Sei sulla pagina 1di 10

This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

Adaptive Disturbance Rejection Control Scheme for


DFIG-Based Wind Turbine: Theory and
Experiments
Akbar Tohidi

Hadi Hajieghrary

M. Ani Hsieh

Electrical and Computer Engineering Mechanical Engineering and Mechanics Mechanical Engineering and Mechanics
Drexel University
Drexel University
K. N. Toosi University of Technology
Philadelphia, PA, USA
Philadelphia, PA, USA
Tehran, Iran
Email: MHsieh1@Drexel.edu
Email: Hadi.Hajieghrary@Drexel.edu
Email: Akbar.Tohidi64@Yahoo.com

AbstractIn this paper a new control structure is presented


to extract maximum power from a wind regime. In this novel
approach a discrete-time higher order sliding mode controller is
designed as an observer to construct the reference value for the
extractable power based on the condition that it operates. This is
possible by tracking the optimal tip speed ratio with manipulating
the voltages of the rotor in the Doubly-Fed Induction Generator
(DFIG) configuration. The presented structure improves performance under abrupt changes in the wind speed, and can be used
for any type of optimum active power tracking algorithms. The
simulations show the significant improvement in performance of
the nonlinear discrete-time backstepping controller utilizing this
technique.
Index TermsDoubly-Fed Induction Generator (DFIG), Wind
Turbine, Digital Control, Sliding Mode Control, Maximum Power
Point Tracking, Nonlinear Observer.

I. I NTRODUCTION
Variable-speed wind turbines achieve better efficiency comparing to the fixed-speed systems, and are more reliable. Using
them improves disturbance rejection characteristics of the grid
and reduces the flicker problem [1] [2]. The aerodynamic
system of a wind turbine operates over a wide range of
speed in order to achieve optimum aerodynamic efficiency by
tracking the optimum tip-speed ratio. Therefore, the rotor of
the electrical generator must be able to operate at a variable
rotational speed to cope with this operating condition. In a
variable-speed power generator the main control objective is to
capture maximum energy by tracking the changes in the wind
speed. There are three common approaches for Maximum
Power Point Tracking (MPPT) control in a variable-speed
wind turbines [3]: Tip Speed Ratio (TSR) control; Power
Signal Feedback (PSF) control; and Hill-Climb Searching
(HCS) control. Each of these methods comes with a great
deal of difficulties. For example, in the TSR control the wind
velocity and the optimal tip speed ratio are to be measured
and calculated accurately in advance [4]. In the PSF strategy
the wind turbine maximum power curve has to be known
beforehand, which must be obtained by costly simulations and
experimental tests [5]. And, the HCS strategy covers a class
of Extreme Seeking Control (ESC) algorithms that needs to

measure the wind velocity or its differentiations, and therefore


severely suffers from noise amplification problem [6].
In this paper, it is tried to exploit the advantages of ESC
algorithm [7], improve its performance by introducing an
online mechanism for estimating the parameters of the system
and the its operating condition, and design a novel Sliding
Mode Controller (SMC) to regulate the state of the system
and keep it at the peak of its performance [8]. In the proposed
control strategy, the wind velocity is treated as an external
disturbance to the control system, and the closed-loop is
designed such that the output of the system tracks the regulated
optimal output power. In the intermediate level of this control
scheme, the speed of the wind is identified online and used
to calculate the maximum achievable power from the wind
regime; and, the optimal tip speed ratio of the turbine is found
accordingly.
Doubly-Fed Induction Generators (DFIG) have winding
on both stationary and rotating parts, and both of them are
connected to electrical sources [9] [10]. The stator of DFIG
is connected directly to the grid and its rotor is fed by
a voltage or current source inverter to operate on variable
speed. AC-DC-AC inverter used on the rotor consists of two
voltage-sourced converters, i.e., rotor-side converter (RSC) and
grid-side converter (GSC), which are connected back-to-back.
Between the two converters is a dc-link capacitor in order to
keep the voltage variations in the dc-link to the minimum.
The overall scheme of the system is shown in figure 2. With
the rotor-side converter it is possible to control the torque
or the speed of the DFIG and also the power factor at the
stator terminals, while the main objective for the grid-side
converter is to keep the dc-link voltage constant regardless of
the magnitude and direction of the rotor power. The converter
rating depends on the range of operating speed. The required
power of the converter is usually 20 to 40 percent of the total
power of the generator [11] [12].
One of the main advantages of a DFIG is that, if rotor
current is governed by applying stator-flux-oriented vector
control then the active and reactive powers on the stator side
can be decoupled [9]. This could be carried out using com-

