Professional Documents
Culture Documents
2007.2 - Martinez Jasmin
2007.2 - Martinez Jasmin
Jasmin Martinez
September 21, 2007
In recent years, wind turbines have become an acceptable alternative for electrical
energy generation by fossil or nuclear power plants, because of the environmental and
economic benefits. Still, much research remains to be done in order to improve wind
turbines behaviour and make them more profitable and reliable.
This work covers the modelling of wind turbines for power system studies. The
operation of horizontal, variable speed wind turbines with pitch control was
investigated. Complexities of various parts of a wind turbine model, such as
aerodynamic conversion, drive train and generator representation were analyzed. The
mathematical equations describing the dynamic behaviour of a wind energy system
were successfully simulated in gPROMS. The wind turbine model was further tested
upon step changes in the wind velocity as well as the blade pitch angle, confirming
the need of power control.
Based on the obtained wind turbine model, a power control structure was developed
that takes into consideration the dynamical aspects of the wind turbine as well as
constraints. An explicit parametric controller, a novel control method, was designed
using MATLAB and the Parametric Optimization (POP) software. A simple explicit
optimal control law was constructed that allows the on-line implementation via simple
linear function evaluations. The controller was implemented using gO:MATLAB and
the simulation results show that the controller accurately adjusts the blade pitch angle
to set the wind turbine power output to its reference value.
2
Acknowledgments
First of all I wish to thank my parents for being an example and encouraging me
throughout my studies, making this thesis possible. I am also grateful to my favourite
sister, Cami.
I acknowledge Prof. Pistikopoulos and Dr. Kouramas for their supervision, advice and
contribution for the success of this work. I am indebted to Mark Pinto and Lin
Zhenhua for their help with gPROMS and gO:MATLAB.
Special thanks to my friends at Imperial College, who made this thesis an enjoyable
experience.
Financial support for this work was provided by Programme Alban.
Finally, I would like to express my deepest gratitude for the constant support, patience
and love that I received from Nico.
3
Contents
ABSTRACT ............................................................................................................................................ 2
ACKNOWLEDGMENTS ...................................................................................................................... 3
CONTENTS ............................................................................................................................................ 4
LIST OF FIGURES................................................................................................................................ 6
LIST OF TABLES.................................................................................................................................. 7
CHAPTER 1. INTRODUCTION .......................................................................................................... 8
1.1 WORK MOTIVATION............................................................................................................... 8
1.2 WORK OBJECTIVE .................................................................................................................. 9
1.3 THESIS OUTLINE .................................................................................................................... 9
CHAPTER 2. BACKGROUND AND LITERATURE REVIEW .................................................... 11
2.1 WIND TURBINE DEVELOPMENT AND TYPES OF TURBINES ............................................................ 11
2.1.1 Wind Turbine History........................................................................................................... 11
2.1.2 State-of-the-art Technologies ............................................................................................... 12
2.1.2.1 Definition of a Wind Turbine ....................................................................................................... 12
2.1.2.2 Aerodynamic Lift and Aerodynamic Drag Wind Turbines .......................................................... 12
2.1.2.3 Horizontal-axis and Vertical-axis Wind Turbines ........................................................................ 12
2.1.2.4 Variable-speed and Constant-speed Wind Turbines ..................................................................... 15
2.1.3 Power Control...................................................................................................................... 17
2.1.3.1 Stall Control ................................................................................................................................. 17
2.1.3.2 Pitch Control ................................................................................................................................ 18
2.1.3.3 Active Stall Control...................................................................................................................... 19
2.2 WIND FARMS ................................................................................................................................ 19
2.2.1 Wind Farm Definition and Siting ......................................................................................... 19
2.2.1.1 Definition of Wind Farm .............................................................................................................. 19
2.2.1.2 Wind Farm Siting ......................................................................................................................... 20
2.2.2 Wind Farm Control .............................................................................................................. 21
2.2.2.1 Centralised Control Structure ....................................................................................................... 22
2.2.2.2 Decentralised Control Structure ................................................................................................... 22
2.2.2.3 Partly Centralised, Partly Decentralised Control Structure .......................................................... 22
2.2.3 Requirements for the Interconnection of Wind Farms to the Power System ........................ 23
2.2.3.1 Active Power Control ................................................................................................................... 23
2.2.3.2 Frequency Control ........................................................................................................................ 23
2.2.3.3 Voltage Control ............................................................................................................................ 23
2.2.3.4 Tap Changers ............................................................................................................................... 24
2.2.3.5 Wind Farm Protection .................................................................................................................. 24
2.2.3.6 Modelling Information and Verification ...................................................................................... 24
2.2.3.7 Communication and External Control .......................................................................................... 24
CHAPTER 3. WIND TURBINE MODELLING ............................................................................... 26
3.1 MODELLING OF THE BLADES ......................................................................................................... 27
3.2 MODELLING OF THE DRIVE TRAIN ................................................................................................. 29
3.3 MODELLING OF THE ASYNCHRONOUS GENERATOR ....................................................................... 30
3.3.1 Model assumptions............................................................................................................... 31
3.3.2 0dq reference frame ............................................................................................................. 32
3.3.3 Per unit system ..................................................................................................................... 32
3.3.4 Asynchronous generator model............................................................................................ 33
3.3.4.1 Model including stator transients ................................................................................................. 34
3.3.4.2 Model neglecting stator transients ................................................................................................ 35
3.4 MODELLING AND SIMULATION IN GPROMS ................................................................................. 36
3.4.1 Modelling in gPROMS ......................................................................................................... 36
3.4.2 Simulation results................................................................................................................. 37
3.4.2.1 Base case ...................................................................................................................................... 38
3.4.2.2 Wind velocity step change ........................................................................................................... 40
3.4.2.3 Blade pitch angle step change ...................................................................................................... 41
4
CHAPTER 4. WIND TURBINE CONTROL .................................................................................... 42
4.1 POWER CONTROL .......................................................................................................................... 42
4.2 EXPLICIT PARAMETRIC CONTROLLER ............................................................................................ 42
4.2.1 Model identification ............................................................................................................. 46
4.2.2 Explicit parametric controller design .................................................................................. 47
4.2.3 Controller implementation (gO:MATLAB) .......................................................................... 50
CHAPTER 5. CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE WORK............. 54
5.1 CONTRIBUTIONS OF THIS THESIS ................................................................................................... 54
5.2 FUTURE WORK RECOMMENDATIONS ............................................................................................. 55
REFERENCES ..................................................................................................................................... 56
APPENDICES ...................................................................................................................................... 61
5
List of Figures
FIGURE 5. C p VS. TIP-SPEED RATIO AND PITCH ANGLE FOR A TYPICAL WIND TURBINE WITH PITCH
FIGURE 26. GPROMS WIND TURBINE MODEL EXECUTING WITHIN MATLAB ........................................ 51
FIGURE 27. GO:MATLAB POWER PROFILE ............................................................................................ 52
FIGURE 28. GO:MATLAB BLADE PITCH ANGLE PROFILE ...................................................................... 53
6
List of Tables
TABLE 1. VARIABLES WITH A KNOWN VALUE, VARIABLES WITH AN UNKNOWN VALUE AND NUMBER OF
EQUATIONS FOR WIND TURBINE MODEL ......................................................................................... 37
TABLE 2. WIND TURBINE PARAMETERS [LUBOSNY, 2003], [MARTINS ET AL, 2007] ............................... 38
TABLE 3. CONTROL DESIGN VARIABLES .................................................................................................. 47
7
Chapter 1. Introduction
Wind energy is one of the fastest growing renewable energies in the world. The
generation of wind power is clean and non-polluting; it does not produce any by-
products harmful to the environment.
Nowadays, modelling is the basic tool for analysis, such as optimization, project,
design and control. Wind energy conversion systems are very different in nature from
conventional generators, and therefore dynamic studies must be addressed in order to
integrate wind power into the power system. According to [Lubosny, 2003], in the
case of power systems with classical sources of energy analysis, the modelling is
relatively simple because the models of objects and controllers are well known and
even standardized; the data are available. But in the case of wind turbine modelling,
researchers meet problems related to the lack of data and lack of control-system
structures due to strong competition between wind turbine manufacturers. This leads
to the situation in which many researchers model the wind energy conversion systems
in relatively simple form, almost neglecting the control systems, which significantly
influence the reliability of the analytical results.
