264 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO.

1, MARCH 2009

Optimal Control of Matrix-Converter-Based

WECS for Performance Enhancement and
Efficiency Optimization
Vinod Kumar, R. R. Joshi, and R. C. Bansal, Senior Member, IEEE

Abstract—This paper describes a new wind energy conversion conversion system (WECS), composed of a wind turbine, an
scheme, where fuzzy logic principles and four-leg-improved novel electric generator, a power electronic converter, and the cor-
matrix converter (MC) model are used for performance enhance- responding control system. Based on the types of components
ment and efficiency optimization, particularly for sites where most
of the time low wind speeds prevail. The MC is intended as the used, different WECS structures can be realized. However, the
interface medium between the induction generator and the utility objective in all structures is the same, i.e., the wind energy at
grid. The power factor at the interface with the grid is controlled varying wind velocities has to be converted to electric power
by the MC to ensure purely active power injection into the grid at the grid frequency [3]–[5]. Wind turbine configurations for
for optimal utilization of the installed wind turbine capacity. The extracting energy from the wind are categorized based on hori-
generation system has fuzzy logic control with vector control in the
inner loops. Fuzzy controller tracks the angular frequency with zontal or vertical axis, number of blades, and power rating.
the wind velocity to extract the maximum power and programs Conventionally, the ac–dc–ac conversion system is adopted
the machine flux for light load efficiency improvement. The com- for wind power applications [6]–[11]. The generator-side con-
plete control system has been developed, analyzed, and validated verter is a diode rectifier, a thyristor rectifier, or a voltage source
by simulation study. converter (VSC). The utility-side converter is often a VSC unit.
Index Terms—Adaptive fuzzy logic controller, matrix converter The main shortcomings of the ac–dc–ac conversion unit are:
(MC), squirrel cage induction generator, wind energy, wind power 1) large physical dimensions, high weight, and excessive vol-
generation. ume/footprint of the dc link component, i.e., the dc capacitor or
the inductor; 2) the low reliability of dc capacitor; and 3) the
I. INTRODUCTION
poor line power factor and harmonic distortion in line and ma-
UE to the increasing demand of electrical energy and envi- chine currents. The IEEE Standard 519 [12] severely restricts
D ronmental concerns, a considerable amount of effort is be-
ing made to generate electricity from renewable energy sources.
line harmonic injection. Recently, matrix converter (MC)-based
WECS has been proposed [13], [14] to overcome some of afore-
The major advantages of using renewable energy sources are mentioned problems, but to the authors’ best knowledge, im-
abundance and lack of harmful emissions. Wind turbines are proved space-vector-pulsewidth-modulation (SVPWM)-based
arguably the most developed source of renewable electrical en- four-output leg MC topology for wind power applications has
ergy with ratings of commercial wind turbines now exceeding neither been proposed nor investigated.
10 MW [1]. Also, it is the most cost competitive of all the en- In this paper, a new idea is explored. This paper proposes a
vironmentally clean and safe renewable energy sources in the new WECS scheme that uses a four-output-leg-improved MC
world. It is also competitive with fossil-fuel-generated power topology. The fourth leg creates a three-phase plus neutral utility
and much cheaper than nuclear power. power supply, which provides the facility to supply unbalanced
It has been estimated that even if 10% of raw wind potential and single-phase loads. This enables the generator to provide
could be put to use, all the electricity needs of the world would stable power over a range of wind speeds and load condi-
be met [2]. Of course, the main drawback of wind power is tions. It is designed to meet tight harmonic specifications along
that its availability is somewhat statistical in nature and must with improved voltage gain and simplified control, free from
be supplemented by additional sources to supply the demand commutation problems, and providing extremely fast transient
curve. The wind energy can be harnessed by a wind energy response.
MC provides direct ac–ac conversion and is considered an
Manuscript received November 26, 2007; revised May 28, 2008. First emerging alternative to the conventional two-stage ac–dc–ac
published January 13, 2009; current version published February 19, 2009. Paper converter topology [15], [16] as it provides a large number of
no. TEC-00456-2007.
V. Kumar and R. R. Joshi are with the Department of Electrical Engi-
control levers that allow for independent control on the output
neering, College of Technology and Engineering, Maharana Pratap Univer- voltage magnitude, frequency, and phase angle, as well as the
sity of Agriculture and Technology (MPUAT), Udaipur 313001, India (e-mail: input power factor. When compared with the ac–dc–ac converter
[email protected]; [email protected]).
R. C. Bansal is with the School of Information Technology and Electrical
system, the bold feature of MC is elimination of the dc-link re-
Engineering, The University of Queensland, St. Lucia, Campus, Qld. 4072, active elements, e.g., bulky electrolytic capacitors and/or induc-
Australia (e-mail: [email protected]). tors, thus allowing for extreme environment storage and wider
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
operating temperature range than with electrolytic capacitors.
Digital Object Identifier 10.1109/TEC.2008.2008921 The MC topology provides a compact solution and yields high