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

mercial double-sided PWM inverters. This is the ground for


the assumption we made in this paper to separate the control
of active and reactive power output of the generator. The
components of the voltages of the rotor in D-Q rotating frame
are chosen as the control inputs. Also, an observer is designed
to construct the reference value for the extractable power based
on the condition the turbine operates and with the optimal tip
speed ratio [13]. It is concluded that the presented structure
improves performance under abrupt changes in the wind speed,
and can be used for any type of optimum active power tracking
algorithms. The simulations show the significant improvement
in performance of the nonlinear discrete-time backstepping
controller utilizing this technique.
The paper is organized as follows: Section II contains the
problem statement, and presents a brief review on modeling
of a DFIG-based wind energy conversion system. A simple
dynamic model of a DFIG is appended by a mechanical
model of a turbine to achieve an overall nonlinear system of
equations suitable to be used for the proposed controller design
method. In Section III an overview of the controller design
method is discussed. In this section it is tried to point out
the fundamental differences of designing discrete controller
over the more prominent continuous counterpart. Section IV
is where the overall structure of the system is discussed.
In Section V the simulation results is presented, and these
results are evaluated in the experimental implementation. And,
Section VI concludes the paper.
II. P ROBLEM S TATEMENT
In this section, a model of a DFIG-based wind turbine is
briefly introduced. The emphasis will be on the relationship
between the wind velocity, the aerodynamics of the turbine,
and the amount of power can be extracted from wind. First,
the model of wind energy conversion system is presented. In
this section the configuration of the wind turbine to extract
maximum power of the wind is discussed. Second, a simple
mechanical model of the turbine is presented. And third, a
model of the electrical conversion system is presented. This
section is concluded by appending the mechanical system to
the electrical one and writing a simplified nonlinear equations
governing a DFIG-based wind turbine. The focus in this
section will be on the properties which are important for us
to design the controller.
A. Modeling of a Wind Energy Conversion System
There is a nonlinear relationship between the wind velocity
and the mechanical power output of the wind turbine, which
can be represented by the following equation [14]:
1
(1)
Pm  Rt2 Vw3 Cp p, q,
2
in which, is the air density, Rt is the radius of the rotor
blades, Vw is the wind velocity, is the tip speed ratio, and is
the pitch angle of the blades of the wind turbine. In general,
the power coefficient, Cp p, q, is defined as the ratio of the
output power of the turbine over the power available in the
wind regime. This coefficient determines the maximum power

can be harvested by the wind turbine at the given wind speed,


and it is the function of the blade pitch angle as well as the
tip speed ratio [4]:
116
p 21 q p6.8e  3q , (2)

0.4 qe i
Cp p, q  0.5176p
i
i
where, i is a constant related to the structural configuration
of the blades of the turbine:


 p

1
0.8 q

3.5e  2
p 3 1q

1

(3)

The values in this paper are obtained for a wind turbine of the
experimental setup, for which the parameters can be found in
the A PPENDIX.
The Pitch control is one of the common trends in the optimal
wind power harvesting. It is fast; it limits the speed of the
rotor by regulating aerodynamic power absorbed from wind;
and, it is effective, especially when the turbine operates at a
speed near its upper limit, higher than the nominal speed which
the turbine is designed for [15] [16]. However, at the lower
speed of wind, the pitch angle must set to zero to obtain the
maximum power coefficient with respect to the tip speed ratio.
Here, the tip speed ratio is proportional to the rotor speed and
the inverse of the velocity of the wind:


Rt r
,
(4)
Vw
and, the maximum amount of power coefficient in zero blade
pitch angle can be achieved by maximizing the equation (1)
with respect to tip speed ratio, i.e.
Cp,max p, 0q  0.48

f or

opt

 8.1,

(5)

This value is going to be used to design the reference power


signal for the control system to follow, regardless of the torque
exerted to the system by the wind.
B. Mechanical Model of a Wind Turbine
The mechanical parts of a wind turbine consist of the rotor
shaft, generator shaft, and a gearbox. A more comprehensive
study on the mechanical model of a wind turbine can be found
in [17] and [18]. The following simplified dynamic model of
wind turbine train is used for the purpose of this paper:
Jt 9 r

 Tl  Kt r  Tr

(6)

Figure 1 shows a simple schematic of a wind turbine drive train


dynamics. The aerodynamic torque, Tl , drives the wind turbine
at the angular velocity, r , and Tr is the reaction torque of the
rotor. The low-speed torque, Tls , has a breaking effect on the
rotor. The generator is driven by the high speed torque Ths
and broken by the generator electromagnetic torque Tg . The
rotor speed is increased through a gearbox with the ratio of ng
to reach the generator speed g while the low-speed torque is
augmented. Br and Bg is the shaft friction torque, which are
negligible in comparison to the scale of the generated torque.
Remark 1: A synchronous electrical machine has to work
at speed less than synchronous speed to act like a generator.
Otherwise, it is going to work like an electrical motor and

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

will absorb power from the network. However, in practical


cases injecting the power of the generator to its own stator
can compensate this to some extent, and the generator can
continue working at sub-synchronous speeds. The same case
happens when the generator is working in super-synchronous
speed. In this paper, to be consistent to our previous works in
[19] and [8], the angular velocity of the rotor is assumed to
be in the range of 0.7 r 1.3.
C. Dynamic Model of a Doubly-Fed Induction Generator
The dynamic model of a DFIG has generally been defined
in synchronous D-Q frame, with constant stator flux and/or
constant stator voltage. Accordingly, the control approach
suggested in this article uses dynamic model written in the
rotating synchronous D-Q frame [20] [21] [22]. The threephase supply voltages, uqs and uds , and the rotor voltages,
uqr and udr , in D-Q reference frame is given by:
$
uqs
'
'
'
&

