Electrical Power and Energy Systems 96 (2018) 347–356

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier.com/locate/ijepes

Sliding mode control of DFIG powers in the case of unknown flux and rotor MARK
currents with reduced switching frequency
⁎
Abdelhak Djoudia,b, , Seddik Bachac, Hossein Iman-Einid, Toufik Rekiouae
a
G2ELAB Laboratoire de Génie Electrique de Grenoble Bâtiment GreEn-ER, 21 avenue des martyrs, CS 90624, 38031 Grenoble CEDEX 1, France
b
CDER, Centre de Développement des Energies Renouvelables, Route de l’Observatoire Bouzeréah, Algiers, Algeria
c
Univ. Grenoble Alpes, CNRS, Grenoble INP1, G2Elab, 38000 Grenoble, France
d
School of Electrical and Computer Engineering, Faculty of Engineering, University of Tehran, Tehran, Iran
e
Laboratoire de Technologie Industrielle et de l’Information (LTII), Faculté de Technologie, University of Béjaia, Algeria

A R T I C L E I N F O A B S T R A C T

Keywords: This paper aims to implement a stator power control for a doubly fed induction generator (DFIG) with unknown
Doubly fed induction generator (DFIG) flux and rotor currents. This goal is achieved based on a high-gain sliding mode control (HG-SMC) and by
Reduced switching frequency exploiting the fact that the flux remains near the nominal value. A new implementation scheme has been de-
Real-time implementation veloped to achieve low switching frequency compared to the conventional method. The advantages of the new
High-gain sliding mode control (HG-SMC)
scheme over the old one are presented. Some interesting features of the new scheme are: (1) elimination of
Unknown flux and rotor currents
rotor’s current sensor, which leads to lower cost and failure rate in the drive system; (2) simpler implementation
because it does not require any modern method for the flux estimation; (3) longer lifetime of the rotor side-
converter. The proposed control scheme is verified by simulations results and experiments on a 7.5 kW DFIG
power prototype.

1. Introduction
Doubly fed induction generators (DFIGs) are mostly used in high-
power wind energy conversion systems (WECSs) due to their salient
features [1,2] and they can be employed in other power generation
systems like hydropower [3]. Several ways of driving a grid connected
DFIG-based WECS have been proposed in literature which includes
controlling the electromagnetic torque [4–10], controlling the rotating
speed [11–13], and controlling the stator active power [2,14–20]. The
main goal of all these three control methods is to track the maximum
attainable power of the wind turbine as long as the wind speed is below
a certain upper limit. These goals are also valid for other grid connected
DFIG-based renewable energy conversion systems.
The main drawback of the torque control is that the electromagnetic
torque is estimated and is not measured. Moreover, the estimated value
changes with the variation of DFIG inductances. On the other side, the
speed control needs the knowledge of some quantities, such as system
inertia, electromagnetic torque, and aerodynamic torque, which are
difficult to estimate. If these quantities are not estimated correctly,
mechanical oscillations may be formed and affect the wind turbine.
Stator active power control is the most studied approach in the litera-
ture [2,14–20] since it has interesting features compared to two other
methods. In this approach, the active power is measured easily. More-
over, it does not dependent on DFIG parameters or on mechanical
quantities, such as total inertia and aerodynamic torque.
One common weakness of the above mentioned approaches [1–20]
is their sensitivity to various errors [21] resulting from the employment
of classical flux estimation or simple integral method. These errors in-
clude DFIG resistance variations, measurement noises, digital approx-
imation errors, DC offset, and initial conditions errors. In addition,
these methods require rotor current sensors, which increase the system
cost and may cause system malfunction under sensor faults.
The sensitivity of classical flux estimation, i.e., voltage model flux,
can be reduced by employing a low-pass (LP) filter instead of a pure
integrator [22,23]. This modification reduces the performance of the
drive system due to insertion of phase and magnitude errors in the
control loops. Refs. [24,25] have proposed two methods to diminish the
effect of this weakness. These methods, however, are not straightfor-
ward and increase the implementation cost. Another solution is based
on the current model flux [26], which depends heavily on the accuracy
of rotor time constant. This technique is difficult to implement, needs a
large set of machine parameters and requires rotor current sensors.
The authors of [21] have proposed a method for stator power
control, without using flux or rotor current measurements. In this

