Professional Documents
Culture Documents
1 Introduction
Wind energy is by far the fastest-growing renewable
energy resource. In 2012, the European Union
installed capacity of wind turbines reached
105 000MW. The energy available in the wind is
obtained through the wind conversion process,
which is strongly nonlinear and very challenging
from the control system viewpoint as the power
obtained from the wind is proportional to the third
power of the effective wind speed. The control
system have to be able to optimize the wind energy
conversion by application of appropriate generator
torque in weak wind and also manipulate the pitch
blade angle to keep the output power at nominal
value in strong winds. The effective control strategy
also reduces structural fatigue and load affecting the
drive-train and tower structure, leading to
potentially longer lifetime of the wind turbine [1].
Several methods have been developed to control the
wind turbines. Conventional PI controllers have
been tested in [2]. In [1] multivariable approach to
the control of wind turbines in the context of gain
scheduling techniques is applied. Soliman et al.
developed a Multiple Model Predictive controller to
control variable speed variable-pitch wind turbine
over its full operating range [3]. Robust control
techniques of the same type of wind turbine have
been investigated in [4], [5]. To cope with nonlinear
behavior of the wind turbine system, adaptive
control approaches have been proposed in [6].
Review of the control techniques for horizontal axis
wind turbines is presented in [7]. This work presents
application of the MIMO predictive control with
ISBN: 978-960-474-316-2
97
(9)
where variable describes the twist of the shaft,
and
are the rotor and generator inertia,
respectively. The effective wind velocity is modeled
superimposed by
as a mean wind velocity
turbulent wind velocity .
(10)
0.5
0.4
cp
0.3
0.2
0.1
0
30
100
20
50
10
0 0
, ,
0
0 0
0 0
where
is the blade pitch angle setpoint. The
mechanistic part of the turbine can be divided into
rotor and generator side connected by a shaft. The
dynamic nature of the shaft is described by the
and spring constant
. The torques
damping
acting on each side of the transmission are related
by the gear ratio
. Using the Newtons second
law for rotating bodies the following equations can
be formulated to describe the dynamics of the
mechanical part of the wind turbine:
0
0
0
0
0
0
(13)
The slow varying element of the wind velocity
is
assumed to be known. The fuzzy global model of
the process is obtained through piecewise
linearization. The nonlinear system:
x (t ) f ( x(t ), u (t ))
(14)
y (t ) f ( x(t ), u (t ))
(7)
(8)
ISBN: 978-960-474-316-2
0 0
0 0
0 0
0
98
0
0
1 0
0 1
0
0
(19)
0 0
(20)
i 1
i 1
i 1
i 1
i 1
i 1
(28)
T
J MPC Y W Q Y W URU +
(29)
U
U
S
U
U
s
s
with input constraints:
(30)
(23)
(31)
1,2, ,
1,
is the control horizon,
where
(24)
is the estimation covariance error matrix,
is the measured output vector,
is the
Kalman gain. The covariance error matrix is
obtained as a solution of algebraic Riccati equation:
where
(25)
ISBN: 978-960-474-316-2
C = pi Ci , D = pi Di , F = pi Fi
,
|
A = pi Ai , B = pi Bi , E = pi Ei
(27)
99
Th
he wind turb
bine can be ooperated in four differennt
reg
gimes (Fig. 3).
3 In the reggion I the rottational speeed
of the rotor is at its low
west allowab
ble level. Thhe
con
ntrol strategy
y is to keep the pitch off the blades at
a
thee optimum value and control thee system by
b
maanipulating the
t generatorr torque . In the regioon
II the rotationaal speed of the rotor
is within thhe
mits ,
and the blade pitch is
i
lim
,
kep
pt at the optimum value prroviding thhe
maaximization of . The m
maximal valu
ue of is foor
blaade pitch
0 and tip sp
speed ratio
7.5.
5 Implem
mentation
The
T proposedd predictive control schem
me is presennted
in
i Fig. 2.
Fig.
