See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/257132143

Modelling and control of wind turbines

Conference Paper · May 2009

CITATIONS READS

3 2,336

3 authors, including:

Pierre Borne Dumitru Popescu

École Centrale de Lille 139 PUBLICATIONS 510 CITATIONS
693 PUBLICATIONS 6,607 CITATIONS
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Andreea Pintea on 19 May 2014.

The user has requested enhancement of the downloaded file.

 Modelling and control of wind turbines
Andreea Pintea, Pierre Borne
Dumitru Popescu

University “Politehnica” of Bucharest

Faculty of Automatic Control and Computer Engineering
Splaiul Independentei 313, Bucuresti
Cod Postal 060042, ROMANIA
dpopescu@indinf.pub.ro

Abstract: Nowadays, the development of electrical power generation from wind currents is a big concern
as well for the society energy issue and for the management of electrical power systems. As wind
turbines prove to be one of the cheapest, cleanest and most efficient sources of energy, it has become of
great necessity to focus on complex algorithms to meet with multiple objectives such as speed regulation,
blade load and mode stabilization with simultaneously maximizing energy capture. In this paper we
present some theoretical results concerning different approaches and control of wind turbines. The study
focuses on variable-speed wind turbines and presents by comparison, the improvement in performance
offered by each control method.
Keywords: Wind power, renewable, energy, turbine, pitch control.

1. INTRODUCTION the rotor is used instead. This leads to the conclusion that a
variable speed turbine is a multivariable system for which a
multi-objective command is required.
Wind energy has proved to be an important source of clean This paper focuses on variable speed turbines control
and renewable energy, as no fossil fuels are burnt to generate methods due to the advantages obtained with such a turbine.
electricity from wind energy. The advantages of wind Among these advantages, the generation of more energy at a
turbines usage made necessary the design of different control given wind speed and lack of grid voltage fluctuations are the
systems in order to improve wind turbines behavior and make most important and determined this paper’s research
them more reliable and efficient. direction.
As the wind is the energy source, it is of great importance to The form, in which the delivered energy is controlled,
be able to generate predictions regarding the wind behavior. depending on the wind speed, determines the working region
In order to better understand the dynamics of a wind turbine of the variable speed turbine. We can classify these working
it is necessary to determine the characteristics of the wind regions into three different areas as shown in Figure 1.
currents that pass through the turbine rotor.
The stochastic nature of the wind determines the necessity of
a wind turbine to be able to work under different wind
velocities, on either which certain conditions are to be
considered.
In order to keep a wind turbine’s performance within these
conditions, controllers must be designed and implemented.
Due to the unpredictable nature of the wind, the controllers
used must be robust enough as to be able to perform properly
even when wind bursts appear.
The output power of the turbine strongly depends in a non-
linear form on the wind speed, the rotation speed of the
turbine and the pitch angle of the blades. The designed
controller must be able to adjust the torque of the generator
and as well the pitch angle of the blades in order to adapt the
rotational speed of the turbine which activates the rotor. It is
well known that wind speed and direction vary according to Figure 1. The working regions of a variable speed turbine
geographical area and surface aspect, that is why it is very The first is region corresponds to the minimum operational
difficult to use exact wind speed measured at a given moment wind speed for the turbine. That is, if the wind speed is lower
in time and so, an average value of the wind speed traversing that a certain threshold (usually around 5m/s), then it is not
worth to turn on the turbine because the energy consumed is
higher than the one produced. (This area is also known as
First Partial Load Area).
The second region covers from its minimum operational
speed up to the speed where the maximum safe electric
power can be given by the generator (around 14m/s). In this
area, the power delivered to the generator is controlled by
adjusting the torque that is given to the generator. This area is
also known as the Second Partial Load zone.
The third region goes from the wind speed where the
maximum safe electrical power is given up to the cut off
speed, which is the speed at which is no longer safe to run the
wind turbine (around 25 m/s). This is the area in which the
pitch angle of the blades is adjusted in order to maintain the
maximum power output without exceeding the threshold
(Full Load Area). Figure 3. Wind speed spectrum
K
2. THEORETICAL BACKGROUND Φ v (ω ) = (1)
(1 + (T )
5