 pqs ds Rs iqs
uds  pds qs Rs ids
'
uqr  pqr p  r qdr
'
'
%
udr  pdr p  r qqr

Rr iqr
Rr idr

(7)

in which, Rs and Rr are stator and rotor resistances of the


winding of each phase of the electrical machine, and r
are the electrical frequency of the machine and mechanical
frequency of the turbine, and p is the number of the pole pairs.
Flux linkages at three-phase rotor and stator in d-q reference
frame can be calculated by knowing the leakage inductances
of the stator and the rotor windings, Llr and Lls , and the
magnetizing inductance, Lm :
$
'
'qs
'
&

ds
'
qr
'
'
%
dr

 pLls
 pLls
 pLlr
 pLlr

Lm qiqs
Lm qids
Lm qiqr
Lm qidr

Lm iqr
Lm idr
Lm iqs
Lm ids

(8)

ds
'
9 qr
'
'
%9
dr

 uds  Rs ids qs
 uqs  Rs iqs  ds
 udr  Rr idr sqr
 uqr  Rr iqr  sdr

(9)

where, s is the generator slip parameter. The active and


reactive power equations of the stator are given by:

 23 puds ids uqs iqs q


3
Qs  puqs ids  uds iqs q
2
Ps

and the active and reactive powers of the stator, Ps and Qs , are
the outputs from the system. The components of the voltage
of the stator are considered to be constant due to the fact that
the stator of the generator is attached to an infinite network.
When qs is assumed constant the produced active and reactive
powers are subordinates of three phase current of the rotor in
synchronize D-Q frame; in other words, the active and reactive
powers of stator can be controlled independently, which is the
recognized concept of vector control.
#

9  F pX q
X
Y  H pX q

 iqs

U  udr
F pX q 

X

(10)

D. Nonlinear Dynamic Model of a Wind Turbine


By appending the mechanical model to the electrical model
of the generator a simple dynamic model of a wind turbine
can be presented as a Multi-Input/Multi-Output (MIMO) state
space equations (11). In this model, the components of the
voltage of the rotor, uqr and udr , are the inputs to the system,

ids
uqr

GpX qU

W di

iqr

T

idr

, di

 Tl ,

T

(11)

a11 iqs a12 ids a13 iqr a14 ids m a15 idr m
a21 iqs a22 ids a23 idr a24 iqr m a25 iqs m 2 


 a31 iqs a32 iqr a33 idr a34 idr m a35 ids m  ,


a41 ids a42 iqr a43 idr a44 iqr m a45 iqs m 4 
b1 iqs idr b2 ids iqr b3 m


The flux linkage and the DFIG equations are as follow [23]:
$
9
'
'qs
'
&9

Fig. 1: The overall structure of a wind turbine.

0
g21

GpX q  
 0
g42
0

g12
3

0 
2 uds iqs



g32 
 , H pX q  3

0
2 uds ids
0

in which X, U , and Y represent the states, inputs, and outputs


of the system; di is the stochastic disturbance signal, modeling
the wind power, and W is a weighting function which indicates
how much this disturbance effects each state of the dynamic
system. In simulation, we have filtered the noise to achieve the
proper bounds on its variation. The coefficients of the dynamic
equations are presented in the A PPENDIX.
III. D ISCRETE -T IME S LIDING M ODE C ONTROL
Continuous Sliding Mode Controller (CSMC) has a strong
theoretical background [24] [25]. A large class of continuoustime systems are controlled with this Lyapunov-based method,
that are in fact being controlled by a computer [26]. In these
cases, the available information about the system (Measurements) is only accessible in the certain time steps, and the
control signal can also be altered only at such ticks. This limits

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

the reaction time of the controller to the sampling speed, and


neglecting this fact can cause a serious performance issue, and
even leads to the instability of the whole closed-loop system.
To design a Discrete-time SMC (DSMC) a discrete-time
model of the system will be considered [27]; and, in case we
choose the sampling period carefully based on the dynamics
of the system, the controller can react to the changes in states
of the system adequately fast. The discrete time equivalent
model of the system is achieved by solving the differential
system of equations (11) for each step of time. The resulting
system is not necessarily a linear one, and the discrete model
is generally non-affine in input U , unlike the continuous time
model:
#

X rk 1s  pX rk s, U rk s, di rk sq
y rk s  hpX rk sq

(12)

in which X P Rn , U P Rm , and Y P Rl , and the system


includes the origin as the equilibrium point. Furthermore, even
when the discrete time model is affine, after state feedback
linearization procedure the linear output is generally a nonaffine function of external input, unlike the case in continuous
time system.
The design procedure follows the same method of dynamic
backstepping used in the continuous case in [8]. To accommodate the differences, the following adjustments are made in
the design method:

in which vxi rk s is the ith component of the vx rk s. The


reaching law of equation (14) can be written as:
#

vxi rk
vxi rk

1s  vxi rk s signpvxi rk sq 0

1s vxi rk s signpvxi rk sq 0

(15)

