You are on page 1of 16

1

Scientific Bulletin - Faculty of Engineering - Ain Shams Uni. Vol. 41, No. 2, June 30, 2006.





MODELING AND CONTROL OF WIND TURBINES INCLUDING
AERODYNAMICS

M. EL-SHIMY

Electrical Power and Machines Department
Faculty of Engineering
Ain Shams University, Cairo, Egypt

ABSTRACT
Wind has proven to be one of the most successful of all available sources of renewable
energy offering relatively high capacity, with generation costs competitive with
conventional energy sources. Therefore, the technologies for generation of electrical
energy from renewable energy sources, especially wind energy, have evolved in recent
years. Most of available wind farm models in literature neither include aerodynamic
torque models nor model for mechanical parts. Therefore, this paper presents a model
for a variable-pitch, constant-speed horizontal-axis wind turbines including
aerodynamics and mechanical parts. Moreover, two techniques of control of the
considered wind turbine are applied. Both techniques objected to keep wind turbine
speed constant for wide range of wind speed variations. Both PID and neural network
NARMA-L2 controllers are used to control rotor blade pitch angle of the wind turbine.
Results show that both controllers are capable to keep wind turbine speed and output
power for various types of disturbances by controlling rotor blade pitch angle.


.
.
.
.
.
- - .

.


2
INTRODUCTION

For thousands of years, man has harnessed energy from wind to sail ships, grind grain
and pump water. Windmills were in use in ancient Egypt fifty centuries ago for
grinding flour. Water pumping windmills have been recorded in Kautalyas
Arthashastra, indicating their existence in India from 400 BC. But it was only as
recently as the late 20
th
Century, that windmills were developed in Europe to generate
electricity.

Wind results from the differential heating of the earth and its atmosphere by Sun. It is
the kinetic energy in the wind that windmills convert into mechanical or electrical
energy. Wind energy is free, there is no cartel that controls its distribution and no
sanction or blockade on wind is possible. Since it is unlimited, renewable and a
pollution free source, there has been a movement the world over to develop highly
sophisticated technology to convert kinetic energy in wind to electrical energy.

Wind has proven to be one of the most successful of all available sources of renewable
energy offering relatively high capacity, with generation costs competitive with
conventional energy sources. Therefore, the technologies for generation of electrical
energy from renewable energy sources, especially wind energy, have evolved in recent
years [1, 2]. In 2002, the total installed capacity of Wind Energy allover the world
reached 27.257 GW [3]. It is predicted that 12% of the total world electricity demands
will be supplied from wind energy by 2020 [4].

In recent years, the cost associated with electric energy derived from fossil and nuclear
fuel, and the increase in environmental regulations continue to constraint the planning
and operation of electric utilities. Furthermore, the global economic and political
conditions that tend to make countries more dependent on their own energy resources
have caused growing interest in the development and use of renewable energy [5, 6].
In terms of its environmental advantages, wind turbines generates electricity with no
contribution of carbon dioxide (CO
2
) or other greenhouse gases to the atmosphere and
they produce no pollutant discharge on water or soil. In terms of economics, the
improvements in technology and the acquired experience with wind power plants have
shown reliability and durability, and their operation and maintenance costs predictable
[5].

Wind farm modeling is very important to investigate the performance of the farm
either with grid connection or as a stand-alone system. During the last decade several
models have been introduced for the wind farm depending on the purpose of the study.
Most of available wind farm models in literature neither include aerodynamic torque
models nor model for mechanical parts.

A third-order dynamic model of induction machine is used in [7] neglecting
aerodynamics and the turbines mechanical coupling on the performance of wind
farms.
3
A PQ-model was used to investigate the operation of a wind farm consisting of parallel
connected doubly fed induction machines [8] and used in calculating the output active
and reactive power of each generator unit. The variation of the wind speed experienced
by each generator was simulated in two different ways; a constant wind speed and a
sinusoidal type wind speed. The conventional PQ-model for the wind farm where the
generators active power and the power factors were assumed and the reactive power
was calculated was modified by considering the steady state model of induction-type
generators that assumed to be identical [9] [12]. Moreover, an RX-model was
developed. In these models, the real power was calculated from the wind speed using
the power curves in the first iteration then it was assumed constant during the iterative
simulation. This made the simulation easier as the reactive power will be dependent
only on the buses voltages. Both PQ-models and RX-models are suitable for load flow
studies. Therefore, these models are not sufficient for studies involving dynamic
operation under various operating conditions. This because the lake of the effect of the
mechanical construction of the turbines as well as the models for the turbine
interconnections.

