You are on page 1of 5

The Online Journal on Power and Energy Engineering (OJPEE) Vol. (2) – No.

(1)

Sensor-Less Maximum Power Extraction


Algorithm for Small Wind Turbine Systems
Medhat. H. Elfar*, Ahmed. E. Kalas*, Soliman. M. Sharaf **
.
*Electrical Engineering Dept., Faculty of Engineering, Port-Said University
**Electrical Engineering Dept., Faculty of Engineering, Helwan University

Abstract--In this paper, the modeling and simulation of Another scheme is constant tip speed ratio scheme [6-11].
sensor-less maximum power extraction algorithm for This scheme is based on wind speed and rotor speed
small wind turbine systems are presented. The algorithm measurements. The wind speed is measured, and the required
is based on the efficiency error to adjust the modulation rotor speed for maximum power generation is computed. The
index of the inverter in order to extract the maximum rotor speed is also measured and compared to the calculated
power from the wind turbine system. The model variables optimal rotor speed, while the resulting error is used to
are expressed in the d-q synchronous reference frame. A control a power interface. The first barrier to implement this
closed loop scheme employing a PI controller has been scheme is the wind and rotor speed measurements, which
simulated by using Matlab/Simulink. The complete system adds extra cost and increases the control system complexity.
has been simulated and tested for different types of
In this paper, an alternative approach for wind turbine
disturbance. The advantages of the proposed algorithm
systems maximum power extraction is described. Fig.1 shows
are that no knowledge of the wind turbine characteristics
a block diagram of the proposed system. The scheme is based
or measurement of the rotor speed is required. Also, the
on monitoring the generator’s output power using
paper demonstrates how to employ the particle swarm
measurements of the generator’s output voltages and currents
optimization (PSO) method to search efficiently the
then the stator copper loss is calculated. By using the power
optimal PI controller.
and the stator copper loss, the actual efficiency is calculated
Index Terms- maximum power extraction, small wind and compared to predetermined reference efficiency. The
turbine, sensor-less algorithm, PSO. modulation index is adjusted according to the error to make
the actual efficiency close to the reference one. Thus, neither
I. INTRODUCTION
knowledge of the wind turbine power versus the rotor speed
Wind turbine is considered today as energy source which of rotation or wind-speed characteristic nor measurements of
allow electrical production with minimum environment the wind speed are required.
perturbations. This energy source is especially suitable for
remote areas, which are not connected to the conventional
electrical grid. Large wind turbines are complex in operation, II. STAND-ALONE WIND TURBINE SYSTEM MODEL
deploy multitude of control methods and operate in grid-
A stand-alone variable speed wind turbine when connected
connected mode. On the other hand, small wind turbines can
to a load should be controlled to extract the maximum power.
be used for stand-alone as well as grid-connected system. To
The configuration of a wind generation system is depicted in
achieve optimum utilization of wind turbine system, the
Fig.1. The system consists of wind turbine, self-excited
maximum power extraction schemes have been used. These
induction generator (SEIG), three phase uncontrolled rectifier
schemes employ a rotor speed and wind velocity sensors.
bridge, DC link, pulse width modulation (PWM) inverter and
However, from aspects of reliability and cost, rotor speed and
local load. The wind turbine and the associated components
wind speed sensors are not usually preferred. There are
have been described in more details in the following
several schemes to extract the maximum power from the wind
subsections.
turbine system such as power control scheme and constant tip
speed ratio scheme. Power control scheme [1-5] is based on
A. The wind turbine characteristics
the wind turbine system optimal power versus the wind speed
The theoretical mechanical power available from the wind
characteristics, which is usually stored in a microcontroller
turbine is given by
memory. The wind speed is measured; the optimal output
power is calculated and compared to the actual output power. P m = 0 . 5  AV 3 C p (1)
The resulting error is used to control a power interface. This
scheme has disadvantage that knowledge of the power-speed Where  is the air density, A is the swept area by the blades,
characteristic is required, this maximum power curve needs to V the wind speed, C P the power coefficient, C P is used to
be obtained via simulations or tests for individual wind
define the ratio between the mechanical power available of
turbines, which makes power control scheme difficult and
the turbine to the available wind power. Maximum power
expensive to implement in small wind turbine systems.