The first condition guarantees that the state of the closed-loop


system always moves toward the sliding surface, while the
second condition sets a boundary on the speed of the convergence. Applying this reaching law guarantees the following
properties for the close loop system: 1) Starting from any
admissible initial condition, the state of the system follows a
monotonic trajectory toward the switching plane, and might
pass beyond it for a limited time; 2) When the trajectory
passes the switching plane for the first time it must pass it
again in each sampling period, and finally creating a zigzag
motion around the switching plane; and 3) The size of each
zigzag step will not increase, and the trajectory remains within
a certain and limited strip region around the sliding surface.
Although these conditions of a control algorithm have been
introduced for a single input system, but they can be used for
a multi-input system by applying them to each components of
the switching function, vx rk s, separately.
In this paper the sliding surface is defined as follow, which
is generally linear and stable:
qr1 erk s

v rk s  pz

A. Sliding Surface

r1

i erk

(16)

is

i 0

Same as CSMC, the controller designed method for the


discrete time counterpart is divided into two stages: first,
calculating the sliding surface, vx rk s  xrk s, on which the
internal dynamics is stable; and second, creating a control law
propelling the state variables of the closed-loop system toward
this sliding surface. The first step of designing procedure can
be performed by following the exact same principles used to
design a controller for a continuous system. The second step in
designing method for DSMC have to be done quite differently,
and considering the limited switching frequency, the closedloop system may not settle on the sliding surface as neat as
in a continuous-time system.
Discretizing the reaching law designed for the CSMC [24]
might seem as an attractive option. Using the first order
forward deference approximate of the derivative, v9x ptq, the
discretized reaching law is obtained as:

pvx rk

vx rk s
1s  vx rk sq 
||vx rks|| .

(13)

However, this does not guarantee the convergence of the states


to the sliding surface; i.e., an input control leading to vx rk s 
cp1qk sequence with large enough coefficient c also satisfies
the above reaching law condition, but clearly does not stabilize
the system. The method used in this paper to obtain a DCMC
is a modified version of one introduced in [28]:

vx k
i

1s vxi rk s,

(14)

in which, is a parameter to adjust the bandwidth of the


controller, and z is the delay parameter.
B. Input-Output Linearization
In this section an adoption of the feedback linearization
for a discrete time system has been developed.
 Assume that
a static state feedback, upk q  T xrk s, v rk s , can solve the
problem of synthesizing linearizing feedback controller; and
also assume that this feedback controller preserves the relative
degree of the system, i.e.,


B
Bv rks T xrks, v rks  0. Without

loosing the generality of the statement, the resulted closedloop system can be taken as a linear system of:
y pk

rq

l y rk

r  l s  0 v rk s

(17)

l 1

in which, l , l  0, 1, ..., r are constant scalars constitutes a


stable characteristic polynomial. This statement is correct as
long as the relative degree of the auxiliary input signal, the
disturbance di rk s, is greater than the relative degree of the
system with respect to the control input, the components of
the voltage of the rotor. Then, the disturbance signal will not
appear in the normal form, and instead appears as an input to
the subsystem of the internal dynamics of the system. If the
system is non-minimum phase, the internal dynamics of the
system is stable, and the external disturbance is digested in
this dynamics and will not appear in the output of the system.

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

IV. D ESIGNING ROBUST N ONLINEAR C ONTROLLER FOR A


DFIG-BASED W IND T URBINE
Figure 3 shows the schematic of the controller structure
designed for a wind turbine system equipped with DFIG. The
ultimate objective of the design procedure is to control the
equation (6) such that the opposing torque of the machine
tracks the torque which wind exerts to the turbine. In this way
the system extracts the maximum power from the wind.
To serve the purpose of the system two control loops are
designed: the first loop is a first degree backstepping for each
output of the system, active and reactive powers, to stabilize
the DFIG-based turbine. The voltage components of the rotor
serve as the control inputs of the system which modify the
speed of the shaft of the generator. The corresponding inputs
appear at the first derivatives of each output and a first-order
controller will suffice to have an acceptable performance. The
objective of this control loop is only to increase the stability
margins of the system. The chattering will not cause a serious
problem due to the inertia of the system.
The second control loop acts as an estimator to generates the
reference signal of the maximum extractable power. Since the
chattering of the reference signal can pose a serious problem to
the stability and performance of the whole system, in this loop
a second order backstepping controller is utilized to filter out
the high frequency noises and generates a smooth reference
signal [8] [29]. In the simulations and the implementation, the
designed controller structure is proven successful in extracting
maximum power from wind despite of severe vacillations of
the wind speed. Additionally, this implementation of the closed
loop structure offers the opportunity of controlling the active
and reactive power output of the turbine independently.
A. Estimating the Maximum Absorbable Power
1) Gain-Scheduling (off-line): A wind turbine with a variable speed DFIG must trace the points of the maximum power
coefficients, Cpmax . These points must be defined highly
accurate and applied to the system as the reference trajectory.
The wind turbine systems have clearly nonlinear features, and
finding these points is hard from practical point of view. In
this off-line method, the maximum absorbable power of the
wind has been defined by fitting a curve to the data points in
hand. A limited points of wind power is really measured, and
the other points in curve must be estimated. By interpolating
these points, shown in figure 4, the static equation for the
reference power signal as a function of the velocity of the
wind is achieved as
Psref f

 105  p0.0627Vw2

0.29Vw  1.4748q. (18)

2) Estimating the maximum absorbable power by the online method: The practical implementation of the off-line
method suffers several drawbacks: 1) Estimating the maximum
wind power point needs an exact model of the system, which
is hard to find. 2) Working points of the system changes as
the parameters of the system change during time; also, unknown perturbations, unknown dynamic models, and erosion
challenges the accuracy of the recognized points. Thus, an

online identification method can be a solution to the problem.


