A23
2011 IEEE Student Conference on Research and Development (SCOReD)
Improving Dynamic Response of Wind Turbine
Driven DFIG with Novel Approach
Mostafa Eidiani*, Natan Asghari shahdehi
Hossein Zeynal
Department of Electrical Engineering,
Bojnourd Branch, Islamic Azad University,
Bojnourd, Iran
*E-mail: eidiani@bojnourdiau.ac.ir
Centre of Electrical Energy Systems (CEES),
Universiti Teknologi Malaysia (UTM),
81310 Skudai, Johor, Malaysia
E-mail: hzeynal@ieee.org
Abstract— The frequency converter is the most sensitive part in
the variable-speed wind turbine generator system equipped with
a double-fed induction generator (DFIG). The frequency
converter is normally controlled by a set of PI controllers. In
order to improve the response of DFIG when subjected to system
disturbances, the best way is to tune the PI controllers of the
frequency converter. Due to the high complexity of the system,
the tuning of these PI controllers is very difficult. In this paper an
approach is offered to improve the response of DFIG when
subjected to system disturbances using Hybrid Particle Swarm
Optimization and Genetic Algorithm (PSO-GA). In this case,
tuning all PI controllers’ parameters is considered. The results
show that the proposed algorithm is well suited in terms of
accuracy and quick response.
at the stator terminals, but the rotor terminals are connected to
the grid via a partial-load variable frequency AC/DC/AC
converter (VFC) and a transformer. The VFC only needs to
handle a fraction (25-30%) of the total power to achieve full
control of the generator. Compared to the fixed-speed wind
turbine with IG, the VSWT with DFIG can provide decoupled
control of the active and reactive power of the generator, more
efficient energy production, improved. The behaviour of the
VFC and the associated WTGS relies on the performance of
its control system. Using well designed controllers, it is
possible to increase the chance of the WTGS to remain in
service during grid disturbances. In the last decade, various
modern control techniques such as adaptive control, variable
structure control and intelligent control [6]-[8], have been
intensively studied for controlling the nonlinear components in
power systems. However, these control techniques have few
real applications probably due to their complicated structures
or the lack of confidence in their stability. Therefore, the
conventional PI controllers, are still the most commonly used
control techniques in power systems due to their simple
structures, as can be seen in the control of the WTs equipped
with DFIGs [9, 10]. Unfortunately, tuning the PI controllers is
tedious and might be difficult to tune the PI gains properly due
to the nonlinearity and high complexity of the system. Over
the years, heuristic search-based algorithms such as genetic
algorithms (GAs), tabu search algorithm and simulated
annealing (SA) have been used for power system stabilizers
(PSS) design [11]. However, when the parameters which have
been optimized are highly correlated, the performance of these
heuristic search algorithms degrades [12]. Recently, a new
technique has been successfully used for single- and
multiobjective nonlinear optimization, based on swarm
intelligence called particle swarm optimization (PSO) [13].
The use of PSO for designing a single PID controller in the
automatic voltage regulator (AVR) system of a conventional
turbo generator has been reported in [14]. This design,
however, is based on the step response and did not investigate
the transient performance of the controller. In [15] an
approach is presented to use the particle swarm optimization
algorithm to design the optimal PI controllers for the rotor side
converter of the DFIG. In [21] it is shown that the particle
Keywords- Double-fed induction generator (DFIG); variablespeed Wind Turbine; Genetic Algorithm (GA); Particle Swarm
Optimization (PSO)
I.
INTRODUCTION
Wind turbines can either operate at fixed speed or variable
speed. For a fixed speed wind turbine, the generator is directly
connected to the electrical grid. For a variable speed wind
turbine, the generator is controlled by power electronic
equipment. There are several reasons for using variable-speed
operation of wind turbines; among those are the possibilities
of reducing stresses of the mechanical structure, acoustic noise
reduction and the possibility of controlling active and reactive
power [1]. Most of the major wind turbine manufactures are
developing new larger wind turbines in the 3-5 MW range [2].
