  −S   
 VY  2  wA SwB −SwC  ⋅  v BN  (5.1.2)
 v WZ   −SwA −SwB SwC   v CN 
where: [ v UX v VY v WZ ] are the phase voltages of the machine; [ v AN v BN vCN ] are the phase-
T T
to-ground supply voltages and SwA , SwB , SwC are the switching functions for each phase.
Based on the phase-to-ground voltages and the phase (line) currents the switching functions
are computed.
The considered load has a power factor of 0.85. Using (5.1.1) a limit angle around 95o will
result. In order to highlight the two possible operation modes for the soft-starters two values
for the firing angle have been used. A value of 70o for the firing angle will correspond to
Mode 1 and 110o to Mode 3 respectively.
The per-phase voltage and the phase current for the considered load are shown in Fig. 5.5 for
different firing angles.
a) b)
Fig. 5.5. Typical waveforms for voltage and currents for a 3-phase star connected load:
a) Mode 1and b) Mode 3 of operation.
84
From Fig. 5.5 it can be observed that when the firing angle α is greater than the limit angle
αlim the phase currents occurs in discontinuous, non-sinusoidal, alternating pulses. Therefore,
a high harmonic content for the phase currents will result.
5.1.6 Delta connected 3-phase load

Fig. 5.6 shows a three-phase resistive-inductive load in delta connection.
Fig. 5.6. Soft-starter-fed 3-phase R-L load in delta connection.
It can easily be demonstrated that the voltages applied to a delta-connected load are given by:
 v UX   SwASwB −0.5 ⋅ SwA −0.5 ⋅ SwB   v AB 

 v  =  −0.5 ⋅ S S S − 0.5 ⋅ S
 
⋅  v BC  (5.1.3)
 VY   wC wB wC wB 
 v WZ   −0.5 ⋅ SwC −0.5 ⋅ SwA SwCSwA   v CA 
 
where: [ v AB v BC vCA ] are the line-to-line supply voltages (or phase voltages for the load).
T
For a delta connection the switching functions are determined based on phase-to-ground
voltages and line currents.
Considering the same load as in the precedent case the applied per-phase voltage, the phase
and the line currents for different firing angles are shown in Fig. 5.7.
a) b)
Fig. 5.7. Typical waveforms for voltage and currents for a 3-phase delta connected load:
a) Mode 1and b) Mode 3 of operation.
Again, it can be observed that when the firing angle α is greater than the limit angle αlim the
line currents occurs in discontinuous, non-sinusoidal, alternating pulses. Therefore, for this
85
topology a high harmonic content for both line and phase currents will result especially in
Mode 3. As a result the electromagnetic torque will have a high harmonic content.
5.1.7 Branch-delta connected load
A soft-starter-fed 3-phase resistive-inductive load in branch-delta connection is presented in

Fig. 5.8.
Fig. 5.8. Soft-starter-fed 3-phase R-L load in branch-delta connection.
The output voltages of the soft-starter in this case are:

 v UX  SwA 0 0   v AB 
v  =  0 S 0  ⋅  v BC  (5.1.4)
 VY   wB
 v WZ   0 0 SwC   v CA 
For a branch-delta connection the switching functions are determined based on the line
voltages and line currents.
Again, a 3-phase resistive-inductive load as in precedent cases has been considered. The per-
phase voltage, the phase and the line current are shown in Fig. 5.9 for the considered firing
angles.
a) b)
Fig. 5.9. Typical waveforms for voltage and currents for a 3-phase branch delta connected load:
a) Mode 1, b) Mode 3 of operation.
Using a branch-delta connection the line current waveform is approximately the same for the
entire range of possible firing angle. Therefore, the harmonic content in the line / phase
86
currents as well as in the electromagnetic torque is not so high comparing with the other two
topologies.
5.1.8 RMS model for soft-starter
In steady-state analysis of the soft-starter-fed induction machine or when reduced order

models of the machine are used, an RMS model for the AC controller should be used. The
expression of the soft-starter output voltage is a function of firing angle, phase angle and so-
called extinction angle.
Since, a SCR will not permit the flow of reverse current, the conduction of it will ceases at a
point ξ, which is called extinction angle or cut-off angle. The extinction angle can be
obtained solving the following transcendental equation [2]:
sin ( ξ − ϕ ) − sin ( α − ϕ ) e − cot ϕ( ξ−α ) = 0 (5.1.5)
where: α is the firing angle and ϕ is the phase angle (power factor)
Only an iterative solution of (5.1.5) is possible. This yields the set of characteristics shown in
Fig. 5.10.
Fig. 5.10. Extinction angle versus firing angle for a series R-L load with different power factors..
When α = ϕ , which represents sinusoidal operation, (5.1.5) reduces to sin ( ξ − ϕ ) = 0 , which

gives ξ = π + ϕ . The values ξ for sinusoidal operation are seen to lie on the dashed linear
characteristic of Fig. 5.10. For a purely inductive load, therefore, the variation of ξ with α is
linear, as shown in the ϕ = 90o characteristic from Fig. 5.10.
Considering a single R-L load as shown in Fig. 5.11
Fig. 5.11. Single-phase controller with a resistive inductive load.
87
the applied voltage at the load terminals is given by [2]:
1
VUX = VAB
2π
( ( ξ − α ) + 0.5sin 2α − 0.5sin 2ξ ) (5.1.6)
Based on (5.1.6) the variation of the fundamental voltage with firing angle is shown in Fig.
5.12 for several fixed values of phase-angle.
Fig. 5.12. Fundamental voltage versus firing angle for single-phase controller with a series R-L load.
Only for a delta-branch connection (5.1.6) can be used. Based on the input voltage and the
phase current in complex form and a look-up table for the extinction angle the output voltage
per-phase can be obtained.
Unfortunately, for the star- or the delta connection due to the complexity of the conducting
mechanism is very difficult to establish an analytical expression for the RMS AC-controller
output voltage.
5.2 Voltage Source Converters
5.2.1 Introduction
Currently, in wind turbine applications mainly the back-to-back voltage source converter
VSC is used as shown in Fig. 5.13.
Fig. 5.13. Structure of the back-to-back voltage source converter.
This topology comprises a double conversion from AC to DC and then from DC to AC. Both
converters can operate in rectifier or inverter mode and therefore a bi-directional power flow
can be achieved.
88
A voltage source converter can be implemented in several ways: six-step, pulse amplitude
modulated (PAM) or pulse width modulated (PWM). Moreover, the implementation of a
PWM VSC may be realized by three methods: harmonic elimination, “sinusoidal” PWM or
space vector strategy (SV-PWM).
Independent on the implementation method the converter can be seen as a black box with
some input-output characteristics as a function of the control strategy. Again, the switching
function concept can be used in modelling of these topologies.
So, in the following paragraphs the input-output characteristics as a function of the switching
function for VSCs, both inverter and rectifier mode of operation will be presented.
The influence of the windings connection for generator will be also taken into account.
5.2.2 Star connected generator
When a VSC is used in conjunction with a star connected machine, (squirrel-cage or wound
rotor) the equivalent diagram of the circuit is like in Fig. 5.14.
Fig. 5.14. A star-connected induction machine with a VSC.
The applied voltages at the machine terminals as a function of the DC-link voltage and the
switching functions are [10]:
 v UX   2 −1 −1 SwA 
 v  = 1 V  −1 2 −1 S  (5.2.1)
 VY  3 DC    wB 
 v WZ   −1 −1 2  SwC 
The DC-link current can be expressed as a function of the input currents and the switching
functions by:
i A 
i DC = [SwA SwB SwC ]  i B  (5.2.2)
 i C 
The voltages and currents waveforms in this case are shown in Fig. 5.15.
89
Fig. 5.15. Typical waveforms for voltages and currents for a Voltage Source Converter with
a star connected generator.
5.2.3 Delta connected generator