Estimating the source signal in online method is done by
measuring and comparing it with its optimal value for the
system. If the measured is equal to its optimal value, then
the electrical network absorbs the maximum power from the
turbine. Otherwise, the source power of the rotors of the
generator must be changed to adjust the tip speed of the blades.
In this paper a simple exponential feedback equation is used
to construct the reference signal, i.e.,
Psref f

 Ps

 k1 p1  epk |
2

initial

opt

|q ,

(19)

in which, k1 and k2 are to be tuned based on the dynamic of


the system, and Psinitial can be found from equation (18).
Although this equation improves the performance of the
total system, it acts slowly in the estimation of the maximum
wind power. Therefore, an auxiliary mechanism is applied to
estimate the adaptive sliding surface. The suggested structure
of the controller is given in figure 3. The control loop consists
of an estimator to adjust the sliding surface. Defining the error
rk s  rk s  opt , a second order sliding mode controller
is designed to regulate the tip speed ratio of the turbine by
manipulating the desired power reference signal given to the
system. The suggested Lyapanov function is considered as
V

 21 c1 2 rks

1
c1 2 rk
2

1s.

(20)

The stability of the point prk s, rk 1sq  p0, 0q can be


achieved in limited time by the reaching conditions:
2prk s

rk

2sq  rk ssat

rk s 

(21)

The sliding mode which has been defined in this way is called
the second order sliding mode (DSOSM) [30]. This strategy
is used as a nonlinear state observer to obtain the value of
unmeasured parameter, here the optimal tip speed ratio.
One of the most important features of the controller structure designed in this paper is that the input obtained from
DSOSM does not affect the real system. Hence, chattering
eventually influence the model but not the system output. In
the following section, the above strategies are applied for wind
turbine system. This method of estimating the reference signal
by DSOSM is introducing for the first time in this paper, and
it shows a promising results. In addition, comparing with the
other estimation methods, one of the distinctive advantages of
this method is explicit and clear stability condition satisfying
during the design phase; also, this algorithm avoids the nonlinear difference conditions for estimation algorithm which is
a vital issue for real time implementation.
By choosing proper parameters of the equivalent linear
system, i s in equation (17), the closed-loop system will be
stable. However, there remains some concern about the internal
stability of the system. Dynamic stability of the generator
limits the current of the stator and the rotor, is and ir .
Therefore, the internal stability of the system will only depend
on the stability of the speed of the shaft coupling generator

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

Fig. 2: The overall schematic of the proposed system structure.

Fig. 3: The Structure of the control system including the control loop of the DFIG (1) and the state estimation loop (2) to
generate the reference trajectory based on the optimal tip speed ratio .

row of the state space equations (11) can be rewritten as


m rk

1s  p1

b3 tqm rk s  Ap ,

(22)

in which Ap constitutes the asymptotically bounded error and


the states of the system, and bounded too. The stability margin
of this equation increases if the inertia of the shaft of the
machine increases.
V. S IMULATION R ESULTS AND E XPERIMENTAL
I MPLEMENTATION

Fig. 4: Power coefficient curves of a wind turbine for different


blade pitch angles.

with wind turbine. Discretized equation (6) or as if at the last

In this section, the simulation results for the proposed


control structure for a wind turbine system equipped with
DFIG generators is presented, and then, this structure is
implemented on a real setup with 1.5M W wind turbine
emulator. The parameters and dimensions of the experimental
setup is given in the A PPENDIX. According to the formulation
of the nonlinear system in equations (12), accessible inputs of
the wind turbine, uqr and udr , are voltages of the rotor, and

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

the unpredictable wind velocity, Vw , is treated as the external


disturbance to the system. The velocity of the wind changes
smoothly. The objective of the suggested controller is to obtain
the maximum power of the disturbance in any condition and
yield the output to the net. On other word, it is the power
of disturbance which determines the working points of the
system. The designed control structure, first, estimates the
power of the wind and devises a reference signal for the output
power of the generator, and second, controls the system with
manipulating the voltages of the rotor to follow this trajectory
in spite of the changes in the wind velocity.
The control inputs appear at the first difference of the
outputs of the system of equations (11):
$
'
'
y1 k
'
'
'
'
'
'
'
'
'
'
'
'
&

'
'
y2 k
'
'
'
'
'
'
'
'
'
'
'
'
%

1q 


3
2 uds t

1q  32 uqs t

a11 iqs pk q

a12 ids pk q

pkq

g12 uqr pk q

a21 iqs pk q

a22 ids pk q

a23 idr pk q

pkq

g12 udr pk q

a14 ids pk qm pk q

1
t iqs

a24 iqr pk qm pk q
1
t ids

The active power output of the wind turbine is given in


figure 6, and the figure 7 shows the error between the output
power and the reference signal. The output power of the
turbine tracks the maximum extractable power of wind with a
less than 2 percent error.

a13 iqr pk q

a15

idr pk qm pk q
.