0885-8969/$25.00 © 2009 IEEE

KUMAR et al.: OPTIMAL CONTROL OF MATRIX-CONVERTER-BASED WECS 265

Fig. 1. Schematic diagram of the proposed wind energy conversion scheme.

input and output quality. It draws sinusoidal input current and,

depending on the modulation technique, can be arranged so that
unity displacement factor is seen at the supply side irrespective
of the type of load.
Fig. 2. Power–speed characteristics of a wind turbine for different wind
This paper presents a simulation analysis to validate the the- velocities.
oretical concepts (the experimental work is in progress and will
be reported later), describes a variable-speed wind turbine sys-
tem, particularly for low-wind-velocity-dominating sites, with
an induction generator and a four-output leg commutation-free
MC where adaptive fuzzy logic control along with improved
SVPWM switching have been used extensively to maximize the
power output and enhance system performance.

II. PROPOSED WIND GENERATION SYSTEM DESCRIPTION

Fig. 1 shows the block diagram of the proposed system par-
ticularly suitable for sites where most of the time low wind Fig. 3. C P − λ characteristics of wind turbine.
speeds prevail. Here, a four-output-leg-improved topology of
MC is used for interfacing WECS with the grid, and fuzzy logic
control has been used for performance enhancement and effi- shown in Fig. 3 [20], for the same wind turbine. The aerody-
ciency optimization. As the entire power generated by the wind namic torque (Tm ) and mechanical power (P0 ) generated by a
turbine is transferred through the MC, this paper targets low-to- turbine are given by (1) and (2) [21], respectively


medium-power wind turbines. As most of the time, low wind ρπRω3
speed prevails, so SCIG is preferred [17]–[19] in the proposed Tm = Cp (λ) 0.5 Vω2 (1)
ηgear
system.
The MC interfaces the SCIG with the grid and implements 1
P0 = ρCp Ar Vw3 (2)
shaft speed control to achieve maximum power point tracking 2
(MPPT) at varying wind velocities. It also performs power factor where P0 is the power (in watts), ρ the air density (in kilograms
control at the grid interface and satisfies the volt ampere reactive per cubic meter), Cp a dimensionless factor called power coef-
(var) demand at the induction generator terminals. ficient, Ar the turbine rotor area (in square meters) (Ar = πRr2 ,
For a particular wind velocity (Vw ), there will be an optimum where Rr is the rotor blade radius), ηgear is the gear efficiency,
setting of angular frequency (wref ). The speed loop will generate and Vw the wind speed (in meters per second).
the torque component of machine current so as to balance the The power coefficient is related to the tip-speed ratio λ and
developed torque with the load torque. The variable-voltage rotor blade pitch angle θ according to (3) [21], as shown in Fig. 3
variable-frequency power from the induction generator will be  
151
pumped to the utility grid through MC. In the following sections, Cp (λ, θ) = 0.73 −0.58θ−0.002θ2.14 − 13.2 e−18.4/ λi
λi
different elements of the system are described. (3)
where
A. Wind Turbine Model 1
λi = (4)
The steady-state wind turbine model at various wind speeds is [1/(λ − 0.02θ)] − [0.003/(θ3 + 1)]
given by the power–speed characteristics shown in Fig. 2 [19]. and
At a given wind speed, the operating point of the wind turbine
ωm R r
is determined by the intersection between the turbine character- λ= . (5)
istics and load characteristics. In practice, the characteristics of Vw
a wind turbine can also be represented in a simplified form of In (5), ωm is the angular speed of the turbine shaft. The
power performance coefficient (CP ) and tip-speed ratio (λ), as theoretical limit for Cp is 0.59 according to Betz’s law [21], but
266 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 1, MARCH 2009

comes a “fictitious link voltage” as an intermediate control vari-

able in the control software. Since input terminal a, b, and c are
connected to voltage sources, they must not be short-circuited
through the MC switches. Similarly, since output terminals A,
B, and C are connected to current sources, they must not be
open-circuited at any instant. Based on these constraints, the
switching function for each switch is

1, switch SK j closed
SK j = (6)
0, switch SK j open
Fig. 4. Schematic diagram of the improved MC topology.
where K = {A, B, C}, j = {a, b, c}, and
SK a + SK b + SK c = 1, K = {A, B, C}. (7)
its practical range of variation is 0.2–0.4. In this paper, the rotor
pitch angle is assumed to be fixed. Equations (1)–(5) suggest The mathematical relationship between input and output in-
that the aerodynamic torque (Tm ∝ Vw2 ) and the mechanical stantaneous phase voltages is
power (Tm ∝ Vw2 ) generated by the wind turbine at a given wind     
vA SA a SA b SA c va
velocity are the functions of the shaft speed. This means that  vB  =  S B a S B b S B c   vb  (8)
at reduced speed-light-load steady-state conditions, generator vC SC a SC b SC c vc
efficiency can be improved by programming the flux [22]. At
any given wind velocity, maximum power can be captured from where vA , vB , vC and va , vb , vc represent instantaneous output
the wind if the shaft speed is adjusted at the value corresponding and input phase voltages, respectively. Based on (8), line-to-line
to the peak power. The idea in this paper is to change the angular voltages and phase currents at the output and input terminals are
    