Classical techniques such as proportional (P), integral (PI) and derivative (PID)
controllers are typically used to regulate wind power. But by assuming the wind
turbine operating in steady state conditions, most of the previous work regarding wind
turbine control does not take into consideration the dynamical aspects of the wind and
the turbine, which have strong non-linear characteristics [Balas et al, 2006]. Advances
in wind turbine technology made necessary the design of more powerful control
systems, to improve wind turbines behaviour and make them more profitable and
reliable [Boukhezzar et al, 2005]. However, as stated in [Balas et al, 2006]
“Controlling modern turbines to minimize the cost of wind energy is a complex task,
and much research remains to be done to improve controllers”. An interesting
characteristic of wind energy systems is that wind speed determines the point of
operation; it simply defines the available amount of energy that can be converted into
electricity. The wind cannot be controlled; in other words the system is driven by
noise, which makes wind turbine systems essentially different from most other
systems. This explains the need for robust controller design [Bongers et al, 1992].
8
On the other hand, theoretically, the electrical output from a wind turbine should be
smooth and non-fluctuating [Butterfield et al, 2001]. But electricity generated from
wind farms can be highly variable on different time scales: from hour-to-hour, daily
and seasonally. This represents a considerable challenge when incorporating wind
power into a grid system, since in order to maintain grid stability energy supply and
demand must remain in balance.
The main objective of this work is to contribute to the topic of wind energy systems
modelling and control by developing an accurate model for a wind turbine and based
on this model contemplate control issues.
The scientific objectives of this research include the following:
• Modelling and dynamic behaviour investigation of the aerodynamic,
mechanical and electrical parts of a variable speed wind turbine equipped with
an induction generator and blade pitch angle control.
• Simulation of the overall wind energy system in gPROMS.
• Formulation of an explicit parametric controller in MATLAB/ Parametric
Optimization (POP) software.
A parametric controller is a novel control method that has been recently applied to a
number of processes. This work aims at exploiting the properties of explicit
parametric control and demonstrate the potential benefits of this control method for
wind turbines.
The wind turbine configuration considered throughout this work is an aerodynamic
lift, 3 blade, horizontal-axis, variable speed, pitch controlled wind turbine.
The thesis is divided into five chapters including this introduction chapter. The paper
is structured as follows.
Chapter 2 contains a background on theoretical fundamentals regarding wind turbines
and wind farms. The first part gives an overview of the wind turbine history and
development. The main types of wind turbines and their configurations are explained
9
in detail. Furthermore, the different power control techniques available to control the
wind turbine power output are exposed. Wind farms are introduced and classified
accordingly to their siting. The main wind farm control structures are described and
the requirements for the interconnection of wind farms to the power system are
discussed.
Chapter 3 presents detailed mathematical models that describe the dynamic behaviour
of a wind energy system, including aerodynamic, mechanical and electrical parts.
Simulation results of the overall wind turbine model are given for a base case, as well
as for wind speed and blade pitch angle step changes.
Chapter 4 contains the formulation of an explicit parametric control strategy for a
wind turbine. The properties and potential benefits of this control method for wind
energy systems are investigated. Moreover, the controller is implemented and tested.
Finally, Chapter 5 provides conclusions on the research done and offers
recommendations for future work.
10
Chapter 2. Background and Literature review
This chapter is aimed at presenting a review on the wind turbines and wind farm state
of the art technologies.
Wind-powered machines have been used by humans for thousands of years. Until the
20th century wind power was used to provide mechanical power to pump water or to
grind grain. The earliest recorded windmills are vertical-axis mills and were used in
Afghanistan in the seventh century BC. Horizontal-axis windmills are found in
historical documents from Persia, Tibet and China around 1000 AD. From Persia and
the Middle-East, the horizontal-axis windmill spread across Europe in the 12th century,
where windmill performance was constantly improved; by the 19th century a
considerable part of the power used in the industry in Europe was based on wind
energy. Industrialisation then led to a gradual decline in windmills, as the use of
fluctuating wind energy was substituted by fossil fuel fired engines which provided a
more consistent power source [Ackermann et al, 2000].
In the 1970s, with the first oil price shock, the modern era of wind turbine generators
began, focusing in producing electricity instead of mechanical energy. Conventional
methods to generate electricity burn fuel to provide the energy to drive a generator,
creating pollution, acid rain and contributing to global warming. In recent years there
has been a growing interest in wind energy power systems because of the
environmental benefits and the economic benefits of fuel savings [Fujita et al, 2006].
The wind is a clean source and it will never run out. Wind energy technology is
developing fast; turbines are becoming cheaper and more powerful, bringing the cost
of renewably-generated electricity down [British Wind Energy Association, 2006];
The cost of generating electricity from wind has fallen almost 90% since the 1980s
[Karrari et al, 2005]. Nowadays, wind energy is one of the most important sustainable
energy resources and has become an acceptable alternative for electrical energy
11
generation by fossil or nuclear power plants [Bongers at al, 1992]. A list of wind
turbines manufacturers can be found in Appendix 1 [Energy Source Guides, 2007].
12
Figure 1. Vertical-axis wind turbine [Wind-Works, 2007]
Vertical-axis wind turbines have the advantages that no tower is needed, they operate
independently of the wind direction (a yawning mechanism is not needed) and heavy
gearboxes and generators can be installed at ground level. But they have many
disadvantages: they are not self-starting, the torque fluctuates with each revolution as
the blades move into and away from the wind, and speed regulation in high winds can
be difficult. Vertical-axis turbines were developed and commercially produced in the
1970s until the end of the 1980s. But since the end of the 1980s the research and
production of vertical-axis wind turbines has practically stopped worldwide
[Ackermann et al, 2000].
At present, horizontal-axis wind turbines dominate the market; Figure 2 illustrates the
different configuration between a horizontal-axis and a vertical axis turbine.
13
A horizontal-axis wind energy conversion system mainly consists of [Harrison et al,
2000]:
• The rotor blades, which extract the kinetic energy present in the wind and
transform it into mechanical power.
• The nacelle, with a power control system that limits and conditions the
extracted power; a gear box that transfers the load and increases the rotational
speed to drive the generator; and an electrical system which converts the
mechanical energy into electrical energy.
• A tower that supports the nacelle.
The major components of a modern horizontal-axis wind turbine are shown in Figure
3.
The yaw mechanism turns the turbine so that it faces the wind. Sensors are used to
monitor wind direction and the tower head is turned accordingly.
Wind turbines can have three, two or just one rotor blades. Two or three blades are
usually used for electricity power generation. Two blades cost less than three blades,
but they need to operate at higher rotational speed than three-bladed wind turbines. As
a result, the individual blades need to be lighter and hence more expensive on a two-
14
bladed turbine [McNerney et al, 1993]. Besides, three-bladed turbines are generally
accepted as more aesthetic than two or one bladed turbines. Hence, turbines with three
blades dominate the wind industry.
vTIP ω⋅R
λ= = (2. 1. 1)
vWIND v
where ω is the blades angular velocity (rad/s), R the rotor radius (m) and v the wind
speed (m/s).
of the power in the wind can be converted to useful energy by a wind turbine. The
power available for a wind turbine is equal to the change in kinetic energy of the air
15
as it passes through the rotor. This maximum theoretical C p was first formulated in
tip speed ratio, λ , rather than against the wind velocity, as this creates a
dimensionless graph. A typical C p vs. λ curve is shown in Figure 4.
Figure 4. C p Vs. λ for a typical wind turbine [De Montfort University, 2007]
This curve illustrates that the maximum value of C p is only reached for a specific λ
speed changes. Hence, variable speed wind turbines are designed to operate at
optimum energy efficiency, regardless of the wind speed.
On the other hand, due to the fixed-speed operation for constant speed turbines, all
fluctuations in the wind speed are transmitted as fluctuations in the mechanical torque
and then as fluctuations in the electrical power grid [Ackermann, 2005]. This,
together with the increased energy capture obtained by using a variable-speed wind
turbine provides enough benefit to make the power electronics (frequency converter)
16
cost effective [Balas et al, 2006]. Therefore, the wind industry trend is to design and
construct variable-speed wind turbines.
The kinetic energy in a flow of air through a unit area perpendicular to the wind
1 2
direction is ⋅ v per mass flow rate. For an air stream flowing through an area A the
2
mass flow rate is ρ ⋅ A ⋅ v , therefore the power in the wind is equal to
1 1
P = (ρ ⋅ A ⋅ v ) ⋅ ⋅ v 2 = ⋅ ρ ⋅ A ⋅ v 3 (2. 1. 3)
2 2
where ρ is the air density (kg/m3), A the area (m2) and v the wind speed (m/s), and
P the power of the wind (watts or J/s).
From equation 2.1.3, the power available from the wind is a function of the cube of
the wind speed. That means that a doubling of the wind speed gives eight times the
power output from the turbine. Therefore, turbines have to be designed to support
higher wind loads than those from which they can generate electricity, to prevent
them from damage.
Wind turbines reach the highest efficiency at a wind speed between 10 and 15m/s.
