You are on page 1of 5

Disturbance Accommodating LQR Method Based Pitch Control Strategy for

Wind Turbines

Jianlin Li
1
,Hongyan Xu
2
,Lei Zhang
1
, Zhuying
1
, Shuju,Hu
1

1
Institute Electrical Engineering of Chinese Academy of Science, Beijing, 100190,China
2
Hebei polytechnic university,Tangshan 063009,China
ljl@mail.iee.ac.cn
Abstract

A disturbance accommodating Linear Quadratic
Regulator (LQR) method was applied in pitch control
system to achieve good performance. The disturbance can
be estimate by designing state estimator, and a feedback
was added into the input to eliminate disturbance effect.
The feed back matrix was calculated in accordance to
LQR theory. A wind turbine dynamic modal was set up,
and simulation of the control system was preformed based
on Matlab7.1/simulink. The simulation results show that
the controller ensure pitch control actuator little fatigue,
and has smaller overshoot. The proposed method is has
better performance and easy to realize.

1. Introduction

Since the 1990s, the wind energy industry has been
growing rapidly. The wind power generation technology
had developed from stall-controlled to variable speed
pitch regulated. And wind turbine has demanded better
performance of controller [1-3].
With the increasing of capacity of wind turbines, pitch-
control technique of large wind turbine has become a key
technique of wind energy. Pitch-control can not only
output power steadily, but also make wind turbine have
better starting and braking performance. Additionally,
using optimized control algorithm can lower load and
torque ripple of wind turbine, extending the life of wind
turbine. At present, in China most wind turbine is
controlled by PID algorithm, which cannot have a satisfy
effect. Abroad researchers have proposed many advanced
control theory and strategy about pitch-control. Senjyu, T
et al had applied GPC control method to pitch-control [1].
This is wind speed predict model based on average wind
speed and standard deviation, having pitch controlled
according to predicted wind speed.
This paper analyzed the process of pitch-control, built a
wind turbine dynamic model and studied a LQR optimal
control algorithm based on disturbance correction.
Adopting this kind of LQR algorithm to perform pitch-
control can not only optimize output power, but also
decrease variable propeller pitch mechanism wear. At last,
this dynamic model was simulated in Matlab 7.1/simulink.

2. Simulation of wind turbine
The equivalent model of wind turbine is shown in Fig.
1
gen
J
rot
J
r
T
shaft
T
gen
T

gen

d
K
d
C
frot
K
fgen
K

Fig .1 Wind Turbine model
The aerodynamic torque gained by blade from wind
energy [5]:
2
2
) , (
2
1
V
C R
T
P
r


=
(1)
in which,

is the density of airKg/m3,R is the


radius of rotor (m),V is the wind speedm/s,

is the
pitch angle degree ,

is tip speed ratio
V R / =
,

is the rotor speed,


p
C
is power conversion
coefficient, which indicates wind turbines efficiency of
converting wind energy to usable mechanism power.
p
C

is function of tip speed ratio

and blade pitch angle

.
p
C
can be written as [5, 6]
i
e
i
P
C


5 . 22
) 5 4 . 0
116
( 22 . 0 ) , (

=
(2)
in which
i

satisfies
1
3
035 . 0
08 . 0
1 1
+

+
=


i

Although wind turbine is a nonlinear model, at some
Second International Symposium on Intelligent Information Technology Application
978-0-7695-3497-8/08 $25.00 2008 IEEE
DOI 10.1109/IITA.2008.247
766
Second International Symposium on Intelligent Information Technology Application
978-0-7695-3497-8/08 $25.00 2008 IEEE
DOI 10.1109/IITA.2008.247
766
Second International Symposium on Intelligent Information Technology Application
978-0-7695-3497-8/08 $25.00 2008 IEEE
DOI 10.1109/IITA.2008.247
766
point near by it can be treated as linear model. Linearizing
torque Tr at point
) , , (
0 0 0
V
nearby :
+ + + = V V T T
r r
) , , (
0 0 0
(3)
In which,
0
=
,
0
V V V =
,
0
=

)
0
,
0
,
0
( )
0
,
0
,
0
( )
0
,
0
,
0
(
, ,




=
V
r
V
r
V
r
T T
V
T

Let state variable q1 and q2 are blade angle and rotor
angle respectively (calculated in low speed shaft. Tshaft is
the reaction torque on the shaft. Then:
) ( ) (
2 1 2 1
q q C q q K T
d d shaft
+ =
(4)
) ( ) (
2 1 2 1
q q C q q K T
d d shaft
+ =
(5)
=
rot
f shaft r rot
K T T q J
1

