Computers and Electrical Engineering 95 (2021) 107410

Contents lists available at ScienceDirect

Computers and Electrical Engineering

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

Fractional order PID controller design for wind turbine systems

using analytical and computational tuning approaches☆
Mohamed Lamine Frikh a, Fatma Soltani a, Nadia Bensiali b, Nadir Boutasseta c, *,
Nadir Fergani c
a
Department of Electrical Engineering, Badji Mokhtar Annaba University, P.O. Box 12, Annaba 23000, Algeria
b
Badji Mokhtar Annaba University, P.O. Box 12, Annaba 23000, Algeria
c
Research Center in Industrial Technologies, CRTI, P.O. Box 64, Cheraga, Algiers, Algeria

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

Keywords: In this paper, a recently proposed analytical fractional order PID controller design approach is
Closed-loop desired model applied to control the speed of a two-mass wind turbine system. The wind turbine complex
Analytical fractional order PID (FOPID) nonlinear system is commonly modeled as a two-mass system because of its inherent advantages.
controller design
On the other hand, wind turbulence and high overshoot are factors that cause stress in wind
Bode’s ideal transfer function
Renewable energy
turbines which may be alleviated using an adequate speed control strategy. The applied technique
Two-mass wind turbine system allows the design of the fractional order PID parameters according to desired specifications in
Computational intelligence (CI) terms of phase margin and unity gain crossover frequency. The main advantage of using such
tuning method is that the closed loop system behaves as Bode’s ideal transfer function. Simulation
results show the higher performance of the fractional order PID controller design technique when
compared with computational tuning approaches in terms of multiple performance evaluation
indices.

1. Introduction

The increasing population growth associated with higher rates of electricity consumption have resulted in a higher risk of natural
resources depletion, but on the other hand an increasing interest for renewable energy [1]. Wind energy generation systems are rapidly
growing sources of electricity that operate using different types of electrical machines that include induction machines with
wound-rotor or cage-type rotor [2], and synchronous or double stator permanent magnet machines [3]. The dynamics of the wind
turbine energy generation systems are commonly modeled as a two-mass system because this latter has the advantage of the general
aspect of its controllers that are suitable for use in different sizes of wind turbines. In addition, the flexibility of the wind turbine is
integrated in the two mass model as the modes are present. Stress in wind turbines is caused by some of the factors such as wind
fluctuations and high overshoot. Because the axis of the wind turbine and generator are coupled with two shafts through a gearbox, the

Abbreviations: PID, proportional-integral-derivative; CI, computational intelligence; PSO, particle swarm optimization; AVR, automatic voltage
regulator; ABC, artificial bee colony; FPA, flower pollination algorithm; FOPID, fractional-order proportional-integral-derivative; FOC, fractional-
order controllers; FOPI, fractional-order proportional-integral.
☆
This paper was submitted for regular issues, but should be included in VSI-srpg. Reviews processed and recommended for publication by
Guest Editor Prof. Ahmad Harb.
* Corresponding author.
E-mail address: [email protected] (N. Boutasseta).

https://doi.org/10.1016/j.compeleceng.2021.107410
Received 25 June 2020; Received in revised form 7 June 2021; Accepted 27 August 2021
Available online 8 September 2021
0045-7906/© 2021 Elsevier Ltd. All rights reserved.
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 1. Two-mass mechanical drive train.

speed fluctuations will transfer to the generator. This shows the importance of the speed control of two-mass wind turbines in reducing
rotor speed fluctuations and high overshoot [4].
Numerous control techniques have been studied during the past decades, conventional and artificial intelligence based methods
have made great advances in modern industry. The Proportional-Integral-Derivative (PID) controller is characterized by its robust
performance and simple structure for a wide range operating conditions that makes it one of the conventional and well-established
control techniques which has been widely implemented in industrial installations. Unfortunately, it has been a hard task to tune
properly the gains of PID controllers because industrial installations are often characterized by a complex behavior. For such reasons,
heuristic methods have been proposed for the purpose of tuning the parameters of PID controllers in the presence of complex models.
The Particle Swarm Optimization metaheuristic algorithm has been formulated in Eberhart and Kennedy [5] to optimize different
types of objective functions. A direct PID parameters tuning method has been proposed in Gaing [6] which uses the PSO algorithm to
find the optimal values of the controller parameters for an Automatic Voltage Regulator (AVR) system. The Artificial Bee Colony (ABC)
based optimization approach proposed in Karaboga and Basturk [7] has been applied in Gozde and Taplamacioglu [8] to optimize the
PID controller parameters for an AVR system and has been compared with PSO based tuning, the obtained results show the superiority
of the transient performance obtained using the parameters resulting from the ABC optimization technique. A PI-PD cascade controller
designed to control an interconnected four areas thermal system has been tuned using a Flower Pollination algorithm (FPA) technique
in Dash et al. [9].
In the last two decades, fractional-order proportional-integral-derivative (FOPID) controllers have received a considerable
attention. Research works on fractional-order controllers (FOC) of Oustaloup [10] and Podlubny [11] for the control of dynamic
systems have demonstrated the superior performance and higher flexibility of the FOC over the classical controllers [12]. FOPID tuning
methods can be divided into three different categories: Rule-based methods, Analytical methods and Numerical methods [13]. In
Fergani and Charef [14], an analytically tuning method is presented. The step response of the plant and the parameters of the desired
closed loop system are used to derive the five parameters of the FOPID controller through six steps of calculations. This technique is
considered in this work as it gives a simple transition between initial specifications and final controller design.
In this paper, the FOPID is used to control the speed of a two-mass wind turbine. The controller is tuned using the recently
developed method in Fergani and Charef [14], which allows the design of its parameters according to the desired specifications in
terms of phase margin and unity gain crossover frequency. The main advantage of using such tuning method is that the closed loop
system behaves as Bode’s ideal transfer function, the iso-damping propriety is verified and a good precision both in phase margin and
unity gain crossover frequency is obtained for a wide variety of linear systems. A comparison with computational tuning approaches is
carried out in terms of multiple performance evaluation indices to verify the results obtained using the FOC design specifications based
technique and optimization based approaches.
The paper is organized as follows: the next section gives a description of the studied two-mass wind turbine system. In Section 3, the
speed control problem is presented. The computational based PID controller design approach is given in Section 4. The fractional order
controller design approach along with the used analytical FOPID technique are detailed in Section 5. In Section 6, simulation results of
the studied approaches implementation are presented, with comparative study and discussions. Conclusions are given at the end.