Above this wind speed, the power output of the rotor must be controlled to reduce
driving forces on the rotor blades as well as the load on the whole wind turbine
structure [Ackermann et al, 2000]. High winds occur only for short periods and hence
have little influence in terms of energy production but, if not controlled, they would
dominate the design and cost of the drive train and the generator [Harrison et al, 2000].
Accordingly, all wind turbines are designed with a type of power control. There are
different ways to control aerodynamic forces on the turbine rotor and therefore limit
the power in high winds in order to avoid damage to the wind turbine [Ackermann,
2005].
Three options for the power output control are currently used:
17
control, since there are no moving parts to adjust: it is the inherent aerodynamic
properties of the blade which determine power output. The twist and thickness of the
rotor blade vary along its length in a way that turbulence occurs behind the blade
whenever the wind speed becomes too high. This turbulence means that less of the
energy in the air is transferred, minimising power output in higher speeds [British
Wind Energy Association, 2006]. In other words, the design of the blades
aerodynamic causes the rotor to stall (lose power) when the wind speed exceeds a
certain level. Thus, the aerodynamic power of the blades is limited.
The disadvantages of this control method are low efficiency at low wind speeds, and
no assisted start-up [Ackermann, 2005]. Besides, this type of control requires the use
of a constant speed turbine which, as already explained, has lower energy efficiency
than the variable speed turbine.
low and medium wind speeds, the pitch angle is regulated to allow the wind turbine to
operate at its optimum condition. In high wind speeds, the pitch angle is increased in
order to shed some of the aerodynamic power and maintain the rotor speed within a
controllable limit. As pitch angle increases, the wind turbine operates at lower
efficiency [Butterfield et al, 2001].
18
Figure 5. C p Vs. Tip-Speed Ratio and Pitch angle for a typical wind turbine with pitch control
[Balas et al, 2006]
Nowadays, large wind turbines are increasingly being operated with pitch control
systems [Harrison et al, 2000].
19
the preparation of the environmental statement, legal fees and project management
costs. They are largely independent of the size of the wind farms, so it is desirable to
spread them over as large a project as possible [Jenkins et al, 1997]. This had led to
the construction of large wind farms, some of them having as many as 150 wind
turbines [EERE, 2007] and an output of hundreds of megawatts [Ackermann, 2005].
• Onshore developments (Figure 6), where wind farms are constructed inland,
usually in hilly or mountainous regions to favour windy conditions.
• Offshore developments (Figure 7), where wind farms are installed in the sea,
at least 10 kilometres away from the land. This is attractive because of higher
wind speeds over the sea but mainly because of the reduced environmental
impact. The major disadvantage of this type of wind farm development is the
higher cost involved in terms of foundations, power collection cables,
installation and maintenance [Jenkins et al, 1997].
20
Figure 7. Offshore wind farm [Arthus-Bertrand, 2007]
The trend in the wind energy industry is to install wind turbines in large
concentrations with hundreds of megawatts of power capacity. Wind farms of this
size are often connected directly to the transmission grid, replacing traditional power
plants. This means that wind turbines are required to behave as active controllable
components in the power system. Such large wind farms need to meet very high
technical demands, such as to perform frequency and voltage control, to regulate
power and provide quick responses during transient and dynamic situations in the
power system. The traditional wind turbines, where the active power is controlled by
a simple pitching of the blades, do not have such control capabilities and cannot
contribute to power system stability as will be required. Power electronic technology
will therefore become more and more attractive for large wind farms that will have to
fulfil future high demands [Ackermann, 2005].
There are currently several research activities in progress in order to develop the
electrical control of such wind farms. Many control methodologies are being
investigated and some are already being implemented in practice. Depending on how
the power electronic devices are used inside a wind farm, there are different topology
options, each with its particular advantages and disadvantages [Ackermann, 2005]:
21
2.2.2.1 Centralised Control Structure
In this type of control structure there is a central power electronic converter. The
advantage of such structure is that the internal behaviour of the wind turbines is
separated form the grid behaviour, and thus the wind farm is robust to possible
failures of the grid. The disadvantage of this concept is that all wind turbines are
rotating with the same average angular speed and not at an individual optimal speed,
therefore giving up some of the features of the variable-speed concept, for each
individual turbine. Furthermore, this type of centralised control structure is difficult to
design, sometimes leading to complex optimisation problems [Ackermann, 2005].
In this type of control methodology, there is a hierarchical structure with both a
central control level and a local control level. The central wind farm control level
controls the power production of the whole farm by sending out reference power
signals to each individual wind turbine, while the local wind turbine control level
ensures that the reference power signal sent by the central control level is reached
[Blaabjerg et al, 2006].
The main trend of wind turbines/wind farms is clearly a variable speed operation and
attention is drawn to wind farms with centralised automatic control, which
intermediates a wind farm production conditioned by the system operators demands
[Blaabjerg et al, 2006].
Another option to these typical wind farm control structures is the use of a high-
voltage DC link as power transmission, where all wind turbines are connected to the
same power converter, and the entire wind farm is connected to the public supply grid
though another power converter. These two converters are connected to each other
through a long link cable [Ackermann, 2005].
22
2.2.3 Requirements for the Interconnection of Wind Farms to the Power System
The integration of large scale wind farms into the grid can have severe impacts on the
power system operation. Traditionally, wind turbines were not required to participate
in frequency and voltage control; however, in recent years, wind farm performance in
the power system has gained attention. Consequently, some grid codes have been
defined to specify the steady state and dynamic requirements that wind turbines and
wind farms must meet in order to be connected to the grid [Chan, 2005].
This section provides an overview of the relevant technical interconnection
regulations for wind power systems. The main requirements, directed towards
distribution network companies, wind turbine manufacturers and network operators,
are:
23
This can be the limiting factor on the amount of wind power which can be installed
[Chen, 2005]. Voltage regulators and the control of reactor power at the generators
and consumption connection points is used in order to keep the voltage within the
required limits and avoid voltage stability problems [Ackermann, 2005].
24
generator transformer tap position, regulation capability, frequency control status,
abnormalities and external control possibilities [Ackermann, 2005].
25
Chapter 3. Wind Turbine Modelling
Modelling is a basic tool for analysis, such as optimization, project, design and
control. Wind energy conversion systems are very different in nature from
conventional generators, and therefore dynamic studies must be addressed in order to
integrate wind power into the power system. Models utilised for steady-state analysis
are extremely simple, while the dynamic models for wind energy conversion systems
are not easy to develop. Dynamic modelling is needed for various types of analysis
related to system dynamics: stability, control system and optimization.
Referring to [Lubosny, 2003], in the case of power systems with classical sources of
energy analysis, the modelling is relatively simple because the models and controllers
of the processes are well known and even standardized; the data are available. But in
the case of wind turbine modelling, researchers face problems related to the lack of
data and lack of control-system structures due to strong competition between wind
turbine manufacturers. This leads to the situation in which many researchers model
the wind energy conversion systems in relatively simple form, almost neglecting the
control systems, which significantly influence the reliability of the analytical results.
Modern wind turbine generator systems are constructed mainly as systems with a
horizontal axis of rotation, a wind wheel consisting of three blades, a high speed
asynchronous generator (also known as induction generator) and a gear box.
Asynchronous generators are used because of their advantages, such as simplicity of
construction, possibilities of operating at various operational conditions, and low
investment and operating costs. The wind turbine under study falls under this category
and is also equipped with a blade pitch angle control system, which enables the power
generated by the wind turbine to be controlled. A typical wind energy conversion
system is displayed in Fig. 8.
26
The wind turbine model, consisting of the aerodynamic, drive train and electrical
generator model is described next. These models are proposed by [Lubosny, 2003],
[Martins et al, 2007] and [Lei et al, 2006]. A list of the wind turbine variables
symbols and units used throughout the model definition can be found in Appendix 2.
As pointed out in Chapter 2, the wind turbine blades extract the kinetic energy in the
wind and transform it into mechanical energy. The kinetic energy in air of an object of
mass m moving with speed v is equal to
1
E= ⋅ m ⋅ v2 (3. 1. 1)
2
The power in the moving air (assuming constant speed velocity) is equal to
dE 1 ⋅ 2
Pw = = ⋅ m ⋅v (3. 1. 2)
dt 2
⋅
where m is the mass flow rate per second. When the air passes across an area A (e.g.
the area swept by the rotor blades), the power in the air can be computed as
1
Pw = ⋅ ρ ⋅ A ⋅ v3 (3. 1. 3)
2
where ρ is the air density. Air density can be expressed as a function of the turbine
elevation above sea level H
ρ = ρ 0 − 1.194 × 10 −4 ⋅ H (3. 1. 4)
where ρ 0 = 1.225 kg/m3 is the air density at sea level at temperature T=298K.