These large wind turbines are all based on variable-speed
operation with pitch control using a directly driven
synchronous generator (without gearbox) or a double-fed
induction generator (DFIG). Fixed-speed induction generators
with stall control are regarded as unfeasible [3] for these large
wind turbines. Today, double-fed induction generators are
commonly used by the wind turbine industry for larger wind
turbines [4,5]. Compared to the variable speed wind turbine
equipped with a synchronous generator (SG), in which a full
load variable frequency control (VFC) is connected directly
between the generator stator and the grid, the VFC of the
DFIG is smaller in size and therefore much cheaper. In the
DFIG concept, the induction generator (IG) is grid-connected
978-1-4673-0102-2/11/$26.00 ©2011 IEEE
425
A23
2011 IEEE Student Conference on Research and Development (SCOReD)
(Pb) as well as the best position of its neighbours (Pg), and
then compute a new position that the ‘‘particle’’ is to fly to.
Assuming the dimension of a searching space to be D, the
total number of particles is n, the position of the ith particle
can be expressed as vector Xi = (xi1,xi2, . . .,xiD); the best
position of the ith particle being searched until now is denoted
as Pib = (pi1,pi2, . . .,piD), and the best position of the total
particle swarm being searched until now is denoted as vector
Pg = (pg1,pg2, . . .,pgD); the velocity of the ith particle is
represented as vector
Vi = (vi1, vi2. . . viD). Then the original PSOA is described as
[19]:
swarm optimization (PSO) was demonstrated to converge
rapidly during the initial stages of a global search, but around
global optimum, the search process will become very slow. In
this paper, the hybrid PSO-GA algorithm is used to find the
optimal parameters of the various PI controllers for the rotor
and grid side converters of the variable frequency converter
simultaneously.
II. GENETIC ALGORITHM (GA)
The GA is a search algorithm based on the mechanism of
natural selection and natural genetics [13]. In a simple GA,
individuals are similar to a chromosome that codes for the
variables of the problem. The strength of an individual is the
objective function that must be optimized. Population of
candidates evolves by genetic operators: mutation, crossover,
and selection. The characteristics of good candidates have
more chances to be inherited, because good candidates live
longer. So the average strength of the population rises through
the generations. Finally, the population stabilizes, because no
better individual can be found. At that stage, the algorithm has
converged, and most of the individuals in the population are
identical, and represent a suboptimal solution to the problem.
A GA is governed by three factors: the mutation rate, the
crossover rate, and the population size. The implementation of
GA is detailed in [14]. GA is one of the effective methods for
optimization problems especially in non-differential objective
functions with discrete or continues decision variables [15].
As with any search algorithm, the optimum solution is
obtained only after much iteration. The speed of the iterations
is determined by the length of the chromosome and the size of
the populations. There are two main methods for GA to
generate itself, namely generational and steady state. In
generational, an entire population is replaced after iteration
(generation). Whereas in steady state, only a few members of
the population are discarded at each generation, and the
population size remains constant [16]. One of the drawbacks
of GA is its possibility to converge prematurely to a
suboptimal solution [17]. Another drawback of this algorithm
is its high sensitivity to initial population [16, 18]. There are a
few main limitations of a GA when applied to problems [16]
vid (t + 1) = vid (t ) + c1 * rand() *[ pid (t ) − xid (t )] +
(1)
c2 * rand() *[ pgd (t ) − xid (t )]
xid (t + 1) = xid (t ) + vid (t + 1)
1≤ i ≤ n
(2)
1≤ d ≤ D
Where c1, c2 are the acceleration constants with positive
values; rand is a random number between 0 and 1; w is the
inertia weight. In addition to the parameters c1, and c2
parameters, the implementation of the original algorithm also
requires placing a limit on the velocity (Vmax). After adjusting
the parameters w and Vmax, the PSO can achieve the best
search ability. The adaptive particle swarm optimization
(APSO) algorithm is based on the original PSO algorithm,
firstly proposed by Shi and Eberhart in 1998 [20]. The APSO
can be described as follows:
vid (t +1) = w*vid (t) + c1 *rand()*[ pid (t) − xid (t)]+
c2 *rand()*[ pgd (t) − xid (t)]
xid (t + 1) = xid (t ) + vid (t + 1)
1≤ i ≤ n
1≤ d ≤ D
(3)
In which w is a new inertial weight. These algorithms can
reduce w gradually as the generation increases by adjusting
the parameter w. In the searching process of the PSO
algorithm, the searching space will reduce gradually as the
generation increases. So the APSO algorithm is more
effective, because the searching space reduces step by step
nonlinearly, so the searching step length for the parameter w
here reduces correspondingly. Similar to GA, after each
generation, the best particle of particles in the last generation
will replace the worst particle of particles in current
generation, thus a better result can be achieved. Generally, in
the beginning stages of algorithm, the inertial weight w should
be reduced rapidly, around optimum, the inertial weight w
should be reduced slowly [20].
