ISA Transactions 88 (2019) 258–267

Contents lists available at ScienceDirect

ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans

Practice article

Deadbeat–fuzzy controller for the power control of a Doubly Fed

Induction Generator based wind power system
∗
C.M. Rocha-Osorio , J.S. Solís-Chaves, Lucas L. Rodrigues, J.L. Azcue Puma,
A.J. Sguarezi Filho
Federal University of ABC - Santo André, SP, Brazil

highlights

• A novel Deadbeat Fuzzy Controller developed for a Wind Power System DFIG based with simulated and experimental results during normal operating
conditions.
• The application of the deadbeat fuzzy controller is a valid alternative to the DFIG Stator Power Control.
• Experimental test bench using a 3 kW DFIG and a DSP TMS320F28335 card.

article info a b s t r a c t

Article history: This paper proposes a Fuzzy Logic Controller for improvement of the steady-state response of a Doubly
Received 21 March 2018 Fed Induction Generator used in a wind energy system, and governed by means of a Deadbeat Power
Received in revised form 20 October 2018 Controller. The generator mathematical model is consistent with the Stator Flux oriented strategy in
Accepted 27 November 2018
the synchronous reference frame. Different simulation scenarios were developed in Matlab/Simulink
Available online 5 December 2018
to evaluate the dynamic and the steady-state responses. In order to obtain experimental results, the
Keywords: simulated scenarios were repeated by means of a test bench and a Digital Signal Processor board. These
Doubly fed induction generator (DFIG) results demonstrate that the response still follows the power references imposed, despite the fact that
Fuzzy control the generator parameters ( Rr , Ls and Lm ) were varied in a 30%. A lower steady-state error is also achieved
Deadbeat controller when compared with a Deadbeat and a classical PI controller. All the aforementioned evidence the proper
Power control
application of this Fuzzy Controller in a wind power system based on a Doubly Fed Induction Generator.
Wind generation
© 2018 ISA. Published by Elsevier Ltd. All rights reserved.
Renewable energy

1. Introduction increase their power levels up to the MW scale. Moreover, a cumu-

Wind Energy (WE) has experienced an enormous growth since
the year 2000, increasing its percentage in the global energy matrix
in countries such as China, USA, India, South-Africa, Brazil, and the
European Union. The WE global total cumulative generated at the
end of 2017 was 539.12 GW [1]. This type of renewable energy
represents a clean generation alternative and can contribute to
resolving the environmental crisis because, during its operation
stage, WE is free of greenhouse gas emissions.
Besides, according to the International Energy Agency, an es-
timated average carbon dioxide reduction of 600 g/kWh is con-
sidered to be obtained from WE [2]. This is possible because of
the development of new technologies for generators and power
converters [3,4] which have allowed wind turbines to considerably
∗ Corresponding author.
E-mail addresses: rocha.carlos@ufabc.edu.br (C.M. Rocha-Osorio),
sebastian.chaves@ufabc.edu.br (J.S. Solís-Chaves), rodrigues.lucas@ufabc.edu.br
(L.L. Rodrigues), jose.azcue@ufabc.edu.br (J.L.A. Puma),
alfeu.sguarezi@ufabc.edu.br (A.J.S. Filho).
lative capacity growth rate increase of up to 8.8% is expected by the
year 2022 [1].
In WE Systems, three well-defined control levels [5–7] are
present:–The first level corresponds to the energy control between
the generator and the power grid, and it is based on the behaviour
of the mechanical power income from the renewable source. The
second level controls the interaction between aerodynamics and
the wind turbine mechanical systems. This level will be responsible
for generating the control input signals for the first level, such as
the power references for the electric control of the wind turbine.
Finally, the third control level aims to command the integration
between the electrical network and the wind turbine. The three
control levels are represented in Fig. 1. In this paper, a first level
control is studied with a focus on the Rotor Side Converter (RSC).
Fuzzy Logic Controllers (FLC) applied to the control of electrical
machines have already been used in different and innovative ways,
along the time [8,9], mainly because of the Fuzzy Logic being
a powerful tool to resolve some control problems through the
emulation of human thoughts when a plant with well-known be-
haviour is present, as in the case of an electric generator. Literature

https://doi.org/10.1016/j.isatra.2018.11.038
0019-0578/© 2018 ISA. Published by Elsevier Ltd. All rights reserved.
 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267 259