Fig. 6: Active power reference (red) signal and output (blue)


for offline method of estimation of the reference signal.

a25 iqs pk

qm pkq

The input signals uqr and udr appear as the control signals for
Pout and Qout respectively. These inputs can be devised such
that for some positive coefficient of c1 and c2 the Lyapunov
function V  12 c1 e12 21 c2 e22 is stabilized, where, e1  Pout 
Psref f and e2  Qout  Qsref f . The linear part of the closedloop system will be as follows:
#

y1 pz q 
y2 pz q 

10
z 11 e1
20
z 21 e2

(23)

Fig. 7: Active power tracking error for offline method of


estimation of the reference signal.

Figure 5 shows the stochastic wind speed profile generated


in a 200-second period, and the reference signal in equation
(18) is the offline estimated power of the wind regime. To illustrate the effectiveness of the proposed MPPT control strategy,
the simulations are designed to compare the performances of
the offline and online strategies.

The high order sliding mode controller is used to regulate


the output such that the tip speed ratio settles on its optimal
value calculated in equation (5). Figures 8 and 9 depict the
variations of this parameter, which is less than 0.5 percent of
the reference value. The small deviation of tip speed ration
from its calculated optimal value contributes to the fact that
our output power tracks the desired value with considerably
small deviation.

A. Offline Calculation of the Reference Signal

B. Online Calculation of the Reference Signal

Fig. 5: Simulated stochastic wind velocity profile used for the


simulation purposes.

In this section the simulation results for the proposed online


method is presented. This method is based on obtaining the
maximum absorbable wind energy using a DSOSMC controller as a disturbance observer. As it is shown in figure 10,
the effect of wind speed volatility on the active power output
is considerably degraded with respect to the offline method.
Also, the improvement in tracking performance is significant,
and the active power output tracking error is decreased to less
than one percent. The improvement in the active power output
is because the online system can adjust its tip speed ratio to
the optimal value. Also, the filtering effect of the second order

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

Fig. 8: Tip speed ratio for offline method of estimation of the


reference signal.

Fig. 11: Active power tracking error for online method of


estimation of the reference signal.

Fig. 9: Tip speed ratio tracking error for offline method of


estimation of the reference signal.

Fig. 12: Tip speed ratio tracking error for online method of
estimation of the reference signal.

a linear approximation with hysteresis has been applied as the


optimum active power generation target function.

Fig. 10: Active power reference (red) signal and output (blue)
for online method of estimation of the reference signal.
backstepping can be seen both in the variations of both output
power and tip speed ratio.
C. Experimental Implementation of the Online Method
The efficiency and reliability of the proposed control strategy is practically examined in a pilot experimental setup. The
hardware used in this experiment consists of a 2Kw induction generator, a 0.7Kw doubly-fed inverter, and necessary
speed, voltage and current sensors. Sensors are connected to
the computer using the Data Acquisition Card 1716, which
operates with Real Time Windows Target of Matlab/Simulink
software. It is important to mention that since the installed
hardware in the generator cannot use the cubic reference curve,

Fig. 13: Wind velocity profile in the experiment.


Figure 13 shows the stochastic wind profile emulated to
be use in the experimental setup. The variation frequency
of this profile is taken less than the one considered for the
simulation purpose due to the technical difficulties on the
interface equipment and limitations on sampling rate.
Figures 14 and 15 are the results of the experimental implementation, and show the tracking performance of the system
controlled with the proposed structure. These results clearly
support the effectiveness of the presented on-line method in
reducing the effect of wind swing on generated power, in

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

Fig. 14: Active power reference signal (red) and output power
(blue) in the experimental setup for online method of estimation of the reference signal.

the control inputs. Also, an observer is designed to construct


the reference value for the extractable power based on the
condition that it operates. The objective is to follow the
optimal tip speed ratio with controlling the voltages of rotor.
Beltran et al in [31] design and implement the same kind
of controller for the power system of the wind turbine. In
their proposed method the reference power is calculated priori
using the MPPT method, assuming to have the wind turbine
maximum power curves beforehand. Comparing to the similar
works, it is concluded that the presented structure improves
performance under abrupt changes in the wind speed, and
can be used for any type of optimum active power tracking
algorithms. The simulations show significant improvements
in performance of the nonlinear discrete-time backstepping
controller utilizing this technique.
ACKNOWLEDGMENT

compliance with our simulation results. The swing of the


output active power is only registered less than 3 percent of
the output power, and this owe to the fact that the tip speed
ratio is regulated more effectively within the proposed control
structure.

Fig. 15: Active power tracking error in the experimental setup


for online method of estimation of the reference signal.