Fig. 5.16 shows a VSC connected to an induction generator with delta-connected windings.
Fig. 5.16. A delta-connected induction machine with a VSC.
The applied voltages at the machine terminals as a function of the DC-link voltage and the
switching functions are:
 v UX   1 −1 0  SwA 
 v  = V  0 1 −1 S  (5.2.3)
 VY  DC    wB 
 v WZ   −1 0 1  SwC 
The DC-link current can be expressed as a function of the input currents and the switching
functions by:
90
i A   1 0 −1  i UX 
i DC = [SwA SwB SwC ]  i B  = [SwA SwB SwC ]  −1 1 0   i VY 
  (5.2.4)
 i C   0 −1 1  i WZ 
The voltages and currents waveforms in the case of a VSC with a delta-connected generator
are shown in Fig. 5.17.
Fig. 5.17. Typical waveforms for voltages and currents for a Voltage Source Converter with
a delta connected generator.
5.2.4 DC-link Circuit
Typically, the DC-link circuit is represented as an ideal capacitor as shown in Fig. 5.18.
Fig. 5.18. Equivalent diagram for DC-link circuit.
The DC-link voltage, which is the voltage at the capacitor terminals, is given by:
1 1
v DC = ∫ i C dt = ∫ ( i R − i I ) dt (5.2.5)
C C
where iR is the current of the generator side converter and iI is the current of the grid side
converter.
91
In s-plane modelling (using transfer functions) (5.2.5) cannot be used due to numerical
problems. Hence, the capacitor bank should be modelled as a series R-C circuit, which is
similar with a real component. A capacitor always will exhibit an internal resistance.
So, (5.2.5) can be written in s-plane as:
 1  sRC + 1
VDC =  R +  ( I R − I I ) = ( IR − II ) (5.2.6)
 sC  sC
In this way the connection between the two considered VSCs could be achieved through the
DC-link circuit as shown in Fig. 5.19.
Fig. 5.19. Simulink model of a back-to back VSC.
5.2.5 Grid-side filter
In order to limit the harmonic stress on the grid caused by power electronic systems an R-L
or an LCL filter can be used at the output of the grid-side converter. The LCL filter is smaller
and less expensive than the R-L filter and has the same attenuation level of current ripple
[11].
In a symmetrical three-phase system the equivalent diagram per phase of the LCL filter can
be represented as shown in Fig. 5.20.
Fig. 5.20. Equivalent diagram per-phase for the LCL filter.
Parasitic serial resistances of the inductances are ignored, instead that a resistance R is added,
which is used for appropriate damping of the filter at its resonance frequency [11],[12].
Based on Fig. 5.20 the following voltage equations can be written in s-plane:
Vin = Z11Iin + Z12 Iout
(5.2.7)
Vout = Z21Iin + Z22 Iout
where:
92
1  1 
Z11 = R + sL1 + Z12 = −  R + 
sC  sC 
(5.2.8)
1  1 
Z12 = R + Z11 = −  R + sL 2 + 
sC  sC 
Since the grid voltage is assumed as an ideal voltage source and represents a short-circuit for
harmonic frequencies, it can be settled to zero:
v out = 0 (5.2.9)
Using (5.2.7) the desired transfer functions in s-plane can be obtained [12]:
Iout Z21
H1 (s) = =
Vin Z12 Z21 − Z11Z22
Iin Z22
H 2 (s) = = (5.2.10)
Vin Z11Z22 − Z12 Z21
Iout Z21 ( Z11Z22 − Z12 Z21 )
H 3 (s) = =
Iin Z22 ( Z12 Z21 − Z11Z22 )
Based on (5.2.10) a Simulink model for this LCL filter can be built.
5.3 Summary
Based on the switching function concept the mathematical models for the most used power
converters in wind turbine applications are presented. The AC-controllers are modelled
taking into account the winding connection type for the generator. Some typical waveforms
for each topology are shown.
5.4 References
[1]. A.M. Trzynadlowski – Introduction to Modern Power Electronics, J. Wiley & Sons, 1998, ISBN 0-471-
15303-6;
[2]. W. Sheperd, L.N. Huley, D.T.W. Liang – Power Electronics and Motor Control, Cambridge University
Press 1995, ISBN 0-521-47241-05;
[3]. S.B. Dewan, G.R. Slemon, A. Straughen – Power semiconductor devices, J. Wiley & Sons, 1984, ISBN-0-
471-89831-7;
[4]. M.H. Rashid – Power electronics, circuits, devices and applications, Prentice Hall, 1988, ISBN 0-13-
686619-0;
[5]. V. Subrahmanyam – Electric drives: concepts and applications, McGraw Hill, 1994, ISBN 0-07-460370-1;
[6]. M.S. Smiai, S. Lesan, W. Shepherd – Analysis and performance of hybrid thyristor controlled three-phase
resistive inductive loads,
[7]. S.A. Hamed, B.J. Chalmers – Analysis of variable-voltage thyristor controlled induction motors, IEE
Proceed., Vol. 137, No. 3, May 1990, pp.184-193;
[8]. G. Zenginobuz, I. Cadirci, M Ermis, C. Barlak – Soft starting of large induction motors at constant
current with minimized starting torque pulsations, IEEE Trans. on Ind. App., Vol. 37, No. 5, Sept/Oct
2001, pp. 1334-1347;
[9]. S.N. Ghani – Digital computer simulation of three-phase induction machine dynamics – a generalized
approach, IEEE Trans. on Ind. App., Vol. 24, No. 1, Jan/Feb 1988, pp. 106-114;
93
[10]. N. Mohan, T.M Undeland, W.P. Robbins – Power Electronics: Converters, Applications and Design, J.
Wiley & Sons, 1995, ISBN 0-471-58408-8;
[11]. M. Liserre, A Dell’Aquila, F. Blaabjerg – Design and control of a three-phase active rectifier under non-
ideal operating conditions, Industry Applications Conference, 2002, Vol. 2, 2002
pp.: 1181 –1188;
[12]. E. Urlep – Control of three-phase active rectifier for wind turbine applications, Aalborg University, 2002;
94
Chapter 6
Control of variable-speed wind energy

conversion systems
connected and stand-alone mode are shown. A controllers design method based on the
presented. The parameters for the controllers will be determined in two cases: a fast control,
which is necessary during the start-up sequence of the doubly-fed induction generator and a
slow control for a normal operation of the wind turbine. The slow rotor control can be used
directly in the active and reactive power control of the wind turbine. The influence of
disturbance in the stator voltage in the rotor current control loops is shown for both
considered cases. Finally, in order to analyse the response of the rotor current control loops in
normal operation of a wind turbine, some simulation results for a 2 MW doubly-fed induction
generator are shown.
6.1 Introduction
High-performance variable-speed drives systems using induction machines have gain more
and more popularity during the last decade due to the raid development of power electronics
and microprocessors. The complex control problems, which arise in these variable-speed
drives, are solved using fast microprocessors and DSPs and new control techniques are added
each year.
Currently, there are two main control technique tendencies: namely the vector-control and the
direct torque control used for high performance applications based on the induction
machines. Derived from the early works of Blashcke and Hasse, and largely due to the
pioneering work of Leonhard, the vector control of AC machines has become the most
adopted control technique worldwide [2]. German and Japanese researchers have first
introduced the direct-torque-controlled (DTC) drives, and currently there are DTC variable-
speed drives on the market [2].
An overview of these control methods for induction machine is presented in Fig. 6.1 (based
on [1]-[4]).
The vector-control can be applied to voltage-, current source converters or cycloconverters,
while the direct-torque control can be used only for the first two power converters topologies.
In the case of an induction generator these converters must supply reactive power for the
machine magnetization.
The torque and flux decoupled control (or the active and reactive power control) is achieved
based on a special reference frame fixed to the stator-, rotor- or magnetizing flux in the case
of the vector control technique or in a reference frame fixed to the stator flux for a direct-
torque controlled drive system.
95
It should be noted that the vector control-techniques could have two implementation methods
namely direct and indirect. The direct method implies measurement of some variables
(voltages-, current- and fluxes sensors) or they are calculated from the machine flux model,
while the indirect method uses the monitored voltages and currents and/or rotor speed.
Both vector- (VC) and direct-torque control (DTC) requires the estimation of the magnitude
and space angle of the flux-based control [2] and several estimation techniques can be used
as shown in Fig. 6.1.
Fig. 6.1. Basic control methods for induction machine.
Since the machine parameters can change, especially the rotor ones, as a result of saturation,
variation of the temperature, and effects of current displacement (deep-bar effect) and hence
the control is sensitive to these parameters, it is preferable to implement control schemes
which incorporate some form of on-line parameter adaptation or insensitive in respect with
this changes.
However, in the future, artificial-intelligence-based control techniques (fuzzy, neural, neural-
fuzzy, etc.) will have a more dominant role in high-performance drives [2].
6.2 Control for WECS

Based on the “basic” control techniques presented in Fig. 6.1, several control techniques for
induction generators used in wind energy conversion systems (WECS) have been developed
in order to achieve a fully decoupled active and reactive power control. There are currently
two main tendencies in the control of the induction generators. One is focused on the control
of stand-alone systems, using especially VSC-fed squirrel-cage induction machine [5]-[22]
and the other one deals with the control of doubly-fed induction generators [23]-[32], with
96
both uni- and bi-directional power flow on the rotor side. Regarding the control of a grid-
connected variable-speed WECS using squirrel-cage machine, only few references can be
found in the literature [18]-[22].
All these control schemes have basically the same layout. The general structure of the control
for a variable speed WECS using a squirrel-cage induction generator both stand-alone or grid
connected is presented in Fig. 6.2
Fig. 6.2. General structure of the control for a variable-speed WECS using VSC-fed SCIG.
The supervisory system for a WECS using a squirrel-cage induction generator should be able
to assures a proper operation for all possible operation mode of the wind turbine:
• Stand by, when the turbine is available to run if external conditions permit;
• Limit the inrush currents and so the electromagnetic torque during start-up sequence;
• Grid synchronization;
• Optimal speed control to obtain optimal efficiency of the wind turbine;
• Control the electrical power;
• As far as possible continue normal operation during transient disturbances in the grid
voltage
• Switch between grid or stand alone operation;
• Shutdown
• Stopped with fault.
All these functions are realized using measured quantities form the system e.g. wind speed,
rotor speed, generator voltages and currents, etc. However these measured variables can be
different for a particular solution.
There are 3 main control modules in such a scheme: the control of the pitch blades, the
control of the generator-side converter and the control of the grid-side converter (active front-
end converter). The reference values for the pitch control as well as the generator control are
realized based on a maximum power point tracker (MPPT) algorithm for a given operating
97
point. Usually, the wind turbine manufacturers implements some look-up tables with the
optimal output power and so the optimal torque. In [16]-[18] and [20] fuzzy-controllers are
used in order to extract the maximum available power from the wind as well as to control the
output voltage of the grid side converter. The generator can be either current- (torque-) or
speed controlled [28].
The general structure of the control for a variable speed WECS using a doubly-fed induction
generator both stand-alone or grid connected is presented in Fig. 6.3.
Fig. 6.3. General structure of the control for a variable-speed WECS using VSC-fed DFIG on the rotor circuit.
It can be observed that the general layout of the control scheme is basically as for a squirrel-
cage machine. The only difference consists in the requirements for the supervisory system,
which should have some extra features as:
• Synchronize to grid by controlling stator voltage phase, frequency and magnitude
through applying a proper rotor voltage;
• Fast re-synchronization after disconnection;
• Protect the converter at line short-circuit close to the stator windings.
6.2.1 Control of squirrel-cage induction generator
Usually, an indirect rotor-flux-oriented control is used for the squirrel-cage induction