for these uncertainties, improving the DFIG steady-state response

Nomenclature in a way similar to the one proposed for the SCIM; though with
an intriguing dynamic response due to the presence of the rotor
Acronyms windings. In this context, the control of the RSC in a Back-to-Back
topology, through a Deadbeat–FLC, can be an alternative to the use
DFIG Doubly Fed Induction Generator
of the classical PI controller for the power control in a WE System.
DSP Digital Signal Processor
In the last ten years, researchers have expanded the scope of FLC
FLC Fuzzy Logic Controller
to DFIG’s WE applications and to three-phase Controlled Voltage
RE Renewable Energy Sources; however, a power control for a DFIG-based WE system,
SFOC Stator Field Oriented Control using a Deadbeat–FLC, has not been reported yet. Nevertheless,
SVM Space Vector Modulation some interesting researches can be cited to demonstrate the FLC’s
RSC Rotor Side Converter vast field of action in renewable energy applications. For instance, a
SCIM Squirrel Cage Induction Machine kind of FLC called Fuzzy-PI Controller, was proposed with success,
mainly, for controlling the active and the reactive powers in WE
Symbols systems [13,14]. Direct Torque Control strategies were depicted
Rs , Rr stator and rotor resistances; in [15] and [16], using different types of generators for testing the
Ls , Lr , Lm stator, rotor, and mutual inductances; projected FLC algorithms. In [17], a Deadbeat–FLC was used for
controlling a Voltage Source (VSR) PWM Rectifier. In 2011 the same
σ total leakage factor;
VSR prototype was reported again with fully simulation results and
ωs , ωr , ωm synchronous, slip, and rotor angular fre-
an extended description of the experimental setup developed, as
quency;
well as the results obtained with this small-scale prototype [18].
θs , θr , θm stator flux, slip, and rotor angles;
Both of the above-mentioned papers present a Deadbeat con-
P, Q Active and reactive power; troller with overshoot and delay control. The FLC is used to obtain
v⃗ , ⃗i, ψ
⃗ voltage, current, and flux space vectors; a good trade-off between the current overshoot and phase delay,
v, i, ψ voltage, current, and flux components; when reference signals undergo a change in their values [18].
p number of poles; Elkhadiri et al. in 2018 present another FLC to govern a DFIG
T, k sampling period and sampling time; wind turbine. The control was performed by way of the rotor flux
e, 1e error and error variation; oriented vector technique [19]. The FLC uses the rotor current
error and the evolution of this error to estimate the control signal
Subscripts Ku . Kalaivani et al. in 2017 propose a FLC for a DFIG powered by
s, r stator and rotor; a Back-to-Back converter and operating under fault conditions.
The error and the change in the error of the rotor voltage signals
α, β direct-and quadrature-axis expressed at
are the inputs for the FLC based on the Mamdani inference [20].
stationary reference frame;
An interesting alternative with a FLC for a RSC in a DFIG-based
d, q direct-and quadrature-axis expressed at
WE system operating with constant power control strategy was
synchronous reference frame;
presented by Mahalakshmi et al. in 2016. This time, the FLC was
ref reference value.
employed to calculate the pulses of the RSC thyristors, by using
the mechanical and the synchronous speed signals. The proposed
controller was also tested in an experimental setup [21].
Similar to fuzzy controllers applied in power electronics, an-
reports some interesting background papers from the nineties.
other non-model based controller (free model controller) is pro-
Among them, the following could be cited: A Deadbeat–fuzzy con-
posed in [22] but applied to an autonomous humanoid robot,
trol algorithm was used as a method to regulate the phase current demonstrating fast and robust convergence to cope with external
in a Squirrel Cage Induction Machine (SCIM), validating it with sim- disturbances. This type of controllers could also be applied to
ulated and experimental results; but keeping in mind that the hys- the wind power system (which have a multi-input, multi-output,
teresis band had to be small in order to obtain a satisfactory perfor- highly coupled, nonlinear and drifting system response) and it
mance. This causes an increase in that the switching losses, being could be the future research direction of this project.
necessary the addition of a high-frequency resonant inverter, thus The fuzzification, decision making, rule base, and defuzzifica-
increasing the cost of the whole system [10]. The Deadbeat–fuzzy tion proposed here, need the rotor current error and the rotor
current regulator and the Fuzzy Speed Controller (FSC) described current error variation to obtain the new rotor voltages that will
in [11] demonstrate a superior performance when they are applied be applied in the next control step (as can be observed in Fig. 3b),
to an Induction Machine drive, obtaining a stable and a faster step the stator-flux oriented equations for the DFIG are modelled in
response. On the other hand, the Deadbeat controller in [11] is the synchronous reference frame, thus easing the complexity of
robust to the parameter variations. However, the deterministic the mathematical dynamic expressions of the generator and the
component in the current loop is difficult to be obtained, because Deadbeat-FLC. Therefore, the steady-state error percentage is null
it heavily relies on many system parameter variations, feedback in the power responses, even when a random wind speed profile
signal errors, and load disturbances. Furthermore, the determin- is considered, and a variation of 30% in the DFIG’s parameters is
istic component is a highly nonlinear term with a huge difficulty programmed in the DSP algorithm. A whole description of the
to be modelled by means of analytical equations. Therefore, the Deadbeat–FLC is depicted in the sections below.
FLC must be employed for this task [11]. In [12], the same authors The Deadbeat–FLC allows a better tracking of the reference
employed the Deadbeat–fuzzy current regulator and expanded the signals, and its superior performance is similar to other novel
experimental results range by doubling the values of the stator and DFIG controllers proposed in the literature, for instance: A DPC-
rotor resistances, testing the speed tracking, as well as the transient Deadbeat proposed in [23] and probed for normal operation, dur-
and the steady-state responses. In the present study, considering ing a balanced voltage sag conditions and a parametric variation
that the Deadbeat controller performance depends on the model up to 7%. Similarly, in [24] a robust finite control is depicted,
accuracy and the parameters evaluation, the FLC can compensate but in that case is used a cost function to minimize error due to
260 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267