frequency of the induction generator through SVPWM control vA B SA a − SB a SA b − SB b SA c − SB c va
of MC to track the shaft speed corresponding to the maximum  vB C  =  S B a − S C a S B b − S C b S B c − S C c   vb 
turbine power at all times. vC A SC a − SA a SC b − SA b SC c − SA c vc
(9)
B. Modulation and Control of MC     
ia SA a SB a SC a iA
By properly operating the switches in the MC, one can achieve  ib  =  SB a SB b SC b   iB  (10)
control on the output voltage magnitude, frequency, and phase
ic SA c SB c SC c iC
angle, as well as control on the input displacement angle. Fig. 4
shows the schematic diagram of the improved MC topology where vA B , vB C , and vC A (iA , iB , and iC ) are the output in-
developed by Yue et al. [23]. stantaneous voltages (currents), and ia , ib , and ic are the input
Considering various advantages, and with the present trend instantaneous phase currents. The instantaneous input source
of decreasing converter and control cost, this type of conversion voltages are given by
system has the potential to be universally accepted in the future.
va = vim sin(wi t) (11)
As shown in Fig. 4, MC offers four control levers that can be  
used to control the input displacement angle, output voltage 2π
vb = vim sin wi t − (12)
magnitude, frequency, and phase angle. 3
Based on the dynamic MC model [24], control and space  
2π
vector modulation (SVM) at both generator and grid sides as vc = vim sin wi t + . (13)
voltage sources, instead of current sources, have been developed, 3
which enable the MC to: In this paper, an input modulation approach is developed to
1) achieve the maximum voltage transfer ratio without uti- work with high frequencies of fundamental voltages. Also, it
lizing the third harmonic component injection method; is desirable to have a capability of operating with reasonably
2) accommodate any input power factor independent of the good stability under significant line imbalances, say over 10%,
output power factor; for a given time period. The fundamental frequency of the input
3) reduce the effective switching frequency in each cycle, power can be up to several 100 Hz. Our objective is to minimize
and thus the switching losses; the switching frequency on the line side and maximize the volts–
4) minimize harmonics. seconds that can be transferred to the second-stage modulation
The MC model used in the proposed system can accommo- to increase the output voltage. The switching function per input
date: 1) large terminal voltage excursions at either side of the phase is given by
MC; 2) large frequency variations at either side of the MC;
Sj i (t)= [Sai (wi t−ϕ0 ) Sbi (wi t − ϕ0 ) Sci (wi t − ϕ0 )] (14)
3) any input (source) to output (grid) frequency ratio; and 4)
unbalanced grid conditions. where j ∈ {a, b, c}, Sj i is a switching matrix function con-
The overall modulation is split into two stages: input-side sisting of a set of switching vectors, and Sai , Sbi , and Sci are
modulation and load-side modulation. In between, there be- the switching vectors in time domain, consisting of a set of
KUMAR et al.: OPTIMAL CONTROL OF MATRIX-CONVERTER-BASED WECS 267

TABLE I
EXAMPLE SWITCHING ORDER

Fig. 6. Example switching pattern (half period).

Fig. 5. Input-side pulsewidth modulation patterns with low switching fre- the line-side PWM is illustrated in Fig. 5(b), which shows that
quency. (a) Line voltage source PWM pattern. (b) Line voltage source PWM the new switching function provides more volts–seconds to the
pattern. grid, thus increasing the average value of RVi (t) and achieving
a higher voltage transfer ratio. The number of the switch angles
switching angles, for the input phase voltage of Va , Vb , and Vc , in the switching vector Sj i can be conveniently increased or
respectively. By premultiplication of the input voltages in (11)– changed in control software implementation, from the pattern
(13) by the switching matrix in (14), we can mathematically given in Fig. 5(a) and (b) to improve the input current quality.
obtain an intermediate variable in the control software that can The second-stage modulation adapts the existing dc–ac in-
be expressed by verter modulation strategy by using the intermediate control
  variable RVi (t) as its input. For modulation index (m) < 1, the
sin (wi t) triple harmonics were removed from the modulation algorithm.
  