generator, due to its simplicity. Using the decoupled control of the flux- and torque-
producing components of the stator current the torque control can be performed as shown in
Fig. 6.4 [1]-[3]. The torque reference is given by the MPPT while the reference for the rotor
magnetizing is a function of the rotor speed or can be kept constant for simplicity. The stator
currents can be independently controlled only if a decoupling circuit is used [2],[3] as shown
in Fig. 6.4.
98
Fig. 6.4. Schematic of the rotor-flux-oriented control of a VSC-fed induction generator.
Based on a rotor-flux model the control variables from the system are obtained. However,
since this rotor flux model is strongly dependent on the rotor parameters, an inaccurate value
for these parameters could deteriorate the dynamic performances of the drive. Using on-line
parameter adaption or estimators this problem can be avoided. A Model Reference Adaptive
Control (MRAC) for a WECS using SCIG is presented in [21].
6.2.2 Control of doubly-fed induction generator
The control for a variable speed doubly-fed induction generator can be realized both with
stator-flux-oriented [23]-[31] or rotor-flux-oriented control [32]. However, since a doubly-
fed machine has the stator windings directly connected to the grid the stator voltage equations
will result simpler as well as the stator-flux-oriented control scheme than the rotor-flux
oriented one as shown in Fig. 6.5. Moreover, the stator active- and reactive power are
controlled directly.
Fig. 6.5. Schematic of the stator-flux-oriented control of a VSC-fed induction generator.
99
A method for the controller design and analysis in this case will be presented in detail in
paragraph 6.3.
6.2.3 Control of grid-side (active front-end) converter
When the wind turbine is grid-connected the active front-end converter should keep the DC-
link voltage constant regardless of the magnitude and direction of the rotor power and to
assure a quality of the output voltage and current according with the standards, which are
imposed for connecting wind turbines and wind farms to the transmission network [36].
Usually, this converter operates at unity power factor. Two control methods should be
implemented in order to achieve grid connected operation mode or stand alone mode.
6.2.3.1 Grid-connected operation
The basic control structure for a grid-connected operation is presented in Fig. 6.6.
Fig. 6.6. Schematic of the control structure for grid-connected operation.
The controller is designed in xy-synchronous reference frame. So, grid voltage angle should
be detected accurately and a common practice is to use a Phase Locked Loop (PLL). The
DC-link voltage is controlled in the outer loop of the x-axis while the current in the inner
loop. In order to compensate the cross coupling of the output filter a decoupling circuit is
used. Since the converter operates at unity power factor, the y-axis current is set to zero. An
R-L or an LCL structure can be used for the output filter.
The grid-connected operation can be applied for WECS using both doubly-fed induction
generator and squirrel-cage machine.
6.2.3.2 Stand alone operation
In this case the wind turbine generator is the only voltage source and therefore the control
strategy for the grid-connected operation cannot be used. A schematic structure of the control
is shown in Fig. 6.7.
100
Fig. 6.7. Schematic of the control structure for stand-alone operation.
The converter should provide a three-phase voltage system with a constant magnitude and
frequency and with a current limitation and DC-link voltage control. Both current limitation
and DC-link voltage control strategy are based on the lowering of the output voltage [12].
When the load demand is lower than the wind turbine production, a dump resistor RD is
activated in order to dissipate extra power by using a DC-link chopper control. However,
when there is not enough produced power for consumers, the DC-link voltage will drop to the
minimum and lead the system to a stop functioning.
6.2.4 Phase-Locked Loop
Since, the PLL is a central component in a control structure for the grid-side converter, it will
be presented in the following. A generalized PLL structure is shown in Fig. 6.8.
Fig. 6.8. General structure of a Phase Locked Loop.
The three-phase voltage inputs are transformed into αβ components referred to the grid
voltage phasor. Using this components sufficient information about the phase angle is
 π π
obtained. A phase detector range of  − ;  is obtained using a simple inverse tangent
 2 2
function, while using a four quadrant inverse tangent function this range can be expanded to
[ −π; π] . The loop filter, contains a low pass filter to suppress noise and high frequency terms
in the signal from the phase detector, e.g. due to asymmetry in the three-phase voltages. To
avoid stationary error in phase after a step in input signal frequency, a PI-controller is added.
The performance of the PLL depends on the phased detector and the selected bandwidth of
the loop filter [12]. With slow dynamic of the PLL loop, higher rejection of the disturbances
can be achieved, although a poor tracking of the grid angle.
101
In the power system applications, the voltage-controlled oscillator usually is implemented as
an integrator. The loop is close by feeding the estimate angle into the αβ-transformation.
As long as the phase angle is correct, the output of the phase detector is zero, thus the
frequency input to the integrator is constant – the PLL is in lock. If the angle estimate is not
correct, the frequency is adjusted. And the grid voltage phasor referred phase angle changed.
6.3 Design case study: Doubly-fed induction generator

controllers
In this paragraph the controller structure for the rotor current loops and the active and
reactive power control in the case of a doubly-fed induction generator is treated in detail.
Based on the linearized equations of the induction machine presented in Chapter 4 the linear
design and analysis techniques are used to investigate the stability and the response of the
system using Control System Toolbox from Matlab.
6.3.1 Stator-flux oriented control
In this paragraph the stator flux oriented control will be briefly presented in order to
understand the basic principle of this control.
Starting from the voltage equations of the induction machine given by (3.2.9) written in
synchronous reference frame the rotor voltage equations will be written as:
di rx di
v rx = R r i rx + L rr + L m sx − ( ωx − ωr ) ( L rr i ry + L m isy )
dt dt
(6.3.1)
di ry disy
v ry = R r i ry + L rr + Lm + ( ωx − ωr )( L rr i rx + L m isx )
dt dt
where: ωx is the angular speed of the synchronous reference frame.
The relation between fluxes and currents are given by:
 ψ sx   Lss 0 L m 0  isx 
    
 ψ sy  =  0 Lss 0 L m  isy 
⋅ (6.3.2)
 ψ rx   L m 0 L rr 0  i rx 
     
 ψ ry   0 L m 0 L rr  i ry 
Fig. 6.9. Explicative to stator-flux oriented control.
Attaching the reference frame to the stator flux as shown in Fig. 6.9, the following
simplifications can be made:
102
ψ sx = L m i ms = Lssisx + L m i rx
(6.3.3)
0 = Lssisy + L m i ry
where ims is the stator magnetizing current.

Based on (6.3.3) the components of the stator current can be expressed as:
Lm
isx = ( i ms − i rx )
Lss
(6.3.4)
L
isy = − m i ry
Lss
Thus, the governing equations for the rotor currents are:
di rx v di
Tr + i rx = rx + ( ωx − ωr ) Tr i ry − ( σ r − Tr ) ms
dt Rr dt
(6.3.5)
di ry v ry
Tr + i ry = − ( ωx − ωr ) ( Tr i rx + ( σ r − Tr ) i ms )
dt Rr
D L
where: Tr = is the rotor time constant, and σ r = rr .
R r Lss Rr
By inspecting (6.3.5) it can be observed some cross coupling between the x- and y-axis due
to the rotational emf terms. The current loop dynamic along the two axes can be made
independent of each other by compensating for these cross coupling terms. However, since
the stator windings are directly connected to the grid the stator flux is almost constant in
normal operation and the contribution on its derivative can be ignored. Moreover, as the slip
range is limited the contribution of these coupling terms is rather weak.
In order to complete the description of the machine behavior the equation of the active and
reactive power at the stator terminals should be added.
Considering that the x-axis component of the stator voltage is zero the general expression of
the active and reactive power in synchronous reference frame is:
3
vsy isy Ps =
2 (6.3.6)
3
Qs = vsy isx
2
Taking into account (6.3.4), (6.3.6) can be written as:
3 Lm
Ps = vsy i ry
2 Lss
(6.3.7)
3 L 
Qs = vsy  m i ms − i rx 
2  Lss 
103
Usually, the rotor controller design is based on (6.3.5). In the following a new design method
for the rotor current controller as well as for the active and reactive power controllers, which
use the complete model of the machine, will be presented.
6.3.2 The rotor current controllers
Considering the control structure presented in Fig. 6.5 and neglecting the cross coupling
circuit the diagram of the rotor current control loops is shown in Fig. 6.10.
Fig. 6.10. Schematic of the rotor current control loops using transfer functions.
It will be shown later that the decoupling circuit for the rotor voltage can be neglected using
this new approach in the controllers design. The influence of the deep-bar effect for the
induction generator will be taken into account in design and analysis of the controllers.
The rotor currents control depends on the operating mode of the wind turbine:
• Start-up and synchronization to grid. The induction generator is magnetized through
the rotor circuit for a given rotor speed – 0.5 value of slip in this case – with the stator
windings open. The rotor currents in this case equal the machine magnetizing current.
Then, by controlling magnitude, frequency and phase of the rotor currents, the stator
voltage is controlled so that its magnitude and phase equal the grid voltage.
• Variable speed operation. The generator is controlled so that it will extract the
maximum available power from the wind for a given rotor speed at the desired power
factor.
Thus, these controllers should operate in any operation mode of the wind turbine. The
controller design is performed for the start-up, when the state variables of the considered
system are zero. Later in Chapter 7, the parameters will be verified for a normal operation of
the entire system.
6.3.2.1 Transfer functions of the induction machine
Starting from the complete linearized 8th order system given by (4.3.5) a multiple input /
multiple output (MIMO) system, which describe the induction machine and the drive-train
model can be build using the specific functions from Control System Toolbox [35]. Then, the
104
desired transfer function between each input/output pair can be extracted as shown in Fig.
6.11.
Fig. 6.11. MIMO system of the induction machine with the complete model of the drive train.
It has been found that the transfer function for the rotor y-axis in the case of a 2 MW DFIG
with the parameters given in Table 4.1 and Table 4.2 at 0.5 slip is:
i ry 2.695s3 + 6.542e 4s 2 + 2.66e8s + 4.666e9
H vry→iry ( s ) = = (6.3.8)
v ry s 4 + 47.84s3 + 1.24e5s 2 + 2.888e6s + 2.466e9
Similar transfer functions can be written for other input/output pairs.
Since the MIMO system is based on the machine equations written in synchronous reference
frame the x-axis component of the stator voltage is zero vsd = 0 and the transfer functions
corresponding with this input will not be considered in the analysis.
In the controllers design will be considered the following transfer functions:
• Transfer functions between the stator voltage and the rotor currents:
H vsy →irx,y (s) (6.3.9)
• Transfer functions between the rotor voltages and the rotor currents.
H vrx,y→irx,y (s) (6.3.10)
First for the rotor current controller design, the influence of the stator voltage will be
neglected. Then, the controller parameters will be modified in order to take into account the
effect of the disturbance from the stator voltage.
6.3.2.2 Transfer function of the converter
The converter can be modelled in synchronous reference frame in terms of its fundamental
components with harmonics treated as a separate issue [3]. So, the transfer function
associated with the fundamental component for a voltage source converter is:
1
H C x,y ( s ) = (6.3.11)
1 + sTC delay
Since this time delay is very small the converter transfer function will be neglected in
analysis.
105
6.3.2.3 Transfer function of the filter
Usually a first-order filter is used for the feedback signals. The transfer function for this filter
is given by:
1
H F xy ( s ) = (6.3.12)
1 + sTF
Depending on the modulation strategy this time constant can have different values in order to
eliminate the ripple in the measured currents. However, for a SV-PWM modulator with a
switching frequency around 4 kHz the current ripple is very small, so the time constant of
this filter will result small. In the design of the rotor current controllers a value of 0.5 msec
has been considered.
6.3.2.4 Open loop analysis
The design starts with the state-space model of the induction generator including the
complete model of the drive train given by (4.3.5). The open-loop eigenvalues of the
electrical system (stator and rotor voltage equations) for a 0.5 slip and considering the initial
values of the fluxes zero are presented in Table 6.1.
Table 6.1. Open-loop eigenvalues at 0.5 slip considering zero initial conditions.
Frequency
Eigenvalue Damping
[rad/sec]
Rotor eigenvalue -11 ± i 158 0.0728 158
Stator eigenvalue -12.4 ± i 313 0.0395 314
First it will be applied a impulse input for the transfer functions between the rotor voltages
and the rotor currents, H vrx,y→irx,y (s) . The response of the rotor system in this case is shown in
Fig. 6.12. It can be observed that the rotor system is lightly damped.
Fig. 6.12. Impulse response of the transfer functions between the rotor voltages and the rotor currents.
106
The rotor current controller design will be performed only for the y-axis, however the
determined parameters are identical for the x-axis controller.
Selecting only the transfer function between the rotor voltage and the rotor current in the y-
axis H vry −>iry (s) , the Bode diagram of it is presented in Fig. 6.13. The other transfer function
between the rotor voltage and the rotor current in the x-axis exhibit the same frequency
response.
Fig. 6.13. Bode diagram for the transfer function voltage and current in y-axis of the rotor.
From this Bode diagram it can be observed that the y-component of the rotor voltage has
significant effect around the rotor eigenvalue frequency, near 158 rad/sec.
6.3.2.5 Root Locus Design
The appropriate gain for the controller can be determined using the root locus technique as
shown in Fig. 6.14. A value of 0.0278 will assure the desired damping of 0.707.
Fig. 6.14. Root locus for the transfer function voltage and current in y-axis.
107
The transfer function of the PI controller can be arranged as [34]:
sTi + 1
H PI (s) = Kα (6.3.13)
sαTi + 1
The integration time constant of the PI controller is selected at 0.005 sec in order to obtain a
bandwidth of 200Hz.
Selecting for α a value of 10, the rotor current controller parameters are:
K p = 0.28
(6.3.14)
Ti = 0.005 sec
The controller parameters can be verified plotting the impulse response for the closed loop as
well as the step response as shown in Fig. 6.15 and Fig. 6.16.
Fig. 6.15. Impulse response for the closed loop system.
Fig. 6.16. Step response for the closed loop system.
108
The impulse response of the closed-loop system settles quickly and the oscillations of the
current are completely eliminated. In Fig. 6.16 can be observed the effect of the cross
coupling between x and y-axis in steady state. This influence can be removed if the transfer
function of the cross coupling is added in the system.
6.3.3 Direct active and reactive power controllers
In paragraph 6.3.1 has been found a direct relation between the active and reactive power of
the stator and the rotor currents. Since the machine is directly connected to the grid, based on
(6.3.7) a very simple control of the active and reactive power can be done. This control is
useful for a reduced model of the machine. The basic control structure is shown in Fig. 6.17.
Fig. 6.17. Control structure for active and reactive power.
By controlling directly the rotor currents the stator active and reactive are controlled. The
MPPT tracker gives both references for the control scheme. The stator magnetizing current
ims is calculated into the flux model of the machine.
6.3.3.1 Open loop analysis