F 2 Predictive control scheme
1.5
Based
B
on thee measured outputs an
nd inputs the
states
s
of the system arre estimated using multiiple
Kalman
K
filterrs whose outtputs are weiighted using the
fuzzy
f
weightts scheduled with mean wind speed
which
w
is asssumed know
wn and estim
mated turbullent
speed
s
. Thhe actual paarameters off the linearizzed
model
m
are obbtained usingg (30). The setpoints
s
aand
steady-state
s
inputs
foor computatiion of criterrion
(31)
(
are alsoo computed for actual effective w
wind
speed.
s
The
T wind tuurbine that is
i being con
nsidered in the
paper
p
is a w
wind turbinee model wid
dely known as
NREL
II.
[rad/s]
I.
IV.
1
III.
0.5
P [[MW]
10
0
6
x 10
10
15
2
20
25
10
15
2
20
25
0
4
x 10
10
15
2
20
25
10
15
2
20
25
[deg]
40
20
0
5
T [N.m/s]
/ ]
4
3
2
1
0
v [m
m/s]
Fig
g. 3 Operatio
on modes of tthe NREL wind
w
turbine
Table
T
1 NRE
EL wind turbine parameteers
parameter
nominal pow
wer ,
rated rotor sppeed ,
drive-train sppring const.
drive-train daamp. const.
generator ineertia
rotor inertia
blade radius
gear ratio
max blade piitch
min blade pittch
max bl. pitchh rate
min bl. pitch rate
max gen. torqque ,
min gen. torqque ,
max
geen.
torquue
rate ,
min
genn.
torquue
rate ,
pitch actuatoor const.
gen. actuatorr const.
ISBN: 978-960-474-316-2
units
[MW]
[rad/s]
[N.m/rad]]
[N.m/rad.s]
[kg.m2]
[kg.m2]
[m]
[-]
[deg]
[deg]
[deg/s]
[deg/s]
[N.m]
[N.m]
[N.m/s]
value
5
1.2671
867.637E
E6
6.215E6
534.116
3.8768E77
63
97
90
0
8
-8
47402.977
0
15000
[N.m/s]
-15000
[s]
[s]
0.12
0.1
he primary ob
bjective is too keep tip speeed ratio at itts
Th
op
ptimal valuee
to
t maximizze
turrbines aerod
dynamic effi
ficiency. In the
t region IIII
thee rotational speed
s
of the rrotor is at its maximum
m
bu
ut generated power
is below its nominal valuee.
In this transition region iss system is controlled
c
thhe
sam
me way as in
i the regionn II. In the top
t region IV
V
bo
oth the rotaational speedd of the rotor
annd
gen
nerated pow
wer
are at their rated values.
v
In thiis
reg
gion the torq
que is kept cconstant and blade pitch
is used
u
to comp
pensate the vvariation in wind
w
power.
Th
he wind turbiine is controolled in all reegions. As thhe
dy
ynamics of the
t system is different for differennt
wiind speeds a local m
model was obtained by
b
lin
nearization of the nonlineear model of the process.
21 local models were deeveloped forr mean winnd
25 . The paarameters of
o the system
m
speeed 4,2
bettween these local modelss are obtaineed using fuzzzy
meembership function
f
as shown on Fig. 4. Thhe
sam
mpling perio
od is choseen to be 0.02
25s.
100
M1
(41)
,
,
(42)
The weights for the MPC were also scheduled using
fuzzy membership function and their values for
different wind speeds are presented in Table 2.
M21
M2
p [-]
0.8
0.6
0.4
0.2
0
-0.2
10
15
20
25
v [m/s]
(32)
where
0.001 0
0 0 0
0
0.1
0 0 0
0
0 1
0
0
(34)
0
0 0 0.001 0
0
0 0
0
0.1
0
0
0
0
0
0
0
0
0
0
0
0
0
/
0
0
0
0
/
(35)
0
0
0
0
0
0
0
0
0
0
/
0
0
/
0
0
0
0 0 0 0 0 0
0 0
(36)
The measurement covariance matrix was set to [13]:
v [m/s]
10
50
100
150
T [s]
200
250
300
200
250
300
200
250
300
1.2
1.1
1
0.9
(37)
0.8
0
0
0
Model predictive controller was employed for
computation of manipulated variables. Prediction
and control horizons were set to:
2 ,
1
(38)
For such values acceptable performance was
obtained and these values does not represent big
computational burden for the processing unit. The
ISBN: 978-960-474-316-2
15
5
0
12-25 m/s
10
0
0
0
10 0
0 0
0 0
0 0
0
10
0
10
0
10
20
r [rad/s]
11m/s
10
0
10
0
10
0
(33)
4-10 m/s
10
0
0
10
10
0
0
10
10
0
0 10
50
100
150
x 10
Pe [MW]
Pe,norm
2
0
-2
0
50
100
150
Time [s]
and rotor
Fig. 6 Wind turbines power output
in simulation with turbulent wind
speed
101
References:
[1] F.D. Bianchi, H. Batista, R.J. Mantz, Wind
Turbine
Control
Systems:
Principles,
Modelling and Gain Scheduling Design,
Heidelberg: Springer, 2007.