⁎
Corresponding author at: CDER, Centre de Développement des Energies Renouvelables, Route de l’Observatoire Bouzeréah, Algiers, Algeria.
E-mail address: ab.djoudi@cder.dz (A. Djoudi).
1
Institute of Engineering Univ. Grenoble Alpes.

http://dx.doi.org/10.1016/j.ijepes.2017.10.009
Received 11 May 2017; Received in revised form 19 September 2017; Accepted 7 October 2017
0142-0615/ © 2017 Elsevier Ltd. All rights reserved.
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

approach, however, the control law requires a high switching frequency Ps = vqs iqs, Qs = vqs ids (4)
for the implementation. This will cause high switching losses in the
From (3) and (4), the dynamics of active and reactive powers are
rotor-side converter and will reduce the converter efficiency and its
obtained as
lifetime. Moreover, the robustness and feasibility of this control method
has not been verified practically. dQ
⎧ dt s = −a1 ids vqs + ωs iqs vqs + a2 φds vqs+
This paper proposes the sliding mode control (SMC) of stator active ⎪ − a3 ωφqs vqs−a4 vdr vqs
⎪
and reactive powers for unknown rotor currents and flux with new
implementation scheme. The target of proposed approach is achieving a ⎨ dPs = −ωs ids vqs−a1 iqs vqs + a2 φqs vqs+
⎪ dt
reduced switching frequency compared to the classical method, even in ⎪ + a3 ωφds vqs−a4 vqr vqs + a3 vqs2
⎩ (5)
the case of applying a high sliding mode gain control. The rest of this
paper is categorized as follows. The DFIG model is presented in Section The sliding surfaces SP ,SQ relative to stator active and reactive
2. Sections 3 and 4 present the SMC for unknown rotor currents and powers are defined as
flux, and the proposed scheme, respectively. In Section 5, simulation
results are presented to verify the viability of new method. Section 6 ⎧ SP = Ps−Pref
⎨ S = Qs−Qref (6)
presents the experimental results and conclusions are given in Section ⎩ Q
7. where Pref and Qref are reference values of Ps,Qs , respectively.
̇
In the sequel, it is assumed that Pref and Qref and their derivatives Pref
2. Model of doubly fed induction generator ̇ are bounded and available.
and Qref
According to (5) and (6), the dynamics of sliding surfaces SP and SQ
In the rotational reference frame (or (d,q) reference frame) and are given by
based on the usual assumptions [27], the DFIG model can be given by
dSQ
the following equations ⎧ dt
= Y1−a4 vdr vqs

dφds
⎨ dSP = Y2−a4 vqr vqs
⎧ vds = Rs ids + −ωs φqs ⎩ dt (7)
dt
⎪ dφqs
⎪ With
⎪ vqs = Rs iqs + dt
+ ωs φds
⎨v = R i + dφdr ̇
⎧Y1 = −a1 ids vqs + ωs iqs vqs + a2 φds vqs−a3 ωφqs vqs−Qref
−(ωs −ω) φqr
⎪ dr r dr dt ⎪
⎪ dφqr Y2 = −ωs ids vqs−a1 iqs vqs + a2 φqs vqs + a3 ωφds vqs+
⎪ vqr = Rr iqr + + (ωs −ω) φdr ⎨
⎩ dt (1) ⎪ +a3 vqs2 −Pref
̇
⎩ (8)