The power extracted from the wind is given by
1
PBLADE = Cp (λ , β ) ⋅ Pw = Cp (λ , β ) ⋅ ⋅ ρ ⋅ A ⋅ v 3 (3. 1. 5)
2
The power factor has a maximum theoretical value equal to Cp=0.593.
The rotor power coefficient is usually given as a function of two parameters: the tip-
speed ratio λ and the blade pitch angle β (in degrees). The blade pitch angle is
defined as the angle between the plane of rotation and the blade cross-section chord.
And the tip speed ratio is defined as
ωm ⋅ R
λ= (3. 1. 6)
v
27
where ω m is the angular velocity of the rotor and R the rotor radius (blade length).
1
⋅ Cp (λ , β ) ⋅ ρ ⋅ A ⋅ v 3
PBLADE
Tw = = 2 (3. 1. 7)
ωm ωm
The area covered by the blades is given by
A = π ⋅ R2 (3. 1. 8)
Substituting Eq. 3.1.8 into Eq. 3.1.7 leads to
1
⋅ π ⋅ Cp (λ , β ) ⋅ ρ ⋅ R 2 ⋅ v 3
Tw = 2 (3. 1. 9)
ωm
Based on previous research, the power coefficient Cp can be defined as a function of
the tip-speed ratio and the blade pitch angle as follows
1
⎛ 1 ⎞ −C 6⋅
Cp (λ , β ) = c1 ⋅ ⎜⎜ c 2 ⋅ − c3 ⋅ β − c 4 ⋅ β x − c5 ⎟⎟e γ (3. 1. 10)
⎝ γ ⎠
with γ defined as
1 1 0.035
= − (3. 1. 11)
γ λ + 0.08β 1 + β 3
while the coefficients c1-c6 are proposed as equal to: c1=0.5, c2=116, c3=0.4, c4=0,
c5=5, c6=21 (x is not used here because c4=0).
According to [Lubosny, 2003] an example of the power coefficient ( Cp (λ , β ) )
characteristics computed taking into account equations 3.1.10 and 3.1.11 and the
above parameters c1-c6 for a given rotor diameter, rotor speed and for various blade
pitch angles β is presented in Figure 9.
28
3.2 Modelling of the drive train
The drive train (mechanical parts) of a wind turbine system in general consists of a
blade pitching mechanism, a hub with blades, a rotor shaft (relatively long in wind
energy conversion systems with asynchronous generators) and a gearbox with
generator. The drive train model presented in this paper includes the inertia of both
the turbine and the generator. The moment of inertia of the wind wheel (hub with
blades) is about 90% of the drive train total moment, while the generator rotor
moment of inertia is equal to about 10%. At the same time, the generator represents
the biggest torsional stiffness.
The acceptable and common way to model the drive train of a wind turbine in power
system operation analysis is based on the assumption of two lumped/masses only: the
generator (with gearbox) mass and the hub with blades (wind wheel) mass [Lubosny,
2003]. The structure of the model is presented in Figure 10.
29
the interaction between the windmill and the rotor, an additional equation describing
the motion of the windmill shaft is adopted
dω m
H m⋅ = Tw − Tm (3. 2. 2)
dt
The mechanical torque Tm can be modelled with the following equation
θ ω g − ωm
Tm = K ⋅ + D⋅ (3. 2. 3)
n n
dθ
= ω g − ωm (3. 2. 4)
dt
where n is the gear ratio, θ is the angle between the turbine rotor and the generator
rotor, ω m , ω g , H m and H g are the turbine and generator rotor speed and inertia
constant, respectively, K and D are the drive train stiffness and damping constants, Tw
is the torque provided by the wind (from section 3.1) and Te is the electromagnetic
torque.
The mechanical power of the wind turbine is converted into electric power by an
alternating current (AC) generator or a direct current (DC) generator. The AC
generator can be either a synchronous machine or an induction (asynchronous)
machine. The latter is most widely used in the wind power industry and was selected
for this project. The electrical machine works on the principle of action and reaction
of electromagnetic induction. The resulting electromechanical energy conversion is
reversible. The same machine can be used as a motor for converting mechanical
power into mechanical power or a as generator for converting mechanical power into
electric power.
As pointed out by [Lubosny, 2003], it is assumed that the asynchronous generator,
also called induction generator, has three-phase stator armature winding (AS, BS, CS)
and a three-phase rotor winding (AR, BR, CR) as shown in Fig. 11. The stator is the
outer stationary member and the rotor is the inner rotating member of the machine.
The rotor is mounted on bearings fixed to the stator. In the electromagnetic structure
of the induction generator, when the stator winding is supplied with three-phase
current (waveforms of equal amplitude, displaced in time by one-third of a period), a
30
rotating magnetic field is produced. The angular speed of the rotating magnetic field
is called the synchronous speed, ω s . The relative speed between the rotating field and
the rotor induces a current in the rotor. The resulting magnetic field interacts with the
stator field to make the rotor rotate in the same direction. In this case, the machine
acts as a motor since, in order for the rotor to rotate, energy is drawn from the electric
power source. However, if an external mechanical torque (in this case the wind torque)
is applied to the rotor to drive it beyond the synchronous speed, then electrical energy
is pumped to the power grid, and the machine will act as a generator [Dorf, 2000].
An induction machine needs no electrical connection between the stator and the rotor.
Its operation is entirely based on electromagnetic induction. The absence of rubbing
electrical contacts and simplicity of its construction make the induction generator a
very robust, reliable, and low-cost machine.
• The stator currents are positive when flowing towards the network.
• The real and reactive power are positive when fed into the grid.
31
• The stator and rotor windings are placed sinusoidally along the air-gap as far
as the mutual effect with the rotor is concerned.
• The stator slots cause no appreciable variations of the rotor inductances with
rotor position.
• The rotor slots cause no appreciable variations of the stator inductances with
rotor position.
• Magnetic hysteresis and saturation effects are negligible.
• The stator and rotor windings are symmetrical.
• The capacitance of all the windings can be neglected.
The set of equations of the asynchronous generator model is usually converted into a
model related to an arbitrarily set reference frame: the machine is converted into the
so-called 0dq reference frame model. The dq axis representation of induction
generator is used for simulation, taking flux linkage as basic variable [Jangamshetti et
al, 2006]. It is based on fifth-order two axis representations. Mathematical
transformations are used in the analysis and simulation of three-phase systems, mostly
to decouple variables, to facilitate the solution of difficult equations with time-varying
coefficients. Park’s transformation [Slemon, 1989] decouples and rotates the stator
variables into a dq reference frame. The positive d-axis of the dq frame is aligned with
the magnetic axis of the field winding, that of the positive q-axis is ahead in the
direction of rotation or lead the positive d-axis by π / 2 . ds and qs correspond to stator
direct and quadrature axes; dr and qr correspond to rotor direct and quadrature axes.
As stated in [Weedy et al, 1998], in electrical engineering the per unit (p.u.) system is
the expression of system quantities as fractions of a defined base unit quantity. These
fractions are called per unit and the p.u. value of any quantity is defined as
32
actual value (in any unit)
value in p.u. =
base or reference value in the same unit
Parameters of electrical generators are often specified in terms of per unit.
Calculations are simplified because quantities expressed as per unit are the same
regardless of the voltage level. Similar types of apparatus will have impedances,
voltage drops and losses that are the same when expressed as a per-unit fraction of the
equipment rating, even if the unit size varies widely. Although the use of p.u. values
may at first sight seem a rather indirect method of expression there are several reasons
for using a per-unit system:
Referring to [Slootweg et al, 2001] it is difficult to calculate the per unit value of the
power extracted from the wind, because aerodynamic and mechanical wind turbine
characteristics such as rotor diameter and wind velocity come into play. Therefore the
asynchronous generator equations are given in the per unit system (p.u.), and the
aerodynamic and drive train equations in the standard international units.
An appropriate model of the induction generator is the most complicated part of the
total wind generation model. The model of such a system is well described in many
books and papers [Karrari et al, 2005].
Two main induction generator models are used when performing power system
dynamic studies [Martins et al, 2007]:
33
• A simplified model which neglects stator transients, containing two
electromagnetic state variables. This model is sometimes referred in the
literature as the third order model, accounting for the two electric state
variables and the generator speed.