v ⋅ ω) 2 6
The wind turbine cannot generate unlimited power due to
physical limitations. Some of these limitations are the Wind currents are considered orthogonal on the turbine’s
tolerable rotational speed of the shafts and the maximum rotor.
power the generator can produce before getting damaged. Control algorithms use a linearization of the turbulent
In Figure 2 one can observe a simplified energy conversion component of the wind, which is given by a first order filter
system scheme [1]. on which the white noise, mv(t) is applied. [3]:
1
v&t = − ⋅ ⋅ vt (t ) + mv (t ) (2)
Tv
This equation shows that the turbulent component can be
modeled as a linear state variable of the system and helps in
generating, in a simple way, the control law.
The power spectrum that corresponds to the linear model is:
K
Φ v (ω ) =
(
1 + (Tv ⋅ ω ) 2 ) (3)

and it represents a suitable approximation on equation (1).

The time constant Tv of the model (2) and the white noise
variance mv(t) depend on the average wind speed vm and the
Figure2. Energy conversion system [1] characteristics of the turbine location. [2]
L
As it can be observed, the energy conversion system is de- Tv =
vm
composed in several sub-components that are to be analyzed
and modeled for a more suitable control of the system. σ m = kσ ,v m
The wind speed in a fixed point in space is characterized in
the frequencies domain and is represented in Figure 3.
The shape models the van de Hoven spectrum and it The mechanical power received by the turbine, Paero, depends
highlights the presence of two components of the wind speed. on the air density, wind speed and power coefficient Cp.
One of the components is the average value, vm(t), it has a 1
low variation in. Paero = ⋅ ρ ⋅ π ⋅ R 2 ⋅ v 3 ⋅ C p (4)
2
The second component is the turbulent component, vt(t). Where R is the radius of the area covered by the blades, v is
The turbulent speed is the high frequency component and its the wind speed and ρ is the air density.
power spectrum model is given by von Karman The power coefficient, Cp is a non-linear function of the
expression[2][3]: blade pitch angle β and the λ parameter, which is the ration
between the peripheral speed of the blades and wind speed. In
Figure 4, one can observe the power coefficient variation
with the relative rotational speed λ.
ωT ⋅ R
λ= (5)
v
where ωT is the rotational speed of the rotor.
The power coefficient Cp(λ, β) of the turbine can be In the second partial load area, the control system must adjust
expressed through a polynomial formula: the rotational speed of the turbine to its nominal value in
C p (λ , β ) = ∑a
i , j =1..4
ij ⋅ λi ⋅ β j
(6) order to obtain a maximum output power. The value of the
pitch angle β is maintained at a fixed value (βopt) for which
the aerodynamic torque has also a maximum value. The
control method only uses the electromagnetic torque of the
generator for the adjustment of the rotational speed.
A possible control block diagram for this situation is
presented in Figure 6.