The open-loop eigenvalues of the induction generator including the complete model of the
drive train in steady state for a 0.55 value of slip are presented in Table 6.2.
Table 6.2. Open-loop eigenvalues in steady-state at 0.55 slip.
Eigenvalue Damping Frequency [rad/sec]

Torque eigenvalue -0.325 1 0.325
Twist eigenvalue -0.639 ± i 16.2 0.0394 16.2
Rotor eigenvalue -28.9 ± i 179 0.159 181
Stator eigenvalue -8.25 ± i 313 0.0264 313
It can be observed that both stator and twist eigenvalues are lightly damped. The influence of
these eigenvalue frequencies are highlighted by the Bode plots as shown in Fig. 6.18.
109
Fig. 6.18. Bode diagram for the transfer function voltage and current in the rotor.
However the interest in the controller design in this case is to compensate the eigenvalue
frequency of the rotor. Thus a relatively fast control of the stator active and reactive power
can be obtained.
Applying an impulse input to the rotor circuit the response of the system is shown in Fig.
6.19.
Fig. 6.19. Impulse response of the rotor circuit.
6.3.3.2 Root locus design
The appropriate gain for the controller can be determined using the root locus plot for the
open loop transfer function between rotor voltage and the rotor current in y-axis as well as for
the x-axis as shown in Fig. 6.20 and Fig. 6.21.
110
Fig. 6.20. Root locus for the transfer function between the rotor voltage and the current in y-axis.
Fig. 6.21. Root locus for the transfer function between the rotor voltage and the current in x-axis.
A gain value around 0.03 will assure a damping greater than 0.707 for both controllers.
The transfer function of the PI controller can be arranged as:
s + ωz  1 
H PI (s) = K p = K p 1 +  (6.3.15)
s  sTi 
Since, the zero of the PI controller should compensate the desired rotor eigenvalue frequency,
which is 181 rad/sec in this case a value around 40 msec has been found.
In order to verify the parameters of the controllers the response of the closed loop system,
both for an impulse and for a step input are shown in Fig. 6.22 and Fig. 6.23.
111
Fig. 6.22. Impulse response for the closed loop system.
Fig. 6.23. Step response for the closed loop system.
Since the cross coupling between axis has not been taking into account an error in steady
state can be observed.
6.3.4 Influence of a stator voltage disturbance in the rotor current

control loops
During the controller design the stator voltage has been considered zero in case of the start-
up or equal with the steady state value in the case of the active and reactive power
controllers.
It is of great interest to analyse the response of the rotor current control loops when a
disturbance in the stator voltage occurs, e.g. a three-phase fault on the stator windings.
112
First, it has been considered the impulse response of the transfer functions between the stator
voltage and the rotor currents of the machine for a fast rotor current loop ( Ti = 5 ms ec ) as
shown in Fig. 6.24.
Fig. 6.24. Response of the system for a disturbance in the stator voltage for a fast rotor current control loop.
Fig. 6.25. Response of the system for a disturbance in the stator voltage for a slow rotor current control loop.
Then the impulse response of the transfer functions between the stator voltage and the rotor
currents of the machine for a slow rotor current loop ( Ti = 40 ms ec ) has been considered as
shown in Fig. 6.25.
With a very fast rotor current loop the system is unstable in respect with a stator voltage
disturbance, while using a relative slow rotor current loop the rotor current oscillations are
damped quickly. So, it is much more convenient to use a slow control for the rotor currents.
113
6.3.5 Simulation results
Using the new Matlab/Simulink “Wind Turbine Blockset”, which it will be presented in
detail in Chapter 7, the parameters of the rotor controllers determined in §6.3.3 have been
tested.
The Simulink model for the active and reactive power control of a 2 MW doubly-fed
induction generator is presented in Fig. 6.26.
Fig. 6.26. Simulink model for validation of the rotor current controller parameters in case of
a 2 MW doubly-fed induction generator.
The model comprises a model of the doubly-fed induction generator in synchronous