Fig. 16: Tip speed ratio tracking error in the experimental


setup for online method of estimation of the reference signal.
VI. C ONCLUSION
It is possible to control a wind turbine in many ways such as:
pitching the blades, controlling the generator torque, applying
the high-speed shaft brake, or deploying the tip brakes. In
this paper, the voltages of the rotor in the DFIG is chosen as

The authors would like to express their appreciation to the


late Prof. Suhada Jayasuriya, whose contribution to this work
was of great significance.
R EFERENCES
[1] M. Nunes, J. Peas Lopes, H. Zurn, U. Bezerra, and R. Almeida,
Influence of the variable-speed wind generators in transient stability
margin of the conventional generators integrated in electrical grids,
Energy Conversion, IEEE Transactions on, vol. 19, no. 4, pp. 692701,
Dec 2004.
[2] W. Qiao, W. Zhou, J. Aller, and R. Harley, Wind speed estimation
based sensorless output maximization control for a wind turbine driving
a dfig, Power Electronics, IEEE Transactions on, vol. 23, no. 3, pp.
11561169, May 2008.
[3] C.-T. Pan and Y.-L. Juan, A novel sensorless mppt controller for
a high-efficiency microscale wind power generation system, Energy
Conversion, IEEE Transactions on, vol. 25, no. 1, pp. 207216, March
2010.
[4] J. Slootweg, S. de Haan, H. Polinder, and W. Kling, General model for
representing variable speed wind turbines in power system dynamics
simulations, Power Systems, IEEE Transactions on, vol. 18, no. 1, pp.
144151, Feb 2003.
[5] W. Lu and B. Ooi, Multi-terminal lvdc system for optimal acquisition
of power in wind-farm using induction generators, in Power Electronics
Specialists Conference, 2001. PESC. 2001 IEEE 32nd Annual, vol. 1,
2001, pp. 210215 vol. 1.
[6] A. Ghaffari, M. Krstic, and S. Seshagiri, Power optimization and
control in wind energy conversion systems using extremum seeking,
Control Systems Technology, IEEE Transactions on, vol. 22, no. 5, pp.
16841695, Sept 2014.
[7] R. Datta and V. Ranganathan, A method of tracking the peak power
points for a variable speed wind energy conversion system, Energy
Conversion, IEEE Transactions on, vol. 18, no. 1, pp. 163168, Mar
2003.
[8] A. Tohidi, O. Abedinia, H. Hajieghrary, and S. Jayasuriya, Experimental study of a disturbance rejection controller for dfig based wind energy
conversion systems, in Dynamic Systems and Control Conference,
ASME 2014, October 2014, p. V003T53A004.
[9] A. Tapia, G. Tapia, J. Ostolaza, and J. Saenz, Modeling and control of a
wind turbine driven doubly fed induction generator, Energy Conversion,
IEEE Transactions on, vol. 18, no. 2, pp. 194204, June 2003.
[10] W. Qiao, Dynamic modeling and control of doubly fed induction
generators driven by wind turbines, in Power Systems Conference and
Exposition, 2009. PSCE 09. IEEE/PES, March 2009, pp. 18.
[11] R. Datta and V. Ranganathan, Variable-speed wind power generation
using doubly fed wound rotor induction machine-a comparison with
alternative schemes, Energy Conversion, IEEE Transactions on, vol. 17,
no. 3, pp. 414421, Sep 2002.

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TIA.2016.2521354, IEEE
Transactions on Industry Applications