Figure 6. Control Structure for Partial Load 2 area

Finally, in the Full Load area, the control system has to

maintain the output power value to the nominal value of the
Figure 4. The power coefficient of a variable speed turbine
generator. Through this, the rotational speed of the turbine is
equal to its nominal value, while the pitch angle and
The power coefficient is very important because it gives
electromagnetic torque are varied in order to obtain:
information upon the aerodynamic efficiency of the turbine.
Given the fact that a wind turbine has three functioning Pnom
C p (λ , β ) =
regimes, each having distinct characteristics, lead to the idea ρ
that the control objectives should be different for each area. ⋅π ⋅ R 2 ⋅ v3
In the first partial load area, the main objective is to maintain
2
This area corresponds to high wind speed values and
the turbine at the maximum yield. The control system must
important mechanical solicitation of the system. This is why,
act on the electromagnetic torque in order to adapt the
in this area, multivariable approaches are used.
rotational speed of the rotor, ωT to the wind speed, in order to
have an optimized power coefficient (Cp, max = Cp(λopt, βopt)) ,
and therefore an optimized specific speed λopt = (ωT R)/v. 3. CONTROL METHODS
One control solution is based on an estimation of wind speed
starting on an estimation of the aerodynamic torque of the 3.1 SISO CONTROLLER WITH POLE PLACEMENT
rotor. This solution is proposed in [7] and allows the
calculation of the torque estimator by taking into Given the complexity of a wind turbine system, many control
methods have been proposed and implemented, each having
consideration the inertial forces of the shaft transmission
system (Figure 5). both advantages and disadvantages.
Thus, the referential rotational speed of the rotor is: The most used controllers for wind turbine control are based
on PI/PID control algorithms, which are calculated from
^
linear models of the system around a functional point,
C aero
ωT ,ref = , where Ĉ aero is the aerodynamic torque. situated on the trajectory characteristics of the turbine (Figure
k 1).
Usually, the aerodynamic torque that drives the wind turbine These controllers usually tend to focus on only one control
rotor and thus the generator is given by: objective and that is obtaining of a maximum energy yield
[1]. PID controllers are one of the most used controllers for
1 ρ ⋅ R 2π ⋅ v 3 ⋅ C P (λ , β )
C aero = ⋅ (7) wind speed and power control. This is due to their simplicity
2 ωT and rather high robustness. The small number of parameters
makes possible for designer to quickly arrive at satisfactory
system behavior.
The PID control method proposed in [8] must ensure the
desired behavior of the closed loop system, so in such a
manner that maximum power output is obtained and a
reducing in structural loads and fatigue is achieved. As
turbine towers grow in height, tower oscillations cannot be
ignored. In this situation, a model of the wind turbine with
Figure 5. Block Diagram representing a control distributed parameters is required and a description in terms
method in Partial Load I area of mass and stiffness distribution. Also, the non-linear model
of the wind turbine takes into account the fact that the wind The pitch controller HPC has to assure that closed loop
speed seen by the rotor is a sum of wind speed (v) and tower transfer function (9) is equal to the chosen model transfer
nodding speed ( x&t ). function H0.
The choice of model transfer function is the crucial step in
By using modal analysis, tower dynamics can be described controller design. The method proposed by [8] is based on
as: increasing the modal damping D up to a desired value D’.
M ⋅ &x&t + D ⋅ x&t + C ⋅ xt = F , where M, D and C are modal This new value is introduced in the model, a new
mass damping and stiffness respectively and F is the linearization is performed and transfer functions (10), (11)
generalized force that causes wind tower oscillation. are re-calculated. Using these transfer functions, a PID
In this case, the expression (7) of the aerodynamic torque can controller was designed, under the following conditions: the
be re-written as: phase margin must be around 60◦, system frequency
1 1 bandwidth is of 1 rad/s.
C aero = ⋅ ρ ⋅ R 3 ⋅ π ⋅ C P (λ , β ) ⋅ ⋅ (v − x&t ) 2 (8) With the controller’s transfer function one can calculate the
2 λ closed loop transfer function, which is regarded as desired
model transfer function H0 (H0 = HD_w).
A pitch controller HPC is designed to assure that the closed The system response with SISO pole placement in simulation
loop system behaves in the manner given by the pole can be seen below (Figure 8, 9 and 10):
placement method. The linearization of the wind turbine
model is needed (Figure 7)

Figure 8. Wind speed used for simulation

Figure 7. SISO control block diagram

From the block diagram given in figure 7, closed loop

transfer function with respect to wind speed change can be
obtained:
Δω ( s ) H w ( s)
HD_w = = (9)
Δv( s ) 1 + H PC ⋅ H SD ⋅ H β
Hβ(s) and Hw(s) are transfer functions resulted by
linearization of the aerodynamic torque expression (8) and Figure 9. Response of pitch angle of the system controlled
the formulas are: with PID controller
Δω ( s )
Hβ = (10)
Δβ ( s )

