I.J.

Intelligent Systems and Applications, 2017, 5, 62-73

Published Online May 2017 in MECS (http://www.mecs-press.org/)
DOI: 10.5815/ijisa.2017.05.08

Comparative Analysis of Pitch Angle Controller

Strategies for PMSG Based Wind Energy
Conversion System
Ramji Tiwari
School of Electrical Engineering, VIT University, Vellore, India-632014
E-mail: [email protected]

Ramesh Babu. N
School of Electrical Engineering, VIT University, Vellore, India-632014
E-mail: [email protected]

AbstractThis paper proposes an advanced pitch angle and grid integration [2].
control strategy based on neural network (NN) for Based on the wind speed range, the operating regions
variable speed wind turbine. The proposed methodology of WECS are classified in to two operating modes as
uses Radial Basis Function Network (RBFN) and Feed- shown in Fig. 1. Below rated wind speed ( Vrated ) region
forward based Back propagation network (BPN) is termed as partial-load region where the turbine speed
algorithm to generate pitch angle. The performance of the of the system is adjusted to the optimal value to extract
proposed control technique is analyzed by comparing the maximum power from the available energy. The region
results with Fuzzy Logic Control (FLC) and Proportional where the wind speed is above the rated wind speed is
- Integral (PI) control techniques. The control techniques full-load region where the generator power is controlled
implemented is able to compensate the nonlinear to maintain the optimum power so that the generated
characteristic of wind speed. The wind turbine is power does not exceed the rated capacity of turbine and
smoothly controlled to maintain the generator power and generator 3. In this paper the pitch angle control is used
the mechanical torque to the rated value without any in both the partial-load and full-load region for power
fluctuation during rapid variation in wind speed. The smoothening and power limiting respectively thus
effectiveness of the proposed control strategy is verified maximizing the power in low wind speed region and
using MATLAB/Simulink for 2-MW permanent magnet stabilizing during high wind speed region.
synchronous generator (PMSG) based wind energy The pitch angle is mostly preferred for the full-load
conversion system. region to limit the aerodynamic torque as suggested in the
previous literature [3]. Several methods have been
Index TermsWind energy conversion system, implemented by researchers in the past. The conventional
Permanent magnet synchronous generator, Pitch angle, pitch angle control such as proportional-integral (PI) or
Fuzzy logic, Back propagation, neural network, Radial proportional-integral-derivative (PID) based controller
basis function network. have been most common and preferred for small-WECS
for power stabilization. The major disadvantage of this
controller is its failure to track the non-linearity of the
I. INTRODUCTION rapid change in the wind speed. The response time of the
Recently, the emphasis on demand for clean and system is very slow which affects the performance of the
sustainable energy increases manifold on the WECS since they are based on the operating points of the
development of renewable energy sources. Among the turbine [4].
renewable energy, wind energy is the most viable and The pitch control strategy based on H [5] and sliding
promising source for its recent advancements and mode controller [6] is implemented in previous literature.
remarkable development in past decade [1]. Especially, They provide a good performance of the output power
the enhancements of wind energy conversion system and robustness towards the variation in wind speed.
(WECS) were more focused on the low-cost, high However, the control strategy of these techniques are
efficient and reliable power generation which led to a complex since they purely depend on the parameter of
development of large scale wind turbines. The variable wind turbine and they need to be redesigned when the
speed variable pitch wind turbine is mostly preferred parameter changes. The other robust pitch angle
wind turbine for its reliability and efficient conversion of controller used in the literature is the linear quadratic
wind power to electrical power. Consequently, the Gaussian (LQG) [7]. The performance of LQG method is
advancement in the control techniques of WECS is deteriorated due to nonlinear parameters of wind turbine.
gaining lot of importance in terms of power stabilization The generalized predictive control (GPC) [8] is proposed

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System 63

to operate in the entire wind speed region. They use the A. Wind Turbine
output power error as the input for the controller, thus
The torque obtained from the wind turbine is expressed
during large output error the controller become unstable.
as [12],
To overcome the above issues, soft computing based
controller is preferred.
1 C p ( , ) 3 2
Tm = R v (1)
2

Partial-load region Full-load region
Where, is the air density kg m3 , C p is power
Nominal Power

coefficient and the parameters of C p are and which

Output power (kW)

Maximum Rotor Reduced Rotor

efficiency efficiency represents the tip speed ratio and pitch angle respectively.
No generation No generation
The wind speed is represented as v (m s ) and R
represents the radius of the blade (m) .