⎧ φds = Ls ids + Lm idr Now, the dynamics of sliding surfaces are chosen as
⎪ φqs = Ls iqs + Lm iqr
dSQ
⎨ φdr = Lr idr + Lm ids ⎧ dt
= −K sign (SQ )−GSQ
⎪φ = L i + L i ; K ,G > 0
r qr m qs
⎨ dSP = −K sign (SP )−GSP
⎩ qr (2) ⎩ dt (9)
where where the gains K and G are positive values.
The terms −GSQ and −GSP are added in (9) to improve the con-
(ids,r ,iqs,r ) are the stator-rotor current components; vergence speed of active and reactive powers to their references Pref and
(φds,r ,φqs,r ) are the stator-rotor flux components; Qref , respectively.
(vds,r ,vqs,r ) are the stator-rotor voltage components; According to (8) and (9), the (dq) components of rotor voltages
Rr ,Rs are the rotor and stator resistances; (vdr ,vqr ) are calculated to satisfy the relation (9):
Lr ,Ls ,Lm are the rotor, stator, and mutual inductances; and
ω , ωs are the rotor and stator angular frequencies, respectively. ⎧ vdr = (K sign (SQ ) + GSQ + Y1)/(a4 vqs ) ; a ,v ≠ 0
4 qs
⎨ vqr = (K sign (SP ) + GSP + Y2)/(a4 vqs ) (10)
⎩
3. High-gain SMC for control of stator’s active and reactive powers When rotor voltage components take the form (10), the quantities
dS dS
SQ dtQ , SP dtP would be negative. In this case, SQ → 0 and SP → 0 . In
This section presents an SMC that allows the drive of DFIG by the other words, the SMC law or (10) ensures the regulation of Ps and Qs to
active and reactive stator powers in the case of unknown rotor currents the reference values Pref and Qref , respectively.
and flux. Unfortunately, the values of vdr ,vqr in (10) cannot be calculated ea-
According to (1) and (2), the dynamics of the stator current com- sily due to several reasons, including DFIG parameters variations,
ponents can be obtained as follows (see Appendix A). modeling errors, and stator flux estimation difficulties. For this reason,
dids the gain K in (10) must be chosen high to ensure the convergence.
⎧ = −a1 ids + ωs iqs + a2 φds −a3 ωφqs −a4 vdr + a3 vds
dt To overcome the problems related to estimation methods and to
⎨ diqs = −ωs ids−a1 iqs + a2 φqs + a3 ωφds −a4 vqr + a3 vqs reach the SMC robustness, the control law in (10) is modified as
⎩ dt (3)

where ⎧ vdr = (K sign (SQ ) + GSQ + Y1)/(a4 vqs )

⎨ vqr = (K sign (SP ) + GSP + Y2)/(a4 vqs ) (11)
⎩
1 1 ⎞ 1 1 (1−σ )
a1 = ⎛ ⎜+ , a2 = ⎟ , a3 = , a4 = ; where
σT
⎝ s σTr⎠ σL T
r s σL r σLm
L2 ⎧Y1 = −a1 ids vqs + ωs iqs vqs + a2 φds vqs+
and σ = 1− L mL is the dispersion coefficient. ⎪
s r
⎪ − a3 ωφqs vqs−Qref ̇
If it is assumed that the q axis of the rotating reference frame co-
incides along the stator voltage vector, the direct component of the ⎨Y2 = −ωs ids vqs−a1 iqs vqs + a2 φqs vqs + a3 ωφds vqs+
⎪
stator voltage will be zero. Then, the expression of stator active and ⎪ +a3 vqs2 −Pref
̇
⎩ (12)
reactive powers, i.e., Ps,Qs can be given by (4)

348
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

Fig. 1. Classical implementation of HG-SMC scheme.

a1,a2,a3,a4 stand, respectively, for a1,a2,a3,a4 when the nominal values

of DFIG parameters (given in Appendix A) are employed. One has to
note that the stator flux components remain close to the nominal values
as long as the stator is directly connected to the grid.
In the following paragraphs, the necessary conditions for the sliding
mode gain K is obtained to guarantee the robustness of SMC. To
achieve this goal, the following Lyapunov function is associated with
the sliding surface dynamics:

1 2 1
VS = SP + SQ2
2 2 (13)

The time derivative of (14) is given by

dSP dSQ
Vṡ = SP + SQ
dt dt (14)

By substituting (7) into (14), one can obtain

Vṡ = SP (Y2−a4 vqr vqs ) + SQ (Y1−a4 vdr vqs ) (14a)

The equivalent quantities of vdr ,vqr in (11) are inserted to (14a) and
the expression of VṠ is rewritten as (15):

⎛ (K sign (SP ) + GSP + Y2) ⎞ ⎞

VṠ = SP ⎜Y2−a4 ⎛⎜ ⎟ vqs
⎟
⎝ (a4 vqs ) ⎠ ⎠
⎝
⎛ (K sign (SQ ) + GSQ + Y1) ⎞ ⎞
+ SQ ⎜Y1−a4 ⎛⎜ ⎟ vqs
⎟
⎝ (a4 vqs ) ⎠ ⎠ (15)
⎝

or

a a a a
VṠ = ⎛Y2− 4 Y2 ⎞ SP + ⎛Y1− 4 Y1 ⎞ SQ− 4 G (SP2 + SQ2 )− 4 K (|SP | + |SQ |)
⎜ ⎟ ⎜ ⎟

⎝ a 4 ⎠ ⎝ a 4 ⎠ a 4 a4
(16) Fig. 2. (a) and (b) Principle of sinusoidal-PWM. (c) Duty cycle in case of classical HG-SMC
implementation.
The time derivative expression of VṠ remains negative if