(6)
gen
gen
f gen shaft gen
K T T q J =
2

(7)
Above,
d
K
is elastic coefficient of propeller shaft,
d
C

is damping coefficient on propeller shaft,
rot
J
and
gen
J

are rotation inera of low speed side and generator (
calculated in low speed side),
rot
f
K

rot
f
K
are friction
coefficient of low speed side and high speed side
respectively.
0
shaft
T
is counter torque at working point
) , , (
0 0 0
V
. The speed acceleration is 0, so
0
0
0 0 0
) , , ( + =
rot
f shaft r
K T V T
(8)
Then:
=
rot
f shaft r rot
K T T q J
1
(9)
Let
2 3
2 1 2
1 1
) (
q x
q q K x
q x
d

=
=
=

Then:
V x C x x K C x J
d
rot
f d rot
+ + + =
3 2 1 1
) (
(10)
) (
3 1 2
x x K x
d
=
(11)
According to the torque equation of generation
gen
gen
f d d gen
T x K C x x C x J + + =
3 2 1 3
) (

(12)
In state equation form

+ =
+ + =
u x y
u u x x
D
D C
B A
(13)
where


=
gen
f d
gen gen
d
d d
rot
d
rot rot
f d
J
K C
J J
C
K K
J
C
J J
K C
gen
rot
1
0
1
) (
A

=
gen
rot
J
J
1
0
0 0
0

=
0
0
rot
J


[ ] 1 0 0 = C

0 D =
input
[ ]
gen
T u = ,
disturbance quantity
V u
D
=

At present pitch actuator has hydraulic and electric two
forms. For simplicity, pitch actuator can be simplified to a
first-order inertia model, no matter it is hydraulic or
electric actuator. The pitch actuator transmission function
is:
1
1
) (
+
=
s
s Act

(14)
3. Pitch control strategy based on LQR
After connected to the grid, wind turbine can work in
two modes: one mode is when wind speed is slower that
rated wind speed, another is when faster. When wind
speed is slow, wind turbine output power is smaller than
rated power. So the pitch angle is set to 0 and wind
turbine runs in optimal tip speed by controlling generator
speed, in order to absorb as much wind energy as possible.
While wind speed is faster than rated speed, the output
power will excess rated power. Because the electrical and
mechanical limitation of wind turbine, the rotator speed
and output power cannot excess rated value. So, when
output power is larger than rated power, pitch angle
should be increased to smaller wind energy utilization
efficiency. When output power is smaller than rated
power, pitch angle will be decreased to maintain the
output power at about rated power nearby.
Nowadays variance speed pitch-control wind turbine
always has its electromagnetic torque given value constant,
maintaining output power by regulating generator speed.
The most common method is adopting PI control to
regulate generator speed. This method is simple and easily
applied in engineering. However, PI control may have
overshoot problems, which makes pitch actuator
complicated and easily fatigued.
LQR is linear quadrics regular, whose control object is
linear system given by state space form in modern control
theory. And its object function is object states and
quadrics function which controls input. LQR optimal
control is designing state feedback controller G. In order
to minimize the quadrics object function J , and also G is
767 767 767
decided only by weight matrix Q and R, the selection of Q
and R is very important. LQR theory is a relatively mature
theory in modern control theory. It provided an efficient
analysis method for multi-variable feedback system.
Object function J included state variable and input
variable, which requires state variable and input variable
to be small. In the pitch-control system, input value is the
error of pitch angle. Because of large inertia of blade,
rapid pitch-control would damage pitch regulated
mechanism and aggravate the friction of pitch-control
shaft. So, having some limitation to input energy will be
reasonable. Additionally, choosing torque variation as
state variable can suppress torque ripple as much as
possible in LQR optimal control. Then the life of wind
turbine can be extanded.
Set linear length-determined systems state function as :

=
+ + =
x y
u u x x
D
C
B A
(15)
Object function
[ ]

+ =
f
t
t
T T
dt t Ru t u t Qx t x J
0
) ( ) ( ) ( ) (
2
1
(16)
where Q is positive semidefinite matrixR is positive
definite matrixQ and R are weighted matrix for state
variable and input variable respectively. x(t) is n-
dimension state variable u(t) is m-dimension input
variable. According to control theory, in order to
minimize object function, optimal control is:
) ( ) ( t Gx t u =