The power coefficient C p of the wind turbine is
expressed as,

Cut-in Speed Rated Speed Cut-out Speed

3 m/s 12 m/s 20 m/s c2 c5
Wind Speed (m/s) C p ( , )= c1 * c3 c4 * exp i + c6
(2)
i
Fig.1. Operating regions of wind turbine
Where,
Fuzzy logic controller (FLC) is gaining lot of interest
for its simplicity and adaptability. The FLC technique
1 0.035
adapts the nonlinearity of the system swiftly and operates i =
in stability region [9]. The FLC is purely depend on the +0.08 1+ 3
prior knowledge of the system, thus a parameter variation
affect the overall performance of the system. Moreover And c1 to c6 are the parameter coefficients with c1=
the information of wind speed is required in FLC which 0.5176, c2=116, c3=0.4, c4=5, c5=21, and c6=0.0065
in turn increases the cost of the system [10]. which is used in this paper.
In this paper, a radial basis function network (RBFN) The power coefficient of wind turbine with respect to
based pitch angle is proposed to limit the output power tip speed ratio is shown in Fig. 3.
and the generator speed in both the operating region. A
B. Pitch Actuator
detailed analysis of fuzzy logic controller and back
propagation neural network based pitch angle controller The pitch angle control is generally used to limit the
is performed and compared with the RBFN control turbine mechanical torque of wind turbine in large wind
strategy. Generated speed and output power are adopted turbine. The blades are adjusted in such a way that the
as the input for the control strategy which eliminates the torque from the wind turbine is always at optimal range
use of anemometer for wind speed measurement. Thus by [13]. The electric pitch angle is preferred over hydraulic
using these parameters as inputs for control technique the pitch controller for its simplicity and efficiency [14]. The
turbine is maintained at the rated values which reduces angle of the blades is adjusted in order to maintain the
the ripple of the generated power. Permanent magnet turbine torque. The pitch actuator system consists of
synchronous generator (PMSG) based wind generator integrator and time constant c . The pitch actuator of
with the capacity of 2MW is used in this study. The wind turbine is expressed as [15],
simulation result shows the effectiveness of the proposed
method. d 1 1
= + ref (3)
dt c c
II. MODELING OF WIND ENERGY CONVERSION SYSTEM
Which is subjected to,
The configuration of gearless wind generation system
is shown in Fig. 2. The direct driven PMSG is preferred d d d
mostly because it eliminates the use of gear box thus
reducing the complexity and overall size of the system dt min
dt dt max
[11]. The speed of the wind generator is controlled using
the pitch actuator. The variable pitch angle is fed into the Where, min and max are the minimum and
wind turbine in order to obtain optimal torque for
stabilised power generation. maximum pitch angle, respectively.
The configuration of pitch control system is shown in
Fig. 4. The time constant of the pitch actuator determines
the response time of the pitch controller. The response

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
64 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System

time of the pitch controller is usually between 0.2s to an adverse effect on the performance of the controller
0.25s. Blade angle for typical WECS lies between -2 to [16].
30 degrees. Thus the variable range in the pitch angle has

v Tr g
Aerodynamic Mechanical
PMSG Pg
model Model
r Tg

v Pg
Pitch angle
control
g

Fig.2. Typical configuration of gearless wind generation system

C. PMSG Generator

Power Coefficient (Cp)

The mechanical torque (Tm ) and electrical torque (Te )
of three phase PMSG generator used in this study are
expressed as [17],

Pm
Tm = (4)
r
Tip speed ratio
Pe
Te = *2 (5) Fig.3. CP vs curve of wind turbine
r n p
Where, Lmd , I fd are the mutual inductance and
Where, Pm , Pe are mechanical and electrical power
magnetizing current in d axis. s is the stator frequency,
obtained respectively. r is the mechanical rotational which is represented as,
speed and n p is the number of pole pair used in PMSG.
In general, the dynamic motion of PMSG generator is s = n pr (9)
given as,
The PMSG generator does not require any drive train
dr and they are used for variable speed variable pitch
Te = J eq + Dr +Tm (6)
dt application [18].

Here, D represents the rotational damping and J eq

III. PITCH ANGLE CONTROL TECHNIQUES
represents the equivalent inertia of wind turbine and
generator. Pitch angle control technique is implemented in order
The machine model of PMSG in d-q reference frame to regulate the output power of wind generator. The pitch
are described as, control provides full control over the mechanical power
of wind turbine in the entire operating region. In order to
q = Riq + pq + s d optimize the output power, four pitch angle control
(7) techniques are implemented and described below.
d = Rid + pd s q
A. PI controller
where, d , id represents the stator voltage and current PI controller is the conventional pitch control strategy
in the d axis and q , iq in q axis respectively. R denotes which is used to regulate the turbine speed to optimize
the power. Generally, the pitch angle ref is set to zero
the stator resistance of PMSG. q , d are the stator flux
for partial load region and it is activated in the full load
linkage of d, q axis respectively given as, region so that the generator speed is regulated based on
the reference values [19]. The block diagram of PI based
q = Lq iq controller used in this study is shown in the Fig. 5. The
(8) generator power is fed as an input to the PI controller
d = Ld id + Lmd I fd
which provides pitch angle reference ref as its output. ( )