A state variable model for wind turbines was developed to investigate the transient
stability analysis of the Cyprus power system with small wind farm connection [13].
Aggregate models of wind farms considering aerodynamics and mechanical coupling
were developed in [14, 15] consisting of aggregated models for wind speed, wind
turbines, and a wind farm layout. A Model suitable for transient stability studies was
developed [16]. Each equivalent wind turbine model consisted of the aerodynamic
torque, drive train, and induction generator model. This model was used to investigate
the effect of short circuit power at bus connection, reactive power compensation, rotor
inertia, and wind speed on the transient stability of the system. It was shown that the
effect of the short circuit power and the variation in the rotor speed is very significant
on the stability, while the reactive power compensation has only a slight effect.

Detailed modeling of wind farms [17, 18] suitable for wide range of studies were
developed consisted of; a model for electric generators, a model for the mechanical
parts of the wind turbine, an aerodynamic model represented by the aerodynamic
power equation, and a wind model.

Several schemes and techniques have been developed to control the active and reactive
power flow to and from the wind farm by controlling each individual unit or a group of
units in order to insure that the reactive power demand by the farm is within the
required limits. While others were devoted to regulate the output voltage of each unit.
In addition to the maximum power tracking control by which each unit was equipped
[19-25].

This paper presents a model for a variable-pitch, constant-speed horizontal-axis wind
turbines including aerodynamics and mechanical parts. Moreover, two techniques of
control of the considered wind turbine are applied. Both techniques objected to keep
wind turbine speed constant for various disturbances. Both PID and neural network
NARMA-L2 controllers are used to control rotor blade pitch angle of the wind turbine.
4
Results show that both controllers are capable to keep wind turbine speed and output
power for various types of disturbances by controlling rotor blade pitch angle.

DESCRIPTION OF WIND TURBINE

The wind turbine rotor is connected to a generator. The generator output can be
controlled to follow the commanded voltage. The wind turbine has pitchable blades to
control the aerodynamic power extracted from the wind. Also there is a mechanical
component (gearbox) between the low-speed rotor shaft and the high-speed generator
shaft. The low-speed shaft is driven by the turbine blades, which generates
aerodynamic power. The high-speed shaft is loaded by the electric generator in the
form of an electrical load. As the wind speed fluctuates, the wind turbine is controlled
by changing the pitch angle to avoid the rotor speed following the variation of the
wind speed. Therefore, the wind-turbine-generator (WTG) system converts rotational
energy to electrical energy, which is usually supplied to the utility grid at the
distribution level.

WIND POWER AVAILABLE

The kinetic energy, U of a packet of wind of mass m flowing at upstream speed u in
the axial direction of the wind turbine is given by [26]:

2 2
2
1
2
1
u Ax mu U ) ( = = (1)

where A is the cross-sectional (swept) area of the wind turbine, is the air density and
x is the thickness of the wind packet.

The wind power, P
w
in the wind, which represents the total power available for
extraction, is given by:

dt
dU
P
w
= (2)
Therefore,
3 2
2
1
2
1
Au
dt
dx
Au P
w
= = (3)

The mechanical power, P
m
extracted from the available power in the wind P
w
is
expressed by the turbine power coefficient of performance C
P
which is a nonlinear
function of tip speed ratio and pitch angle |. Therefore,

( )
w p m
P C P | , = (4)

In ideal conditions [26], the turbine cannot extract more than 59% of the total power of
undisturbed tube of air with cross sectional area equals to wind turbine swept area.
5
The tip speed ratio is a variable that combines effect of rotational speed of the
turbine and wind speed. It is defined as the ratio between the rectilinear speed of the
turbine tip (eR) and the wind speed (u).

u
R e
= (5)

where R is the maximum radius of the wind turbine swept area.

The following equation can be used to approximate the C
P
(, |) curve [27]:

( )
K
p
e K C
5 12
5 4 0 116 22 0
.
. . ) , (

= | | (6)
where
1
035 0
08 0
1
3
+

+
=
| |
.
.
K (7)

Variation of C
P
with tip speed ratio at constant pitch angle and variation of C
P
with
pitch angle at constant tip speed ratio are shown in Fig. 1 and Fig. 2 respectively.