Fig. 1. Wind power system based on DFIG.

parametric variations. To evidence the superior performance of

the Deadbeat–FLC presented in this study, comparative simulation
⃗ir ,dq = ψr ,dq − Lm ψsd
⃗
results were carried out via Matlab/Simulink. The two controllers (6)
chosen for comparing are a Deadbeat Power Controller proposed σ Lr σ Ls Lr
by Sguarezi [25], and a classical DFIG PI Controller adjusted as In steady state, and neglecting the voltage drop in stator re-
reported in [26], both working under the same operating condi- sistance, the stator flux is proportional to the network voltage vs .
tions than the Deadbeat-FLC herein proposed. For obtaining the Thus, the stator voltage components are:
experimental results, a 3 kW DFIG-based wind energy system, con-
trolled via a TMS320F28335 DSP system from Texas Instruments, vsd = 0 (7)
was used. In order to explain the Deadbeat–FLC, the present paper
is organized as follows: First, a fully description of the DFIG power
stator control strategy is resumed in Section 2.1. Next, Section 2.2, vsq = vs ≈ ωs ψs (8)
depicts the Deadbeat Controller theory. Then, the compensation
technique applied using the FLC is thoroughly explained in Sec-
tion 2.3. Later, the Simulation Results are presented in Section 3. 2.1. Power control using deadbeat–fuzzy controllers
After that, in Section 4, the experimental results are presented and
compared with the same scenarios simulated via Simulink. Finally, According to (5), (7) and (8) the active and reactive powers of
some conclusions for this work are enlisted in Section 5. the stator are calculated as follows :
3 Lm
2. Power control of doubly fed induction generator Ps = − vs irq (9)
2 Ls
Fig. 2 shows the equivalent circuit of a DFIG in dq reference
frame. Using the stator-flux-oriented vector control (SFOC) [6], the ψs
( )
3 Lm
generator voltages fluxes and currents can be expressed as follows: Qs = vs − ird (10)
2 Ls Ls

dψsd In (9) and (10), it can be realized that the active Ps and reactive
v⃗s,dq = Rs⃗is,dq + + jωs ψsd (1) Qs power can be controlled independently, as well as propor-
dt
tionally, according to the rotor current components irq and ird ,
respectively.
dψ
⃗ r ,dq
v⃗r ,dq = Rr⃗ir ,dq + + jωr ψ
⃗ r ,dq (2) Fig. 3(a) illustrates the rotor side power control with deadbeat–
dt fuzzy controllers. This control scheme consists of an estimation
stage (please see the box entitled as Estimator), where the active
ψ
⃗ s,dq = ψsd = Ls⃗is,dq + Lm⃗ir ,dq (3) and reactive power are estimated from stator voltages and currents
using the following equations:
3(
vsα isα + vsβ isβ
)
ψ
⃗ r ,dq = Lr⃗ir ,dq + Lm⃗is,dq (4) Ps = (11)
2

⃗is,dq = ψsd − Lm ψ ⃗ r ,dq (5) Qs =

3(
vsβ isα − vsα isβ
)
(12)
σ Ls σ Ls Lr 2
 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267 261

Fig. 2. Equivalent circuit of a DFIG in dq reference frame.

Fig. 3. Rotor side control using deadbeat–fuzzy controllers. (a) General scheme, (b) deadbeat–fuzzy loop.