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System 65

The difference between obtained power and reference (P g Pnom )

power is calculated, which generates the relative error. = (10)
Pnom
The control gain parameter of the PI controller is chosen
using Ziegler- Nichols equation. In this paper, the PI
controller is implemented for both the operating region. The following limits for pitch angle reference are used
During partial load region the PI controller is used to in controller,
extract the maximum power from the available wind
speed. The aerodynamic torque is limited in the full load ref = 0 o , for partial load region (11)
region in order to maintain constant and optimum
generator power. The relative error of the PI controller
is determined using,

ref max
Reference power
Pref
error
Pitch
- 1
Wind
turbine and
Pg

- +
+ control c generator

Generator power min

Pg

Fig.4. Pitch actuator of wind energy conversion system

Pref = K opt*ropt
ref = (
P P
P g nom
)
, for full load region (12) (13)

PI controller is generally used for linear system where Where, ropt is the optimum generator speed and K opt
they are subjected to nominal variations. Wind speed is defined as,
being highly non-linear, induces various damping for
different operating points which causes disturbance and * * C p * R 5
instability in system. Thus the power and voltage K opt = (14)
oscillates unpredictably [20]. Moreover, PI controller 2* 3opt
fails to track the rapid and vast variation in wind, which
leads to huge change in power output making the 45
Pg ref
unsuitable for grid integration. +- PI Servo Limiter 0

B. Fuzzy Logic Controller Pnom

The fuzzy logic control (FLC) is implemented to Fig.5. PI pitch angle control strategy
overcome the drawbacks of PI based controller. FLC
strategy is employed in both I and II operating regions. The optimal tip speed opt is determined by the power
The FLC strategy is designed using the prior knowledge coefficient C p . The above equations are used to calculate
of the system and human experience [22]. The important
feature of FLC is that the use of linguistic variables Pref during low wind region. In full load region where
instead of numerical variables, which provide imprecise wind speed is higher than that of rated wind speed, Pref
and qualitative communication [23]. The FLC strategy
consists of fuzzification, rule based inference system and is chosen as the rated power of wind turbine. The pitch
defuzzification process [24]. angle reference is calculated using the error in turbine
The main aspect of FLC is to normalize the output power and the rotor speed. The error of generator power
power fluctuation and to limit the turbine torque during and error of rotor speed error is calculated as,
full load region. The reference pitch angle is generated in
order to optimize the generator power and generator P = Pg Pref
(15)
speed. The error of generator power and generator speed r = r r ref
is considered as the input variables for the fuzzy control.
The FLC strategies do not require wind turbine parameter
values to estimate the pitch angle of blades. Where, P is the error of the generator power Pg and
The block diagram of FLC based control strategy is r is the error in rotor speed r . The pitch angle

shown in Fig. 6. The reference power Pref for FLC is
reference obtained as the output of fuzzy controller is
expressed as, considered as the pitch input of the wind turbine.
The triangular membership function is used to design

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
66 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System

the input and output fuzzy sets. The linguistic variable is feed-forward network which uses gradient descent based
illustrated as Negative Big (NB), Negative Medium (NM), delta learning rule [22]. The total squared error of the
Negative Small (NS), Null (ZE), Positive Small (PS), output is minimized using computation by net technique.
Positive Medium (PM), and Positive Big (PB) for input Supervised learning method is used to train the network,
membership functions. The output membership function which improves the ability to response correctly to the
has linguistic variables as Extra Small (XS), Small (S), input characters [25]. The BPN follow data normalization
Medium (M), Large (L), Extra Large (XL) and Maximum and early termination method to increase its ability and
Large (MXL). The rules for FLC used in this paper are implement approximation nonlinear function and to
presented in Table. 1. improve the network convergence performance [26].
The grade of membership functions of input variables In this paper, BPN uses wind speed and generator
is calculated using Eqn. (16) [3], speed as the input variable and generates pitch angle
which is fed to turbine in order to obtain desired
z m performance. The BPN is trained with two hidden layers
(z )=1 (16) thus they have four layers: Input layer, hidden layer I,
0.5
hidden layer II and output layer. The nodal operation of
BPN is processed in these layers. The modeling of each
layer is described as follows.
The input signal given to the BPN is denoted as x and y
as the targeted output. Then the matrix of the input and
Defuzzification output signal is represented as,
Fuzzy Output
x=[x1 x2 x3 ......xn ] (17)
Inference system Control rules