reference frame, the complete model of the drive train including torsional torques and the
control for the stator active and reactive power. The deep-bar effect is taken into account both
for the machine and in the control block. The cross coupling between the axis is taken also
into account.
The system has been considered in steady state at 0.5 slip and a step of 0.07 pu in the
reference of the active power has been applied, while the reference of the stator reactive
power is set to zero. This step in the reference of the active power corresponds to a sudden
change in the generator speed from 740 rpm to 1040 rpm or in terms of the average wind
speed from 3.6 to 5 m/sec. Since, the variations in the reference power are usually slow in a
normal operation of the wind turbine, this can be a severe test of the rotor controllers.
The references for rotor currents as well as the measured ones are shown in Fig. 6.27.
114
Fig. 6.27. Response of the rotor currents control loops for a sudden change
in the reference of the stator active power.
It can be observed that the response of the y axis controller follows the reference while the x-
axis current is set to zero.
The reference signals for the stator active and reactive power and the measured ones as well
as the rotor active and reactive power are shown in Fig. 6.28 and Fig. 6.29.
Fig. 6.28. Reference and measured signals for the stator active and reactive power for a sudden change
in the active power reference.
115
Fig. 6.29. Measured active and reactive power in the rotor side of the machine.
The rotor current loop responds as expected. Since the machine is supplied with the reactive
power from the rotor side, the reactive power on the stator side is set to zero, while the active
power follows the reference signal.
Thus, a relative slow control of the rotor currents can provide an effective control of the
active and reactive power on the stator side of the machine.
6.4 Summary
connected and stand-alone mode are shown. A controllers design method based on the
presented. The parameters for the controllers have been determined in two cases: a fast
control, which is necessary during the start-up sequence of the doubly-fed induction
generator and a slow control for a normal operation of the wind turbine. The slow rotor
control can be used directly in the active and reactive power control of the wind turbine. The
influence of disturbance in the stator voltage in the rotor current control loops is shown for
both considered cases. Finally, in order to analyse the response of the rotor current control
loops in normal operation of a wind turbine, some simulation results for a 2 MW doubly-fed
induction generator are shown.
6.5 References
[1]. W. Leonhard – Control of electric drives, Springer Verlag, 2001, ISBN 3540418202;
[2]. P. Vas – Sensorless vector and direct torque control, Oxford University Press, 1998, ISBN 0-19-856465-
1;
[3]. D.W. Novotny, T.A. Lipo – Vector control and dynamics of AC drives, Clarendon Press, Oxford, 1996,
ISBN 0-19-856439-2;
[4]. P.C. Krause, O. Wasynczuk, S.D. Sudhoff – Analysis of electric machinery and drives systems, IEEE
Press and J. Wiley & Sons, 2nd edition, 2002, ISBN 0-471-14326-X;
116
[5]. E. G. Marra, J. A. Pomilio – Self-Excited Induction Generator Controlled by a VS-PWM Bi-directional
Converter for Rural Applications, Proceedings of APEC ’98, 15-19 Feb. 1998, Annaheim, California,
U.S.A;
[6]. Castelli Dezza, F., DiGerlando, A., Perini, R. - Modelling and Experimental Investigation of Energy
Conversion Systems employing Wind-Driven Induction Generators with Electronically Controlled Self-
excitation and Power Output. EPE ‘97, Trondheim, Norvegia, vol.2, pp.652-658 ;
[7]. Colliez, C., Tounzi, A., Piriou, F. - Vector Control of a Autonomous Induction Generator connected to a
PWM Rectifier. EPE ‘97, Trondheim, Norvegia, vol.2, pp.711-716;
[8]. Miranda, M.S., Lyra, R.O.C., Silva, S.R. -Wind Powered Pumping System using Induction Machine with
PWM Excitation. EPE ‘97, Trondheim, Norvegia, vol.2, pp.637-641;
[9]. Castelli Dezza, F., DiGerlando, A. Perini, R. - Performance Comparision among different Converters Fed
by Self-excited Wind Driven Induction Generators. Electrical Machines & Drives, 11-13 sept 1995,
Conference Publication No. 412, pp.438-443;
[10]. J. Sallan, E. Muljadi, M. Sanz – Control of Self-Excited Induction Generators Driven by Wind,
Proceedings of EPE 99, Laussane, Switzerland;
[11]. F. Caricchi, V. Callea, F. Crescimbini – Power Electronic Interface for Wind Energy Conversion Systems
Supplying Remote Installations in Antarctica, Proceedings of EPE 99, Laussane, Switzerland;
[12]. J. Sallan, E. Muljadi, M. Sanz – Control of Self-Excited Induction Generators Driven by Wind,
Proceedings of EPE 99, Laussane, Switzerland;
[13]. R. Leidhold, G. Garcia, M.I. Valla – Induction generator controller based on the instantaneous reactive
power theory, IEEE Tans. on Energy Conversion, Vol. 17, No.3, September 2002, pp. 368-373;
[14]. R. Leidhold, G. Garcia, M.I. Valla – Field oriented controlled induction generator with loss minimization,
Trans. on Industrial Electronics, Vol. 49, No. 1,February 2002, pp. 147-156;
[15]. LoBianco, G., Caricchi, F., Crescimbini, F., Solero, L. - Power Electronic Interface for Unconventional
Small Size Stand Alone Wind Energy Conversion Systems. EPE ‘97, Trondheim, Norvegia, vol.2, pp.659-
663;
[16]. Marcelo Godoy Simoes, Bose, B.K., Spigel, R.J. - Design and Performance Evaluation of a Fuzzy Logic
Based Variable Speed Wind Generation Systems. IEEE Trans. On Ind. App., vol.33, no.4, july/august
1997, pp.956-965;
[17]. Hilloowalla, R.M., Sharaf, A.M. - A Rule- Based Fuzzy Logic Controller for a PWM Inverter in a Stand
Alone Wind Energy Conversion Scheme. IEEE Trans. On Ind. App., vol.32, no.1, Jan/Feb 1996, pp.57-65;
[18]. B. Robyns, M. Nasser, F. Bertheau – Equivalent continuous dynamic model of a variable speed wind
generator – Proceed. of Electromotion 2001;
[19]. R. Pena, R. Cardenas et all – Vector controlled induction machines for stand alone wind energy
applications, Proceed. of IAS 2000, pp. 1405-1415;
[20]. R. Pena, R. Cardenas et all. – A cage induction generator using back-to back PWM converters for variable
speed grid connected wind energy systems, Proceed. Of IECON 2001, pp. 1375-1381;
[21]. R. Cardeans, R. Pena – Sensorless control of induction machines for wind energy applications, Proceed. of
PESC 2002, pp.265-270;
[22]. R. Teodorescu, F. Blaabjerg, F. Iov – Control strategy for stand-alone wind turbines, Proceed. of PCIM
2003 Power Quality, 20-22 May 2003, Nuremberg, Germany, pp. 201-206;
[23]. V. Akhamtov - Modelling of Variable-Speed Wind Turbines with Double-fed Induction Generators in
Short Term Stability Investigations, Proceed. of 3rd Int. Workshop on Transmission Networks for Offshore
Wind Farms, April 11-12, 2002, Stockholm, Sweden, pp. 1-23;
[25]. E-H. Kim, S-B. Oh, Y-H. Kim C-S. Kim – Power control of a doubly-fed induction machine without
rotational transducer pp.951-955;
[26]. M Yamamoto, O. Motoyoshi – Active and reactive power control for doubly-fed wound rotor induction
generator;
[27]. R. Data, V.T. Ranganathan – Decoupled control of active and reactive power for a grid connected doubly-
fed wound rotor induction machine without position sensor, pp. 2623-2630;
[28]. R. Pena, J.C. Clare, G.M. Asher – Doubly-fed induction generator using back-to back PWM converters
and its application to variable-speed wind-energy generation, IEE Proc.-Electr. Power Appl., Vol. 143,
No. 3, May 1996, pp. 231-241;
[29]. C.R. Kelber, W. Schumacher – Control of doubly-fed induction machines as an adjustable speed
motor/generator;
117
[30]. J.R. Saenz, A. Tapia, et. all – Reactive power regulation in wind farms: Control strategies, Proceed. Of
EPE 2001, Gratz;
[31]. H-J Conraths – Rotor-controlled generator system for wind energy applications, Proceed. Of EPE 2001,
Gratz;
[32]. U. Radel, D. Navarro, G. Berger, S. Berg – Sensorless field-oriented control of a slip ring induction
generator for a 2.5 MW wind power plant from Nordex Energy GmbH, Proceed. Of EPE 2001, Gratz;
[33]. E. Urlep – Control of three-phase active rectifier for wind turbine applications, Aalborg University, 2002;
[34]. G. F. Franklin, J.D. Powell, A. Emami-Naeini – Feedback Control of Dynamic Systems, 3rd Edition,
Adisson-Wesley Publishing Company, 1994, ISBN 0-201-53487-8;
[35]. *** - Using Control System Toolbox, Mathworks Inc, 2000;
[36]. *** - Specifications for connecting wind farms to the transmission network, 2nd Edition, ELTRA, April
2000;
118
Chapter 7
Wind Turbine Blockset

In this chapter the structure of a new Matlab/Simulink Toolbox for wind turbine applications
as a part of a new simulation platform dedicated to model, optimize and design wind turbines
is presented. The content of the main libraries from this toolbox is briefly presented. Then
some simulation demos are shown.
7.1 Simulation Platform Concept

Simulation is the most important technique used today for evaluation of engineering
solutions. In general different simulation tools exist to simulate power converters, generators,
power systems, mechanical parts and aero elastic behaviour of wind turbines. No simulation
tools can be stated to handle all issues in detail at the same time. Therefore when developing
improved electrical models, it is important that they can be used in the selected tools and the
user knows what are inside the models. The tools should be used according to the focus area
of the designer. Independent of tools, the structure of a complete simulation platform for
wind turbine should be as shown in Fig. 7.1.
Fig. 7.1. Structure of a complete simulation platform.
The main idea of this new simulation platform is to extend the ability of the existing wind
turbine design tools to simulate the dynamic behaviour of the wind turbines and the wind
turbine grid interaction. The model database must be able to support the analysis of the
interaction between the mechanical structure of the wind turbine and the electrical grid
during different operational modes.
The models are the core of a simulation platform including the parameters or alternatively
methods to determine the parameters based on measurements or data sheets. The new
119
simulation platform contains models for four different simulation tools: HAWC, DigSilent,
Saber and Matlab / Simulink. Fig. 7.2 illustrates the systems used in this new simulation
platform.
Fig. 7.2. Simulation tools used in the new simulation platform dedicated to
model, optimize and design wind turbines.
HAWC, an aeroelastic tool developed at RISØ National Laboratory, is used in the calculation
of the dynamic loads on the structure of the wind turbine. It currently has focus on the
aerodynamic and mechanical parts of the wind turbine, and it does not provide sufficient
modeling details on the electrical system for assessing highly controllable wind turbines.
DigSilent is a dedicated electrical power system simulation tool used for assessment of power
quality and analysis of the wind turbine interaction with the grid. It has focus on the power
system and not on the mechanical aspects of the wind turbine. Saber is a simulation tool used
in circuit and systems power design including electrical, thermal, magnetic and mechanical
components. However, this tool is currently not focused on wind turbine applications.
Matlab / Simulink is used as a general model developer tool and also for validation of the
models. The models are thus first checked in Matlab and then implemented in the other three
tools.
The overall demands to all the models should be as follows:
• models should be open;
• parameters easy to determine (physical based most preferable) e.g. datasheets;
• simulation speed important;
• models based on common literature. Next improvement may be added;
• user friendly: documentation, reliable.
• easy to extend with extra features.
The idea of this new simulation platform is thus, to provide an extensive model database to
perform simulations and performance studies of the wind turbines, both in continuous and
fault operation.
A long-term purpose is to use these models for a system optimization of the entire wind
turbine, namely in a simultaneous optimization of the aerodynamic, mechanical, electrical
and control systems over the whole range of wind speeds and grid characteristics.
120
7.2 Matlab/Simulink “Wind Turbine Blockset”
A new Matlab/Simulink Toolbox for wind turbine applications is developed. In order to
analyze the dynamic and/or steady-state behaviour of a wind turbine, the basic components of
a wind turbine have been modelled and structured in seven libraries as shown in Fig. 7.3.
Fig. 7.3. Structure of the new Matlab/Simulink Toolbox for

wind turbine simulations.
A short description of these libraries is presented in the following paragraphs.
7.2.1 Mechanical Components Library