 2π   This reduces the common mode voltage presence in the system,
 sin wi t −  which is desirable. This differentiates the proposed SVPWM
RVi (t) = (Sj i (t))  3 . (15)
   control from the existing SVM where the common mode volt-
 2π 
sin wi t + age exists in the whole operating range.
3 For the standard 3 × 3 MC, there are 27 (33 ) switching
In fact, RVi (t) is a modulated dc voltage. It is sometimes states [25]. However, with the extra output leg, this converter
referred to as a “fictitious link voltage.” Because it will become has 81 (34 ) switching states. This set of six switching states can
the input variable (voltage) for the second-stage PWM, the av- then be used to create any space vector with the input and output
erage value of RVi (t) plays a key role for achieving a higher sectors. Therefore, for a certain input and output sector, there
voltage transfer ratio of this MC. are two fixed input space vectors and three fixed output space
An example of the proposed new switching matrix function vectors created by using the six switching states, which gives
and its switching vectors Sai , Sbi , and Sci is shown in Fig. 5(a) two switching states per output space vector and three switching
where states per input space vector. Then, it is calculated how much
each space vector contributes to creating the output, or control-
Sai = {1, 0} , 0 < wi t ≤ π ling the input. Once the duty ratios for the six switching states
Sai = {0, −1} , π < wi t ≤ 2π are known, it is then simple to design switching strategy. Tak-
ing an example, the six switching states +1, −3, −4, +6, +13,
where Sbi and Sci are lagging Sai by 120 and 240 electrical and −15 can be arranged as shown in Table I and the result-
degrees, respectively. In comparison with Fig. 5(a), a new set ing switching pattern, using three zero states per half period, is
of modulation patterns with a reduced switching frequency for shown in Fig. 6.
268 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 1, MARCH 2009

Fig. 7. Fuzzy-logic-based control block diagram of a wind generation system.

C. Control System
It appears that fuzzy-logic-based intelligent control [22], [26],
[27] is most appropriate for performance improvement of wind
generation systems. Fig. 7 shows the control block diagram of
the system that uses the power circuit of Fig. 1.
The MC uses vector control in inner current control loop to
permit fast transient response for the system. For a particular
wind velocity, there will be an optimum setting of generator Fig. 8. Angular frequency regulator control system (AFCS).
speed. The speed loop will generate the torque component of
machine current so as to balance the developed torque with the
load torque. In order to reduce the time-consuming process of and FC2 , which dynamically detects online angular frequency
the medium frequencies (MFs) tuning or ameliorate the perfor- that corresponds to the maximum aerodynamic efficiency of the
mance when it does not satisfy the specification, we can apply an wind turbine for a specific wind speed ωref (see Fig. 8). The
online-tuned adaptive fuzzy control system (AFCS). An AFCS following are the meaning of the symbols shown in Fig. 8:
can adapt to its environment and acquire new knowledge by 1) P0 is the output real power;
itself through learning. A possible arrangement of such a sys- 2) ωe is the angular frequency at the terminal of the generator;
tem is the implementation of a fuzzy controller (FC) to adjust 3) ∆ωref is the angular frequency for maximum aerodynamic
the parameters of another FC. This adjustment is accomplished efficiency;
online. The main FCs MFs are tuned online through the super- 4) L∆ωref represents last change in angular frequency for
vised FC that follows the reasoning of an expert, which would maximum aerodynamic efficiency;
manually tune the MFs. AFCS are thus very suitable for the 5) δ represents the phase angle of PWM firing pulse genera-
control of systems, which are fluctuating strongly, such as the tion;
wind turbine generation system. 6) ∆δ represents the phase shift of phase angle δ (positive
for increment of δ and negative for decrement of δ);
III. AFCS DESCRIPTION 7) RF is the regulating factor of FC1 MFs;
8) ∆RF is the step change of RF;
The system has the following fuzzy logic controllers. 9) σ is the variance of ωe1 ;
10) |ωe − ωref | is the absolute mean deviation of ωe from ωref .
A. Angular Frequency Regulator The frequency at the point ωe is compared to ωref , and the
In order to achieve maximum wind power absorption ac- error is passed through FC1 , which produces a signal ∆δ (see
cording to the current wind speed, MC regulates the angular Fig. 8). By accumulating the successive values of ∆δ, the value
frequency of the generator. This is managed through an AFCS of δ comes up according to
(see Fig. 8). This system consists of FC1 , which is the main con-
troller of the AFCS, FC3 , whose main role is to fine tune FC1 , δ news = δ old + ∆δ (16)
KUMAR et al.: OPTIMAL CONTROL OF MATRIX-CONVERTER-BASED WECS 269

Fig. 10. Membership functions for the fuzzy set derivative of ω e − ω re f .