a4 ⎛ ⎛ a4 ⎞ a 4. HG-SMC implementation with the reduced switching frequency

K> ⎜ Y2−
⎜ Y2 SP + ⎛Y1− 4 Y1 ⎞ SQ⎞⎟
⎟ ⎜ ⎟

a4 ⎝ ⎝ a4 ⎠ ⎝ a4 ⎠ ⎠
In this case, VS → 0 , which means that SQ → 0 and SP → 0 .From
(11), one can note that the rotor voltage components, which are ne-
cessary for the control, do not depend on rotor current components,
which means that the developed control law in (11) or high-gain sliding
mode control (HG-SMC) ensures the control of stator active and re-
active powers without the rotor current measurement or flux estimation
under the condition of (17). This fact confirms that a better margin of
robustness is achieved if a greater value for K is selected.
(17)
In this section, a new implementation scheme is developed for HG-
SMC approach. The main objective is to reduce the switching frequency
of the rotor-side converter (RSC), compared to the conventional one.
The classical implementation scheme is described in Fig. 1. The
DFIG stator circuit is directly connected to the power grid. The DFIG
rotor circuit is driven by the RSC which in this paper is a two-level
three-phase converter.
At time kT (T represents the sampling time and k is an integer
value), the dq components of the rotor voltage references are calculated

349
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

using HG-SMC and are given in (18)–(19).

⎧ ⎛ K sign (SQ (kT ))+ ⎞

⎪ vdr (kT ) = ⎜+GS (kT ) + Y (kT ) ⎟ (a4 vqs (kT ))
⎪ ⎝ Q 1 ⎠
⎨ K sign (S (kT )) + ⎞
⎛
⎪ vqr (kT ) = ⎜
P
⎟ (a4 vqs (kT ))
⎪ ⎝+GSP (kT ) + Y2 (kT ) ⎠ (18)
⎩

where

⎧Y1 (kT ) = −a1 ids (kT ) vqs (kT ) + ωs iqs (kT ) vqs (kT )+
⎪ + a φ v (kT )−a ωφ v (kT )−Q̇ (kT )
2 ds qs 3 qs qs ref
⎪
⎪Y2 (kT ) = −ωs ids (kT ) vqs (kT )−a1 iqs (kT ) vqs (kT )+
⎪
⎪ + a2 φqs vqs (kT ) + a3 ωφds vqs (kT ) + a3 vqs (kT )2+
⎪
̇ (kT )
−Pref
⎨
S
⎪ P (kT ) = Ps (kT )− P ref (kT )
Fig. 3. Update of reference value in the proposed HG-SMC implementation. ⎪ SQ (kT ) = Qs (kT )−Qref (kT )
⎪
̇ (kT ) = Pref (kT ) − Pref (kT − T )
⎪ Pref
⎪ T
⎪Q̇ (kT ) = Qref (kT ) − Qref (kT − T )
⎩ ref T (19)

Fig. 4. Investigation of proposed approach, (a) Stator active

power and the reverence active power and (b) Stator re-
active power and the reference reactive power.

350
A. Djoudi et al.
Then, time domain references of three phase rotor voltages
(var (kT ),vbr (kT ),vcr (kT )) which are required to control the DFIG are
obtained from dq/ abc transformation (Inverse Park transformation).
Based on these references, a pulse-width modulation (PWM) algorithm
is used to synthesize the desired voltages (var (kT ),vbr (kT ),vcr (kT )) on the
rotor-side converter (RSC). At time k T, the terminal voltages
(va (kT ),vb (kT ),vc (kT )) of the RSC are function of switching states of
converter switches, i.e., L (kT ) and the DC bus voltage Vdc .
The PWM strategies for voltage source converters are mainly di-
vided into two categories: sinusoidal-PWM and space vector PWM
[28–31]. In these methods, as it is shown in Fig. 2a for a single phase
system, the variation of duty cycle in a switching period, i.e., NT , is
neglected. It is worth noting that the duty cycle is proportional to the
shape of desired voltage. Thus, the reference waveform is compared to a
351
Electrical Power and Energy Systems 96 (2018) 347–356
Fig. 5. Investigation of VC approach, (a) Stator active power and the
reverence active power and (b) Stator reactive power and the re-
ference reactive power.
saw-tooth carrier and the corresponding switching states of the con-
verter are obtained. Moreover, during the switching period, the mean
value of output voltage would be equal to the reference value which is
shown in Fig. 2b.
In this study, the rotor-side converter is a three phase inverter and
the corresponding three phase duty cycles are expressed as
v
Dl (x ) = ⎛ xr + 1⎞
⎜ ⎟ 2, x = a,b,c
⎝ Vdc ⎠ (19.a)
But in the classical HG-SMC approach, due to high gain of K , the
assumption of constant duty cycle or voltage reference during a
switching period is not correct (see Fig. 2c). To overcome this issue, one
may reduce the switching period T significantly to reach a similar op-
erating condition as sinusoidal-PWM. However, this will cause a high
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