[12] R. Pena, J. Clare, and G. Asher, Doubly fed induction generator using
back-to-back pwm converters and its application to variable-speed windenergy generation, Electric Power Applications, IEE Proceedings -, vol.
143, no. 3, pp. 231241, May 1996.
[13] A. Tohidi, H. Hajieghrary, and M. Hsieh, Adaptive disturbance rejection
control scheme for dfig-based wind turbine, in Industry Applications
Society Annual Meeting, 2015 IEEE, Oct 2015, pp. 18.
[14] S. Benelghali, M. El Hachemi Benbouzid, J. Charpentier, T. Ahmed-Ali,
and I. Munteanu, Experimental validation of a marine current turbine
simulator: Application to a permanent magnet synchronous generatorbased system second-order sliding mode control, Industrial Electronics,
IEEE Transactions on, vol. 58, no. 1, pp. 118126, Jan 2011.
[15] E. Muljadi and C. Butterfield, Pitch-controlled variable-speed wind turbine generation, Industry Applications, IEEE Transactions on, vol. 37,
no. 1, pp. 240246, Jan 2001.
[16] B. Biegel, M. Juelsgaard, M. Kraning, S. Boyd, and J. Stoustrup, Wind
turbine pitch optimization, in Control Applications (CCA), 2011 IEEE
International Conference on, Sept 2011, pp. 13271334.
[17] Y. Guo, S. Hosseini, C. Y. Tang, J. Jiang, and R. Ramakumar, An
approximate wind turbine control system model for wind farm power
control, Sustainable Energy, IEEE Transactions on, vol. 4, no. 1, pp.
262274, Jan 2013.
[18] I. Girsang, J. Dhupia, E. Muljadi, M. Singh, and L. Pao, Gearbox and
drivetrain models to study dynamic effects of modern wind turbines,
Industry Applications, IEEE Transactions on, vol. 50, no. 6, pp. 3777
3786, Nov 2014.
[19] A. Tohidi, A. Shamsaddinlou, and A. Sedigh, Multivariable inputoutput linearization sliding mode control of dfig based wind energy
conversion system, in Control Conference (ASCC), 2013 9th Asian,
June 2013, pp. 16.
[20] R. Pena, R. Cardenas, J. Proboste, G. Asher, and J. Clare, Sensorless
control of doubly-fed induction generators using a rotor-current-based
mras observer, Industrial Electronics, IEEE Transactions on, vol. 55,
no. 1, pp. 330339, Jan 2008.
[21] B. Beltran, T. Ahmed-Ali, and M. Benbouzid, Sliding mode power
control of variable-speed wind energy conversion systems, Energy
Conversion, IEEE Transactions on, vol. 23, no. 2, pp. 551558, June
2008.
[22] C. Evangelista, F. Valenciaga, and P. Puleston, Active and reactive
power control for wind turbine based on a mimo 2-sliding mode
algorithm with variable gains, Energy Conversion, IEEE Transactions
on, vol. 28, no. 3, pp. 682689, Sept 2013.
[23] P. Ledesma and J. Usaola, Doubly fed induction generator model for
transient stability analysis, Energy Conversion, IEEE Transactions on,
vol. 20, no. 2, pp. 388397, June 2005.
[24] H. K. Khalil, Nonlinear systems; 3rd ed. Upper Saddle River, NJ:
Prentice-Hall, 2002, the book can be consulted by contacting: PH-AID:
Wallet, Lionel.
[25] J. Hung, W. Gao, and J. Hung, Variable structure control: a survey,
Industrial Electronics, IEEE Transactions on, vol. 40, no. 1, pp. 222,
Feb 1993.
[26] K. J. Aom and B. Wittenmark, Computer-controlled Systems: Theory
and Design (2Nd Ed.). Upper Saddle River, NJ, USA: Prentice-Hall,
Inc., 1996.
[27] A.-M. Lienhardt, G. Gateau, and T. Meynard, Digital sliding-mode
observer implementation using fpga, Industrial Electronics, IEEE
Transactions on, vol. 54, no. 4, pp. 18651875, Aug 2007.
[28] S. Sarpturk, Y. Istefanopulos, and O. Kaynak, On the stability of
discrete-time sliding mode control systems, Automatic Control, IEEE
Transactions on, vol. 32, no. 10, pp. 930932, Oct 1987.
[29] A. Zinober, J. Scarratt, R. Mills, and A. Koshkouei, New developments
in dynamical adaptive backstepping control, in Nonlinear control in
the year 2000 volume 2, ser. Lecture Notes in Control and Information
Sciences, A. Isidori, F. Lamnabhi-Lagarrigue, and W. Respondek, Eds.
Springer London, 2001, vol. 259, pp. 565622. [Online]. Available:
http://dx.doi.org/10.1007/BFb0110326
[30] H. Elmali and N. Olgac, Robust output tracking control of nonlinear
{MIMO} systems via sliding mode technique, Automatica, vol. 28,
no. 1, pp. 145 151, 1992.
[31] B. Beltran, M. El Hachemi Benbouzid, and T. Ahmed-Ali, Secondorder sliding mode control of a doubly fed induction generator driven
wind turbine, Energy Conversion, IEEE Transactions on, vol. 27, no. 2,
pp. 261269, June 2012.

A PPENDIX
Parameters of the Simulations:
Sampling time  104 sec
$
T urbine diameter
'
'
'
&

 60m
N umber of blades  3
'
Hub height  85m
'
'
%
Gearbox ratio  90

$
DF IG : 1.5M W, 690V, 50Hz, two pole pairs
'
'
'
'
'
Rs 0.012, Ls 13.732mH
'
'
'
'
'
0.021, Lr 13.703mH
&Rr





Lm  13.528mH
'
'
'
J
: inertia pturbine
'
'
'
'
'
f
'
'
%

DF IGq  50kgm2
: viscous coef f icient pturbine
DF IGq

2
 7  10 N ms{rd

Parameters of the Experimental Setup:


Sampling time  103 sec
#

T urbine diameter  4m
N umber of blades  3

$
'
DF IG : 2KW, 690V, 50Hz, two pole pairs
'
'
'
'
'
0.2, Ls 3.58mH
&Rs



Rr  0.21, Lr  1.43mH
'
'
'
Lm  18.45mH
'
'
'
%
J : inertia pturbine
DF IGq 

7kgm2

DC Motor: 6.5KW, 310V, 24.8A


Parameters of the Model:

1

 1  LM RL
2   LML kudr

b1

s
r

 3P
4J

,
,

b2

2f

  3P
4J

b3

  PJ

 RL

s
s

g12

aij

 LML  1
2   L1 kudr
z

zk

zk

1
k
Ls
M Rs
Ls Lr

 RL

M
Ls Lr

M Rs
k
Lr Ls

g21

  LML
s

r
r

M Rr
Ls Lr

 2LP ML

 RL

M Rr
Ls Lr

s
s

P
2

zk

zk

  2LP ML

 RL

r
r

g32

PM
2Ls

PM
2Ls

PM
2Lr

 P2 k

  L1 k
r

g42

  P2LM k
r

0093-9994 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Potrebbero piacerti anche