• Magnetic fluxes
ϕ ds = X s ⋅ I ds + X m ⋅ I dr (3. 3. 1)
ϕ qs = X s ⋅ I qs + X m ⋅ I qr (3. 3. 2)
ϕ dr = X r ⋅ I dr + X m ⋅ I ds (3. 3. 3)
ϕ qr = X r ⋅ I qr + X m ⋅ I qs (3. 3. 4)
• Voltages
dϕ ds
Vds = − Rs ⋅ I ds + ω s ⋅ ϕ qs − (3. 3. 5)
dt
dϕ qs
Vqs = − Rs ⋅ I qs − ω s ⋅ ϕ ds − (3. 3. 6)
dt
dϕ dr
0 = − Rr ⋅ I dr + s ⋅ ω s ⋅ ϕ qr − (3. 3. 7)
dt
dϕ qr
0 = − Rr ⋅ I qr − s ⋅ ω s ⋅ ϕ dr − (3. 3. 8)
dt
where the sub indexes (s,r) stand for the stator and rotor quantities, respectively, and
the sub indexes (d, q) stand for the components aligned with the d- and q- axis in a
34
synchronous rotating reference frame. Variable ϕ represents the magnetic flux
linkage, V the voltage and I the current. In the case of the traditional induction
machine, the rotor voltage Vdr and Vqr is equal to zero, since the current is only fed
into the stator. Variables ω s and ω g are the synchronous and generator rotor speed,
respectively.
The slip of the rotor, s, is defined as follows
ωs − ω g
s= (3. 3. 9)
ωs
The slip is positive in the motoring mode and negative in the generating mode.
The electric parameters of the machine Rs , X s , X m , Rr and X r stand for the stator
resistance and reactance, mutual reactance and rotor resistance and reactance,
respectively.
The electrical torque is given by
Te = ϕ qr ⋅ I dr − ϕ dr ⋅ I qr (3. 3. 10)
The developed torque Te is positive for motoring operation and negative for
generation operation.
Finally, the wind turbine active, reactive and apparent power output are given by the
following equations
Pactive = Vds ⋅ Ids + Vqs ⋅ I qs (3. 3. 11)
35
neglected. This is common when performing stability simulations [Martins et al,
dϕ ds dϕ qs
2007]. It is done by neglecting terms and in Equations 3.3.5-3.3.6, which
dt dt
is equivalent to assuming infinitely fast electromagnetic transients in the stator
windings.
Rearranging equations 3.3.1-3.3.13 leads to the following simplified model
ϕ ds = X s ⋅ I ds + X m ⋅ I dr (3. 3. 14)
ϕ qs = X s ⋅ I qs + X m ⋅ I qr (3. 3. 15)
ϕ dr = X r ⋅ I dr + X m ⋅ I ds (3. 3. 16)
ϕ qr = X r ⋅ I qr + X m ⋅ I qs (3. 3. 17)
dϕ dr
0 = − Rr ⋅ I dr + s ⋅ ω s ⋅ ϕ qr − (3. 3. 20)
dt
dϕ qr
0 = − Rr ⋅ I qr − s ⋅ ω s ⋅ ϕ dr − (3. 3. 21)
dt
ωs − ω g
s= (3. 3. 22)
ωs
Te = ϕ qr ⋅ I dr − ϕ dr ⋅ I qr (3. 3. 23)
The mathematical equations above consist of a mixed set of ordinary differential and
algebraic equations that express the wind turbine’s physical laws of conservation of
energy and momentum.
36
A state of the art software application, gPROMS, which enables the user to specify
the order of polynomial and the number of points for discretisation of the spatial
domain was used here for dynamic simulation. The software gPROMS is an equation
oriented modelling system used for building, validating and executing models. The
wind turbine model described in sections 3.1, 3.2 and 3.3 was implemented in
gPROMS. Details can be found in Appendix 3.
In table 1, the variables with a known value, the variables with an unknown value and
the number of equations for the wind turbine are given.
Unknown variables Te , Tm , Tw , ω g , ω m , ϕ ds , ϕ qs , ϕ dr ,
ϕ qr , I ds , I dr , I qs , I qr , s , θ , Cp ,
λ, γ , P
Number of equations 19 (Appendix 3)
Table 1. Variables with a known value, variables with an unknown value and number of
equations for wind turbine model
The table above shows that the number of unknown variables equals the number of
equations for the wind turbine model. Therefore the number of degrees of freedom is
equal to zero, and a simulation of the wind energy conversion system can be run.
The wind turbine parameters used for the simulation are given in Table 2.
Parameter Value
Rotor radius, R 25
37
Aerodynamic coefficients, c1-c6 c1=0.5, c2=116, c3=0.4, c4=0, c5=5,
c6=21
Gear ratio, n 65.27
Damping, D 1E6
Stiffness, K 6E7
Synchronous speed, ω s 1
It is important to note that for simulation purposes, the initial conditions were taken as
steady state (all time derivatives equal to zero).
and Vqs were assigned a value of 1000 V. The blade pitch angle β was set to zero in
the base case, which translates in capturing all the available power from the wind. The
simulation was run in gPROMS for 10 hours; results are shown in the following
graphs.
38
W i nd turbi ne currents
400 30
300
20
S t at o r cu r r e n t ( A )
R o t o r cu r r e n t ( A )
200
100 10
0 0
-100 0 5000 10000 15000 20000 25000 30000 35000
-10
-200
-300 -20
-400 -30
T ime ( s)
The electromagnetic torque and the generator rotor slip are plotted in Fig. 13. As
expected, they have negative values in the generating mode.
0 0
0 5000 10000 15000 20000 25000 30000 35000
E le ct r o m ag n e t ic
-1000 -0.2
t o r q u e ( N .m )
S lip r at io
-2000 -0.4
-3000 -0.6
-4000 -0.8
T ime ( s)
39
The wind turbine output power is presented in the figure bellow.
60000
50000
Power (W)
40000
30000
20000
10000
0
0 5000 10000 15000 20000 25000 30000 35000
Time (s)
Power
12 49300
10 49200
W in d V e lo c it y
P o w er (W)
8 49100
( m /s )
6 49000
4 48900
2 48800
0 48700
0 10000 20000 30000 40000 50000 60000 70000
T im e ( s )
40
3.4.2.3 Blade pitch angle step change
The response of the wind turbine towards a blade pitch angle step change is shown in
Figures 16 and 17. Figure 16 illustrates how an increment in the blade pitch angle β
accurately translates in a reduction of the wind power coefficient.
12 0.25
10 0.2
Beta (degrees)
8
0.15
Cp
6
0.1
4
2 0.05
0 0
0 20000 40000 60000 80000 100000 120000
Time (s)
Beta Cp
Figure 16. Power coefficient towards blade pitch angle step change
Figure 17 shows that the angle of the rotor blades can be adjusted in order to shed the
unwanted power. When the wind speed becomes too high, a control structure could
increase the blade pitch angle in order to reduce the aerodynamic power.
12 49500
10 49000
B e t g a ( de g r e e s)
Powe r (W)
8 48500
6 48000
4 47500
2 47000
0 46500
0 20000 40000 60000 80000 100000 120000
T ime ( s)
Beta Power
Figure 17. Power output towards blade pitch angle step change
41
Chapter 4. Wind Turbine Control
As explained in section 2, the energy from the wind is not constant and the wind
turbine power output is proportional to the cube of the wind speed, which causes the
generated power to fluctuate. In general, the electrical power should be smooth and
non-fluctuating. Therefore, in order to reduce fluctuation, a blade pitch control
strategy can be developed. With pitch control, the power captured ( PBLADE ) from the
wind power ( Pw ) can be controlled by a pitch actuator. As the wind speed increases,
the power generated by the wind turbine also increases. Once the maximum rated
power is reached, the pitch angle is increased (pitch-to-feather) to reduce the power
coefficient and hence the aerodynamic power.
As stated in section 3, the inputs for the wind turbine model are the wind velocity v
and the blade pitch angle β . The wind velocity is a disturbance variable; it changes
constantly and cannot be controlled. On the other hand, the blade pitch angle is a
manipulated variable. It can be adjusted in order to reduce the unwanted power when
the wind speed becomes too high.
Classical techniques such as proportional (P), integral (PI) and derivative (PID)
controllers are typically used to regulate the pitch angle of a wind turbine.
Nevertheless, as stated in [Balas et al, 2006], controlling modern turbines to minimize
the cost of wind energy is a complex task, and much research remains to be done to
improve controllers. Referring to [Bemporad et al, 2002], model predictive control
has become the accepted standard for complex constrained multivariable control
problems in the process industries. Although in the 1970s industry started
implementing model predictive control for many type of processes, considerable
research still needs to be carried out regarding the implementation of this type of
control for wind turbines [Brosilow et al, 2002]. The main advantage of this control
strategy is that it takes into account constraints (such as economic considerations or
operating conditions), which are usually not considered by the investigations done in
42
wind energy systems. A model predictive controller predicts the future behaviour of a
process using a reduced model, and finds the control actions necessary for regulating
the process solving an optimal control problem on a receding horizon. The goal of the
optimization problem is to bring the system to the target regulation set point y * in an
optimal way, while satisfying all constraints. The optimization satisfies possible
constraints and bounds on the states y and inputs u . These constraints can include,
for example, limited control authority and bounds on the blade pitch command and its
rates, or maximum values of important parameters of the system that need not be
exceeded. Model predictive control is a form of control in which the current control
action is obtained by solving, at each sampling instant, a finite horizon open-loop
optimal control problem, using the current state of the process as the initial state
[Mayne et al, 2000].