Fig. 6. Steady-state behavior of proposed approach, (a) stator current

waveforms and (b) rotor current waveforms.

switching frequency which is not good for medium and high power ⎧ Dn (kT ) = Dl (kT ) where: Sw (kT ) = 0
converters. The alternative approach is to decrease the gain K sig- ⎪ Dn (kT + mT ) = Dn (kT )
nificantly, but this solution will reduce the control system robustness ⎨ Sw (kT ) = Sw (kT + NT ) = 0
which is not desirable. ⎪ m = 1: N −1, where : S (kT + mT ) ≠ 0
⎩ w (20)
Here, another solution is proposed to keep the switching frequency
same as the conventional PWM approaches. The main idea is to update Accordingly, the proposed scheme is similar to the classical PWM
the reference value for the PWM algorithm just at the beginning of each methods, but the duty cycle is used to find the states L (kT ) of converter
switching period instead of each sampling period which is shown in switches M based on (20). It is evident that the switching frequency is
Fig. 3. kept lower than the classical implementation of HG-SMC approach.
In the following, the duty cycle Dn is obtained as a function of
classical duty cycle Dl , where the saw-tooth carrier Sw has the switching
5. Simulations results
periodic of NT :

In order to verify the effectiveness of new method, two simulations

352
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

Fig. 7. Photo of hardware prototype, (a)

Electrical part of the test bench. (b)
Electromechanical part of the test bench.

Table1
Utilized DFIG parameters in the study.
Quantity
Nominal power
Nominal stator resistance
Nominal rotor resistance
Nominal stator leakage inductance
Nominal rotor leakage inductance
Nominal magnetizing inductance
Number of pair poles
Grid magnitude and frequency
DC bus voltage
are carried out. In the first simulation, the effectiveness of new method
against the parameters errors is investigated and in the second simu-
lation, it is compared to the classical vector control (VC) approach. In
Value
these simulations, the switching frequency of the rotor side converter is
7.5 kW set to 8 kHz.
0.5 Ω The parameters of the studied DFIG are given in Appendix A. The
1.2 Ω gains of the control system are also selected as k = 10 4,G = 50 . The
4.71 mH
1.13 mH
stator resistances and inductances are changed deliberately +50%
57.71 mH and ± 40%, respectively. The reference of stator reactive power is set to
2 zero and is demonstrated in Figs. 4b and 5b and the profile of active
140 √ 2 V,50 Hz power is demonstrated in Figs. 4a and 5a (by the red color).
350 V The corresponding waveforms of stator’s active and reactive powers
for the proposed approach and the VC approach are given respectively

Fig. 8. Active and reactive stator powers, ampli-

1div=2kW tude of rotor current, and speed.

1div=2kVA

1div=85 tr/min

1div=5.48A

353
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

Fig. 9. Experimental investigation of proposed technique, (a) tracking

errors for active power control, (b) tracking errors for reactive power
control, and (c) dq components of stator current.