[2] A.D. Hansen, P. Sorensen, F. Iov, F. Blaabjerg,
Control of variable speed wind turbines with
doubly-fed induction generators, Wind Eng.,
Vol. 28, pp. 411443, 2004.
[3] M. Soliman, O.P. Malik, D.T. Westwick,
Multiple Model predictive Control for Wind
Turbines With Doubly Fed Induction
Generators, IEEE Transactions on Sustainable
Energy, Vol. 2 (3), 2011, pp. 215-225.
[4] M. Mirzaei, N.K. Poulsen, H.H. Niemann,
Robust Model Predictive Control of a Wind
Turbine, Proceeding of the American Control
Conference, 2012, pp. 4393-4398.
[5] H. Camblong, Digital robust control of a
variable speed pitch regulated wind turbine for
above rated wind speeds, Control Engineering
Practice, vol. 16, no. 8, 2009, pp. 946-958.
[6] S.A. Frost, M.J. Balas, A.D. Wright, Direct
Adaptive Control of a Utility-Scale Wind
Turbine for Speed Regulation, Int. Journal of
Robust Nonlinear Control, Vol. 19 (1), 2009,
pp. 5971.
[7] J.H. Laks, L.Y. Pao, A. D. Wright, Control of
Wind Turbines: Past, Present, and Future,
Proceedings of the American Control
Conference, 2009, pp. 2096-2103.
[8] J.D. Grunnet, M. Soltani, T. Knudsen, M.
Kragelund, T. Bak, Aeolus Toolbox for
Dynamic Wind Farm Model, Simulation and
Control, Proceedings of the 2010 European
Wind Energy Conference, 2010
[9] S. Thomsen, Nonlinear Control of a Wind
Turbine. Masters thesis, Technical University
of Denmark, 2006.
[10] D. Simon, Optimal State Estimation: Kalman,
H-Infinity, and Nonlinear Approaches, WileyInterscience, 2006.
[11] J. Maciejowski, Predictive Control With
Constraints, Prentice-Hall, 2000.
[12] J.M. Jonkman, S. Butterfield, W. Musial, G.
Scott, Definition of a 5-MW reference wind
turbine for offshore system development,
Technical report, NREL, 2009.
[13] A. Gosk, Model Predictive Control of a Wind
Turbine, Masters thesis, Technical University
of Denmark, 2011.
30
25
[deg]
20
15
10
5
0
0
4
x 10
50
100
150
200
250
300
50
100
150
Time [s]
200
250
300
Tg [N.m/s]
4
3
2
1
0
0
opt
[-]
7
6
5
4
0
50
100
150
T [s]
200
250
300
6 Conclusion
In the paper, fuzzy predictive control concept is
applied to the simulated wind turbine model. It was
assumed that measurement noise was present and
not all the states are measurable to make the control
more realistic. The weight scheduled predictive
controller was implemented for control of the wind
turbine in all operating regimes in the turbulent
wind conditions. The MPC controller enables the
desired tradeoff between power smoothing and
speed regulation while reducing generator torque
fluctuations and pitch actuator activity. The use of
future wind speed measuring device as LIDAR
might improve the performance of the control
system when implemented in the predictive control
structure. The simulation on 5MW wind turbine
model shows that the strategy can be applied to the
wind turbine effectively.
Acknowledgement
The authors kindly appreciate the financial support
which was provided by the European Regional
Development Fund under the project CEBIA-Tech
No. CZ.1.05/2.1.00/03.0089.
ISBN: 978-960-474-316-2
102