An important drawback of model predictive control is that it requires on-line
computational effort, which limits its applicability to relatively slow and/or small
problems. A solution to the implementation problem of model predictive control is
given by using novel parametric programming techniques. Parametric programming is
an advanced mathematical programming method for solving optimization problems;
model predictive controllers obtained via parametric programming are usually
referred to as parametric controllers or explicit controllers. This technique [Bemporad
et al, 2002] allows to move all the computations necessary for the implementation of
the model predictive control off-line, while preserving all its other characteristics,
thus increasing the range of applicability of model predictive control. The
optimization problem of the model predictive control is solved off-line and the
optimization variable (in this case the blade pitch angle) is obtained as a function of
the parameters of the process such as the output, states and set points. The feasible
region of the parameters space is also obtained.
In general any optimization framework in engineering problems can be described
mathematically as
z (θ ) = min f ( x ) + d T y (4. 2. 1)
x, y
s.t. g (x ) + Ey ≤ b + Fθ
θ min ≤ θ ≤ θ max
43
x ∈ X ⊆ Rn
y ∈ Y ⊆ {0,1}
m
θ ∈Θ⊆ Rs
where y is a vector of binary variables, x a vector of continuous variables, f a
scalar of continuous differentiable function of x , g a vector of continuous
differentiable function of x , θ is a vector of parameters, b and d are constant vectors.
Parametric optimization obtains z and x as explicit functions z (θ ) , x(θ ) of θ and
the regions CR (θ ) where these functions are valid. By treating the process operating
variables as parameters, the optimization problem of the model predictive control is
solved off-line by parametric optimization to obtain the optimal solution as an optimal
mapping of the operating variables. The on-line implementation of model predictive
control is then reduced to a simple function evaluation at each sampling time instant.
The parametric controller design strategy is shown in Figure 18. The main advantage
of using explicit parametric controllers is that as the operating conditions of the
process fluctuate there is no need to re-optimize since the optimal solution is already
given as a function of the set of operating conditions. Hence, parametric controllers
can be implemented on inexpensive hardware with inexpensive and less complicated
software. This has been successfully demonstrated on several continuous processes
[Panga et al., 2005]. More details of parametric control can be found in [Bemporad et
al, 2002] and [Pistikopoulos et al, 2002].
44
This work attempts to design a parametric controller for the power output of a wind
turbine, by regulating the pitch angle of the blades. The parametric controller was
built in MATLAB. For this reason, the wind turbine model was directly transposed to
the MATLAB/Simulink software. An overview of the implemented
MATLAB/Simulink model is given in Figures 19-22. All blocks can be internally
described with their mathematical relations, defined in section 3.
45
Figure 21. Simulink drive train model
The parametric controller was developed by first obtaining a linear input/output ARX
polynomial representation of the wind energy system. This was done by importing the
simulation data of the wind turbine power and blade pitch angle into MATLAB and
representing the data in the system identification toolbox. The sampling time for the
data acquisition was 1 second. The mathematical description of the resulting ARX
model was:
A(q) ⋅ y (t ) = B(q) ⋅ u (t ) + e(t ) (4. 2. 2)
46
q −1 ⋅ y (t ) = y (t − 1) (4. 2. 3)
The modelling mismatch, e(t ) , was neglected for the design of the controller.
The following model predictive controller formulation is considered for the wind
energy system:
Q ( y t − y ref ,t ) + ρ ∑t =1 RΔu t2
2
∑
N N
J = min t =1
(4. 2. 6)
Ut +1,...UN
0 ≤ u t ≤ 90
0 ≤ y t ≤ 50000
Q = 50 , R = 1 , ρ = 10 −3 , N = 5
where y is the controlled variable (wind turbine power output P in W), u is the
manipulated variable (blade pitch angle β in degrees), ρ is a positive optimization
variable incorporated to perform constraint softening in the presence of output
infeasibility, and N (prediction horizon), Q (output cost), R (control moves
suppression weight) are the tuning parameters. The values of the constraint bounds
(determined by the wind turbine design specifications) along with the tuning
parameters are given in Table 3.
Values 0 50000 0 90
Tuning Q R ρ N
parameters
Values 50 1 10 −3 5
47
The specifications of the parametric controller are to reduce the tracking error
between the output and the optimal power and satisfy the constraints imposed on the
parameters of the wind energy system. The only information needed to solve equation
4.2.6 and compute the control action u (t ) ( β (t ) ) are the values of the elements of the
state vector x . This vector comprises the past input and output values and the future
set-point of the output power spanned until the end of the prediction horizon
⎡ β t −1 ⎤
⎢ ⎥
x = ⎢ Pt −1 ⎥ (4. 2. 7)
⎢ Pset − po int ⎥
⎣ ⎦
The solution of this problem is obtained using recently developed algorithms [Dua et
al, 2002] and the Parametric Optimization (POP) software (property of ParOS Ltd). It
consists of a set of explicit linear expressions for the optimal value of the optimization
variable in terms of the parameters and a set of regions in space where those
expressions are valid. This mapping features an output feedback control law for the
wind turbine as it directly relates the control action (current β (t ) ) to the controlled
output ( P ). The solution gave a polyhedral partition of the state-space into 21 regions.
For example in region 1, the expression for the control law (manipulated variable) is:
β = [1 0.0045579 - 0.0045579 ] ,
⎡ 1 0.0045579 - 0.0045579⎤ ⎡ 90 ⎤
⎢ -1 - 0.0045579 ⎥
0.0045579⎥ ⎢ 0 ⎥⎥
⎢ ⎢
⎢ - 2.4358e - 014 1 -1 ⎥ ⎢ 219.4 ⎥
⎢ ⎥ ⎢ ⎥
⎢ 2.4358e - 014 -1 1 ⎥ ⎢ 219.4 ⎥
⎢ -1 0 0 ⎥ ⎢ 0 ⎥
if ⎢ ⎥*x ≤ ⎢ ⎥ (4. 2. 8)
⎢ 1 0 0 ⎥ ⎢ 90 ⎥
⎢ 0 -1 0 ⎥ ⎢ 47034⎥
⎢ ⎥ ⎢ ⎥
⎢ 0 1 0 ⎥ ⎢ 2966 ⎥
⎢ 0 0 -1 ⎥ ⎢ 47034⎥
⎢ ⎥ ⎢ ⎥
⎢⎣ 0 0 1 ⎥⎦ ⎢⎣ 2966 ⎥⎦
A two dimension projection of the controller regions, by fixing Pset − po int , is given in
Figure 23.
48
Figure 23. Projection of the controller regions in Pt −1 − β t −1 space
Similar 2D plots of the polyhedral space partition are given in Figures 24 and 25, by
fixing Pt −1 and β t −1 respectively.
49
Figure 25. Projection of the controller regions in Pref − Pt −1 space
For the controller implementation, the measured output value y (t ) and the values of
the past output/input variables and future set-point are substituted into the set of the
inequalities for each region. If all the inequalities are satisfied, the region where the
parameters lie is identified. Then, the processor substitutes the parameter values to the
obtained optimal control function to calculate the value of the blade pitch angle that
needs to be adjusted.
Since the wind turbine model was built in gPROMS and the parametric controller was
designed in the MATLAB environment, a link was created in order to connect these
two programmes and implement the controller. gO:MATLAB, licensed as an optional
component of the gPROMS family, allows an entire gPROMS model to be called as a
single function from inside MATLAB, enabling to solve a complex set of algebraic
and ordinary differential equations within a single call. gO:MATLAB avoids the need
for existing models to be simplified and rewritten for the MATLAB environment.
Therefore, the validated, detailed gPROMS wind turbine model was called as a
50
MATLAB function for use in control analysis and design. The model was exported
from gPROMS using a simple export facility, which packages the model and all the
solvers and support software required for its solution within MATLAB. For this
purpose, a schedule was created in the gPROMS process entity that introduces the
communication protocols that defined the data to be exchanged with MATLAB
(Appendix 5).
Data exchange was implemented on the gPROMS side using the gPROMS Foreign
Process Interface (FPI). This was carried out by using the FPI GET and SEND
statements:
• The GET statement is used to read data from MATLAB for input to the
gPROMS simulation.