(a)

200

-200
VAR

200

150
-400 100

50

-600 0

-50

-100

-800 581 581.05 581.1 581.15 581.2

100 200 300 400 500 600 700 800 900 1000
Time (s)
(b)

(c)

in Figs. 4 and 5.
As it is seen from Figs. 4 and 5, the active and reactive powers well
track the reference values in both steady state and dynamic conditions.
The RMS of active power error is 0.35 kW in Fig. 4a and 0.32 kW in
Fig. 5a. In addition, the RMS of reactive power error is 0.25 kVAR in
Fig. 4b and 0.38 kVAR in Fig. 5b.
The waveforms of stator and rotor currents of the proposed ap-
proach are given in Fig. 6, where the computed THDs are 4.4% and
4.88%, respectively. These quantities well satisfy the standard limits
[32].
According to the simulation results, the proposed approach could
control the active and reactive powers with a negligible error while the

354
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

system parameters have been changed deliberately. Moreover, the THD
of stator current is less than 5% and satisfies the grid code [32].
6. Experiment results
6.1. Test bench components
The presented test bench in Fig. 7 allows emulating the behavior of
wind energy conversion systems based on the DFIG. This system can be
divided into two parts. The first part consists of electrical components
(Fig. 7a) which enable the control of DFIG and the DC motor. A back-to-
back drive and a four-quadrant converter are utilized for the drive of
DFIG and DC motor, respectively. In addition, the DFIG is controlled by
dSPACE DS1005 control board and the DC motor is controlled by Texas
Instruments DSP card, i.e., TMS 320F240.
Fig. 7b shows the second part of the test bench or the electro-
mechanical chain which contains a 7.5 kW DFIG with the parameters
given in Table 1 and a 5.5 kW Siemens DC motor to produce the desired
turbine torque. The DC motor and the DFIG are coupled via a multiplier
with a ratio of 1.41 and an encoder is utilized for the speed estimation.
Two computers are utilized for the control of whole test-bench. The
first and main computer is utilized for the control of DFIG through
Matlab/Simulink environment and dSPACE tool boxes. The second
computer is utilized for programming the torque of DC machine via DSP
board and the external link from main computer.
6.2. Implementation procedure
The proposed control system for the DFIG has been implemented in
Matlab/Simulink environment and the dSPACE tool boxes have been
utilized for performing necessary ADC, DAC, PWM and Sensing func-
tions. Moreover, the turbine behavior is simulated in this program and
the corresponding torque value is generated and sent to the DC machine
control board via an external link. This torque is proportional to the
square of the rotor speed. In addition, before running the proposed
algorithm, the synchronization of the DFIG with the grid is done using
the control of stator voltage, which is achieved by the rotor-side
converter.
6.3. Test results
In the experimental tests, the same gain values and switching fre-
quency as the simulation study are utilized. The reactive power is set to
zero and the stator active power is changed in a step-wise manner while
it is filtered by a low-pass filter same as the simulation study. The ro-
bustness of the proposed method is tested against parameter changes by
changing the values of DFIG resistances and inductances in the control
law similar to the simulation part, i.e., + 50% and ± 40% for resistances
and inductances, respectively. The obtained experimental results based
on HG-SMC are shown in Fig. 8.
Tracking errors of the proposed approach are demonstrated in
Fig. 9a and b. As it is seen, the maximum tracking errors for active and
reactive powers are about 350 W and 280 VAR , respectively. These er-
rors are acceptable for a 7.5 kW DFIG system.
The viability of the proposed approach is also confirmed by de-
monstration of stator dq current components in Fig. 9c. It is seen that
these quantities remain inside the acceptable bounds for the considered
DFIG system.
7. Conclusions
This paper presented a new stator power control based on HG-SMC
for a DFIG without using any sensor for rotor currents and flux variable.
The presented implementation scheme allows achieving the control
objectives with a lower switching frequency compared to classical
method. The developed approach provides several advantages, such as
robustness, sensor-less operation, lower converter losses and better life
time. The validity of new approach was verified through simulations
and real-time experiments.
Acknowledgment
This work has been funded by the project PSP, Innov’hydro.

Appendix A

In order to find the dynamics of (dq) stator currents in (3), Eqs. (1) and (2) are used in the mathematical calculations. By assuming that the
variations of inductances are negligible compared to electrical variables and rewriting (1) and (2), one can obtain:
dφ
⎧ dtds = vds−Rs ids + ωs φqs
⎪ dφ
⎪
⎪ dt
qs
= vqs−Rs iqs−ωs φds
⎨ dφdr = vdr −Rr idr + (ωs −ω) φqr
⎪ dt
⎪ dφqr
⎪ dt = vqr −Rr iqr −(ωs −ω) φdr (A.1)
⎩
And,
dφ dids didr
⎧ dtds = Ls dt
+ Lm dt
⎪ dφ
diqs diqr
⎪
⎪ dt
qs
= Ls + Lm
dt dt
⎨ dφdr = Lr
didr
+ Lm
dids
⎪ dt dt dt
⎪ dφqr diqr diqs
⎪ dt = Lr + Lm (A.2)
⎩ dt dt