Furthermore, the stator position (13) and stator angular fre- voltage 1v
⃗r ,dq . The deadbeat controller and the fuzzy controller are
quency (14) were estimated by way of the stator magnetic flux discussed in Sections 2.2 and 2.3.
components from (15).
ψsβ
( )
2.2. Deadbeat controller
θs = tan−1 (13)
ψsα
Using the equations of a discretized continuous linear system,
the Deadbeat controller calculates the input u(k), to ensure that the
vsβ − Rs isβ ψsα − (vsα − Rs isα ) ψsβ
( )
ωs = ( )2 (14) output x(k) reaches the reference values xref .
(ψsα )2 + ψsβ A continuous linear system is represented as follows:
∫ ( x̄˙ = Ax̄ + Bū + Gw̄
) (18)
ψ
⃗ s,αβ = v⃗s,αβ − Rs⃗is,αβ dt (15) ȳ = C x̄
where A, B and G are matrices n × n, w̄ (k) is the perturbation vector,
The mechanical rotor position, θm , can be calculated from:
∫ and C is a matrix identity.
p thus, discretizing (18) :
θm = ωm dt (16)
2
x̄(k + 1) = Ad x̄(k) + Bd ū(k) + Gd w̄ (k) (19)
and the slip position θr , is calculated using (13) and (16): ∫τ
∫where, Ad = eAT = ∼ I + AT , Bd =
0
eAT Bdτ ∼
= BT , Gd =
θr = θs − θm (17) τ AT
e Gd τ ∼ GT , T (1/fm = 1/10, 000 Hz) is the sampling period
=
0
In addition, another stage of the control scheme regards the and k the sampling time.
proposed controller. In Fig. 3(b), the deadbeat control loop com- In order to guarantee a null error, the input is calculated in the
pensated with a Mamdani type fuzzy logic controller (FLC) is de- following way [27]
tailed. It can be noticed that the deadbeat controller has the role ū(k) = F (x̄ref − x̄) (20)
of generating the rotor voltage v ⃗r′ ,dq , and the fuzzy control aims
at compensating the steady-state error by providing the rotor where, x̄ref represents the reference, while F is the gain matrix.
262 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267

Fig. 4. The block diagram of a deadbeat controller. Fig. 5. Input–output structure of the Deadbeat controller.

L
Substituting (20) in (19), and considering that x̄ref = x̄(k + 1),  
the input ū(k) can be written using the following equation [25]:
] [
ωr L m
[ ]
0 σ Lr isd
F + −ωr Lm (31)
0 isq
   [ σ Lr
ū(k) = Bd Ad Ad x̄ref − x̄(k) − Ad Gd w̄ (k)
−1 −1 −1
]
(21)
Expressing (31) in form of a discretized differential equation
The block diagram of the deadbeat controller can be observed where T is a sampling period at time k, and renaming x̄ = ⃗ir ,dq ,
in Fig. 4 Ad = H, Bd = K , ū = v
⃗r ,dq , Gd = L and w̄ = ⃗is,dq .
By applying the deadbeat control to the control of the DFIG The vector for the rotor current can be estimated:
powers, the voltage of the rotor expressed in (2) is rewritten in ⃗ir ,dq (k + 1) = Ad⃗ir ,dq (k) + Bd v⃗r ,dq (k) + Gd⃗is,dq (k) (32)
terms of the currents, as follows:
ψ
⃗ r ,dq ψ
⃗ r ,dq and its components can be written according to:
⃗ir ,dq Ad
d

    
ref
v⃗r ,dq = Rr⃗ir ,dq + (Lr⃗ir ,dq + Lm⃗is,dq ) + jωr (Lr⃗ir ,dq + Lm⃗is,dq ) (22) [  ] [
  ] [
dt ωr T
1 − RσrLT
]
ird (k + 1) σ ird (k)
In other words, the rotor voltage components can be expressed = −ωr T
r
irq (k + 1)
σ
1 − RσrLT irq (k)
as [28]: r

Bd
d
vrd = Rr ird + (Lr ird + Lm isd ) − ωr (Lr irq + Lm isq )
  ] [
(23) T
vrd (k)
[ ]
dt σ Lr
0
+
0 T
σ Lr
vrq (k)
d Gd
vrq = Rr irq + (Lr irq + Lm isq ) + ωr (Lr ird + Lm isd ) (24)
dt
[  ] [
ωr Lm T ]
ψ 0 σ Lr isd (k)
but considering that isd = L s − LLm ird , isq = − LLm irq and substituting + −ωr Lm T (33)
s s s 0 isq (k)
them in the derivative of (23) and (24): σ Lr