Δω ( s )
H w ( s) = (11)
Δv( s )
These transfer functions are of the third order. [8] HSD is the
transfer function of the system that controls the position of
the servo drives that orient the blades through pitch control.
Position control of servo drives is usually achieved using
frequency converters. For simulation analysis, HSD is
modeled as a second order system:
β (s) ω 2 nβ
H SD ( s ) = = 2 .
β ref ( s) s + 2 ⋅ ζ β ⋅ ω nβ ⋅ s + ω 2 nβ Figure 10. Response of rotor speed of the system controlled
with PID controller
 3.2 FULL STATE FEEDBACK CONTROLLER u = (r − K ⋅ x) , where r(t) is a vector of desired state
variables and K is referred to as the state feedback gain
Another method presented in [8] uses a full state feedback
matrix.
controller instead of a PID controller. The closed loop desired
If we make the substitutions, we obtain:
behavior is also chosen with the pole placement method.
For this purpose, equation (8) has to be re-written in the state x& = A ⋅ x + B ⋅ (r − K ⋅ x)
space form: x& = ( A − BK ) ⋅ x + B ⋅ r where (A-BK) is the closed loop
x& = A ⋅ x + B ⋅ u system matrix.
y = C ⋅ x + D ⋅u The characteristic equation of the closed loop system is given
by:
Then, the poles of the open loop system are the roots of the
characteristic equation, given by: s ⋅ I − A + B ⋅ K = 0 and its roots are the closed loop poles
s⋅ A− I = 0 or eigenvalues.
In order to compute the matrix K, we use the Ackerman
In order to find an appropriate state-controller one must first equation and so, the matrix of feedback gains for selected
choose the state variables of the system. states is calculated [8].
Figure 11 shows a classical diagram of a full state control Ackermann’s formula gives:
system.
K = [0 0 ... 0 1] ⋅ M −1 ⋅ Φ( A) , where M is the
controllability matrix of the system
M = [B AB ... A n−1 B]
and:
Φ ( A) = A n + α n−1 A n−1 + ... + α1 A + α 0 I (12)
In equation (12), A is the system matrix and αi are the
coefficients of the desired closed loop characteristic equation.
The values of these coefficients are given by the poles
placement method and are the same values that were used
with the PID controller method presented at the beginning of
this section.
Figure 11. Block Diagram of full state control system The desired controller is characterized by this vector of
feedback gains, and its design is representing the major
State variables used for system description and control are difficulty of the method.
rotor speed ωT, tower top speed x&t and tower top acceleration The state-feedback control scheme assumes complete
knowledge of the system state. In a real world application,
&x&t .The reason why these parameters are chosen is because is system state knowledge will have to be obtained by use of a
important to have both tower oscillations information (given state estimator like a Kalman filter, for instance.
by top speed and acceleration) and about rotor speed (which Simulation results [8] show that full state feedback controller
also needs to be controlled and is represented by the rotor maintains good rotor speed regulation and also offers a good
speed parameter ωT). response of the pitch angle of the system (Figures 12, 13). As
The next step is establishing the input parameters of the a plus, this controller is achieving better damping and it
system and they are wind speed v (that determines tower practically removes the oscillatory movement of the tower.
oscillations and the rotor to move), pitch angle β (that
influences the incidence of the rotor with the wind direction)
and electromagnetic generator torque Mg, which opposes the
driving torque Caero. We put all these parameters in two
arrays as it is shown below:
⎡ω ⎤
⎢ ω& ⎥ ⎡ v ⎤
x = ⎢ ⎥ , u = ⎢⎢ β ⎥⎥
⎢ x&t ⎥
⎢ ⎥ ⎢⎣ Mg ⎥⎦
⎣ &x&t ⎦
Rotor speed and tower acceleration are measured variables
while the other two states are derived from them. This
concludes in full access to the state of the system.
We go further and we propose a control law of the form
Figure 12. Response of rotor speed of the system controlled
with a full state feedback controller
 This equation serves for R and S polynomial computation.
This is done through factorization and matrix computation in
order to obtain performances in regulation.
For the calculation of T(q-1), one must assure an unitary static
gain between the generated trajectory and the output of the
system.
Finally, the RST command will result in:
T (q −1 ) R ( q −1 )
u (k ) = ⋅ r (k ) − ⋅ y (k ) , where r(k) is the
S (q −1 ) S (q −1 )
discrete reference and y(k) represents the output of the
system.
The goal was to maintain a constant electrical voltage,
produced by the turbine, and this can be expressed in terms of
constant angular velocity of the turbine rotor. Therefore, we
Figure 13. Response of pitch angle of the system controlled used the angular position of
with full state feedback controller. We wanted a robust controller for our system and so we
improved the classical RST controller. In a first stage we
As it can be observed from the equations above, the model of have computed a controller with imposed tracking (ω = 1.21
the open loop system (fixed part HSDHβ) is of 5-th degree. It rad/s ξ = 0.8) and then we have established the regulation
is well known that PID solutions are suitable for maximum performances which include open loop (ω = 2.51 rad/s ξ = 1
second order systems. 0.8) and an additional real pole with multiplicity 5.
This makes the classical PID controller insufficient for the As one can observe from Figure 15, the system tracks the
closed loop control of the turbine structure and an RST reference, the perturbation is rejected. The overshot is very
controller is proposed. This type of controller offers good small and can be neglected. The raising time is approximately
performances both with reference trajectory tracking of 3.6s. As for the robustness margins, the obtained values
(reference variation) and with regulation. were: Gain = 11.88dB, Phase =67.8◦, and Modulus = -
The classical scheme of a system with an RST controller is 3.19dB.
(Figure 14):