Fig. 9. Membership functions for the fuzzy set ω e − ω re f .

where δ new is the new value of the phase angle δ and δ old is the
previous value of the phase angle δ.
As was previously mentioned, ωref is the angular frequency,
where the wind turbine operates at the maximum aerodynamic
efficiency for a current wind speed. The value of ωref is dy-
namically approached in real time from FC2 (see Fig. 8), us-
ing an MPPT technique. The realization of this technique is
Fig. 11. Membership functions for the fuzzy set phase angle δ.
achieved by changing the reference value of the frequency by
∆ωref and monitoring the corresponding change of the out-
put power ∆P0 . With an increment (or decrement) of ωref , the TABLE II
corresponding increment (or decrement) of output power P0 is FUZZY RULES FOR FC 1
estimated.
If ∆P0 is positive with last positive ∆ωref in per unit (p.u.)
value by L∆ωref (PU) the search is continued in the same di-
rection. If, on the other hand, +∆ωref causes ∆P0 , the di-
rection of search is reversed. MC achieves maximum wind
power acquisition from the wind turbine by driving the an-
gular frequency ωe to its optimal reference value ωref . This is
accomplished by regulating the active power absorbed by MC,
through the modulation of signal, δ of its SVPWM reference
signal. The initial MFs for the input are ωe − ωref shown in Fig. 9.
The variables ∆P0 , ∆ωref , and L∆ωref are described by The term initial is used because these MFs are tuned online
membership functions and rule table. In the implementation by FC3 . These MFs are the initial MFs before their tuning.
of fuzzy control, the input variables are fuzzified, the valid con- Of course, the general form of the final MFs will remain the
trol rules are evaluated and combined, and finally, the output is same. As shown in Fig. 9, five fuzzy subsets are needed for the
defuzzified to convert to the crispy value. The wind vortex and input ωe − ωref : negative big (NB), negative small (NS), OK,
torque ripple can lead the search to be trapped in a minimum positive small (PS), and positive big (PB). For the derivative of
that is not global, so the output ∆ωr is added to some amount ωe − ωref , the fuzzy sets needed are NB, NS, PS, and PB, as
of L∆ωref in order to give some momentum to continue the shown in Fig. 10.
search and avoid such local minima. The controller operates The fuzzy sets required for the phase angle ∆δ are: decrease
on a p.u. basis so that the response is insensitive to system very fast (DVF), decrease fast (DF), decrease slowly (DS), de-
variables and the algorithm is universal to any system. The ad- crease very slowly (DVS), OK, increase very slowly (IVS),
vantages of fuzzy control are obvious. It provides adaptive step increase slowly (IS), increase fast (IF), and increase very fast
size in the search that leads to fast convergence, and the con- (IVF), as shown in Fig. 11. The fuzzy control rules are shown
troller can accept inaccurate and noisy signals. The AFCS oper- in Table II.
ation does not need any wind velocity information, and its real- 2) Structure of FC2 : The role of FC2 is to compute the
time-based search is insensitive to system parameter variation. angular frequency reference value online ωref used by FC1 .
The structure of the FCs is described in detail in the following The inputs of this controller are the following:
sections. 1) the last change of ωref , L∆ωref ;
1) Structure of FC1 : FC1 is the main FC of the angular 2) the corresponding change of real power ∆P0 ; its output is
frequency regulator control system. Its inputs are the deviation the current change of ∆ωref . Gaussian MFs are used for
of ωe from its reference ωe − ωref and its derivative. Its output both inputs and outputs. The following fuzzy sets are used
is the phase shift of δ, i.e., ∆δ. by inputs and outputs:
270 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 1, MARCH 2009