ψs At last, the rotor voltage components can be calculated using

[ ( )]
d Lm
vrd = Rr ird + Lr ird + Lm − ird −ωr (Lr irq + Lm isq ) (25) (21) and (33):
dt Ls Ls
ird,ref − ird (k)
[ ( )] vrd (k) = σ Lr + Rr ird (k) − Lr ωr irq (k) − Lm ωr isq (k)
d Lm T
vrq = Rr irq + Lr irq + Lm − ird + ωr (Lr ird + Lm isd ) (26)
dt Ls (34)
( )
d ψs
and taking in-to account that dt Ls
= 0 because of SFOC. By
doing so, the voltage rotor component can be defined as irq,ref − irq (k)
vrq (k) = σ Lr + Rr irq (k) + Lr ωr ird (k) + Lm ωr isd (k)
(
Ls Lr − L2m
)
dird T
vrd = Rr ird − Lr ωr irq + − ωr Lm isq (27) (35)
Ls dt
where the rotor current references from (9) and (10) can be ex-
(
Ls Lr − L2m
)
dirq pressed as:
vrq = Rr irq + Lr ωr ird + + ωr Lm isd (28)
2Qs,ref Ls ψs
Ls dt ird (k + 1) = ird,ref = − + (36)
From the rotor components expressed in (27) and (28) the rotor 3vs Lm Lm
voltage vector is:
2Ps,ref Ls
(
L2m
) ⃗
dir ,dq irq (k + 1) = irq,ref = − (37)
v⃗r ,dq = (Rr + jLr ωr ) ⃗ir ,dq + jLm ωr⃗is,dq + Lr − (29) 3vs Lm
Ls dt Regarding (34) and (35), it can be perceived that the Deadbeat
thus, writing (29) in state–space form: controller can adjust the rotor voltage employing the rotor current
(⃗ir ,dq ), its references (⃗ir ,dqref ), the stator current (⃗is,dq ), and the
d⃗ir ,dq angular speed (ωr ). In Fig. 5, the input–output structure of the
= H⃗ir ,dq + K v⃗r ,dq + L⃗is,dq (30)
dt deadbeat controller is illustrated.
rewriting (30), the terms H, K , and L are defined as follow:
2.3. Design of fuzzy logic controller
H K
[  ] [ Fuzzy logic controllers (FLCs) are widely used in various sys-
[  ] [
ωr
[ ]
dird −Rr 1
vrd
] ]
σ Lr σ ird σ Lr
0
dt = + tems, being one of their main advantages the offering of solu-
dirq −ωr
σ
−Rr
σ Lr
irq 0 1
σ Lr
vrq tions based on experience. This allows a performance which is
dt
 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267 263

equivalent to a human operator, without using the mathematical Table 1

model of the system. For this reason, this type of controller is used Fuzzy rules, 1vrd and 1vrq [29].

in complex systems in which the mathematical model is unable 1vrd 1vrq

to respond with accuracy (e.g. parametric variation and external 1eird /eird NB NM NS ZE PS PM PB 1eirq /eirq NB NM NS ZE PS PM PB
disturbances). NB NB NB NB NM NS NS ZE NB NB NB NB NM NS NS ZE
In order to compensate the deadbeat controller steady state NM NB NM NM NM NS ZE PS NM NB NM NM NM NS ZE PS
error, two FLCs are used to estimate the rotor voltage components NS NB NM NS NS ZE PS PM NS NB NM NS NS ZE PS PM
ZE NB NM NS ZE PS PM PB ZE NB NM NS ZE PS PM PB
1vrd and 1vrq . The input of each controller are the rotor current PS NM NS ZE PS PS PM PB PS NM NS ZE PS PS PM PB
error eird and eirq and its variations 1eird and 1eirq , where: PM NS ZE PS PM PM PM PB PM NS ZE PS PM PM PM PB
PB ZE PS PS PM PB PB PB PB ZE PS PS PM PB PB PB
ird,ref (k+1) ird (k)
 
eird = ird,ref − ird (38) Table 2
Network conditions.
Condition Stator power Power Time
irq,ref (k+1) irq (k) P [W] Q [var] Factor [s]
 
eirq = irq,ref − irq (39) 1 −1000 −620 0.85 1.1–1.2
2 −1500 930 0.85 1.2–1.5
3 −2000 0 1 1.5–1.6