1. The fitness function must be well-written
2. It is a blind and undirected search
3. It is a stochastic search
4. It is sensitive to initial parameters
5. It is computationally expensive
6. What is the stopping criterion?
III. PARTICLE SWARM OPTIMIZATION (PSO)
Particle swarm optimization (PSO) is a kind of algorithm to
search for the best solution by simulating the movement and
flocking of birds. The algorithm works by initializing a flock
of birds randomly over the searching space, in which every
bird is called a ‘‘particle’’. These ‘‘particles’’ fly at a certain
velocity and find the global best position after some iteration.
At each iteration, each particle can adjust its velocity vector,
based on its momentum and the influence of its best position
IV. TUNING THE PARAMETERS OF THE PI
CONTROLLERS USING HYBRID PSO-GA:
In the wind turbine equipped with the DFIG, the variable
frequency converter and its power electronics (IGBT-
426
A23
2011 IEEE Student Conference on Research and Development (SCOReD)
Figure 1. Effect of grid-side PI controllers
transient disturbances [22]. However, tuning controllers is
difficult to achieve a set of optimal parameters manually. In
this section, the PSO-GA algorithm is applied to find the
optimal parameters of the RSC and GSC controllers
simultaneously. The genetic algorithm is very sensitive to the
initial population. In fact, the random nature of the GA
operators makes the algorithm sensitive to initial population
[21]. This dependence to the initial population is in such a
manner that the algorithm may not converge if the initial
population is not well selected. However if the initial
population is well selected, the performance of the algorithm
may enhance. PSO, on the other hand, is not as sensitive as
GA to initial population. One of the characteristics of PSO is
its fast convergence towards global optima in the early stage
of the search and its slow convergence near the global optima.
The idea behind this paper is the combination of the PSO and
GA algorithm in such a way that the performance of the newly
switches) are the most sensitive part. The converter action will
probably determine the operation of the wind turbine during
transient disturbances in the power grid. Grid faults, for
example, even far away from the location of the wind farm,
can cause voltage sags at the connection point of the wind
turbine. This voltage sag will result in an imbalance between
the turbine input power and the generator output power and
therefore a high current in the stator windings of the DFIG.
Because of the magnetic coupling between stator and rotor,
this current will also flow in the rotor circuit and the
converter. Since the power rating of the IGBT converter is
only 25-30% of the induction generator power rating, this
over-current can lead to the destruction of the converter. On
the other hand, the behavior of the converter depends on the
control system. If the controllers are tuned properly, it is
possible to limit the over current of the rotor circuit and
therefore improve the converter’s performance during the
TABLE I.
PI PARAMETERS OF CASE1
Initial
design
Optimal
design
kP
Vdc
kI
Vdc
kP
kI
kP
kI
kP
kI
grid
grid
Q
Q
rotor
rotor
0.0001
0.15
3
80
0.07
7
0.4
0.07
0.0278
0.4957
1.9234
269.01
0.07
7
0.4
0.07
427
A23
2011 IEEE Student Conference on Research and Development (SCOReD)
Figure 2. Effect of rotor-side PI controllers
established algorithm is better than the PSO or GA algorithm.
The objective of the PSO is to find the optimal parameters of
the PI controllers. Generally, the PI controller performance in
the time domain can be measured by a set of parameters: the
overshoot Mp, the rise time tr, the settling time ts, and the
steady-state error Ess. In this paper, the objective is to reduce
the over-current in the rotor circuit during the grid faults.
Therefore, a performance measure function is defined as
follows [23]:
Case1. Effect of grid-side PI controllers
In this case the effect of grid side PI controllers on rotor
current overshoot is considered. By tuning the parameters of
grid side PI controllers ([Kp,vdc, KI,vdc, KP,grid, KI,grid])
using PSO-GA a 30.04% reduction is obtained (Fig.6).
Following Table 1 shows the tuned parameters.
Case2. Effect of rotor-side PI controllers
In this case the effect of rotor-side PI controllers on rotor
current overshoot is considered. By tuning the parameters of
rotor side PI controllers ([Kp,Q, KI,Q, KP,rotor, KI,rotor])
using PSO-GA a 29.86% reduction in rotor current overshoot
is obtained (Fig.7). Following Table 2 shows the tuned
parameters.