The first library “Mechanical Components” contains: wind model, aerodynamic models of
the wind turbine rotor, different types of drive train models as shown in Fig. 7.4.
Fig. 7.4. Mechanical Components Library.
The wind model has been developed as RISØ National laboratory based on the Kaimal
spectra. The wind speed is calculated as an average value of the fixed-point wind speed over
the whole rotor, and it takes the tower shadow and the rotational turbulences into account. To
gain flexibility in modelling, the equation of motion is not included in the machine model.
121
Different types of drive train models are available e.g. one-mass model, two-mass model of
the drive train with torsional torques. Since the last model involves 4 differential equations an
alternative implementation using C S-Function is available.
7.2.2 Electrical Machinery Library
The “Electrical Machinery Library” contains dynamic and steady state models for electrical
machines: squirrel-cage induction machine, wound rotor induction machine, salient-poles
synchronous machine and permanent magnet synchronous machines. The models are written
in dq-reference frame as well as in the natural reference frame (ABC/abc models).
All the machine mathematical models are in the state space form with both fluxes and
currents as state variables. Since all these models involves a relatively big number of
derivates for the state variables two methods have been chosen in order to implement the
models in Simulink: using Simulink blocks and C S-Function. In Chapter 3 it is shown that
using a C S-Function instead of Simulink blocks at least a factor of two in the simulation
speed is gained.
However, the implementation method using Simulink blocks is much more open and the user
is able to “see” the model and to modify it. These Simulink blocks based models are shown
in Fig. 7.5
Fig. 7.5. Electric machinery Simulink blocks based models.
122
In the first stage the focus has been in modelling of the induction machine and some special
models including deep-bar effect, reduced order models (neglecting the stator transients) and
steady state models have been developed. Later on, the corresponding models for the
synchronous machine both field winding and permanent magnet in ABC/abc reference frame
will be added.
In order to gain simulation speed and to increase the numerical stability C S-Function based
models for the induction machine have been implemented as shown in Fig. 7.6.
Fig. 7.6. Electric machinery C S-Function based models.
7.2.3 Power Converters Library
The third library contains models for power converters based on switching functions. The
following models are available at this moment: 3-phase diode bridge rectifier, voltage source
converter, soft-starters and modulation strategies for power converters as shown in Fig. 7.7.
Fig. 7.7. Content of the Power Converters Library.
123
Using two VSC blocks connected via a DC-link circuit a back-to-back topology can be
obtained as shown in § 5.2.4.
The soft-starter type can be selected from the block interface, e.g. star connection, delta or
branch delta connection. Two types of modulation strategy for the VSC are available at this
moment, namely sinusoidal PWM and Space Vector PWM.
7.2.4 Common Bocks Library
This library contains models for grid, 3-phase distribution line, DC-link circuit, capacitor
bank, three-phase 2 winding transformer and circuit breaker as shown in Fig. 7.8.
Fig. 7.8. Content of the Common Blocks Library.
Since one of the key elements in the grid connection of the electrical system of a wind
turbine is the transformer, a special model, which take into account the core geometry as well
as the iron losses has been developed. The model is implemented both using Simulnik blocks
and C S-Function.
The distribution line model is based on the dynamic equations of the short line without
distortion (Heaviside condition), which is a relatively simple model with good results in the
case of a distribution line.
A switch model (on-off) of the circuit breaker, which take into account different opening
time moments for poles based on the zero crossing for the phase current has been developed.
7.2.5 Transformations Library

Since in a dynamic model of a wind turbine there are several transformations for the signals,
a special library, which contains the main transformations, has been implemented as shown
in Fig. 7.9. The basic transformations are from abc to dqo arbitrary reference and the
opposite one, as well as from a dq signal to the complex representation of it. In order to take
into account the connection type some transformations from a three-phase system in phase
quantities to a three-phase one but in line-to-line quantities have been implemented. In order
to find the magnitude and the phase for a three-phase system the corresponding block can be
used.
124
Fig. 7.9. Content of the Transformations Library.
7.2.6 Measurements Library
This library contains some special blocks as: calculation of the period for a sinusoidal
variable, calculation of the angle using a phase locked loop, different modes of active and
reactive power calculation.
Fig. 7.10. Measurements Library.
7.2.7 Control Library
Since the main component of a control scheme is the PI Controller, a model with an anti
wind-up structure has been implemented in Simulink. The maximum power point tracker
MPPT block is based on a look-up table obtained from the wind turbine characteristics.
The active and reactive power control for a doubly-fed induction generator is based on the
control structure presented in Chapter 6.
Fig. 7.11. Content of the Control Blocks Library.
125
In order to complete the control for a wind turbine other control blocks as the active stall
control, etc will be added later.
7.3 Wind Turbine Blockset Demos
7.3.1 Start-up sequence of a fixed-speed wind turbine
The start-up sequence of a soft-starter-fed squirrel-cage induction machine with a delta

connection for the stator windings, which is used in wind turbine applications, has been
simulated as shown in Fig. 7.12. The induction machine has 2 MW rated power, 690 V /
1700 rated phase-voltage and rated line current respectively (delta connection). The induction
machine is connected via soft-starter to the supply voltage below synchronous speed (1450
rpm). The starting firing angle for the soft-starter is 120o.
Fig. 7.12. Equivalent diagram of the system.
Since the electromagnetic torque during start-up sequence and the envelope of the line
current are very important in such a system, these variables are shown in Fig. 7.13.
Fig. 7.13. Electromagnetic torque and envelope of the line current during start-up sequence
for 2 MW soft-starter-fed induction machine in wind turbine applications.
126
It can be observed that using a soft-starter the inrush currents and therefore the high-starting
torque are limited. In order to highlight the different operation modes of the soft-starter
during the start-up sequence the phase voltage and the corresponding line current for
different firing angles are shown in Fig. 7.14.
a) b)
Fig. 7.14. Phase voltage and line current for different firing angles during the start-up sequence:
a) Mode 3 of operation and b) Mode 1 of operation.
It can be observed the two possible modes of operation of a soft-starter fed induction
machine during the start-up sequence, namely: Mode 3 of operation, which correspond to a
firing angle greater than the limit angle, and the Mode 1 of operation respectively.
7.3.2 Control of active and reactive power for a WECS using DFIG
Using the available blocks from the “Wind Turbine Blockset” the control of active and
reactive power for a 2 MW WECS using doubly-fed induction generator has been studied.
The Simulink diagram of the system is shown in Fig. 7.15.
Fig. 7.15. Simulink diagram of a 2 MW wind energy conversion system using DFIG.
127
The simulation structure comprises the wind model, drive train model, an DFIG model
written is synchronous reference frame, the control block for active and reactive power and
the optimal control of the entire system.
The optimal control blocks calculate de references for the pitch angle and the power
reference. The optimal characteristics of the control in terms of optimal pitch angle, optimal
tip speed ratio, optimal power coefficient and therefore optimal power reference are shown in
Fig. 7.16 and Fig. 7.17.
Fig. 7.16. Optimal pitch angle and optimal tip speed ratio versus average wind speed for
a 2 MW variable-speed WECS using DFIG.
Fig. 7.17. Optimal power coefficient and optimal power versus average wind speed for
a 2 MW variable-speed WECS using DFIG.
There are three zones in the optimal control for this system.
• Mode 1 – correspond for values of speed less than the rated speed, which is this case
1.7 rad/sec. In order to obtain maximum efficiency the tip speed ratio is kept constant
128
as well as the pitch angle (β=0). The power coefficient will have its maximum value
in this control mode; therefore the maximum available power will be extracted from
wind.
• Mode 2 – correspond for value of speed greater than the rated speed, however the
available power is less than the rated power. The pitch angle is kept at zero value,
while the optimal tip speed ratio starts to decrease. In this mode the turbine is
controlled so that the speed is kept constant at its rated value.
• Mode 3 – corresponds for values of speed and power greater than the rated values.
The speed is kept constant at the base value while the power is limited at its base
value by modifying the pitch angle.
In order to analyse the optimal control the system has been simulated with an average value
of 10 m/sec for the wind profile. The PI rotor current controllers uses the parameters
determined in §. The synchronous speed of the machine has been considered as the base
value for speed, while the rated power of the machine is the base for the active and reactive
power.
The simulations results in terms of the wind profile, shaft speed, active and reactive power,
both for the stator and the rotor circuit, are shown in Fig. 7.18.
Fig. 7.18. Simulation results.

129
It can be observed that the power is limited at the rated value, while the speed is lower than
the rated value. The reference of the stator reactive power and the measured one are zero in
the entire simulation horizon. Since, the wind speed has been acquired with the inherited
sample time from simulation (0.05 sec) and used in the control algorithm, the reference for
the stator power is not so smooth and the produced active power follows identically this
reference. Into a real system the wind speed is acquired with a bigger sample time and some
calculation in order to find the average wind speed at each 1 min are performed. Due to this
filtering of the wind speed the reference is much more smooth and the output power due not
present this fast variations. It has been omitted the averaging block of the wind speed in order
to study the dynamics performances of the control loops.
7.4 Summary
In this chapter the structure of a new Matlab/Simulink Toolbox for wind turbine applications
as a part of a new simulation platform dedicated to model, optimize and design wind turbines
has been presented. The content of the main libraries from this toolbox is briefly presented.
Then, in order to highlight the capabilities of this Toolbox some simulation demos are shown.
First the start-up sequence of an soft-starter fed squirrel cage induction generator, which is
used in fixed-speed WECS is presented. Then the control of active and reactive power for a
variable speed WECS using DFIG is shown.
130
Chapter 8
A flexible control strategy for small

stand alone wind turbines
In this chapter a flexible control strategy for an 11 kW wind energy conversion system
capable of working both in stand-alone and grid-connected configuration is presented. The
prototype has been developed at Department of Energy Conversion, Aalborg University
during a research project. In the simulation stage of this system some models from the “Wind
Turbine Blockset” have been used.
An adjustable-speed cage-induction generator controlled by a back-to-back power-processing
converter is used. The control of grid-side converter has been developed implemented and
successfully tested using a dSPACE prototype that provides a high degree of flexibility as
programming is performed entirely in Simulink and code generation is fully automated.
8.1 Background
Small efficient wind turbines in the range of 11-22 kW are used to supply private houses or
farms in remote locations from developing countries where low-cost access to the electrical
power grid is impractical.
The trend is that they should be able to work in stand-alone mode but also connected to
isolated local load in parallel with other generators such as other wind turbines, photo-voltaic
generators or diesel generators in the so called hybrid generator systems (HGS). Thus, some
new challenges on the control side of these WT occurred like reasonable voltage regulation in
stand-alone mode with non-linear load, automatic speed control according to the wind speed
and grid-connection mode.
However, in order to achieve these goals, relative complex control strategies for the entire
structure need to be developed. The structure of this system comprises a generator-side
converter and a grid-side converter as shown in Fig. 8.1.
Fig. 8.1. Back-to-back VSC used in an 11 kW WECS with squirrel-cage induction generator.
131
Both converters are three-phase bridge IGBT inverters and thus commercial “on-the-shelf”
frequency converters can be used, at least for the generator.
The control for the grid converter has been developed entirely in Simulink and implemented
on a dSPACE DS 1103 board that features a very high computational power and flexibility
degree.
A control strategy, which allows both stand-alone and grid-connected operation modes, has
been implemented and tested.
8.2 System description