Reference Number: W10-0022 168


The Online Journal on Power and Energy Engineering (OJPEE) Vol. (2) – No. (1)

coefficient, CP max , is considered in mechanical design of the PVds =  b X c i dc +  s V qs (7)


wind turbine. That mean, usually, the generated mechanical
PVqs =  b X c i qc −  s V ds (8)
power from the wind turbine is a maximum at all wind speed
values. The maximum power generated from the wind turbine Since Vqs=0, due to the choice of axis alignment, the above
is
equations reduce to
Pm(max) = 0.5AV3Cp max (2) PVds =  b X c i dc (9)
B. Self-Excited Induction Generator Model b X ciqc
The dynamic model of induction generator can be s = (10)
Vds
simulated by using Equation (3). To simplify the analysis, the
initial orientation of the q-d reference is taken such that the d- The above equation can be used to determine the electrical
axis and stator terminal voltage are initially aligned, this frequency (  s ) of the voltage generated by the induction
assumption result in Vqs being zero initially. Also the rotor is
shorted, hence Vqr=Vdr=0. The state-space form of the generator.
induction generator in the q-d synchronously rotating
reference frame is given by D. Uncontrolled Three Phase Rectifier
The three phase uncontrolled rectifier is used to convert the
variable voltage, variable frequency alternating current out
from the induction generator to direct current. The rectifier
 − rs  b X r − ( s D +  r X m2 ) rr  b X m − r X m X r 
  output voltage in our case is given by [12].
1  s D +  r X m2 − rs  b X r r X m X r rr  b X m 
A=
D  rs  b X m s X m X s − rr  b X s X s X r  r −  s D 3 3
  Vr = V ds (11)
 −  r X m X s rs  b X m − ( X s X r  r −  s D) − rr  b X s  

I • = AI + BV (3) Assuming that the output current of the rectifier is


continuous and ripple-free, the DC output power of the
Where: rectifier is given by
 0 
  Pdc=VrIr (12)
1 − X
B =  b r, Where:
D 0  Vr : Rectifier output DC voltage
 
 b X r  Ir : Rectifier output DC current.
[
I = iqs ; ids ; iqr ; idr , ] V = [Vds ]
Neglecting losses in the rectifier, the DC output power of the
rs , rr Stator and rotor resistance (Ω). rectifier is equal to the AC input power to the rectifier and
Xs, Xr Stator and rotor reactance (Ω). given by
Xm Magnetisation reactance (Ω). 3
Vr I r = (V qs I qL + V ds I dL ) (13)
s , b , r Angular speeds of synchronous reference frame, 2
base, and rotor (rad/sec.) Since Vqs is zero, (13) becomes
And D = X s X r − X m2 3
Vr I r = (V ds I dL ) (14)
The rotor speed r is governed by the following differential 2
equation: The direct axis of rectifier input current may be expressed by
2 (5)
Tin = J   p  r + Te 2 3
 p I dL = Ir (15)

Where Tin is the input torque from the prime-mover and Te is
electromagnetic torque and given by E. DC Filter Model
The DC filter is comprised of a series reactor and shunt
 3  p 
Te =     Lm (ids iqr
'
− iqs idr
'
) (6) capacitor as shown in Fig. 1. The dynamics introduced by the
  
2 2
DC filter are as follow:
The DC voltage, which is the voltage at the capacitor
C. The Excitation Capacitor Model terminals, is given by:
The capacitor bank equation in the d-q axes can be written as

Reference Number: W10-0022 169


The Online Journal on Power and Energy Engineering (OJPEE) Vol. (2) – No. (1)

Figure (1): The proposed maximum power extraction scheme

dV dc
=
1
(I r − I in ) (16)
V qinv = Vlphm = m
V dc
(19)
dt C
2
The DC current, which is the current through the inductor, is
given by: Vdinv = 0 (20)

dI r
=
1
(V − V ) (17)
dt L
r dc Assuming the inverter to be lossless, and neglecting the
harmonic components in the output waveform. Consequently,
Where: the DC and AC powers at the inverter sides are equal
Vin : The input voltage of the inverter
Vr : Rectifier output voltage 3
Vdc I in = Vqinv I qinv (21)
Ir : Inductor current 2
iin : Current of inverter side
C : Capacitance of capacitor The q- axis of inverter output current may be expressed by
L : Inductance of the inductor. 4
I qinv = I in (22)
3m
F. Pulse Width Modulation Inverter Model
The sinusoidal PWM inverter converts the DC bus voltage For resistance load, RL, transforming the load to the DC side
to that of a fixed frequency of 50Hz. The modulation phase of the inverter we get
voltage can be analyzed by applying a Fourier analysis. In
8
this thesis only the fundamental harmonic of the inverter is R Ldc = RL (23)
considered. The fundamental harmonic of the inverter output 3m 2
voltage can be shown to be of the following form [13].
Vdc
V1 ph (t ) = m sin(t ) (18) III. MAXIMUM POWER EXTRACTION SCHEME
2
When the wind speed increases, the input mechanical
Where: power increases. The different between the input power and
m is the modulation index. output power will convert to an increasing in the kinetic
V1 ph the fundamental harmonic of the phase voltage energy storied in the rotor. To benefit from this energy, the
more power should deliver to the load to reduce this
To simplify the analysis the initial orientation of q axis is accelerating.
taken such that it coincides with the output voltage, hence, the The objective of this work is to extract approximately the
q and d axis components of the inverter output voltage (line to maximum available power generated from the induction
neural) generator driven by the wind turbine without wind speed or
rotor speed measuring. To achieve this objective, sensor-less