Case3. Effect of All PI controllers
f ( x) = c1.ΔI r , max + (1 − c1 ).(t s − t0 ) + c2 . E ss
V. SIMULATION AND DISCUSSIONS
TABLE II.
PI PARAMETERS OF CASE2
Initial design
kP
Vdc
kI
Vdc
kP
kI
kP
kI
kP
kI
grid
grid
Q
Q
rotor
rotor
0.0001
0.15
3
80
0.07
7
0.4
0.07
0.0001
0.15
3
80
0.0032
8.12
1.72
9.072
Optimal design
428
A23
2011 IEEE Student Conference on Research and Development (SCOReD)
[9]
In this case the effect of rotor-side and grid side PI controllers
on rotor current overshoot is considered. By tuning the
parameters of rotor-side and grid-side PI controllers
simultaneously ([Kp,vdc, KI,vdc, KP,grid, KI,grid , Kp,Q,
KI,Q,KP,rotor, KI,rotor]) using PSO-GA a 30.06% reduction
in rotor current overshoot is obtained (Fig.8).
[10]
[11]
V. CONCLUSION
Wind farm is represented by an aggregated model in
which hundreds of individual wind turbines and DFIGs are
modelled as one equivalent DFIG driven by a single
equivalent wind turbine. In this paper, the effect of rotor-side
controllers, grid-side controllers, and all PI controllers on over
current in the rotor circuit is studied. The hybrid particle
swarm optimization and Genetic algorithm (PSO-GA) is then
used to find the optimal parameters of the PI controllers for
both the RSC and GSC. The PSO-GA is approximately 3
times faster than RCGA and the accuracy of PSO-GA is
better. The results show that PI controller of Vdc regulator is
the most effective part in reducing the over current of the rotor
circuit. And PI controllers of the grid-side converter reduced
the over current by 30.04% while PI controllers of the rotorside converter reduced the over current by 29.86%.
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
ACKNOWLEDGMENT
[20]
This work was partially supported by the Malaysian
Government under FRGS grant Vot No. 78460.
[21]
REFERENCES
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
V. Akhmatov, Analysis of Dynamic Behavior of Electric Power
Systems with Large Amount of Wind Power, Ph.D. thesis,
Technical University of Denmark, Kgs. Lyngby, Denmark, Apr.
2003.
M. V. A. Nunes, J. A. Pecas Lopes, H. H. Zurn, U. H. Bezerra, and
R. G. Almeida, “Influence of the variable-speed wind generators in
transient stability margin of the conventional generators integrated
in electrical grids,” IEEE Trans. Energy Conversion, vol. 19, no. 4,
Dec. 2004, pp. 692-701.
T. Ackermann and L. S¨oder, “An overview of wind energy-status
2002” Renew. Sustain. Energy Rev., vol. 6, no. 1–2, pp. 67–128,
Feb./Apr. 2002.
Nordex. (2005, Jan.) N80/2500 kW N90/2300 kW. Brochure.
[Online]. Available: http://www. Nordex –online .com /e /online
service/ download/ dateien/PB N80 GB.pdf [5] (2005, Jan.) V1204.5 MW Offshoreleadership. Brochure. [Online]. Available:
http://www.vestas.com/pdf/produkter/AktuellBrochurer/v120/V12
0%20UK.pdf
M. Lown, E. Swidenbank, B. W. Hogg, “Adaptive fuzzy logic
control of a turbine generator system,” IEEE Trans. Energy
Conversion, vol. 12, no. 4, Dec. 1997, pp. 394-399.
Y.-Y. Hsu, C.-H. Cheng, “Variable structure and adaptive control
of a synchronous generator,” IEEE Trans. Aerospace and
Electronic Systems, vol. 24, no. 4, July 1988, pp. 337-345.
G. K. Venayagamoorthy, R. G. Harley, and D. C. Wunsch,
“Comparison of heuristic dynamic programming and dual heuristic
programming adaptive critics for neurocontrol of a
turbogenerator,” IEEE Trans. Neural Networks, vol. 13, no. 3, May
2002, pp. 764- 773
J. Morren and S. W. H. de Haan, “Ridethrough of wind turbines
with doublyfed induction generator during voltage dip,” IEEE
Trans. Energy Conversion, vol. 20, no. 2, Jun. 2005, pp. 435-441.