eird − z −1 eird
1eird = (40) Table 3
T
Parameters of the DFIG.
Parameter Value
eirq − z −1 eirq 220/380 1–Y
1eirq = (41) Rated stator voltage Vs,n
T Rated stator current Is,n 12 A
Rated power Pn 3 kW
where T denotes the sample time. Rated speed 1800 rpm
The operation of the FLC can be summarized in three main Rated frequency fs 60 Hz
stages, as presented in Fig. 6. Stator resistance Rs 1
(1) Fuzzification: the crisp inputs (in this case eird , eirq , 1eird , and Rotor resistance Rr 3.1322 
Mutual inductance Lm 0.1917 H
1eirq ) are transformed into fuzzy inputs using the input member- Stator inductance Ls 0,2010 H
ship sets presented in Fig. 7. Each FLC has 5 triangular member- Rotor inductance Lr 0.2010 H
ship functions and 2 trapezoidal membership functions for each Number of poles p 4
linguistic variable. Lumped inertia constant J 0.05 kgm2
(2) Fuzzy processing: the fuzzy inputs are processed according Frequency modulation, fm 10,000 Hz
to the base rules presented in Table 1 [29]. In order to obtain the
fuzzy outputs, 49 fuzzy rules and a Mamdani type fuzzy infer-
ence were used; owing to the simplicity in the formulation of the error shown in (38) and (39), and the error variation shown in (40)
fuzzy rules, when compared with the Sugeno type fuzzy inference; and (41) are adjusted in such a way that the proposal achieves a
whereas the Sugeno type fuzzy inference requires a detailed math- satisfactory performance even in conditions of parametric varia-
ematical analysis for adjusting the several parameters. tion.
(3) Defuzzification: the fuzzy outputs are transformed into crisp
outputs using the traditional Centre of Gravity (COG) method, 3. Simulation results
which calculates the gravity centre of the area under the member-
ship function and the output membership sets illustrated in Fig. 7.
In order to validate the control scheme proposed in Fig. 3, a
The control outputs 1vrd and 1vrq , considering the COG method
simulation model developed in Matlab/Simulink was tested under
can be written by means of (42) and (43), respectively:
different power conditions and parameters, according to Tables 2
n
∑ and 3, respectively.
µRi 1vrd Ri Fig. 8 shows the simulation results of the wind power system
1vrd =
i=1
(42) when the generator operates under variable speed, according to
n
∑ the rotor speed profile illustrated in Fig. 8(a).
1vrd Ri
The results obtained were compared with the deadbeat con-
i=1 troller considered in [28,31], in which the rotor voltages were
calculated according to (34) and (35) based on the dynamic math-
n
∑ ematical model of the generator and without using a cost func-
µRi 1vrq Ri tion. Similarly, the results were compared with the PI controller
i=1
1vrq = n
(43) proposed in [26] whose controller gains were adjusted by the
∑ pole-placement technique and without using any cost function.
1vrq Ri The stator active power is presented in Fig. 8(b); where it can
i=1 also be observed a similar performance, in terms of steady-state
where, n is the number of rules, µ is the membership degree, and error and rise time, between the deadbeat–fuzzy controller (line
Ri is the rule evaluated. red) and the deadbeat (line blue), as well as a better performance
The FLCs setting is based on the assumption that the generator when compared with PI controller (green line). Similarly, the above
parameters are not error-prone, and that the errors that may behaviour is presented by observing the rotor current component
occur are reflected in the current variations [30]. Therefore, the irq shown in Fig. 8(d). By comparing the behaviour of the reactive
fuzzy controller was adjusted so that those variations would be power and the rotor current component shown in Fig. 8(c) and (e),
compensated, thus improving its performance. To that end, the respectively, a superior performance of the Deadbeat–FLC can be
264 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267

Fig. 6. Operation of the FLC.

Fig. 7. Fuzzy logic controller.

Table 4 3 kW DFIG coupled to a DC motor that emulates the mechan-

Controllers comparison. ical system of the wind turbine (rotor blades, shaft). The con-
Controller Parameters troller was implemented by way of a Digital Signal Processor (DSP)
Error [%] ts [s] tr [s]
TMS320F28335, and a signal acquisition board. Fig. 9 presents (a)
Deadbeat–fuzzy 4.3 0.0011 0.0008
the laboratory setup, and (b) the connection diagram.
Deadbeat 17 0.0008 0.0004
PI 5.3 0.035 0.048 The tests, which were performed at constant speed, variable
speed, and parametric variations; were implemented as follows:

4.1. Constant rotor speed test

perceived. In Table 4, the percentage of the steady-state error, the
settling time (ts ), and the rise time (tr ) are compared for all the
alternatives.
4. Experimental results
The deadbeat–fuzzy controller proposed in Fig. 3 was exper-
imentally validated using a small-scale setup compounded of a
In this test, the deadbeat–fuzzy controller was tested under
constant rotor speed (ωm = 1700 rpm). In addition, the following
rotor current references: ird,ref : 3A to 1A and 1A to 3A; irq,ref : 1A to
3A and 3A to 1A, were used In Fig. 10(a) and (b) the response of the
deadbeat–fuzzy controller, according to the conditions previously
described, is presented. Besides, the experimental results of the
stator current and the stator voltage in phase A were presented
 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267 265

Fig. 8. Response of the deadbeat, deadbeat–fuzzy and PI controllers under rotor speed variation : (a) rotor speed profile, (b) stator active power, (c) stator reactive power,
(d) rotor current component irq (e) rotor current component ird . (For interpretation of the references to colour in this figure legend, the reader is referred to the web version
of this article.)