In other words, the rotor currents and flux value are extracted from (2) as:
1 L
⎧idr = φ − si
Lm ds Lm ds
⎨ iqr = 1 L
φ − si
Lm qs Lm qs
⎩ (A.3)
And,

355
A. Djoudi et al. Electrical Power and Energy Systems 96 (2018) 347–356

Lr Lm σ
⎧ φdr = φ
Lm ds
+ i
σ − 1 ds
⎨φ = Lr Lm σ
φ + i
σ − 1 qs
⎩ qr Lm qs (A.4)
After taking into account (A.3) and (A.4); the formulas (A.1) and (A.2) can be rewritten as:
di di
⎧ Ls dtds + Lm dtdr = vds−Rs ids + ωs φqs
⎪ diqs diqr
⎪ Ls dt + Lm dt = vqs−Rs iqs−ωs φds
⎪

⎪ dt dt Lm Lm ( Lm )
⎨ Lr didr + Lm dids = vdr −Rr 1 φds − Ls ids + (ωs −ω) Lr φqs + Lm σ iqs
σ−1 ( )
⎪
⎩
diqr diqs 1 Ls
( Lr
)
Lm σ
⎪ Lr dt + Lm dt = vqr −Rr Lm φqs − Lm iqs −(ωs −ω) Lm φds + σ − 1 ids ( ) (A.5)
dids diqs
The variables and are found by solving the last system equation. So,
dt dt
di
⎧ dtds = −a1 ids + ωs iqs + a2 φds −a3 ωφqs −a4 vdr + a3 vds
⎨ diqs = −ωs ids−a1 iqs + a2 φqs + a3 ωφds −a4 vqr + a3 vqs
⎩ dt (A.6)

where a1 = ( 1
σTs
+
1
σTr ), a2 =
1
σLr Ts
, a3 =
1
σLr
, a4 =
(1 − σ )
σLm
.