TABLE III TABLE IV

FUZZY RULES FOR FC 2 FUZZY RULES FOR FC 3

PVB positive very big PB positive big Fig. 12. AC voltage regulator (FCS).
PM positive medium PS positive small
P positive ZE zero
B. AC Voltage Regulator
NVB negative very big NB negative big
NM negative medium NS negative small The ac voltage regulator compares the amplitude of the ac
N negative. voltage generated by the MC with the reference voltage, and that
applies error into fuzzy controller FC5 (see Fig. 12). The output
The fuzzy control rules used by FC2 are shown in Table III. of the regulator is the modulation index “m” of the sinusoidal
3) Structure of FC3 : The role of FC3 is to fine-tune online PWM reference signal, and consequently, the modulation signal
the MFs of FC1 . The online tuning of the MFs is a good way of the magnitude of the ac voltage generated by the MC. The
to deal with the continuing variation of the system parameters. structure of the FC5 is similar to that of FC1 .
Obviously, there is no combination of parameters that ensures
optimum performance under any operating conditions. Through
IV. CONTROL IMPLEMENTATION IN dSPACE 1104 Kit
FC3 , it is possible to online tune FC1 MFs in order to optimize
their parameters under any circumstances. In fact, FC3 acts sim- Control algorithm has been implemented in dSPACE, which
ilar to an experienced control system designer, who continuously is very flexible and powerful system featuring both high compu-
monitors the system and modulates the MFs in order to make tational and comprehensive I/O periphery. Additionally, it fea-
the system more or less “strict” according to the circumstances. tures software SIMULINK interface that allows all applications
The following are the inputs of FC3 : to be developed in the MATLAB/SIMULINK friendly environ-
1) the absolute mean deviation of ωe from its reference ment. All compiling and downloading processes are carried out
|ωe − ωref |; automatically in the background.
2) the variance of ωe , σ, and the output of this controller is the The duty ratios are calculated by using angle “δ” and modu-
signal ∆RF (see Fig. 8). By accumulating the successive lation index “m” produced online by the AFCS and ac voltage
values of ∆RF, according to (17), the RF is produced, regulator (Figs. 8 and 12), which are also developed in MAT-
which will optimize FC1 MFs. The variance of ωe , σ, and LAB/SIMULINK programming environment using dSPACE
the output of this controller are 1104 kit for the proposed WECS system. Using duty ratios
and sector numbers, 81 switching signals have been constructed
(not shown). The results for different values of modulation in-
RFnews = RFold + ∆RF. (17) dex “m” are shown in Fig. 5, to show the effectiveness of the
control algorithm implemented in dSPACE. These waveforms
Specifically, RF increases or decreases the range of the MFs are captured using a data acquisition system in dSPACE 1104
that correspond to the output δ in order to render FC1 “stricter” kit.
or “looser.” For example, in case of a large variation of the wind
velocity, the system must become “stricter.” To achieve this, the
range of δ is increased, producing bigger variations of δ. The V. RESULTS AND DISCUSSION
priority of this controller is to regulate the FC1 MFs in order Simulations were performed in MATLAB/SIMULINK.
to drive ωe as close to its reference as possible. When this goal Through C++ programming, the design of AFCS was accom-
is achieved, FC1 can be further fine-tuned in order to minimize plished, converting the simple FCs of the fuzzy logic toolbox
the width of fluctuations of ωe near its reference. into AFCS, which can be self-tuned online. The objective of
Gaussian MFs are used for both inputs and outputs. Three simulation is to illustrate interacting phenomena that have not
fuzzy subsets are needed for each input: big (B), medium (M), been understood before and verify the control strategies (Fig. 8)
and small (S). The fuzzy sets required for are PB, PM, PS, OK, proposed for the new WECS system. The system is simulated
NS, and NB. The fuzzy control rules are shown in Table IV. under a steady state and two disturbances, a sudden decrease,
KUMAR et al.: OPTIMAL CONTROL OF MATRIX-CONVERTER-BASED WECS 271

Fig. 13. Wind velocity.

Fig. 15. RMS ac voltage at the grid.

Fig. 14. Electrical frequency and its reference.

and a sudden increase of the wind speed. Response of the system

with a wind velocity, which fluctuates with a mean value near
9 m/s and at 250 s, following an almost step change, drops to a
mean value near 7.5 m/s (see Fig. 13). After a few seconds, the
wind velocity rises suddenly up to 12 m/s. Fig. 16. Behavior of rotor speed during transient conditions.
With the decreases or increases of wind velocity, the values
of output real power from the generator decreases or increases
correspondingly as the changes of system frequency are largely
constrained and the control system gives an order for corre-
sponding decrease or increase of power. The constant local fre-
quency and voltage are predominantly due to the control effect
of fuzzy logic controller, which adjusts the firing delay angle of
MC according to the change of system frequency. From Fig. 14,
it can be observed that ωe is continuously attached to ωref that
keeps on increasing, trying to reach its optimal value for the
specific wind velocity. During the disturbance, ωe stays close
to ωref and increases until it reaches its optimal value. When
this happens, output power and ωe reach a new steady state that
corresponds to the maximum absorption of real power from the
wind turbines.
Fig. 15 shows the rms ac voltage at the grid. We can observe
that the ac voltage stays between the acceptable limits, even
during the disturbances. It starts deviating from its reference
for a few seconds after the wind speed changes, but the control
Fig. 17. Aerodynamic torque developed by wind turbine during transient
system quickly detects this deviation and drives the voltage to conditions.
its initial level.
Figs. 16–18 demonstrate the comparative results of rotor
speed, torque, and active power for transient behavior of wind
generation system with different control algorithms, i.e., PIs After the fault, the system enters once again in operation and
with conventional ac–dc–ac converter, simple fuzzy control sys- it has been found that the proposed WECS system regains its
tem with conventional ac–dc–ac converter, and the proposed operation more quickly as compared to others, which proves the
AFCS with improved SVPWM switching-based four-leg MC. superiority of the proposed WECS system over conventional
A fault has been applied at t = 1 s, with a duration of 100 ms. ones.
272 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 1, MARCH 2009