Fig. 9. Setup of a 3 kW DFIG system (a) Laboratory Setup, (b) Connection Diagram.

in Fig. 10(c); while, in Fig. 10(d), an experimental comparison
between the deadbeat–fuzzy and the PI controllers is also pre-
sented. The experimental results indicated an adequate behaviour
of the controller, with a settling time equal to 1.8 ms, without
overshooting and a null steady-state error, in spite of the noise
caused by the switching of the back to back converter.
4.1.1. Variable rotor speed test
In this test the same rotor current references presented in
the constant rotor speed test were considered. However, it was
applied a variable rotor speed according to Fig. 11 (purple line), in
which, roughly speaking, the speed varied from 2000 rpm to 1670
rpm where the machine synchronous speed was 1800 rpm. The
rotor current variation, in the reference frame fixed in the rotor
physical structure (ir ,α r ), can be observed in the blue line in the
same figure. Finally, an invariant behaviour can be perceived in the
current components (ird —green line and irq —orange line), despite
the change of the rotor speed.
4.1.2. Parameter variation test
In this test, the same rotor current references used in the pre-
vious tests were also considered, as well as a constant rotor speed.
Nevertheless, a variation of 30% in the rotor resistance Rr , stator
inductance Ls , and magnetization inductance Lm was applied. In
Fig. 12, it can be observed that the response of the Deadbeat
controller compensated with fuzzy, presented a behaviour similar
to the constant speed test, thus demonstrating the advantages of
the proposed control to counteract the negative effects due to the
machine parametric variation.
266 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267
Fig. 10. Experimental results under constant rotor speed (a) ird and irq (100 ms/div.), (b) Detail in the rotor current step ird and irq (2 ms/div.), (c) stator current and stator
voltage in phase A (10 ms/div.). (d) Experimental comparison between deadbeat–fuzzy and PI controllers.
Fig. 11. Rotor current response considering a variation of rotor speed from 2000
rpm to 1670 rpm (1 s/div.). (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
Fig. 12. Rotor current response considering parameters variation of 30% (100
ms/div.).
5. Conclusion
A Deadbeat-fuzzy Controller designed to decrease the steady
state error and improve the robustness of the machine param-
eter variations has been presented in this paper. This one was
used for controlling the stator power of a DFIG-based wind gen-
eration system. This new controller had been simulated using
Matlab/Simulink and tested experimentally under different power
conditions, mechanical speed, and a parametric variation of up to
30% on Rr and Lr , in an experimental manner.
The simulation results corroborate that the compensation made
by the fuzzy logic controller allows a more precise DFIG power
control, reducing the steady-state error that the deadbeat con-
troller by itself cannot avoid. These results were also demonstrated
in an experimental way by using a small-scale prototype at the
same conditions. Regarding all the above mentioned, the proposed
Deadbeat Fuzzy Controller is a suitable solution to the rotor side
control in a DFIG-based wind energy system.
Acknowledgements
The authors would like to thank to CAPES, CNPQ(405757/2018-
2) and FAPESP (17/04623-3) from Brazil for their financial support.
References
[1] GWEC. Global wind report - annual market update 2017. Technical report,
Global Wind Energy Council; 2018.
[2] GWEC. Wind power is crucial for combating climate change. Technical report,
Global Wind Energy Council; 2008.
[3] Blaabjerg F, Ma K. Future on power electronics for wind turbine systems.
Emerg Sel Top Power Electron, IEEE J 2013;1:139–52.
[4] Oliveira Filho ME, Gazoli Jonas R, Sguarezi Filho AJ, Ruppert Filho E. A control
method for voltage source inverter without dc link capacitor. In: Power
Electronics Specialists Conference, 2008. PESC 2008. IEEE, 2008, Rhodes. Anais
PESC. 2008, p. 4432–7.
[5] Asuhaimi A, Zin BM, Pesaran M, Khairuddin A, Jahanshaloo L, Shariati O. An
overview on doubly fed induction generators controls and contributions to
wind based electricity generation. Renew Sustain Energy Rev 2013;27:692–
708.
[6] Abad G, Lopez J, Rodriguez M, Marroyo L, Iwanski G. Doubly fed induction
machine: Modeling and control for wind energy generation, vol. 85. John
Wiley & Sons; 2011.