P B R G
T 1
=

Where P is Riccati function
0
1
= +

Q P B PBR P A PA
T T
(17)
Positive definite symmetric solution. The LQR control
diagram is shown as Fig. 2. In engineering application,
state variable cannot be measured usually. So it needs to
design a state observer to estimate state variable value.
Fig. 2(b) is the diagram used in actual application.
x

Fig. 2 LQR control theory diagram
Because there is a disturbance variable ud in wind
turbine model, only using LQR control cannot regulate
generator speed very well. And the disturbance from
disturbance variable should be minimized as much as
possible. Disturbance Accommodating Control (DAC) is a
good method to solve this problem. DAC control method
was proposed by Johnson (1976), DAC control is a
reconstructed disturbance model method based on state
observer. The disturbance variable is reconstructed and is
part of state feedback, can decrease or neutralize the
disturbance effect. This paper adopted LQR method with
DAC, which means that through LQR optimal control
having a optimal feedback matrix G, then using DAC
method to estimate disturbance variable and eliminating
the disturbance from disturbance variable. DAC diagram
is shown as Fig. 3Using state observer to estimate state
variable and disturbance variable, disturbance can be
eliminated.
x

D
z

u input

Fig. 3 disturbance correction control diagram
Presume the disturbance variable has forms as below:

= =
=
D
D D D
D D
z z t z t z
t z u
0
) 0 ( ); ( ) (
) (
F

(18)
z0D is unknowpresume

and F is already known.


According to DAC control theory, state feedback should
contain the feedback of disturbance:
) ( ) ( ) ( t z t x t u
D D
G G + =
(19)
Replace u(t) in the state function with the up function,
we have:
) ( ) ( ) ( ) ( ) ( t z t x t x
D
BG BG A
D
+ + + =
(20)
To elimilate the disturbance, it requires
0 BG
D
= +
then it can be considered as a system
without disturbance. If system cannot
satisfy
0 BG
D
= +
, then choosing GD to make
BG
D
+
minimum.
Because state variable x(t) and zD(t) cannot be
measured directly, designing state observer is needed to
predict state variable and disturbance variable. Wind
turbines state observers math model:

= =
+ + + =
0 ) 0 ( ); (
)) ( ) ( ( ) ( ) ( ) (
x t x y
t y t y t u t u t x x
d
C
K B A
x

(21)
Disturbance observer:

=
+ =
) (
)) ( ) ( ( ) (
t z u
t y t y t z z
D D
D D

K F
D

(22)
Designing appropriate Kx and KD can let:
768 768 768
0 )) ( ) ( ( ) (
0 )) ( ) ( ( ) (
= =
= =


t z t z im l t e im l
t x t x im l t e im l
D D
t
D
t
t
x
t

(23)
Disturbance state function can be written as:
) ( ) ( ) ( t e t e C K A =
(24)
Where,
[ ]
T
T
D
T
x
e e t e = ) (

=
F 0
A
A

[ ] 0 C C =

=
D
x
K
K
K

According to the formula above, errors expression can
be solved:
) 0 ( ) (
) (
e e t e
t C K A
=
(25)
If system
) ( C A
is measurable, then
) ( C K A
can
have any poles configuration, letting
) (t e
damping to 0
rapidly. Feedback control principal became:
) ( ) ( ) ( t z t x t u
D D
G G + =
(26)
Simulation result
To verify the control performance of LQR algorithm
based on disturbance correlation, a numeric simulation
was performanced on Matlab 7.1/simulink. The wind
turbine model parameter is : rated power 650kW, rotor
diameter 43m, gear box transmission ratio 43.16, rotor
rated speed 42 rpm. LQR algorithm based on disturbance
correction and PI regulation method were simulated.
Choosing work point at
s m V / 17
0
=

rpm 42
0
=

5 3 . 13
0
=
in LQR algorithm and linearizing at this point.
Then wind turbines state function is function (13), where:




=


0624 . 0 10 56 . 1 10 56 . 1
10 69 . 2 0 10 69 . 2
10 108 . 3 10 108 . 3 198 . 0
5 4
7 7
5 6
A

0
0
10 5 . 7
3
B

choosing
1 = R

=

50 0 0
0 10 1 0
0 0 1
12
Q

From matrix A, B, Q and R, state feedback matrix:
[ ] 1.3289 - 10 1.6905 2.2219
-8
= K
In the simulation, wind speed stepped from 17m/s to
18m/s at t=0 moment. In PI regulation, K
p
=8, K
I
=1.5,
simulation result is shown as Fig. 4.