In addition to the turbine power coefficient of performance C
P
, the wind turbine rotor
performance can also be evaluated as function of the coefficient of torque C
q
. As the
wind power P
w
is equal to the product of the aerodynamic torque T
A
and the rotor
rotational speed e, the torque coefficient can be related to the power coefficient by:

) , ( ) , ( | |
q p
C C = (8)

Using (3), (4), (5), and (8), the aerodynamic torque that turns the rotor shaft takes the
form:

2
2
1
u C AR T
q A
) , ( |
|
.
|

\
|
= (9)

WIND TURBINE PLANT MATHEMATICAL MODELING

The most special feature about wind turbines is the fact that, unlike other generation
systems, the power inflow rate is not controllable. In most power generation systems,
the fuel flow rate, or the amount of energy, applied to the generator controls the output
voltage and frequency. However, wind speed varies with time and so does the power
demand. Therefore, other generation systems can be referred to as controlled energy
sources, whereas the wind is an uncontrolled energy resource and the power demand is
an uncontrolled energy sink.
On occasion, the wind speed can be very high resulting in power generation that
exceeds the demand of the load. This might lead to the turbine exceeding its rotational
6
speed rating and subsequent damage to the turbine. On the other hand, the wind speed
can be too low for any power production and therefore alternative energy sources
should be used.

The fact that one has no control over the energy source input, the unpredictability of
wind and the varying power demand are more than enough concerns to justify the need
for a controller, which will regulate all the parameters that need to be controlled for a
matched operation of the wind turbine.

The wind-turbine-generator (WTG) model is divided into two main parts. The first
part is the wind turbine, which included a turbine rotor on a low-speed shaft, a gearbox
and high-speed shaft. The inputs for this part of the plant are the wind speed and the
blade pitch angle while the outputs are the high-speed shaft angular rotation and the
mechanical power, P
m
. The second part is the electric generator whose input is
constant angular rotation from the turbine plant and whose output is electrical power.
Fig. 3 shows a block diagram of the wind turbine system.

The equation of motion of wind turbine system is given by:

L A T
T T
dt
d
J =
e
(10)

where J
T
is the equivalent combined moment of inertia of the rotor, gear reducer and
both the low-speed and high-speed shafts, T
L
is the wind turbine load torque
representing the input torque to the electrical generator and opposed by its electrical
torque.

For the purpose of dynamic analysis and for designing a linear controller, such as PID
controller, equation (10) is linearized around an initial operating point (u
o
, e
o
, |
o
).
Substituting for T
A
in (10) using (9), the linearized form of (10) takes the form:

T
L
J
T
u
dt
d A
A + A + A =
A
| u e o
e
(11)

The parameters o, , and u are calculated at the initial operating conditions (u
o
, e
o
, |
o
)
and are given by:

|
|
.
|

\
|
=
o
q
o qo o
T
d
dC
C ARu
J
o 2
2
1
(12)

o
q
o
T
d
dC
u AR
J

2
2
1
= (13)

7

o
q
o
T
d
dC
ARu
J |
u
2
2
1
= (14)

The magnitude of o and u respectively show the relative weight of the effect of wind
speed and pitch angle on the wind turbine dynamics.

In s-space, (11) takes the form:

|
|
.
|

\
| A
A + A

= A
T
L
J
T
s s u
s
s ) ( ) ( ) ( | u o

e
1
(15)

Equation (15) represents the linearized form of the wind turbine transfer function.

However, the turbine power output is given by:

e
A m
T P = (16)

The linearized form of output power equation (16) takes the form:

e e A + A = A
Ao A o m
T T P (17)

Based on (11) and (17), the block diagram of wind turbine plant is given in Fig. 4.

Wind Turbine Control

A general block diagram for wind turbine control system is shown in Fig. 5.

The transfer function of a hydraulic actuator that changes the blade pitch angle can be
represented by first-order transfer function:

A
A
c
A
k s
k
s
s
s G
+
=
A
A
=
) (
) (
) (
|
|
(18)

In this paper two controllers are considered; a PID and a neural network NARMA-L2
controllers [28].

PID controllers regulate the error, or difference between the measured input and the
desired input. This error value along with its derivative and integral with respect to
time provides a signal to the actuator(s), which affects the controlled plant. The PID
controller is a linear, single-input single- output controller limited to three gains. The
transfer function of the PID controller is given by:

8
s k
s
k
k
s
s
s G
D
I
p
c
C
+ + =
A
A
=
) (
) (
) (
e
|
(19)

The central idea of the NARMA-L2 neurocontroller [28] is to transform nonlinear
system dynamics into linear dynamics by canceling the nonlinearities. There are
typically two stages involved when using neural networks for control. The first stage is
system identification in which a neural network model of the plant to be controlled is
developed by training a neural network to represent the forward dynamics of the
system. The second stage is control design in which the neural network plant model is
used to design (or train) the controller.