Fuzzy Inputs
y=[y1 y2 y3 ......yn ] (18)
Membership
Fuzzification The samples of the matrix of input signal are taken as
Function
820 by 2, and output signal as 820 by 1.
The input signals of the system are processed in this
Pg Pref r r ref layer. This layer calculates the value z which is fed in to
next layer. This layer finds out minimum and maximum
Fig.6. Fuzzy logic based pitch angle control strategy values of x and y.
Table 1. Rules for fuzzy logic controller
z x xmin k ymin (19)
P /
NB NM NS ZE PS PM PB
r where, k is the gain which is given by,
NB ZE ZE ZE ZE XS XS S
NM ZE ZE ZE XS XS S M y max y min
k= (20)
NS ZE ZE XS XS S S M xmax xmin
ZE ZE XS XS S S M L
PS XS XS S S M M L The output of input layer z is given as input in this
PM XS S S M M L XL layer. The weight w1 and bias b1 is calculated in the
PB S S M M L L MXL training period. The output of hidden layer I h1 is given
as
Where, z is the numerical data of input variable, m is
the coordinate point where the grade is 1, and is the h1 = (zw1 +b1 )log sigmoid (21)
width. The fuzzy control inputs and output which are
related to the membership function are shown in Fig. 7. This control strategy is designed for 3 neurons, thus the
The control rules are formulated using the input dataset dimension of weight matrix is 820 by 3. And the bias
through previous knowledge and experience of pitch matrix is calculated as 2 by 3.
based control system. Takagi-Sugeno type inference
The output of hidden layer h1 is given as the input for
system is used in this paper to set the rules for obtaining
precise pitch angle. hidden layer II. The weight and bias of this layer is
considered as w2 and b2 respectively. The output of
C. Feed forward back propagation
hidden layer II h2 is given as,
Feed forward neural network is most significant and
widely used artificial neural network. The most preferred h2 = (h1 w2 +b2 )log sigmoid (22)
type of feed-forward neural network type is back
propagation neural network (BPN). BPN is a multilayer

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System 67

The output layer of BPN determines the desired pitch process of the RBFN network is carried in two stages. In
angle of the system. The input of this layer h2 is first stage the unsupervised method is implemented where
subjected to reverse mapping to achieve the desired the parameter are governed by the radial basis function.
In the second stage, supervised training method in
output. The output of this layer p is processed to fed as employed to train the weights [28]. The supervised
the input to the wind turbine. training method is same as back propagation algorithm
[29].
p = (h2 ymin )k + xmin (23) In this paper, RBFN controller is implemented to
control the turbine of the wind energy conversion system
Thus, from this layer the optimised and desired pitch by alternating the pitch angle based on the wind speed.
angle output is achieved. The proposed RBFN controller consists of three layers:
an input layer, a hidden layer with nonlinear RBF
D. Radial basis function network activation function and a linear output layer [30] as
RBFN is a type of feed forward neural network which shown in Fig. 8. Wind speed and generator speed are fed
uses radial basis network as an activation function [25]. to the input neurons of the RBFN which is used to
The radial basis network is determined by the distance compute the pitch angle as the output neuron.
between the input and the prototype vector. The training
Pg Pref pu
NB NM NS ZE PS PM PB

-1.2 -9 -6 -3 0 3 6 9 1.2
(a) r rref
NB NM NS ZE PS PM PB

-6 -4.5 -3 -1.5 0 1.5 3 4.5 6

(b)

ZE XS S M L XL MXL

-5 0 5 10 15 20 27 35 45
(c)

Fig.7. Membership functions of Fuzzy logic control (a) Input error of generator power in pu. (b) Input error of generator speed in rad/s.
(c) Output Pitch angle in degree.

The basic nodes of operation are characterized into Input Hidden Output
three layers [31], layer layer layer
i j k
The inputs of two neurons in this layer are transmitted
directly to the next layer. The net input and output are
represented as, yi1

v x11 wj
net i1 = xi1 ( N )
(24)
( )
yi1 ( N ) = f i1 net i1 ( N ) = net i1 ( N ) yk3
i =1,2
y 2j
1 1 g x 21
Where, xi is the input layer which consists of x1 as
.
1 1
the wind speed and x2 as the generator speed. The net i x i1 .
.
1
represents the net sum of nodes of input layer and y i is Radial Basis Function
the output of input layer which is fetched to hidden layer
with respect to node i.
Fig.8. Architecture of RBFN

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
68 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System

The neurons in the hidden layer perform Gaussian The number of neuron in each layer is determined by
function which is used as the membership function in taking average for consecutive five trials. The least
RBFN. The net input and output of hidden layer are square error is achieved while using two layer three
represented as, neurons in layer 1. The sampling time for fuzzy logic
controller is 1ms in order to track the non-linearity of
net 2j ( N ) = (X Mj ) (X M )
T
j
wind speed.
j
(25)
y 2j ( N ) = f j2 (net 2
j
(N )) = exp (net (N ))
2
j
j =1,2 ,...,800
Table 2. Parameters of Wind Turbine.