0 20 40 60 80
1810
1815
1820
1825
1830
1835
1840
t/s
g
e
n
e
r
a
t
o
r

s
p
e
e
d
(
r
p
m
)


LQR
PI

(a) Generator rotating speed
0 20 40 60 80
3.4
3.45
3.5
3.55
3.6
3.65
3.7
3.75
3.8
x 10
4
t/s
d
r
i
v
e
-
t
r
a
i
n

t
o
r
s
i
o
n
a
l

s
p
r
i
n
g

f
o
r
c
e
(
N
)


LQR
PI

(b) Elastic force on drive link
0 20 40 60 80
13
14
15
16
17
18
t/s
p
i
t
c
h

a
n
g
l
e

(

)


LQR
PI

(c)Pitch angle
Fig .4 simulation waveform of LQR algorithm base on disturbance and
PI control
From the simulation we can tell, PI regulation method
has a lager overshoot, while LQR algorithm has a much
smaller one. In Fig. 4(b), LQR algorithm can decrease the
elastic force on drive link. In Fig.4(c), after adopting LQR
algorithm, the overshoot can be very small, which can
reduce the action of pitch actuator. While PI regulation
has a larger overshoot, pitch angle fluctuated for a
moment, which is harmful for pitch actuator.
Conclusion
So as to enhance pitch control performance of large
769 769 769
wind turbine, this paper constructed wind turbines
dynamics model, giving a LQR pitch-control algorithm
based on disturbance correction according to LQR control
theory. This method can reduce pitch actuators
movement efficiently and has good control performance
on generator rotate speed. The simulation results showed
that this method has good dynamics performance and it is
simple and effective. So it has the great potential to be
applied into engineering.
Acknowledgement
It is a project supported by China Postdoctoral Science
Foundation (No. 20060390092).
References
[1] Senjyu, T.; Sakamoto, R.; Urasaki, N.; Funabashi, T.
Output power leveling of wind turbine Generator for all
operating regions by pitch angle control [J]. Energy
Conversion, IEEE Trans. 2006(21) , pp. 467 475
[2] Ryosei Sakamoto, Tomonobu Senjyu, Tatsuto Kinjo.
Output Power Leveling of Wind Turbine Generator for All
Operating Regions by Pitch Angle Control[C]. Power
Engineering Society General Meeting, 2005. IEEE, 2005, 1,
pp. 45-52.
[3] Ye Hangzhi. Control technology of wind turbine [M]
Beijing: Mechanics Industry Press2002
[4] Fu Wangbao, Zhao Dongli, Pan Lei. Cutting-in Control of
the VSCF Wind-power Generator Based on Auto-
disturbance Rejection Controller [J]. Proceedings of the
CSEE,2006, 26(3), pp.13-18.
[5] Rajib Datta V. T. Ranganathan. Variable-Speed Wind
Power Generation Using Doubly Fed Wound Rotor
Induction MachineA Comparison With Alternative
Schemes[J]. IEEE Trans. on Energy Conversion, 2002, 17,
pp. 414-421.
[6] Kanellos, F.D.; Hatziargyriou, N.D. A new control scheme
for variable speed wind turbines using neural networks[C].
Power Engineering Society Winter Meeting, 2002. IEEE,
2002, 1: 260-365.
[7] Schinas N. A., Vovos N. A., Giannakopoulos G. B. An
Autonomous System Supplied Only by a Pitch-Controlled
Variable-Speed Wind Turbine. IEEE Trans. on Energy
Conversion, 2006, 21(1), pp. 1-7
[8] H.S.Ko , K.Y. Lee, H.C. Kim. An intelligent based LQR
controller design to power system stabilization[J]. Electric
Power Systems Research. vol. 71. 2004, pp. 19
[9] Chao Hu, Max Qinghu Meng, Peter Xiaoping Liu ,
Observer Based LQR Control of Shaping Process of
Automobile Belt. Proceedings of the 5th World Congress
on Intelligent Control. June 2004, China, pp. 3310-3314.



770 770 770

You might also like