APPLICATIONS

The mathematical model in previous sections is applied to develop the response of
controlled and uncontrolled wind turbine plant. The wind turbine parameters [29] are
given in Table 1.

UNCONTROLLED RESPONSE OF WIND TURBINE PLANT

With initial operating point is (u
o
= 7 m/s, e
o
= 10.5 rad/s, |
o
= 9 deg.) the parameters
o, , and u calculated using equations (12), (13), and (14) and given as 0.2071, -
0.0668, and 0.0298 respectively. With uncontrolled wind-turbine plant, the response
to a unit step in wind speed, a unit step in pitch angle, and a 10% step increase in load
torque are determined and shown in Fig. 6, Fig. 7, and Fig. 8 respectively. These
responses are mainly dependent on the rotational inertia of the wind turbine plant, the
scaling factors o and u, and the parameter . The parameter represents the wind
turbine aerodynamic characteristics and it does not affect the wind turbine plant inputs.

CONTROLLED RESPONSE OF WIND TURBINE PLANT

PID and neural network NARMA-L2 controllers are used to compensate the wind
turbine speed deviations by changing the pitch angle |.

Based on Routh-analysis of the wind turbine transfer function and trial and error
approach, the gains of PID controller are selected to be (k
P
= 60, k
I
= 50, and k
D
= 20).
The plant response to a unit step in wind speed, and a 10% step increase in load torque
are shown in Fig. 9, and Fig. 10 respectively. It is shown that PID controller succeeded
in keeping wind turbine speed and output power. In order to show the validity of the
PID controller, a variable wind speed is assumed as shown in Fig. 11, the plant
response to this variable wind speed is shown in Fig. 12. It is clear that plant response
suffers from a small amount of control errors.

Neural network NARMA-L2 controller with on-line training is used instead of PID
controller. The plant response to a unit step in wind speed, a 10% step increase in load
torque, and variable wind speed are shown in Fig. 13, Fig. 14 and Fig. 15 respectively.

9
Although a small amount of control errors are obtained when using PID controllers,
high accuracy of plant response to follow the objective of zero speed deviation is
obtained with neural network NARMA-L2 controller.

CONCLUSION

In this paper a model for variable-pitch, a constant-speed horizontal-axis wind turbine
is given including aerodynamics. In order to control the wind turbine speed for various
disturbances in wind speed and load torque two controllers are presented classical PID
controller and neural network NARMA-L2 controller. Although PID controllers are
considered as the industry standard for blade-pitch control, the neural network
NARMA-L2 controller gives better system response than the PID controller. However,
both controllers act well in keeping the wind turbine speed and output power.

REFERENCES

[1] Castro Sayas F., Allan R.N., Generation Availability of Wind Farms, IEE
Proceedings on Gen., Trans., and Distr., Vol. 143, No. 5, 1996, pp. 507-508.
[2] Demoulias C., Dokopoulos P., Electrical Transients if Wind Turbines in Small Power
Grid, IEEE Trans. On Energy Conversion, Vol. 11, No. 3, Sept. 1996.
[3] Wind Force 12, European Wind Energy Association (EWEA) briefing, November
2002.
[4] Wind Force 12, Report by the European Wind Energy Association (EWEA), October
2002.
[5] National Renewable Energy laboratory, Wind Energy Information Guide,
Department of Energy, Report No. ED96000474, April 1996.
[6] Current Technologies, Available at:
http: // www. cranfield.ac.uk/sme/ppa/wind/lectintro.html
[7] Saad-Saoud Z.,Jenkins N., Simple Wind Farm Dynamic Model, IEE Proceedings on
Gen., Trans., and Distr., Vol. 142, No. 5, 1995, pp. 545 -548
[8] Tapia A., Tapia G., Ostolaza X., Fernandez E., Saenz J.R., Modeling and Dynamic
Regulation of a Wind Farm, Proc. of the VII IEEE International Power Electronics
Congress, 15-19 Oct. 2000, pp.: 293 297.
[9] A.E. Feijoo, J. Cidras, Modeling of Wind Farms in the Load Flow Analysis, IEEE
Trans. on Power Syst, Vol. 15, No. 1, 2000, pp. 110 115.
[10] C.R. Fuerte-Esquivel, J.H. Tovar-Hernandez, G. Gutierrez-Alcaraz, F. Cisneros-
Torres, A.E. Feijoo, J. Cidras, Discussion of Modeling of Wind Farms in the Load
Flow Analysis, IEEE Trans. on Power Syst, Vol. 16, No. 4, 2001.
[11] A.E. Feijoo, J. Cidras, Closure to Discussion of Modeling of Wind Farms in the Load
Flow Analysis, IEEE Trans. on Power Syst, Vol. 16, No. 4, 2001..
[12] A. Feijoo, J. Cidras, Corrections to Modeling of Wind Farms in the Load Flow
Analysis, IEEE Trans. on Power Syst, Vol.16, No. 4, 2001.
[13] A. Papantoniou, Modelling and Simulation of the Effects of Grid Connected Wind
Farms on the Cyprus Electricity Grid, Proc. of the 10
th
Mediterranean Electro-
technical Conference, No. 3, 2000, pp. 1145 1148.
[14] J.G. Slootweg, W.L. Kling, Modeling of Large Wind Farms in Power System
Simulations, Proc. of the IEEE/PES Summer Meeting, 1, 2002, pp. 503 -508.
[15] J.O.G. Tande, Grid Integration of Wind Farms, Review Article in the Int. Jour. of
Wind Energy Vol. 6, No. 3, 2003, pp. 281 -295.
10
[16] P. Ledesma, J. Usaola, J.L. Rodriguez, Transient Stability of a Fixed Speed Wind
Farm, Int. Jour. of Renewable Energy, Vol. 28, No. 9, 2003, pp. 1341 - 1355.
[17] A.D. Hansen, P. Sorensen, L. Janosi, J. Bech, Wind Farm Modelling for Power
Quality, Proc. of the 27
th
Annual Conference of the IEEE Industrial Electronics
Society, 3, 2001, pp. 1959 -1964.
[18] V. Akhmatov, H. Knudsen, A.H. Nielsen, J. K. Pedersen, N.K. Poulsen, Modelling
and Transient Stability of Large Wind Farms, Int. Jour. of Electrical Power and
Energy Systems, Vol. 25, No. 2, 2003, pp. 123-144.