[7] BlaaBjerg F, Ma K. Wind energy systems. Proc IEEE 2017;PP:1–16.
[8] Song Y-H, Johns AT. Application of fuzzy logic in power systems. ii. com-
parison and integration with expert systems, neural networks and genetic
algorithms. Power Eng J 1998;12:185–90.
[9] Suganthi L, Iniyan S, Samuel AA. Applications of fuzzy logic in renewable
energy systems ? A review. Renew Sustain Energy Rev 2015;48:585–607.
[10] Bu J, Xu L. An effective deadbeat fuzzy algorithm for current regulation in
stationary reference frame. In: Power electronics in transportation. p. 151–7.
 C.M. Rocha-Osorio, J.S. Solís-Chaves, L.L. Rodrigues et al. / ISA Transactions 88 (2019) 258–267
[11] A high performance full fuzzy controller for induction machine drives. 2,
1998.
[12] A new deadbeat fuzzy algorithm for current regulated PWM without rotating
reference frame transformation. 1, 1998, RB1.
[13] Xing-jia Y, Zhong-liang L, Guo-sheng C. Decoupling control of doubly-fed
induction generator based on fuzzy-PI controller. In: International conference
on mechanical and electrical technology (ICMET 2010). 2010, p. 226–30.
[14] Louarem S, Belkhiat S, Belkhiat D. A control method using PI/fuzzy controllers
based DFIG in wind energy conversion system. In: 2013 IEEE grenoble confer-
ence. p. 1–6.
[15] Yao X, Jing Y, Xing Z. Direct torque control of a doubly-fed wind generator
based on grey-fuzzy logic. In: Proceedings of the 2007 IEEE international
conference on mechatronics and automation.
[16] Tamalouzt S, Rekioua T, Abdessemed R. Direct torque and reactive power
control of grid connected doubly fed induction generator for the wind en-
ergy conversion. In: 2014 international conference on electrical sciences and
technologies in maghreb (CISTEM). p. 1–7.
[17] Bo Q, Xiao-yuan H, Wen-xi Y, Zheng-yu L, Guerrero JM. An optimized dead-
beat control scheme using fuzzy control in three-phase voltage source pwm
rectifier. In: 2009 Twenty-fourth annual IEEE applied power electronics con-
ference and exposition. p. 1215–9.
[18] Hang L, Liu S, Yan G, Qu B, LU Zy. An improved deadbeat scheme with
fuzzy controller for the grid-side three-phase PWM boost rectifier. IEEE Trans
Power Electron 2011;26:1184–91.
[19] Elkhadiri S, Elmenzhi PL, Lyhyaoui PA. Fuzzy logic control of DFIG-based
wind turbine. In: 2018 International conference on intelligent systems and
computer vision. p. 1–5.
[20] Kalaivani S, Karthick T, Raja SC, Venkatesh P. Mitigation of voltage distur-
bances using fuzzy logic controller in a grid connected DFIG for different types
of fault. In: 2017 Innovations in power and advanced computing technologies.
p. 1–7.
267
[21] Mahalakshmi R, et al. Fuzzy logic based rotor side converter for constant
power control of grid connected DFIG. In: 2016 IEEE international conference
on power electronics, drives and energy systems. p. 1–6.
[22] Tutsoy O, Barkana DE, Colak S. Learning to balance an nao robot using rein-
forcement learning with symbolic inverse kinematic. Trans Inst Meas Control
2017;39:1735–48.
[23] Sols-Chaves J, Barreto MS, Salles MB, Lira VM, Jacomini RV, Filho AJS. A
direct power control for DFIG under a three phase symmetrical voltage sag
condition. Control Eng Pract 2017;65:48–58.
[24] Filho AJS, Oliveira AL, Rodrigues LL, Costa ECM, Jacomini RV. A robust finite
control set applied to the DFIG power control. IEEE J Emerg Sel Top Power
Electron 2018. 1–1.
[25] Sguarezi Filho AJ, Ruppert E. A deadbeat active and reactive power control for
doubly fed induction generator. Electr Power Compon Syst 2010;38:592–602.
[26] Murari ALLF, Altuna JAT, Jacomini RV, Rocha-Osorio CM, Solis-Chaves JS,
Filho AJS. A proposal of project of PI controller gains used on the control of
doubly-fed induction generators. IEEE Lat Am Trans 2017;15:173–80.
[27] Franklin GF, Workman ML, Powell D. Digital control of dynamic systems, 3rd
ed.. Boston, MA, USA: Addison-Wesley Longman Publishing Co., Inc.; 1997.
[28] Filho AS, Filho ER. Experimental results of deadbeat power control imple-
mentation for DFIG. In: 10th international conference on environment and
electrical engineering. p. 1–4.
[29] Leonid R. Fuzzy controllers. 1, 1997.
[30] Filho AS, Filho O, Filho ER. A digital active and reactive power control for
doubly-fed induction generator. In: 2008 IEEE power electronics specialists
conference. p. 2718–22.
[31] Almeida LAL, Sguarezi Filho AJ, Capovilla CE, Casella IRS, Costa FF. An im-
pulsive noise filter applied in wireless control of wind turbines. Renewable
Energy 2016;86:347–53.