[22]
[23]
[24]
[25]
R. Pena, J. C. Clare, and G. M. Asher, “Doubly fed induction
generator using backto- back PWM converters and its application
to variable-speed wind-energy generation,” IEE Proceedings –
ElectRic Power Applications, vol. 143, no. 3, May 1996, pp. 231241.
M. A. Abido, “Robust design of multimachine power system
stabilizers using simulated annealing,” IEEE Trans. Energy
Conversion, vol. 15, no. 3, Sept. 2000, pp. 297-304
D. B. Fogel, Evolutionary Computation: Toward a New
Philosophy of Machine Intelligence, New York: IEEE Press, 2000,
ISBN 0-7803-5379-X.
K.F.Man, K.S.Tang and S.Kwong, "Genetic Algorithm concepts
and application", IEEE Trans.On Industrial Electronics, vol 43
no.5, pp.519-534,Oct 1996.
J.H. Holland, Adaptation in Natural and Artificial Systems. Ann
Arbor, MI: The University of Michigan Press, 1975.
T.S. Chung, Y.Z. Li," A Hybrid GA Approach for OPF with
Consideration of FACTS Devices", IEEE Power Engineering
Review, February 2001
[ Betram Koh Lin Hon "Accelerated Genetic Algorithm in Power
System Planning" Electrical Engineering thesis 2003
Stephane Gerbex, Richard Cherkaoui and Alain.J.Germond
"Optimal location of multi-type FACTS devices by means of
Genetic algorithm" IEEE Trans. Power system Vol.16 pp 537-544
August 2001
Jose Miva, Jose Ramon, Alvarez "Artificial Intelligence and
Knowledge Engineering Applications" Ebook
James Kennedy and Russel Eberhart, "Particle Swarm
Optimization" Proc. of IEEE International conference on neural
networks, Vol 14, pp 1942 – 1948 December 1995
Yuhui Shi, Russel.C.Eberhart," Empirical study of particle swarm
optimization", Proc. of the congress on Evolutionary computation,
Vol.13, pp 1945-1950, July 1999
Mohammadi, M. Jazaeri, "A hybrid particle swarm optimizationgenetic algorithm for optimal location of SVC devices in power
system planning",
M. Fendereski, A. Mohammadi, M.T. Arabyarmohammadi,
S.Amirkhan, M. Sadighi, "Response Improvement of Doubly Fed
Induction Generators Driven by Wind Turbines Using PSO and
RCGA Algorithms"
Wei Qiao; Venayagamoorthy, G.K.; Harley, R.G. "Design of
Optimal PI Controllers for Doubly Fed Induction Generators
Driven by Wind Turbines Using Particle Swarm Optimization",
Neural Networks, 2006. IJCNN; International Joint Conference on
Neural Networks, Page(s):1982-1987, 10.1109/IJCNN 2006.
M. Eidiani and M.H.M. Shanechi, “FAD-ATC, A New Method for
Computing Dynamic ATC”, International Journal of Electrical
Power & Energy System,vol. 28, n. 2, pp. 109-118, Feb. 2006.
M. Eidiani, “A Reliable and efficient method for assessing voltage
stability in transmission and distribution networks”, International
Journal of Electrical Power and Energy Systems, vol.33, n.3, pp.
453-456, March 2011.
H.Zeynal, A.K.Zadeh, K. M. Nor, M. Eidiani, “Locational
Marginal Price (LMP) AssessmentUsing Hybrid Active and
Reactive Cost Minimization”, International Review of Electrical
Engineering (I.R.E.E.),vol.5, n.5, pp.2413-2418, October 2010
APPENDIX
Induction generator: rated power =10MW,
Rated voltage = 575 V, Rs = 0.00706 pu,
Lls = 0.171 pu, Rr = 0.005 pu, Llr = 0.156 pu, Lm = 2.9 pu, H = 5.04 sec.
Converter: PWM frequency = 1620 Hz,
Rated voltage DC = 1200 V,
Capacitor DC link= 6e-4 Farads
Power network: 2500MVA, 120KV base,
R1 = R2 = 0.1153 ohms/Km, L1 = L2 = 1.05 e-3 H/Km, line Length = 30 km.
429