Reference Number: W10-0022 170


The Online Journal on Power and Energy Engineering (OJPEE) Vol. (2) – No. (1)

algorithm is proposed, where the system is controlled to The performance index in the time domain is presented to
operate at maximum available efficiency of the system. estimate the parameters of PI controller. The performance
index in the time domain is given by:
The reference efficiency is calculated according to the open
loop response of the system at different wind speed. By
Table (1): The summary of parameters tuning
neglecting the rotor copper loss, the actual efficiency can be
PI controller
calculated as Performance
parameters
index
Pout Kp Ki
 = (22) PSO algorithm 10 220 0.0582
Pout + 3 rs i s2 ZN method 11.25 34 0.3396
The actual efficiency is compared with that reference Trial and error 1 10 3.1901
efficiency and any mismatch is used to adjust the modulation
index of the inverter. The algorithm is so simple that it needs 0.014
only an online values of induction generator voltages and
currents which can be obtained easily by using just current 0.012

and voltage sensors. The actual efficiency calculation is 0.01


shown in Fig. 2.
0.008
vq
P-index

vd 3
(v q i q + v d i d ) 0.006
iq 2
id
+
÷ 
0.004
+
3 0.002
R s (iq2 + id2 )
2
0
0 10 20 30 40 50 60 70 80 90 100
Figure (2): The efficiency calculation PSO iterations No.

Figure (4): The performance index with the number of PSO


IV. PSO BASED PI CONTROLLER PARAMTERS TUNNING iterations
In this section, the PI controller’s parameters will be tuned
by using PSO algorithm to find optimal PI controller’s K
parameters. Over the years, several methods have proposed
for the tuning of PI controllers, one of these methods is
P _ index = ∑ e 2 (k ) (23)
1
Ziegler-Nichols (Z-N) method. In general, it is often had to
determine optimal or near optimal PI parameters with Z-N In order to emphasize the advantage of the PSO-PI controller,
formula in many industrial plants [14]. the tuning of PSO based PI controller results will be
Particle swarm optimization (PSO), first introduced by compared to those obtained from Z-N method and trial and
Kennedy and Eberhart, is one of the modern heuristic error method. Fig. 4 shows the performance index with the
algorithms. It was developed through simulation of a number of PSO iterations. As can be seen from the figure, the
simplified social system, and has been found to be robust in performance index value reduces as the number of iterations
solving continuous nonlinear optimization problems [15]– increase. The PSO algorithm can prompt convergence and
[18]. The PSO technique can generate a high-quality solution obtain good performance value.
within shorter calculation time and stable convergence Fig. 5 presents the system response for different PI
characteristic than other stochastic methods. The PI controller controller. The wind speed changes from 10 m/sec to 7 m/sec
will be tuned offline using PSO as shown in Fig.3 at t=3 sec then the load is changed at t= 6 sec. The figure
-
Objective function illustrates three responses, the first response with PI controller
tuned by PSO algorithm, the second response with PI
controller tuned by trial and error and the third response with
PSO
PI controller tuned by Z-N method. It should be noted that the
Updated PI values
response of the PSO-PI is better than the other two responses.
It is clear that the PSO-PI overshoot is the smallest one and
Output
Reference
PI Plant
the efficiency settles to steady state value faster than the two
+ -
controller responses. Table 1 summarizes the tuned parameters of each
of controlled systems and shows the performance index for
each one. It is clear that the PSO based PI controller has the
smallest performance index.
Figure (3): The structure of PSO of PI tuning algorithm

Reference Number: W10-0022 171


The Online Journal on Power and Energy Engineering (OJPEE) Vol. (2) – No. (1)

Efficiency [6] Control Loop,” IEEE Trans. Energy Conversion, Vol. 9,


86
PSO N-Z No. 3, Sept.1994.
84
[7] T. Thiringer and J. Linders, “Control by Variable Rotor
82 Speed of a Fixed-Pitch Wind Turbine Operating in a
82.5 Trial-error Wide Speed Range,” IEEE Trans. Energy Conversion,
80 82