Figure 14. System controlled with a RST controller

This controller design method makes it possible to specify

the desired tracking behavior (changing of reference) by the
choice of the tracking dynamics (Am, Bm polynomials) and
the desired regulation behavior (rejection of an output
disturbance) by the choice of the regulation dynamics, Figure.15 Pitch angle of the system controlled with RST
defined by the system poles in closed-loop. controller
The R, S, T blocks of the controller can be written in a
polynomial form:
4. CONCLUSIONS
−1 −1 − nr
R (q ) = r0 + r1 ⋅ q + ... + rnr ⋅ q
This paper has presented modeling techniques and control
S (q −1 ) = s0 + s1 ⋅ q −1 + ... + sns ⋅ q − ns methods proposed for variable speed wind turbines.
Given the fact that the obtained model for such a turbine is
T (q −1 ) = t 0 + t1 ⋅ q −1 + ... + t nt ⋅ q − nt nonlinearly dependant on wind speed, three different
operation regimes are to be considered. Each regime has its
The design method remains the poles placement with own particularities and this leads to specific demands for
independent objectives in tracking and reference trajectory. wind turbine control. The difficulties in wind turbine control
This is realized through the characteristic polynomial P(q-1): involve both the necessity of maintaining the output of the
generator at a value which must correspond to maximization
P (q −1 ) = A(q −1 ) ⋅ S (q −1 ) + B (q −1 ) ⋅ R(q −1 ) of captured energy and reducing mechanical oscillations of
the structure that supports the turbine.
All this makes the controller design a very difficult task. In
this paper two approaches were presented, with respect to
classical control methods: a SISO control method and a state
analysis and controller design. Both methods have shown a
good regulation of rotor speed and a good response of the
pitch angle, but what differentiates the two methods is the
improvement offered by the full-state feedback controller on
the tower oscillations of the structure.
In order to compensate un-linear behavior and parameter
variation of HSD and Hβ in time, we propose the use of the
RST controller, which is designed in a way that is specific to
robust-adaptive systems.

REFERENCES

[1] LESCHER F., BORNE P., ZHAO J.Y, Commande LQG

multimodele d’une turbine eolienne a vitesse variable, PhD
Thesis, LAGIS – Ecole Centrale de Lille, 2006.

[2] EKELUND T., Modeling and linear quadratic optimal

control of wind turbines, PhD thesis, Chalmers University of
Technology, 1997.

[3] NICHITA C., LUCA D., DAKYO B., CEANGA E.,

Large band simulation of the wind speed for real time wind
turbine simulators, IEEE Transactions on Energy Conversion
vol.17, 2002

[4] KHETTACHE L., Etude et Commande d’un Systeme

Eolien a Base d’une Machine Electrique Double Alimentee,
Laboratoire d’Electronique de Batna (LEB), 2007.

[5] BOUKHEZZAR B., SIGUERDIDJANE H., Robust

multi-objective control of a variable speed wind turbine,
Master Thesis, SUPELEC, 2007.

[6] VANEGAS F.A., ZAMACONA M., Robust Control

Solution of a wind turbine, A simulation Study, International
Master’s Thesis in Information Technologies, Halmstad
University, February 2008.

[7] LEITHEAD W.E., CONNOR B., Control of variable

speed wind turbines: design task. Int. Journal of Control,
13:1189-1212, 2000
[8] JELAVIC M., PETROVIC I., Design of a Wind turbine
pitch controller for loads and fatigue reduction, Electrical
Engineering Institute, Proceedings of the European Wind
Energy Conference & Exhibition - EWEC, Milan, Italy,
2007.

View publication stats