References
[1] Xiong Pinghua, Sun Dan. Backstepping-based DPC strategy of a wind turbine-driven
DFIG under normal and harmonic grid voltage. IEEE Trans Power Electron
2016;31(6):4216–25.
[2] Sun Dan, Wang Xiaohe. Low-complexity model predictive direct power control for
DFIG under both balanced and unbalanced grid conditions. IEEE Trans Industr
Electron 2016;63(8):5186–96.
[3] Elghali SE Ben et al. High-order sliding mode control of DFIG-based marine current
turbine. In: 34th Annual Conference of Industrial Electronics, IECON 2008, IEEE;
2008.
[4] Chen SZ, et al. Integral variable structure direct torque control of doubly fed in-
duction generator. IET Renew Power Gener 2011;5(1):18–25.
[5] Mondal Suman, Kastha Debaprasad. Improved direct torque and reactive power
control of a matrix-converter-fed grid-connected doubly fed induction generator.
IEEE Trans Industr Electron 2015;62(12):7590–8.
[6] Martinez M Itsaso, et al. Sliding-mode control for DFIG rotor-and grid-side con-
verters under unbalanced and harmonically distorted grid voltage. IEEE Trans
Energy Convers 2012;27(2):328–39.
[7] Martinez Miren Itsaso, et al. Sliding-mode control of a wind turbine-driven double-
fed induction generator under non-ideal grid voltages. IET Renew Power Gener
2013;7(4):370–9.
[8] Nian Heng, Cheng Peng, Zhu ZQ. Independent operation of DFIG-based WECS using
resonant feedback compensators under unbalanced grid voltage conditions. IEEE
Trans Power Electron 2015;30(7):3650–61.
[9] Fateh Fariba, White Warren N, Gruenbacher Don. A maximum power tracking
technique for grid-connected DFIG-based wind turbines. IEEE J Emerg Sel Top
Power Electron 2015;3(4):957–66.
[10] Morfin Onofre A, et al. Torque controller of a doubly-fed induction generator im-
pelled by a DC motor for wind system applications. IET Renew Power Gener
2014;8(5):484–97.
[11] Moradi Hamed, Vossoughi Gholamreza. Robust control of the variable speed wind
turbines in the presence of uncertainties: a comparison between H∞ and PID
controllers. Energy 2015;90:1508–21.
[12] Zou Yu, Elbuluk Malik E, Sozer Yilmaz. Stability analysis of maximum power point
tracking (MPPT) method in wind power systems. IEEE Trans Ind Appl
2013;49(3):1129–36.
[13] Yang Bo, et al. Nonlinear maximum power point tracking control and modal ana-
lysis of DFIG based wind turbine. Int J Electr Power Energy Syst 2016;74:429–36.
[14] Song Yipeng, Nian Heng. Modularized control strategy and performance analysis of
DFIG system under unbalanced and harmonic grid voltage. IEEE Trans Power
Electron 2015;30(9):4831–42.
[15] Hu Jiabing, Xu Hailiang, He Yikang. Coordinated control of DFIG's RSC and GSC
under generalized unbalanced and distorted grid voltage conditions. IEEE Trans
Industr Electron 2013;60(7):2808–19.
[16] Cheng Peng, et al. Direct stator current vector control strategy of DFIG without
phase locked loop during network unbalance. IEEE Trans Power Electron
2017;32(1).
[17] Zhang Yongchang, Hu Jiefeng, Zhu Jianguo. Three-vectors-based predictive direct
power control of the doubly fed induction generator for wind energy applications.
IEEE Trans Power Electron 2014;29(7):3485–500.
[18] Hu Jiefeng, Zhu Jianguo, Dorrell David G. Predictive direct power control of doubly
fed induction generators under unbalanced grid voltage conditions for power
quality improvement. IEEE Trans Sustain Energy 2015;6(3):943–50.
[19] Hu Jiefeng, Zhu Jianguo, Dorrell David G. Model-predictive direct power control of
doubly-fed induction generators under unbalanced grid voltage conditions in wind
energy applications. IET Renew Power Gener 2014;8(6):687–95.
[20] Liu Xiangjie, Kong Xiaobing. Nonlinear model predictive control for DFIG-based
wind power generation. IEEE Trans Autom Sci Eng 2014;11(4):1046–55.
[21] Djoudi Abdelhak, Chekireb Hachemi, Bacha Seddik. Low-cost sliding mode control
of WECS based on DFIG with stability analysis. Turk J Electr Eng Comput Sci
2015;23(6):1698–714.
[22] Salvatore Luigi, Stasi Silvio, Tarchioni Lea. A new EKF-based algorithm for flux
estimation in induction machines. IEEE Trans Industr Electron
1993;40(5):496–504.
[23] Hu Jun, Wu Bin. New integration algorithms for estimating motor flux over a wide
speed range. IEEE Trans Power Electron 1998;13(5):969–77.
[24] Comanescu Mihai, Xu Longya. An improved flux observer based on PLL frequency
estimator for sensorless vector control of induction motors. IEEE Trans Industr
Electron 2006;53(1):50–6.
[25] Cirrincione Maurizio, et al. A new adaptive integration methodology for estimating
flux in induction machine drives. IEEE Trans Power Electron 2004;19(1):25–34.
[26] Rehman Habib-ur, et al. A new current model flux observer for wide speed range
sensorless control of an induction machine. IEEE Trans Power Electron
2002;17(6):1041–8.
[27] Lei Yazhou, Mullane Alan, Lightbody Gordon, Yacamini Robert. Modeling of the
wind turbine with a doubly fed induction generator for grid integration studies.
IEEE Trans Energy Convers 2006;21(1):257–64.
[28] Bowes SR, Lai YS. Investigation into optimising high switching frequency regular
sampled PWM control for drives and static power converters. IEE Proc – Electr
Power Appl 1996;143(4):281–93.
[29] Dang Dong Quang, et al. Experimental validation of a fuzzy adaptive voltage con-
troller for three-phase PWM inverter of a standalone DG unit. IEEE Trans Industr Inf
2015;11(3):632–41.
[30] Grandi Gabriele, Loncarski Jelena, Dordevic Obrad. Analysis and comparison of
peak-to-peak current ripple in two-level and multilevel PWM inverters. IEEE Trans
Industr Electron 2015;62(5):2721–30.
[31] Wei Lixiang, McGuire Jeffrey, Lukaszewski Richard A. Analysis of PWM frequency
control to improve the lifetime of PWM inverter. IEEE Trans Ind Appl
2011;47(2):922–9.
[32] IEEE recommended practice and requirements for harmonic control in electric
power systems; 2014. p. 519.

356