78
81.5 Vol. EC-8, pp. 520–526, Sept. 1993
%

81

80.5
[8] Miller, E. Muljadi, and D. Zinger, “A Variable Speed
76 80

79.5
Wind Turbine Power Control” IEEE Trans. Energy
74 79
6 6.2 6.4 6.6 6.8 7
Conversion, Vol. 12, pp. 181-186, June1997.
72 [9] Alan Mullane, ”Adaptive Control of Variable Speed
70
Wind Turbines” renewable energy 2001.
0 1 2 3 4 5 6 7 8 9 [10] R. J. Cárdenas, R. S. Peña, “ Sensorless Vector Control
Time(sec)
of Induction Machines for Variable-Speed Wind
Energy Applications” IEEE Trans. Energy Conversion,
Figure (5): The comparison between different responses Vol. 19, No. 1, pp.196-205,March 2004.
[11] R. S. Peña, R. J. Cárdenas, G. M. Asher, R.B. Gimenez
and J. C. Clare, “ Control Strategies for Power
V. CONCLUSION Smoothing Using a Flywheel Driven by a Sensorless
This paper presents a simple sensor-less maximum power Vector-Controlled Induction Machine Operating in a
algorithm employing the efficiency error to extract maximum Wide Speed Range,” IEEE Trans. Ind. Electron., Vol.
power for small wind turbine system. The advantages of the 51, pp.603-614, June 2004.
algorithm are the wind speed or rotor speed sensors are not [12] Tapia, G. Tapia, and J. Saenz, “Modeling and Control of
required as in the previous algorithm. How to determine the a Wind Turbine Driven Doubly Fed Induction
PI controller parameters using PSO method has been Generator” IEEE Trans. Ind. Electron., Vol. 18,
demonstrated. Through the simulation of the system, the pp.194-204, June 2003
results show that the PSO method can perform an efficient [13] B. K. Bose, Modern Power Electronics and AC Drives.
search for the optimal PI parameters. Prentice-Hall, 2002.
[14] B. S. Borowy, Z.M. Salameh, ”Dynamic Response of a
Stand-Alone Wind Energy Conversion System with
REFERENCES Battery Energy Storage to a Wind Gust” IEEE Trans.
Energy Conversion, Vol. 12, No. 1, pp73-78, March
[1] E. Muljadi and C. P. Butterfield, “Pitch-Controlled 1997.
Variable-Speed Wind Turbine Generation,” IEEE [15] Astrom, K., and Hagglund, T., PID controllers: Theory,
Trans. Ind. Appl., Vol. 37, No. 1, pp. 240–246, Jan. design, and tuning. 2nd Edition, Instrument Society of
2001. America, 1995.
[2] A. M. De Broe, S. Drouilhet, and V. Gevorgian, “A [16] J. Kennedy and R. Eberhart, “Particle swarm
Peak Power Tracker for Small Wind Turbines in optimization,” in Proc.
Battery Charging Applications,” IEEE Trans. Energy IEEE Int. Conf. Neural Networks, vol. IV, Perth,
Conversion, Vol. 14, No. 4, pp. 1630–1635, Dec. 1999. Australia, 1995, pp. 1942–1948.
[3] A. Z. Mohamed, M. N. Eskander, and F. A. Ghali, [17] Y. Shi and R. Eberhart, “A modified particle swarm
“Fuzzy Logic Control Based Maximum Power optimizer,” in Proc. IEEE Int. Conf. Evol. Comput.,
Tracking of a Wind Energy System,” Renewable Anchorage, AK, May 1998, pp. 69–73.
Energy, Vol. 23, No. 2, pp. 235–245, Jun. 2001. [18] Y. Shi and R. C. Eberhart, “Empirical study of particle
[4] R. M. Hilloowala and A. M. Sharaf, “A Rule-Based swarm optimization,” in Proc. IEEE Int. Conf. Evol.
Fuzzy Logic Controller For a PWM Inverter in a Stand Comput., Washington, DC, July 1999, pp. 1945–1950.
Alone Wind Energy Conversion Scheme,” IEEE Trans. [19] R. C. Eberhart and Y. Shi, “Comparison between
Ind. Appl., Vol. 32, No. 1, pp. 57–65, Jan./Feb. 1996. genetic algorithms and particle swarm optimization,” in
[5] R. M. Hilloowala and A. M. Sharaf, “A Utility Proc. IEEE Int. Conf. Evol. Comput., Anchorage, AK,
Interactive Wind Energy Conversion Scheme with an May 1998, pp. 611–616.
Asynchronous DC Link Using a Supplementary

Reference Number: W10-0022 172

You might also like