variable speed wind turbine,” Inst. Electr. Eng. Proc.—Electr. Power

Appl., vol. 143, no. 5, pp. 380–387, Sep. 1996.
[7] R. Teodorescu and F. Blaabjerg, “Flexible control of small wind turbines
with grid failure detection operating in stand-alone and grid-connected
mode,” IEEE Trans. Power Electron., vol. 19, no. 5, pp. 1323–1332, Sep.
2004.
[8] W. Lu and B. T. Ooi, “Multiterminal LVDC system for optimal acquisition
of power in wind-farm using induction generators,” IEEE Trans. Power
Electron., vol. 17, no. 4, pp. 558–563, Jul. 2002.
[9] O. Ojo and I. Davidson, “PWM-VSI inverter assisted stand-alone dual
stator winding induction generator,” in Proc. IEEE-IAS Conf. Rec. Annu.
Meeting, Oct., 1999, pp. 1573–1580.
[10] D. Seyoum, F. Rahman, and C. Grantham, “Terminal voltage control of
a wind turbine driven isolated induction generator using stator orientated
field control,” in Proc. IEEE Appl. Power Electron. Conf., Miami Beach,
FL, Feb. 2003, vol. 2, pp. 846–852.
[11] M. Naidu and J. Walters, “A 4-kW 42-V induction-machine-based auto-
motive power generation system with a diode bridge rectifier and a PWM
inverter,” IEEE Trans. Ind. Appl., vol. 39, no. 5, pp. 1287–1293, Sep./Oct.
2003.
[12] IEEE Recommended Practices and Requirements for Harmonic Control
in Electric Power Systems, Project IEEE-519, Oct. 1991.
Fig. 18. Behavior of the generated active power during transient conditions. [13] L. Zhang, C. Watthanasarn, and W. Shepherd, “Application of a matrix
converter for the power control of a variable-speed wind-turbine driving a
doubly-fed induction generator,” in Proc. Ind. Electron., Control Instrum.
VI. CONCLUSION (IECON 1997), Nov. 9–14, vol. 2, pp. 906–911.
[14] K. Ghedamsi, D. Aouzellag, and E. M. Berkouk, “Application of matrix
In this paper, a new wind turbine generation system based converter for variable speed wind turbine driving a doubly fed induction
on a four-leg MC topology with AFCS and improved SVPWM generator,” in Proc. IEEE Int. Symp. Power Electron., Electr. Drives,
Autom. Motion (SPEEDAM 2006), May, pp. 1201–1205.
switching code developed in dSPACE1104 kit is presented for [15] M. Venturini and A. Alesina, “The generalized transformer: A new bidi-
performance enhancement and efficiency optimization, partic- rectional sinusoidal waveform frequency converter with continuously
ularly for sites where most of the time low wind speeds pre- adjustable input power factor,” in Proc. IEEE PESC 1980, pp. 242–
252.
vails. Fuzzy logic controllers search online the optimum an- [16] P. W. Wheeler, J. Rodrı́guez, J. Clare, L. Empringham, and A. Weinstein,
gular frequency so that aerodynamic efficiency of the wind “Matrix converters: A technology review,” IEEE Trans. Power Electron.,
turbine is optimum and program the machine flux by an on- vol. 49, no. 2, pp. 276–289, Apr. 2002.
[17] R. C. Bansal, “Three-phase self-excited induction generators (SEIG): An
line search so as to optimize the machine-converter system overview,” IEEE Trans. Energy Convers., vol. 20, no. 2, pp. 292–299,
efficiency. Jun. 2005.
The novelty of the proposed model is that it makes no as- [18] F. J. Lin, L. T. Teng, P. H. Shieh, and Y. F. Li, “Intelligent controlled-wind-
turbine emulator and induction-generator system using RBFN,” Inst.
sumptions on the input and output frequencies of the MC Electr. Eng. Proc.—Electr. Power Appl., vol. 153, no. 4, pp. 608–618, Jul.
and so is applicable under unbalanced input and/or output 2006.
conditions. MC along with fuzzy control controls the ter- [19] T. Ahmed, K. Nishida, and M. Nakaoka, “Advanced control of PWM
converter with variable-speed induction generator,” IEEE Trans. Ind.
minal voltage and frequency of the induction generator in Appl., vol. 42, no. 4, pp. 934–945, Jul./Aug. 2006.
such a way that the wind turbine is operating at its maxi- [20] R. C. Bansal, T. S. Bhatti, and D. P. Kothari, “On some of the design
mum power point for all wind velocities. Finally, compara- aspects of wind energy conversion systems,” Energy Convers. Manage.,
vol. 43, no. 16, pp. 2175–2187, 2002.
tive results for transient behavior of the proposed wind gen- [21] J. G. Slootweg, S. W. H. de Haan, H. Polinder, and W. L. Kling, “General
eration system with conventional systems have been pre- model for representing variable speed wind turbines in power system
sented to demonstrate the superiority of the proposed WECS dynamics simulations,” IEEE Trans. Power Syst., vol. 18, no. 1, pp. 144–
151, Feb. 2003.
system. [22] G. C. D. Sousa, B. K. Bose, and J. G. Cleland, “Fuzzy logic based on-line
efficiency optimization control of an indirect vector controlled induction
REFERENCES motor drive,” in Proc. IEEE IECON Conf., Maui, HI, Nov. 1993, pp. 1168–
1174.
[1] H. Polinder, D. Bang, R. P. J. O. M. van Rooij, A. S. McDonald, and M. A. [23] F. Yue, P. W. Wheeler, and J. C. Clare, “A novel four-leg matrix converter,”
Mueller, “10 MW wind turbine direct-drive generator design with pitch or in Proc. 32nd IEEE Annu. Conf. Ind. Electron. (IECON), Nov. 2006,
active speed stall control,” in Proc. IEEE Int. Electr. Mach. Drives Conf. pp. 2694–2699.
(IEMDC 2007), May 3–5, vol. 2, pp. 1390–1395. [24] F. Gao and M. R. Iravani, “Dynamic model of a space vector modulated
[2] “Time for action: Wind energy in Europe,” Eur. Wind Energy Assoc., matrix converter,” IEEE Trans. Power Del., vol. 22, no. 3, pp. 1696–1705,
Rome, Italy, Rep. 01, Oct. 1991. Jul. 2007.
[3] L. L. Freries, Wind Energy Conversion Systems. New York: Prentice- [25] J. Chang, T. Sun, and A. Wang, “Highly compact AC–AC converter achiev-
Hall, 1990. ing a high voltage transfer ratio,” IEEE Trans. Ind. Electron., vol. 49, no. 2,
[4] L. S. T. Ackermann, “Wind energy technology and current status. A re- pp. 345–352, Apr. 2002.
view,” Renew. Sustainable Energy Rev., vol. 4, pp. 315–375, 2000. [26] M. G. Simoes, B. K. Bose, and R. J. Spiegel, “Design and performance
[5] R. C. Bansal, A. F. Zobaa, and R. K. Saket, “Some issues related to power evaluation of a fuzzy-logic-based variable-speed wind generation system,”
generation using wind energy conversion systems: An overview,” Int. J. IEEE Trans. Ind. Appl., vol. 33, no. 4, pp. 956–965, Jul./Aug. 1997.
Emerging Electr. Power Syst., vol. 3, no. 2, pp. 1–19, 2005. [27] R. C. Bansal, “Bibliography on the fuzzy set theory applications to power
[6] R. Pena, J. C. Clare, and G. M. Asher, “A doubly fed induction generator systems (1994–2001),” IEEE Trans. Power Syst., vol. 18, no. 4, pp. 1291–
using back-to-back PWM converters supplying an isolated load from a 1299, Nov. 2003.
KUMAR et al.: OPTIMAL CONTROL OF MATRIX-CONVERTER-BASED WECS 273