The control strategy as well as the hardware structure for this WECS is shown in Fig. 8.2.
Fig. 8.2. Overall control strategy for the wind turbine test system.
The wind speed is calculated as an average value of the fixed-point wind speed over the
whole rotor, and it takes the tower shadow and the rotational turbulences into account. The
wind turbine mechanical shaft is emulated using a standard 11kW induction motor controlled
by a torque-regulated flux vector commercial frequency inverter.
The generator-side converter, which is a commercial PWM inverter with a built-in sensorless
vector flux control, is used to regulate the speed of the turbine and provide the magnetizing
flux of the generator.
132
The grid-converter is a standard three-phase PWM active rectifier connected to the grid
through an LCL filter.
The grid-side converter works with a near unity power factor. It is controlling the DC voltage
and the active/reactive power flow in the grid-connected mode and the amplitude and
frequency of the output voltage in the stand-alone mode.
The control strategy is implemented using a dSPACE DS 1103 system. The virtual control
panel developed in Control Desk is shown in Fig. 8.3.
Fig. 8.3. ControlDesk Panel for testing the wind turbine control strategy.
The control layout is divided in two parts: generator control and grid converter control. In the
generator control panel, the reference for the generator speed and the mechanical power can
be set. The speed and torque are shown in both graphs and digital displays.
The operation mode of the grid-side converter can be switched between stand-alone and grid-
connected modes of operation.
A PLL structure is used to synchronize the inverter voltage phase with the grid voltage phase
before starting in the grid-connected mode.
In grid-connected mode, the reference for the active and reactive power transferred to the
grid as well as the DC voltage can be set.
133
In stand-alone mode, the output voltage reference is set. The grid currents and voltages are
displayed in a scope-like graph. Digital displays show the power balance, including the
dumped power in the DC-chopper, as well as the output voltage and currents.
A test control panel for on-line tuning of the controllers has also been done. It has been
proved that this is a very convenient method for fine-tuning of the PI controllers.
So, the developed control panel is a very flexible tool for testing and tuning of the control
system.
The control algorithm has been developed entirely in Simulink, which is a very friendly,
block-oriented, dynamic simulation environment. First, an off-line simulation model is
developed, and then with only minor changes, the model has been implemented into the
dSPACE controller and tested using the ControlDesk panel. The Simulink model of the
control structure is shown in Fig. 8.4.
Fig. 8.4. Simulink diagram of the real-time model.
8.3 Grid-connected mode of operation

The grid-connected control is depicted in Fig. 8.5. In this mode of operation, basically all the
available power, which can be extracted from the wind turbine, is transferred into the grid.
Additionally, the static reactive power compensation is possible by adjusting the reference of
the reactive current iq.
Using standard PI controllers, the grid currents in dq-synchronous reference frame in the
inner loop and the DC voltage in the outer loop can be controlled. The cross coupling
between the axis is implemented in order to compensate the output of the controllers.
Reference current in the q-axis of the current loop is set to zero in order to achieve a zero
phase angle between the voltage and the phase current and therefore a unity power factor.
134
The output of the current regulators set the voltage reference for a standard space vector
modulation (SVM), which controls the switches of the grid-side converter via an optic link.
Fig. 8.5. Control structure in grid-connected mode of operation.
Synchronism with grid is achieved by using the calculated grid voltage phase angle from
measurements of the grid voltage
8.4 Stand-alone mode of operation

The control structure for the stand-alone mode of operation is shown in Fig. 8.6.
Fig. 8.6. Control structure in stand-alone mode of operation.
135
Now there is no grid so the output voltages need to be regulated in terms of amplitude and
frequency and by doing so, the reactive and respectively the active power flow is regulated.
The control structure comprises an output voltage controller, a DC-link voltage controller, a
DC-link chopper control and the current limiter. The output voltage controller is aiming to
control the output voltage with minimal influence of the shape of the non-linear load currents
or load transients. Again, standard PI controllers are used.
A small DC power supply of 600V/0.5A is necessary to provide the magnetizing energy for
the induction generator during start-up.
In the case of an unbalance between the produced power and the load-required power, by
adjusting the generator speed the produced power will be regulated. Eventual excess in the
produced energy will be quickly dumped in the dump resistor by starting the chopper.
8.5 Experimental results

A physical layout of the developed system is presented in Fig. 8.7. Two Danfoss
VLT5022/400V/18.5kVA/37.5A converters have been used.
Fig. 8.7. Experimental setup.
The current controller of the grid converter have been tested at 10 A step reference and
reduced voltage. The system response depicted in Fig. 8.8 demonstrates the good
performances of the current controller
136
Fig. 8.8. Current response for 10 A step in reference.
The DC-voltage controller has been tested for 100% (16A) current step loading, as depicted
in Fig. 8.9 and it shows good dynamic response.
Fig. 8.9. DC-voltage controller response for a 100% step in load.
The wind turbine has been tested in stand-alone mode at rated nominal power (11 kW) with a
non-linear load. The DC-voltage has been set at 700 V while the output phase voltage at
230V. The measured voltage and current are shown in Fig. 8.10. The measured voltage THD
was 9.8% at a current THD of 34.7%, which indicates a reasonable voltage regulation. The
switching ripple in the waveforms is removed as they are sampled once in the middle of the
switching period using the dSPACE controller. However, the THD is measured with a
Voltech PM3000 power analyzer.
137
Fig. 8.10. Voltage and currents waveforms with a non-linear load.
8.6 Summary
In this chapter a flexible control strategy for an 11 kW wind energy conversion system
capable of working both in stand-alone and grid-connected configuration is presented. The
prototype has been developed at Department of Energy Conversion, Aalborg University
during a research project. The control of grid-side converter has been developed implemented
and successfully tested using a dSPACE prototype that provides a high degree of flexibility
as programming is performed entirely in Simulink and code generation is fully automated.
Some performances of the developed system under different working condition are shown.
138
Chapter 9
Conclusions and Future work

Currently, the MW-sized wind turbines are available on the market and it is expected that in
the near future the size will increase further, as well as the application of power electronics in
wind turbines. Since a wind turbine contains mainly two types of systems: namely
mechanical system and electrical ones, it is of great interest to analyse and predict the
interaction between them. Moreover, the impact of these wind turbines into the main power
system became more and more important. The challenges here are to identify, to predict and
to improve the power system stability limit in these future power systems with a high
penetration of wind energy.
Therefore, simulation will play an important role in the design and analysis of the large wind
turbines. While, the current simulation tools, which are used by the wind turbines
manufacturers, cannot cover all the modelling and analyse aspects a simulation platform
project has been initiated in order to improve and to extend their capabilities. Among the
goals of this simulation platform project, the development of new models for electrical
generators as well as models for the power converters is one of the main targets. A longer-
term goal is the development of dynamic and steady state models for all components within a
wind turbine, which can be used in a complete optimisation of a wind turbine system.
It has been chosen that Matlab/Simulink will be used as a general model-developing tool and
also for validation of these models. After validation of the models they are implemented in
the other tools. Finally, all these models are collected into a dedicated Simulink Toolbox for
wind turbine applications.
The work presented in this dissertation deals with the modelling and analysis of the electrical
generators and power converters within a large wind turbine as a part of the above mentioned
simulation platform project. A new Matlab/Simulink Toolbox for wind turbine applications is
developed and some applications of it are shown. Some models from this toolbox, e.g. three-
phase two-winding transformer model and the cable model have been used and validate in a
particular analysis for the Middelgrunden Windfarm (near Copenhagen), Denmark.
A summary of the results of this thesis is presented here.
Induction machine modelling. A unified approach in dynamic modelling of induction
machine, both squirrel-cage and doubly-fed is presented. The dynamic equations are derived
in the state space form in the abc/abc reference frame as well as in the dqo-dqo arbitrary
reference frame both with currents and fluxes as state-variables.
The abc/abc model of induction machine is very useful in analysis of unbalanced and fault
operation as well as in the analysis of AC controller-fed induction machine.
Based on the dynamic equations in state-space form the linearized equation of the machine
can be obtained.
Since the deep-bar effect has an important influence in a large induction machine, much
attention has been paid in modelling of it based only on data-sheet parameters.
139
As simulation speed is a very important issue in any simulation tool, especially when the
considered system should be analysed in a long time horizon, the implementation methods in
Matlab/Simulink have been investigated in detail. It has been found that using C S-Functions
the simulation speed can be increased with at least a factor of two, while the numerical
stability increases.
Reduced order modelling and eigenvalue analysis of induction machine. One of the main
targets has been the development of a new model for induction generator, both squirrel-cage
and doubly-fed machine for the aero elastic simulation tools. Since here the time scale is in
the range 10-1 – 104 sec the reduced order modelling of the induction machine has been
investigated.
Starting from the complete linearized model of the induction machine, the reduced order
models, which neglect only stator transients and both stator and rotor transients, have been
extracted. One-mass model as well as the two-mass model of the wind turbine drive train
including torsional torques has been taken into account. Moreover the influence of the deep-
bar effect is considered.
A comparison between the response of the complete order model and the reduced order
models, which neglect only stator transients and both stator and rotor transients, has been
performed under different operating modes.
Then, these reduced order models have been implemented the advanced aero elastic tool
HAWC. It has been found that due to the dynamics of the induction generator some
resonances between the mechanical part and the electrical part of the wind turbine arises. The
same behaviour has been observed also in the field measurements.
This interaction between the mechanical part and the electrical one has been found also in the
eigenvalue analysis.
Power converters modelling. Based on the switching function concept the mathematical
models for the most used power converters in wind turbine applications, namely AC-
controllers and back-to-back voltage source converter, are presented. The AC-controllers are
modelled taking into account the winding connection type for the generator. Some typical
waveforms for each topology are shown. The steady state models of AC controllers have
been investigated.
Control of variable-speed wind energy conversion systems. An overview of the control
methods for induction generators, both squirrel-cage and doubly-fed machine is presented.
The basic control structures for these generators, both grid-connected and stand-alone mode
are shown. A controllers design method based on the complete state-space model of the
generator including the two-mass model of the drive train is presented. The parameters for
the controllers have been determined in two cases: a fast control, which is necessary during
the start-up sequence of the doubly-fed induction generator and a slow control for a normal
operation of the wind turbine. The slow rotor control can be used directly in the active and
reactive power control of the wind turbine. In both cases the influence of the deep-bar effect
has been taken into account.
The influence of disturbance in the stator voltage in the rotor current control loops is shown
for both considered cases. Finally, in order to analyse the response of the rotor current control
loops in normal operation of a wind turbine, some simulation results for a 2 MW doubly-fed
induction generator are shown.
140
Again one of the main targets here was to develop a simple control structure for the active
and reactive power for the doubly-fed induction generator, which will be implemented in the
aero elastic simulation tools.
Wind Turbine Blockset. The new Matlab/Simulink Toolbox for wind turbine applications as
a part of a new simulation platform dedicated to model, optimize and design wind turbines
has been presented. The content of the main libraries from this toolbox has been briefly
presented. Then, in order to highlight the capabilities of this Toolbox some simulation demos
are shown. Since, the Danish concept of fixed speed wind turbines is widely used, the start-
up sequence of a soft-starter fed squirrel cage induction generator has been presented. Then a
simple control of active and reactive power for a variable speed WECS using DFIG, which
will be implemented in HAWC, is shown.
One of the main contributions of the thesis is in modelling of the components within a wind
turbine. A collection of these mathematical models is available in a Matlab/Simulnk toolbox
dedicated for wind turbine applications. Using the available models from the Toolbox the
main concepts for large wind turbines, namely fixed speed wind turbines with squirrel-cage
induction generator and variable-speed ones using doubly-fed machines, can be analyzed and
simulated under different operation conditions.
Future work
Concerning the future work in the field of modelling and analyses of the electrical part of
large wind turbines the following steps should be done:
• Development of reduced order models for synchronous machine, both field winding
and permanent magnet.
• Development of mathematical models based on the switching function concept for
multilevel and matrix converters, as well as the steady-state ones.
• Development of an advanced model for three-phase five-limb transformer.
• Development of advanced models for cables and distribution lines.
• Developments of new control strategies for wind turbines.
• Development of a Saber Toolbox dedicated for wind turbine applications.
141
142
Appendix A
Typical parameters for induction