Rated power 2MW

Blade radius 35.78 m
[ T ]
Where, M j = m1 j , m2 j ,......,mij is the mean of the Air density 1.225 kg/m3
Gaussian function and the standard deviation of Gaussian Max. power coefficient 0.47
function is denoted as Cut-in speed 3 m/s
Cut-out speed 20 m/s
[
= diag 1 12j ,1 22 j ,.........,1 ij2 .
j
]T
Rated wind speed 12 m/s
Blade inertia 6.3*106 kg.m2

The output layer computes single neuron which is Table 3. Parameters of PMSG Generator
determined by node k. The pitch angle control signal is
generated in this layer by summing the all the incoming Rated power 2 MW
signals with linear activation function. Stator resistance 0.00856
Stator inductance 3.75 mH
net k3 = w j y 2j ( N ) Friction factor 0.001189 Nms
j Pole pair 32
(26)
y k3 (N ) = f ki3 (net k3 (N )) = net k3 (N ) = ref Generator inertia 48 000 kg.m2

The controllers adapted in this paper are approximated

Where, w j is the weight which connects the hidden
to overcome the nonlinearity in wind turbine model.
layer and output layer. Wind speed, which is the input to the control system and
The supervised learning is implemented once the intermittent in nature, is shown in Fig. 10. The average
RBFN is initialized to train the system. The training wind speed is specified to 12m/s. The pitch angle control
method is same as the back propagation algorithm which of wind turbine is mostly tested for full load region where
is used to adjust the RBFN parameters using the training the wind speed is higher than that of rated wind speed.
patterns. The error of each layer is calculated and updated Fig. 11 shows the performance of PI based pitch
by the supervised learning algorithm in order to track the controller. The generator speed and the output power are
performance of wind system and act appropriately. In this used as the input for the controller. The gains of the PI
control strategy wind speed and generator speed is controller is fixed as kp=5 and kI=50.Fig. 11(a), shows
considered as the input for RBFN which generates the generator power of the wind turbine with PI based
respective pitch angle for wind turbine feedback. pitch angle controller. The generator power is not well
maintained and has ripple during the transition period.
The rotor speed with PI controller is shown in Fig. 11(b).
IV. RESULTS AND DISCUSSIONS There is a large amount of sudden spike in the generator
To validate the performance of the proposed control which may damage the system. The mechanical torque of
strategy, the simulation is performed using wind turbine with PI based pitch angle controller is
MATLAB/Simulink software for 2 MW PMSG wind shown Fig. 11(c). The power coefficient conversion of
energy conversion system. The parameters of wind the wind turbine is shown in Fig. 11(d). From the above
turbine and PMSG used for this study are listed in Table result it is clearly stated that PI controller fails to track the
2 and 3 respectively. The overall simulation diagram of rapid variation in wind speed.
RBFN control strategy with wind turbine and generator is The performance of FLC based controller strategy is
shown in Fig. 9. The RBFN control strategy uses gradient shown in Fig. 12. The inputs for the FLC control strategy
descent algorithm to train the system. The value of spread are generator speed error and generator power error
constant which determines the performance of RBFN which are shown in Fig.12(a) and Fig. 12(b) respectively.
network is analyzed by training the system in random The generator power with FLC control strategy is shown
constant value and the best value for computation is in Fig. 13(a). The power obtained is far better than PI
chosen for the simulation. The feed forward neural controller. During time period 1s and 6s when there is
network strategy along with back propagation training large variations in wind speed, the FLC control strategy
algorithm is used perform non-linear mapping to achieve requires a time interval to obtain the result thus an
corresponding pitch angle for the available wind speed. overshoot occur in that transition period. Similar spike
and swell is present in the generator speed and

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System 69

mechanical torque as shown in Fig. 13(b) and Fig. 13(c) it complex. The power conversion coefficient with the
respectively. However, these spikes can be reduced by FLC control is shown in Fig. 13(d). The FLC control
further decreasing the sampling time and using an strategy explicit the topology for small WECS with more
eminent strategy for FLC rules but this may increase the efficiency and where there is no huge variation in wind
computational time of the overall system and thus making speed rapidly.