Vinod Kumar received the Bachelor’s degree in R. C. Bansal (S’99–SM’03) received the M.E. degree
electrical engineering from the College of Technol- from Delhi College of Engineering, Delhi, India, in
ogy and Engineering, Maharana Pratap University 1996, the M.B.A. degree from Indira Gandhi National
of Agriculture and Technology (MPUAT), Udaipur, Open University, New Delhi, India, in 1997, and the
India, in 2002, and the M.Tech. degree in power elec- Ph.D. degree from the Indian Institute of Technology
tronics from Vellore Institute of Technology (VIT), (IIT)-Delhi, Delhi, India, in 2003.
Vellore, India, in 2004. During June 1999–December 2005, he was an
He is currently a faculty member in the Depart- Assistant Professor in the Department of Electrical
ment of Electrical Engineering, MPUAT. His current and Electronics Engineering, Birla Institute of Tech-
research interests include power electronic convert- nology and Science, Pilani, India. During February
ers and their application in wind power generation. 2006–June 2008, he was with the School of Electrical
and Electronics Engineering Division, School of Engineering and Physics, The
University of the South Pacific, Suva, Fiji. He was with the Civil Construction
Wing, All India Radio, for nine years. He is currently a faculty member in the
School of Information Technology and Electrical Engineering, The University
of Queensland, St. Lucia Campus, Qld., Australia. He has authored or coau-
R. R. Joshi received the Bachelor’s degree in electri- thored more than 90 papers in national/international journals and conference
cal engineering from Mugneeram Bangur Memorial proceedings. His current research interests include reactive power control in
(MBM) Engineering College, Jodhpur, India, and the renewable energy systems and conventional power systems, power system opti-
M.Tech. degree from the Indian Institute of Technol- mization, analysis of induction generators, and artificial intelligence techniques
ogy (IIT)-Delhi, Delhi, India. applications in power systems.
He is currently a faculty member in the De- Dr. Bansal is an Editor of the IEEE TRANSACTIONS OF ENERGY CONVER-
partment of Electrical Engineering, Maharana Pratap SION AND POWER ENGINEERING LETTERS and an Editorial Board member of the
University of Agriculture and Technology (MPUAT), IET, Renewable Power Generation, Electric Power Components, and Systems
Udaipur, India. His current research interests include Energy Sources, Part B: Economics, Planning and Policy. He is a member of
electrical drives, power quality, and solar-based hy- the Board of Directors of the International Energy Foundation (IEF), Alberta,
brid vehicles. He has authored or coauthored more Canada. He is also a member of the Institution of Engineers (India) and a Life
than 35 research papers published in refereed journals and conferences. Member of the Indian Society of Technical Education.