• The SEND statement is used to send simulation results from gPROMS to
MATLAB.
The FPI communication schedule was implemented using a task entity which is itself
called by the Process Entity. Details can be found in Appendix 6.
The gO:MATLAB function was called inside the MATLAB environment using the
syntax: gOMATLAB. The script file is shown in Appendix 7. Figure 26 presents the
gPROMS wind turbine model executing correctly within MATLAB.
51
The explicit parametric controller was successfully implemented using gO:MATLAB.
Results are plotted in Figures 27 and 28. Once the maximum rated power is reached,
the blade pitch angle is increased to reduce the power coefficient and hence the
aerodynamic power. The figures illustrate how in less than one minute the blade pitch
angle is adjusted by the controller in order to set the wind turbine power output to the
reference value.
As explained in section 2, as the wind speed increases, the turbine extracts more
power from the wind. Above 10-15 m/s wind speeds, the power output of the rotor
must be controlled to reduce driving forces on the rotor blades as well as the load on
the whole wind turbine structure.
As shown in the graphs bellow, once the rated power is achieved, the controller
accurately increments the blade pitch angle to shed some of the unwanted power and
prevent the wind turbine from damage.
Power profile
49,500
49,000
48,500
Power (W)
48,000
47,500
47,000
46,500
0 20 40 60 80 100 120 140 160 180 200
Time (s)
52
Blade pitch angle profile
0.012
0.01
Blade pitch angle (degrees)
0.008
0.006
0.004
0.002
0
0 20 40 60 80 100 120 140 160 180 200
Time (s)
53
Chapter 5. Conclusions and Recommendations for future work
A wind energy conversion system consisting of the blades, mechanical parts and
induction generator was modelled. Using the presented model, the output power for
wind turbines was simulated in a simple way in gPROMS. To test the performance of
the proposed model, wind turbine responses both to a step increase in wind speed and
blade pitch angle were simulated. In both cases, the proposed model gave valuable
insight into the performance of the variable speed wind turbine. As expected, the
power generated increases with the wind speed, confirming the need of some sort of
power control. On the other hand, an increment in the blade pitch angle proved to
shed the aerodynamic power. As a normal dynamic simulation time step was adopted,
this model was proven to be computationally efficient.
Based on the obtained rigorous wind turbine model, a blade pitch angle control
strategy for output power levelling was developed. An explicit parametric controller
was formulated using MATLAB and the Parametric Optimization (POP) software.
The controller design was based on an input-output ARX model and a predictive
optimization problem formulation. The controller was derived off-line by recasting
the current and past control input and process output together with the future
reference trajectory as a set of parameters. Then parametric optimization was used to
derive a mapping of the control actions in the parameter space. The solution gave a
polyhedral partition of the state-space into 21 regions. Hence, a simple explicit
optimal control law was constructed that allows the on-line implementation via simple
linear function evaluations. The controller was successfully implemented using
gO:MATLAB, a tool for calling the wind turbine gPROMS model from inside
MATLAB. The simulation results show that the parametric controller performs well,
accurately adjusting the blade pitch angle in order to set the power output to the
reference value, in less than one minute.
Although a comparison of the parametric controller with classical control techniques
such as PI was not made, an important advantage of the parametric controller over
standard controllers is worth mention. The PI controller is designed to operate for the
nominal operating conditions. When a perturbation of the nominal operating
conditions occurs, re-optimization of the PI parameters is required to cope with the
54
new scenario. On the contrary, the explicit parametric controller does not need re-
optimization since the optimal solution is already a function of the set of operating
conditions. Thus, for any change in the operating conditions the explicit parametric
controller can successfully produce a control action to counteract for these changes.
The results obtained during this investigation are encouraging. The potential benefits
of parametric control for wind energy systems were analysed; it appears that this
control method allows for smooth wind turbine operation. However, a comparison of
this control method with classical PI controllers would be desirable to demonstrate if
substantial performance improvement of the parametric controller approach over
standard controllers is possible.
Furthermore, validating the developed wind turbine model and parametric controller
with an experimental real-time implementation on an actual turbine would be very
valuable. Moreover, this study is limited to the available data, and further validation
of the model with other wind turbines, other operating points and different
disturbances would be advisable. For instance, this study can be improved by taking
into account wind gusts and other practical problems.
The inclusion of the power electronics, converter, inverter, rectifier, capacitor,
transformer and grid connection would make the wind energy system model more
complete. Further research includes modelling and control of a group of
interconnected wind turbines or wind farm.
It should be noticed that the design of the parametric controller did not take into
account any robustness requirements. Since this work is a first attempt to develop an
explicit controller for a wind turbine, the design of a robust explicit controller is
recommended as future work.
55
References
56
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
December 2006].
Bouscayrol, A. Delarue, P. Guillaud, X. Lancigu, G. Tounzi, A. (2002) Modelling,
control and simulation of an overall wind energy conversion system.
Renewable Energy, [Online] 28, 1169-1185 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
February 2007].
British Wind Energy Association (2006) The technology of turbines. [Online]
Available from: www.bwea.com/ref/tech.html [Accessed in October 2006].
Brosilow, C. Joseph, B. (2002) Techniques of Model-Based Control. Prentice Hall,
New Jersey.
Butterfield, C. Muljadi, E. (2001) Pitch-Controlled Variable-Speed Wind Turbine
Generation. IEEE Transactions on Industry Applications, [Online] 37, 240-
246 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
December 2006].
Chen, Z (2005) Issues of Connecting Wind Farms into Power Systems. IEEE
Transmission and Distribution Conference & Exhibition: Asia and Pacific.
[Online] 1-6 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
February 2007].
De Montfort University (2007) Wind Energy Training Course. [Online] Available
from: www.iesd.dmu.ac.uk/wind_energy/m32extex.html [Accessed in
February 2007].
Dorf, R. (2000) The Electrical Engineering Handbook. CRC Press LLC, Florida.
EERE (2007) Wind Power. [Online] Available from:
www1.eere.energy.gov/windandhydro/wind_how.html [Accessed in January
2007].
Dua, V. Bozinis, N. Pistikopoulos, E. (2002) A multiparametric programming
approach for mixed integer and quadratic engineering problems. Comp. Chem.
Eng., [Online] 26, 715-733 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed June
2007].
Ekanayake, J. Holdsworth, L. Jenkins, N. (2003) Comparison of 5th and 3rd order
57
machine models for doubly fed induction generator (DFIG) wind turbines.
Electric Power Systems Research, [Online] 67, 207-215 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed July
2007].
Energy Source Guides (2007) Large Wind Turbine Manufacturers in the World by
Business Name. [Online] Available from:
http://energy.sourceguides.com/businesses/byP/wRP/lwindturbine/byB/mfg/b
yN/byName.shtml [Accessed in February 2007].
Fujita, H. Funabashi, T Sekine, H. Senjyu, T. Sakamoto, R Urasaki, N. (2006) Output
Power Levelling of Wind Turbine Generator for All Operating Regions by
Pitch Angle Control. IEEE Transactions on Energy Conversion, [Online] 21,
467-475 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
December 2006].
Gerogiorgis, D. Kouramas, K. Bozinis, N. Pistikopoulos, E. (2006) Design and
Performance of an Explicit Parametric Controller for a Solid Oxide Fuel Cell
(SOFC) System. Presented at the AIChE Annual Meeting. San Francisco,
California, USA.
Harrison, R. Hau, E. Snel, H. (2000) Large Wind Turbines. Wiley, England.
Jangamshetti, S. Naik, R. (2006) Dynamic Simulation Model of Wind Energy
Conversion System. WWEA, [Online] Available from: www.wwec2006.com
[Accessed June 2007].
Jenkins, N. Walker, J. (1997) Wind Energy Technology. Wiley, England.
Karrari, M. Malik, O. Rosehart, W. (2005) Comprehensive Control Strategy for a
Variable Speed Cage Machine Wind Generation Unit. IEEE Transactions on
Energy Conversion, [Online] 20, 415-423 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
January 2007].
Lei, Y. Mullane, A. Lightbody, G. Yacamini, R. (2006). IEEE Transactions on
Energy Conversion, [Online] 21, 257-263 Available from
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed May
2007].
Lubosny, Z. (2003) Wind Turbine Operation in Electric Power Systems. Springer,
Germany.
58
Martins, M. Perdana, A. Ledesma, P Agneholm, E. Carlson, O. (2007) Validation of
fixed speed wind turbine dynamic models with measured data. Renewable
Energy, [Online] 32, 1301-1316 Available from: www.sciencedirect.com
[Accessed March 2007].