Fig.9. Simulation of proposed system

Power coefficient
Wind speed
(m/s)

Ti me (s)

Time (s) (d)

Fig.10. Input pattern of Wind speed in (m/s) Fig.11. Simulation results for PI controller (a) Generated output power
in MW. (b) Rotor speed in (rad/s). (c) Mechanical Torque in NM. (d)
Power conversion coefficient of wind turbine
Generator Power

Output power error

(MW)

(pu)

Time (s) Ti me (s)

(a) (a)
Generator speed error
Rotor speed

(rad/s)
(rad/s)

Time (s)
Time (s)
(b)
(b)
Fig.12. Input for Fuzzy logic controller (a) Error of generated power in
pu. (b) Error in rotor speed in rad/s.
Mechanical Torque
(NM)

Generator Power
(MW)

Ti me (s)

(c)

Time (s)
(a)

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
70 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System

Rotor speed

Rotor speed
(rad/s)

(rad/s)
Time (s)
(b)
Time (s)
(b)
Mechanical torque
(NM)

Mechanical torque
(NM)
Time (s)
(c)
Time (s)
(c)
Power coefficient

Power coefficient

Time (s)
(d)
Time (s)
Fig.13. Simulation results for fuzzy logic controller (a) Generated (d)
output power in MW. (b) Rotor speed in (rad/s). (c) Mechanical Torque
in NM. (d) Power conversion coefficient of wind turbine Fig.14. Simulation results for Feed-forward back propagation controller
(a) Generated output power in MW. (b) Rotor speed in (rad/s). (c)
Mechanical Torque in NM. (d) Power conversion coefficient of wind
The results of feed-forward back propagation neural turbine
network based pitch angle control strategy are shown in
Fig. 14. Wind speed and the generator speed are used as The performance of proposed pitch angle controller
the input for the BPN control strategy. The generator which consists of RBFN topology is shown in Fig. 15.
power obtained with BPN strategy is almost kept at rated The generator output power with high quality and without
value as shown in Fig. 14(a). The BPN shows better any distortion is obtained in this strategy as shown in Fig.
performance in the transition period when compared to 15(a). The ripple and overshoot content in the generator
FLC and PI controller. The adaptation rate in BPN speed and the mechanical torque is also minimised to
control strategy is very fast in transition period. The BPN greater extent in this proposed control strategy as shown
is pre-trained in the transition network range to overcome in Fig. 15(b) and Fig. 15(c) respectively. The power
the overshoot issue which exists in FLC strategy. The conversion coefficient of wind turbine is also smoothened
generator speed and mechanical torque of wind system in this strategy as shown in Fig. 15(d). The generator
with BPN as pitch control technique is shown in Fig. 14(b) speed is kept around the rated speed of 17 rad/s and rated
and Fig. 14 (c). The overshoot in transition region is torque of 1500 Nm. The variation in the amplitude of
almost minimized. The power conversion coefficient these parameters is less than the BPN, FLC and PI
obtained with this control strategy is also similar to FLC controller.
except during the transition period as shown in Fig. 14(d).
Generator Power
Generator power

(MW)
(MW)

Time (s)
Time (s) (a)
(a)

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System 71

The main aim of the pitch angle controller is to limit

the aerodynamic torque to the rated value during high
wind speed because power output higher than the rated
Rotor speed
(rad/s)

value may damage the generator. Thus, the proposed

pitch controller manipulates the pitch angle of blade in
order to maintain the rated power capacity and rated
generator speed of the wind energy conversion system.
Time (s)
(b)
V. CONCLUSION
In this article, a radial basis function network and feed-
Mechanical torque

forward back propagation based pitch controller is

designed for PMSG wind turbine and analyzed for their
(NM)

performance. The main aim of the pitch controller is to

maintain the generator power and the mechanical torque
to their rated value. The control system is developed for
all operating region in order to improve the quality of
Time (s) output power. The proposed controller is then compared
(c) with fuzzy logic controller and conventional PI controller
for their performance. The results indicate that the RBFN
based pitch angle controller smoothes the output power
Power coefficient