generators and drive train
Table A.1. Typical parameters of a drive train for 2 MW wind turbine

Parameters
Moment of inertia of the wind turbine rotor [kgm2] ≅ 90 ⋅105
Equivalent spring constant indicating the torsional stiffness of the shaft
≅120 ⋅106
[Nm/rad]
Equivalent damping coefficient of the shaft ≅3.6 105
Gear box ratio ≅100
Table A.2. Typical electrical parameters for 2 MW induction machines

in wind turbine applications
Parameters at rated speed SCIG DFIG
Rated power [MW] 2 2
Stator resistance [Ω] 0.005 0.001
Stator leakage inductance [mH] 0.4 0.07
Rotor resistance [Ω] 0.01 0.0013
Rotor leakage inductance [mH] 0.3 0.08
Magnetizing inductance [mH] 16 3
Number of pole pairs 2 2
Moment of inertia [kgm2] 90 90
Rated speed [rpm] 1512 1512
143
144
Appendix B
List of publications
• F. Iov, F. Blaabjerg, A.D. Hansen, P. E. Sørensen - Analysis of reduced order

models for large squirrel-cage induction generators in wind turbine applications, EPE
2003 Conference, 2-4 September 2003, Toulouse, France, (in press);
• R. Teodorescu, F. Iov, F. Blaabjerg - Flexible development and test system for
11kW wind turbine, PESC 2003 Conference, 15-19 June, 2003. Acapulco, Mexic (in
press);
• A.D. Hansen, P. E. Sørensen, F. Iov, F. Blaabjerg – Initialisation of grid-
connected wind turbine models in power-system simulations, Journal of Wind
Engineering, Vol. 27, No. 1, 2003, pp. 21-28;
• F. Iov, F. Blaabjerg, A.D. Hansen, R. Teodorescu – Modeling of soft-starters for
wind turbine applications, Proceed. of PCIM 2003 Power Quality, 20-22 May 2003,
Nuremberg, Germany, pp. 179-184;
• F. Iov, F. Blaabjerg, A.D. Hansen, T.J. Larsen – Analysis of reduced order
models for large wound-rotor induction generators in wind turbine applications,
Proceed. of PCIM 2003 Power Quality, 20-22 May 2003, Nuremberg, Germany, pp.
213-218;
• R. Teodorescu, F. Blaabjerg, F. Iov – Control strategy for small stand-alone
wind turbines, Proceed. of PCIM 2003 Power Quality, 20-22 May 2003, Nuremberg,
Germany, pp. 201-206;
• T.J. Larsen, M.H. Hansen, F. Iov – Generator dynamics – analysis and
simulations, RISØ National Laboratory, Roskilde, RISØ-R-1395(EN), ISBN 87-550-
3188-9;
145
• F.Iov, F. Blaabjerg, A.D. Hansen, P.E. Sørensen, Z. Chen – A New Simulation
Platform to Model, Optimize and Design Wind Turbines, Proceed. of IECON 2002
Conference, 5 – 8 November, Sevilla, Spain, 6 pp., ISBN 0-7803-7475-4.
• F.Iov, F. Blaabjerg, A.D. Hansen, Z. Chen – Comparative study of different
implementations for induction machine model in Matlab/Simulink for wind turbine
simulations. Proceed. Of COMPEL 2002, Mayaguez, Puerto Rico.
• F.Iov, F. Blaabjerg, A.D. Hansen, P.E. Sørensen, Z. Chen – A Case Study
regarding Influence of Solvers in Matlab/Simulink for Induction Machine Model in
Wind Turbine Simulations, Proceed. of NORPIE 2002, 12-14 August, Stockholm,
Sweden.
• F. Iov, F. Lungeanu, A-M Tataru, D. Calueanu - Study of a SVM-FOC command
technique without capacitor bank for an autonomous induction generator, National
Conference in Electrical Drives, CNAE 2000, Iasi, Romania;
• F. Iov, N. A. Cutululis, M. Culea, A-M Tataru, D. Calueanu - A Neural-Fuzzy
Approach of Wind Turbine Simulators, Proceed. of SIELMEC ’00 Conference,
Chisinau, R Moldova;
• F. Iov, R. Teodorescu, F. Lungeanu, A.M. Tataru, T. Dumitriu - A Simplorer
Approach Of SVM Command Using Petri Nets, Proceed. of ELECTROMOTION ’99
Conference, Patras, Grecia;
• F. Iov, D. Calueanu, A.M. Tataru, T. Dumitriu - Mathematical Modelling of
Self-Excited Squirell Cage Induction Generator, Proceed. of SIELMEC ’99 Conference,
Chisinau, R Moldova;
• R. Teodorescu, F. Iov, F. Lungeanu, T. Dumitriu, A.M. Tataru, R. Panaitescu -
The Development of a Simulink Toolbox for Advanced Electrical Drives Simulation,
Proceed of OPTIM’98, Brasov, Romania;
• R. Teodorescu, F. Lungeanu, F. Iov, A.M. Tataru, T. Dumitriu - A New
Parameters Identification Technique for Induction Machine Diagnosis, Proceed. of
OPTIM’98, Brasov, Romania;
146

Contributions To Modelling, Analysis and Simulation of AC Drive Systems-Application To Larg

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Contributions To Modelling, Analysis and Simulation of AC Drive Systems-Application To Larg

Uploaded by

Copyright:

Available Formats

“Dunarea de Jos” University - Galati

Department of Electrotechnics, Electrical Machinery and

Contributions to Modelling, Analysis and

Ing. FLORIN IOV

Faculty of Electrical Engineering and Computer Science

Aalborg, spring 2003

Table of contents .......................................................................................................................7

Chapter 1 Introduction ...........................................................................................................9

Chapter 2 Large Wind Turbines...........................................................................................15

Chapter 3 Induction Machine Modelling .............................................................................29

Chapter 5 Power Converters Modelling .............................................................................. 81

Chapter 6 Control of variable-speed wind energy conversion systems............................... 95

Chapter 7 Wind Turbine Blockset ..................................................................................... 119

Chapter 9 Conclusions and Future work ........................................................................... 139

Appendix B List of publications .......................................................................................... 145

1.2 Scope of work

Modelling aspect Time scale Tool

1.3 Review of related research in modelling

Large Wind Turbines

2.1 Proprieties of the wind

2.1.1 Wind distribution

2.1.2 Synoptic and diurnal variations

2.2 Wind turbine performances

2.2.1 The performance curves

2.2.1.1 Tip speed ratio

The tip speed ratio of a wind turbine is defined as:

2.2.1.3 The CQ-λ performance curve

2.2.1.4 The CT-λ performance curve

2.2.2 Constant rotational speed operation

2.2.2.1 The Kp-1/λ curve

2.2.2.2 Stall regulation

2.2.2.3 Effect of rotational speed change

Fig. 2.5. Effect on extracted power of rotational speed.

2.2.2.4 Effect of blade pitch angle change

Fig. 2.6. Effect on extracted power of blade pitch set angle.

2.2.2.5 Pitch regulation

2.2.2.6 Pitching to stall

2.2.2.7 Pitching to feather

2.2.3 Variable speed operation

2.2.4.1 Passive stall control

2.2.4.2 Active stall control

2.2.4.3 Passive pitch control

2.2.4.4 Active pitch control

2.3.1 Present status

Fig. 2.9. Basic topologies for wind turbines:

Switch type GTO IGCT BJT MOSFET IGBT

2.3.2 Future trends in wind turbines technology

Middelgrunden, Baltic Sea, Denmark 2001 20 * 2

Lubeck, Germany 2004 100

Induction Machine Modelling

3.1 The voltage equations formulated in their natural

3.1.1 Fluxes as state variables

or in the compact form:

the inductance matrix [ L ] can be rewritten as:

3.1.2 Inductance matrix analytical inversion

3.1.3 Currents as state variables

where the coefficients g1,…, g3 are defined as follows:

3.1.4 Expression of the electromagnetic torque

Pcopper = [ I ] [ R ][ I ] - copper losses in machine windings

3.2 The voltage equations in arbitrary reference frame

Fig. 3.2. Induction machine windings in the dqo-arbitrary reference frame.

 vsd   R s  isd   Ψ sd   0 −ωg 0 0 0 0 Ψ 