Mayne, D. Rawlings, J. Rao, C. Scokaert, P. (2000) Constrained model predictive
control: Stability and optimality. Automatica [Online] 36, 789-814 Available
from www.elsevier.com/locate/automatica [Accessed May 2007].
McNerney, G. Richardson, D. (1993) Wind Energy Systems. IEEE, [Online] 81, 378-
389 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
December 2006].
Npower s (2007) Project Components. [Online] Available from:
www.natwindpower.co.uk/gwyntymor/components.asp [Accessed in February
2007].
Panga, A. Sakizlis, V. Kosmidis, V. Boustras, G. Ross, R. Charles, G. Kenchington, S.
(2005) Output feedback Parametric Controllers for an Active Valve Train
Actuation System. ECC/CDC Conference, [Online] Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed July
2007].
Pantelides, C. (2005) The Mathematical Modelling of the Dynamic Behaviour of
Process Systems. Imperial College Lecture Notes, London.
Pistikopoulos, E. Dua, V. Bozinis, N. Bemporad, A. Morari, M. (2002) On-line
optimization via off-line optimization tools. Computers and Chemical
Engineering, 26 (2), 175 – 185.
Rawlings, J. (1999) Tutorial: Model Predictive Control Technology. ACC, [Online] 1,
662-676 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed
February 2007].
Scottish Executive (2007) [Online] Available from: www.scotland.gov.uk [Accessed
in January 2007].
Slemon, G. (1989) Modelling Induction Machines for Electric Drives. IEEE
Transaction on Industry Application, [Online] 25, 1126-1131 Available from:
www3.imperial.ac.uk/library/digitallibrary/electronicjournals [Accessed June
2007].
59
Slootweg, J. Polinder, H. Kling, W. (2001) Initialization of Wind Turbine Models in
Power System Dynamic Simulations. IEEE Power Tech Proceedings, [Online]
4, 6-11 Available from: www.ieeexplore.ieee.org
Wasynezuk, O. Yi-Min, D. Krause, P. (1985) Theory and comparison of reduced
order models of induction machines. IEEE Transaction Power Apparatus
Systems, [Online] 3, 598-606.
Weedy, B. Cory, B. (1998) Electric power systems. Wiley, England.
Wind-Works (2007) [Online] Available from: www.wind-
works.org/photos/PhotosVAWTs.html [Accessed in February 2007].
60
Appendices
Appendix 1
61
Appendix 2
D damping Nm.rad-1.s-1
K stiffness Nm.rad-1
θ angle between turbine rotor and generator rotor rad
n gear ratio no units
ϕ magnetic flux p.u.
I current amperes
V voltage volts
s slip ratio no units
62
X inductance p.u.
R reactance p.u.
ωs synchronous speed p.u.
P apparent power W
63
Appendix 3
#####################################################################
Model Entity
###################################################################################
PARAMETER
Rho AS REAL
R AS REAL
Rs AS REAL
Xs AS REAL
Xm AS REAL
Rr AS REAL
Xr AS REAL
Ws AS REAL
Hg AS REAL
Hm AS REAL
K AS REAL
D AS REAL
C1 AS REAL
C2 AS REAL
C3 AS REAL
C5 AS REAL
C6 AS REAL
VARIABLE
Te AS Torque
Tm AS Torque
Tw AS Torque
Wg AS AngularSpeed
Wm AS AngularSpeed
FLUXds AS Flux
FLUXqs AS Flux
FLUXdr AS Flux
FlUXqr AS Flux
Ids AS Current
Idr AS Current
Iqs AS Current
Iqr AS Current
Vds AS Voltage
Vqs AS Voltage
S AS Ratio
Theta AS Angle
Cp AS Ratio
V AS Speed
Lamda AS Ratio
Beta AS Ratio
P AS Power
Q AS Power
64
Gama AS Ratio
EQUATION
# Induction Generator model
FLUXds=Xs*Ids+Xm*Idr;
FLUXqs=Xs*Iqs+Xm*Iqr;
Vds=-Rs*Ids+Ws*FLUXqs;
Vqs=-Rs*Iqs-Ws*FLUXds;
FLUXdr=Xr*Idr+Xm*Ids;
FLUXqr=Xr*Iqr+Xm*Iqs;
0=-Rr*Idr+S*Ws*FLUXqr-$FLUXdr;
0=-Rr*Iqr-S*Ws*FLUXdr-$FLUXqr;
Te=FLUXqr*Idr-FLUXdr*Iqr;
S=(Ws-WG)/Ws;
P=Vds*Ids+Vqs*Iqs+Vqs*Ids-Vds*Iqs;
#####################################################################
Process Entity
###################################################################################
UNIT
Turbine AS windturbine
SET
Turbine.Rs :=0.0121;
Turbine.Xs :=0.0742;
Turbine.Xm :=2.7626;
Turbine.Rr :=0.0080;
Turbine.Xr :=0.1761;
Turbine.Ws :=1;
Turbine.Hg :=35.184;
Turbine.Hm :=1.6E6;
Turbine.K :=6E7;
Turbine.D :=1E6;
Turbine.R :=25;
Turbine.n :=65.27;
Turbine.Rho :=1.225;
Turbine.C1 :=0.5;
Turbine.C2 :=116;
Turbine.C3 :=0.4;
Turbine.C5 :=5;
65
Turbine.C6 :=21;
ASSIGN
Turbine.V :=10;
Turbine.Beta :=0;
Turbine.VQS :=1000;
Turbine.VDS :=1000;
INITIAL
STEADY_STATE
SOLUTIONPARAMETERS
REPORTINGINTERVAL :=100;
SCHEDULE
Continue for 3600*10
66
Appendix 4
% Number of inputs
mpc.nu = 1;
mpc.bMismatch = 0;
mpc.bLimitDU =[1];
mpc.DU_up = [1];
mpc.DU_dn = [1];
mpc.wu = [0];
mpc.yridx=[1];
% Q value
mpc.wy=[50];
% R value
mpc.wdu = [1];
% Rho value
mpc.rho = [1e-3];
mpc.termul = 10;
mpc.lin_mv= [0];
% mpc.e_penalty = [0];
% mpc.sp_prof{1}=[{1:mpc.hy}];
% mpc.sp_prof{1}=[{1:mpc.hy}];
67
m.B = [-219.4 219.4]; % Enter B(q) here
m.na = [1]; % Enter order of A(q)
m.nb = [2]; % Enter order of B(q)
m.nk = [0]; % Enter time delay here
arxm = m; % The arx model is stored with the name arxm: this is needed later
68
Appendix 5
UNIT
Turbine AS windturbine
SET
Turbine.Rs :=0.0121;
Turbine.Xs :=0.0742;
Turbine.Xm :=2.7626;
Turbine.Rr :=0.0080;
Turbine.Xr :=0.1761;
Turbine.Ws :=1;
Turbine.Hg :=35.184;
Turbine.Hm :=1.6E6;
Turbine.K :=6E7;
Turbine.D :=1E6;
Turbine.R :=25;
Turbine.n :=65.27;
Turbine.Rho :=1.225;
Turbine.C1 :=0.5;
Turbine.C2 :=116;
Turbine.C3 :=0.4;
Turbine.C5 :=5;
Turbine.C6 :=21;
ASSIGN
Turbine.V :=10;
Turbine.Beta :=0;
Turbine.VQS :=1000;
Turbine.VDS :=1000;
INITIAL
STEADY_STATE
SOLUTIONPARAMETERS
gRMS := OFF;
FPI := "eventFPI";
ReportingInterval := 1 ;
SCHEDULE
sequence
gMATLAB(wind is Turbine)
end
69
Appendix 6
PARAMETER
wind as MODEL windturbine
SCHEDULE
SEQUENCE
while True Do
sequence
GET
wind.Beta;
END
continue for 1
SEND
wind.A;
END
end
end
END
70
Appendix 7
gOMatlab('startONLY');
gOMatlab('select', 'varpitch', 'varpitch');
gOMatlab('simulate', 'varpitch');
Time=0;
tempu=u;
tempy=y+47033.97;
while Time <=200
i=1;
while i<=M
if psol(i).cr.A*[u;y;2193]<=psol(i).cr.b
Uout=psol(i).X*[u;y;2193;1];
disp ('***********find feasible region************')
count=count+1;
break;
end
i=i+1;
end
ReturnMatrix(1,:)=gOMATLAB('evaluate',[Uout(N)],L);
y=ReturnMatrix(1,1)-47033.97; % -min(y)
u=Uout(N);
tempu=[tempu u];
tempy=[tempy y+47033.97];
Time=Time+ee;
end
plot(tempu);
figure
plot(tempy);
count
Time
gOMATLAB('stop');
71