and reduces the fluctuations significantly. The BPN

method is also considerably beneficial since they have
only small amount of average output power drop when
compared with FLC and PI controller. The FLC based
Time (s) control strategy provides better result during normal
(d) operation but during high transients the tracking speed of
Fig.15. Simulation results for RBFN controller (a) Generated output FLC is fails to overcome the fluctuations. Overall the
power in MW. (b) Rotor speed in (rad/s). (c) Mechanical Torque in NM. RBFN and BPN controllers provide effective result. The
(d) Power conversion coefficient of wind turbine major advantage of RBFN controller is the smaller
settling time when compared to other controllers so there
Therefore, from the above simulation results it is is less oscillations during the high transients in wind
observed that RBFN based pitch angle controller has speed. Hence, the proposed controller can be
more effective performance in comparison with BPN, implemented for real time for better and improvised
FLC and PI controller when parameters like generator operation of wind turbines.
power, generator speed and mechanical torque is analyzed.
Table. 4 shows the average power of proposed RBFN,
ACKNOWLEDGMENT
BPN, FLC and conventional PI controller based on their
operating region. It is evaluated that, proposed RBFN The authors will like to thank School of Electrical
controller gives 2.021%, 4.623% and 9.893% more power Engineering, VIT University, Vellore, India for constant
than the BPN, FLC and conventional PI controller support to carry out the research work.
respectively during below rated wind speed. And during
above rated wind speed the RBFN produces 0.187% of REFERENCES
rated generator power whereas BPN, FLC and PI
[1] K.Y. Oh, J. Y. Park, J. S. Lee, and J. Lee, "Implementation
controller produces 1.67%, 3.67% and 5.38% of rated of a torque and a collective pitch controller in a wind
generator power. The overshoot in generator speed may turbine simulator to characterize the dynamics at three
damage the entire system thus proposed controller control regions," Renewable Energy 79, 150-60 (2015).
produces smooth generator speed in variable wind speed. [2] R. Tiwari, and N. R. Babu, "Recent developments of
control strategies for wind energy conversion system,"
Table 4. Performance analysis of Pitch angle control strategy. Renewable Sustainable Energy Rev. 66, 268-85 (2016).
[3] T. L. Van, T. H. Nguyen and D. C. Lee, "Advanced pitch
Average output power angle control based on fuzzy logic for variable-speed wind
Control Strategy Below rated Above rated wind turbine systems," IEEE Trans. Energy Convers. 30, 578-87
wind speed speed (2015).
PI controller 1.7926 MW 2.1076 MW [4] M. Q. Duong, F. Grimaccia, S. Leva, M. Mussetta and E.
Fuzzy logic controller 1.8974 MW 2.0734 MW Ogliari, "Pitch angle control using hybrid controller for all
Feed-Forward back operating regions of SCIG wind turbine system,"
1.9492 MW 2.0338 MW Renewable Energy 70, 197-203 (2014).
propagation controller
Radial basis function [5] H. Moradi, and G. Vossoughi, "Robust control of the
1.9894 MW 2.00374 MW
network variable speed wind turbines in the presence of
uncertainties: A comparison between H and PID
controllers," Energy 90, 1508-21 (2015).

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
72 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System