[19] J.L. Rodriguez-Amenedo, S. Arnalte, J.C. Burgos, Automatic Generation Control of
a Wind Farm with Variable Speed Wind Turbines, IEEE Trans. Energy Conversion,
Vol. 17, No. 2, 2002, pp. 279 -284.
[20] J.R. Saenz, A. Tapia, G. Tapia, F. Jurado, X. Ostolaza, I. Zubia, Reactive Power
Control of a Wind Farm through Different Control Algorithms, Proc. of the4
th
IEEE
International Conference on Power Electronics and Drive Systems, 1, 2001, pp. 203 -
207.
[21] G. Tapia, A. Tapia, J.R. Saenz, A New Simple and Robust Control Strategy for Wind
Farm Reactive Power Regulation, Proc, of the 2002 International Conference on
Control Applications, 2, 2002, pp. 880 -885.
[22] A. Tapia, G. Tapia, X. Ostolaza, J. Molina, J. Saenz, Digital Simulation Performance
of a Wind Farm, Proc. of the 10
th
Mediterranean Electro-technical Conference, 3,
2000, pp.1153 -1156.
[23] Z. Saad-Saoud, M.L. Lisboa, J.B. Ekanayake, N. Jenkins, G. Strbac, Application of
STATCOMs to Wind Farms, IEE Proceedings on Gen. Trans., Distr, Vol. 145, No. 5,
1998, pp. 511 -516.
[24] W. Lu, B.T. Ooi, Multi-Terminal DC Transmission System for Wind Farms, Proc.
of the 2001 IEEE PES Winter Meeting, 3, 2001, pp.1091 -1096.
[25] W. Lu, B.T. Ooi, Optimal Acquisition and Aggregation of Offshore Wind Power by
Multi-Terminal Voltage-Source HVDC, IEEE Trans, Power Delivery, Vol. 18, No.
1, 2003, pp. 201 -206.
[26] Johnson, Gary L., Wind Energy Systems, Englewood Cliffs, NJ: Prentice Hall,
1985.
[27] Federico Milano, Documentation for PSAT version 1.3.3, January 26, 2005.
[28] Howard Demuth, and Mark Beale, Neural Network Toolbox For Use with Matlab
Users Guide Version 4, Available at: http://www.mathworks.com
[29] European Wind Energy Association, Wind Force 12 - A Blueprint to Achieve 12% of
the World's Electricity from Wind Power by 2020, Technical report, 2001.


Table 1: Wind turbine parameters
Rated Power (kW) 20
Radius (m) 5
Drive train inertia (kg.m
2
) 1270
Gear ratio 11.43
Operation angular speed (rpm) 105
Rated wind speed (m/s) 11.7
Cut-in speed (m/s) 6.5
Furling speed (m/s) 23


11












12

















13












14














15













16

You might also like