[6] R. Saravanakumar and D. Jena, "Validation of an integral Syst. Appl. 6, 29-35 (2014).
sliding mode control for optimal control of a three blade [24] M. T. Makhloufi, M. S. Khireddine, Y. Abdessemed and A.
variable speed variable pitch wind turbine," Int. J. Elect. Boutarfa, Tracking Power Photovoltaic System using
Power Energy Syst. 69, 421-9 (2015). Artificial Neural Network Control Strategy, Int. J. Intel.
[7] X. Yao, S. Liu, G. Shan, Z. Xing, C. Guo and C. Yi, "LQG Syst. Appl. 6, 17-26, (2014).
controller for a variable speed pitch regulated wind [25] C. H. Chen, C. M. Hong and T. C. Ou, "Hybrid fuzzy
turbine," IEEE Int. Conf. Intelligent Human-Machine control of wind turbine generator by pitch control using
Systems Cybernetics (2009), pp. 210213. RNN," Int. J. Ambient Energy 33, 56-64 (2012).
[8] T. Senjyu, R. Sakamoto, N. Urasaki, T. Funabashi, H. [26] N. R. Babu and P. Arulmozhivarman, "Improving forecast
Fujita and H. Sekine, "Output power leveling of wind accuracy of wind speed using wavelet transform and
turbine generator for all operating regions by pitch angle neural networks," J. Elect. Eng. Tech. 8, 559-64 (2013).
control," IEEE Trans Energy Convers. 21, 467-75 (2006). [27] E. Assareh and M. Biglari, "A novel approach to capture
[9] R. Tiwari and N. R. Babu, "Fuzzy Logic Based MPPT for the maximum power from variable speed wind turbines
Permanent Magnet Synchronous Generator in wind Energy using PI controller, RBF neural network and GSA
Conversion System," IFAC-PapersOnLine 49, 462-7 evolutionary algorithm," Renewable Sustainable Energy
(2016). Rev 51, 1023-37 (2015).
[10] M. A. Chowdhury, N. Hosseinzadeh and W. X. Shen, [28] N. R. Babu and P. Arulmozhivarman, "Forecasting of
"Smoothing wind power fluctuations by fuzzy logic pitch wind speed using artificial neural networks," Int. Rev.
angle controller," Renewable Energy 38, 224-33 (2012). Mod. Sim. 5, 2276-80 (2012).
[11] Y. Daili, J. P. Gaubert, L. Rahmani, "Implementation of a [29] S. Saravanan and N. R. Babu, "RBFN based MPPT
new maximum power point tracking control strategy for algorithm for PV system with high step up converter,"
small wind energy conversion systems without mechanical Energy Convers. Manag. 122, 239-51 (2016).
sensors," Energy Convers. Manag. 97, 298-306 (2015). [30] I. Poultangari, R. Shahnazi and M. Sheikhan, "RBF neural
[12] E. G. Shehata, "A comparative study of current control network based PI pitch controller for a class of 5-MW
schemes for a direct-driven PMSG wind energy generation wind turbines using particle swarm optimization
system," Electr. Power Syst. Res. 143, 197-205 (2017). algorithm," ISA Trans. 51, 641-8 (2012).
[13] Y. Zhang, Z. Chen, W. Hu and M. Cheng, "Flicker [31] W. M. Lin, C. M. Hong, T. C. Ou and T. M. Chiu,
mitigation by individual pitch control of variable speed "Hybrid intelligent control of PMSG wind generation
wind turbines with DFIG," IEEE Trans. Energy Convers. system using pitch angle control with RBFN," Energy
29, 20-8 (2014). Convers. Manag. 52, 1244-51 (2011).
[14] X. X. Yin, Y. G. Lin, W. Li, Y. J. Gu, X. J. Wang and P. F.
Lei, "Design, modeling and implementation of a novel
pitch angle control system for wind turbine," Renewable
Energy 81, 599-608 (2015).
Authors Profiles
[15] V. Akhmatov, "Analysis of dynamic behaviour of electric
power systems with large amount of wind power," Ph.D.
Ramji Tiwari received his Bachelors
dissertation (Denmark: Technical University, 2003).
degree in Electrical and Electronics
[16] Y. Zhang, Z. Chen and M. Cheng, "Proportional resonant
Engineering in the year 2012 from Kathir
individual pitch control for mitigation of wind turbines
College of Engineering, Coimbatore and
loads," IET Renewable Power Gener. 7, 191-200 (2013).
Masters Degree in Power Electronics and
[17] A. Beddar, H. Bouzekri, B. Babes and H. Afghoul,
Drives from Sri Krishna College of
"Experimental enhancement of fuzzy fractional order PI+ I
Engineering and Technology, Coimbatore
controller of grid connected variable speed wind energy
in the Year 2014. He is currently working
conversion system," Energy Convers. Manag. 123, 569-80
towards his Ph.D degree at VIT University, Vellore. His
(2016).
research includes Renewable energy, Power Electronics and
[18] S. M. Muyeen, R. Takahashi, T. Murata and J. Tamura,
Control Optimization techniques for renewable energy.
"A variable speed wind turbine control strategy to meet
wind farm grid code requirements," IEEE Trans. Power
Syst. 25, 331-40 (2010).
Dr. N. Ramesh Babu is a Associate
[19] Y. Ren, L. Li, J. Brindley and L. Jiang, "Nonlinear PI
Professor and Head in Department of
control for variable pitch wind turbine. Control Eng.
Electrical and Instrumentation Engineering
Practice 50, 84-94 (2016).
at VIT University, Vellore. He completed
[20] R. M. Kamel, A. Chaouachi, K. Nagasaka, "Enhancement
his B.E. degree in Electrical and Electronics
of micro-grid performance during islanding mode using
Engineering from Bharathiyar University
storage batteries and new fuzzy logic pitch angle controller.
and M.E. degree in Applied Electronics
Energy Convers. Manag. 52, 2204-16 (2011).
from Anna University. He completed his
[21] V. S. Raviraj and P. C. Sen, "Comparative study of
Ph.D degree from VIT University. He has authored and
proportional-integral, sliding mode, and fuzzy logic
coauthored more than 40 publications in reputed International
controllers for power converters," IEEE Trans. Ind. Appl.
Journal. His research area includes Wind speed forecasting,
33, 518-24 (1997).
optimal control of wind energy conversion system, solar energy,
[22] A. S. Yilmaz and Z. zer, "Pitch angle control in wind
Power Electronics, and Application of Soft Computing
turbines above the rated wind speed by multi-layer
techniques in Electrical Engineering. He is an Associate Editor
perceptron and radial basis function neural networks,
of IEEE Access Journal and editorial board member of three
Expert Syst. Appl. 36, 9767-75 (2009).
other journals.
[23] H. Sefidgar and S. A. Gholamian, Fuzzy logic control of
wind turbine system connection to PM synchronous
generator for maximum power point tracking, Int. J. Intel.

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73
 Comparative Analysis of Pitch Angle Controller Strategies for PMSG Based Wind Energy Conversion System 73

How to cite this paper: Ramji Tiwari, Ramesh Babu. N,

"Comparative Analysis of Pitch Angle Controller Strategies for
PMSG Based Wind Energy Conversion System", International
Journal of Intelligent Systems and Applications(IJISA), Vol.9,
No.5, pp.62-73, 2017. DOI: 10.5815/ijisa.2017.05.08

Copyright 2017 MECS I.J. Intelligent Systems and Applications, 2017, 5, 62-73