2. Wind turbine system description

2.1. Dynamic modeling of wind turbine

Based on aerodynamics, the mechanical power captured by a wind turbine is given by Boukhezzar and Siguerdidjane [15,16]:
1
Pt = ρπ R2 Cp (λ, β)υ3 (1)
2

where R,ρ and υ are the wind turbine radius, air density and wind speed, respectively.

2
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 2. Power coefficient Cp for different values of β.

The efficiency of the wind turbine blades power coefficient Cp can be analytically estimated as [16]:
( )
116
(2)
− 21
Cp (λ, β) = 0.5176 − 0.4β − 5 e λi + 0.0068λ
λi

1 1 0.035
= − (3)
λi λ + 0.08β β3 + 1

The expressions in Eqs. (2) and (3) show that Cpdepends on the blade pitch angle β, and the tip speed ratio λ, which is defined as
ωt R
λ= (4)
υ

where ωtis the angular speed of the wind turbine’s shaft.

The torque produced by the wind turbine Tt is given by Liu et al. [17]:

Pt ρπR3 υ2 Cp (λ, β)
Tt = = (5)
ωt 2λ

2.2. Mechanical drive train model

The dynamics of the wind turbine are commonly modeled as a two-mass model as in Novak et al. [18]. The two-mass model of the
wind turbine is characterized by the advantage of the general aspect of its controllers that are suitable for usage in different sizes of
wind turbines. Furthermore, the flexibility of the wind turbine is integrated in the two mass model as the modes are present [19,4].
The two-mass wind turbine mechanical model is shown in Fig. 1, where Jt represents the turbine’s side inertia driven at speedωt
using a torque Tt as follows [4]:
dωt
Jt = Tt − Tls − Kt ωt (6)
dt

where Kt is the wind turbine’s external damping and Tls is the low speed shaft torque given as follows
Tls = Kls (θt − θls ) + Bls (ωt − ωls ) (7)

Kls and Bls represent low speed shaft stiffness and damping coefficients, respectively, ωls is the low speed shaft, θt and θls are the wind
turbine’s and low speed shaft’s angular deviations, respectively.
The high speed shaft’s torque Ths drives the generator inertia Jg and is braked by the electromagnetic torque Tem to give the
following:
d ωg
Jg = Ths − Tem − Kg ωg (8)
dt

3
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 3. PID control of a tow-mass wind turbine system.

Fig. 4. Block diagram of the two-mass wind turbine system.

with

Tem = K1 ω2g − K2 ωg (9)

where K1, K2and Kg are electromagnetic torque coefficients and generator external damping, respectively.
The transmission ratio for an ideal gearbox ng is given by:
Tls ωg θg
ng = = = (10)
Ths ωls θls

3. Wind turbine two-mass system speed control

The electric system in wind turbines has much faster time response compared to other parts of the wind turbine. For this reason,
control designs of the generator and turbine side (aero-turbine) can be dissociated, and thus, two control loops are adopted [15]. The
inner control loop controls the electric generator through the power converters while the outer control loop controls the turbine side to
force the rotor speed to its optimal value. In this work, the electrical loop is not taken into account, making the supposition that it is
well controlled, only the outer control loop will be considered.
Below rated wind speed, the mechanical power should be maximized. The coefficient curve Cp(λ,β) is given in Fig. 2. It shows that
for an optimal pitch angle βopt = 0o and a tip speed ratio λopt = 8.1, Cphas maximum value Cpmax(λopt,βopt) = Copt = 0.48. At this value the
turbine is most efficient and maximum power production is achieved and corresponds to the optimum operating point ωopt = opt R .
λ υ

This optimum operating point is taken as reference, compared with the actual speed and the difference is fed into a PID controller to
obtain the torque Tt as shown in Fig. 3.
The Proportional, Integral and Derivative PID control is one of the most popular methods in control systems which is used in two
mass wind turbines to control the rotor speed of the wind turbine. The transfer function of the most commonly used and practical PID
controller has the following form:
U(s) ki s
C(s) = = kp + + kd (11)
E(s) s Tf s + 1

where kp, ki and kd denote the proportional, integral and derivative gains, respectively. s is the Laplace variable. Tf is the derivative
term filter parameter.
In order to find the controller parameters kp, ki,kd,the transfer function of the controlled system is needed. From the block diagram
shown in Fig. 4 obtained from Eqs. (6)–(10), the transfer function of the wind turbine two-mass system is derived in the following form:

ωt (s) n2 Jg s2 + Bls s + Kls

Gt (s) = = (g ) ( ) (12)
Tt (s) n2g Jt Jg s3 + Bls Jt + n2g Jg s2 + Kls Jt + n2g Jg s

4. Controller design using computational intelligence approaches

As computational intelligence (CI) based techniques are gaining more attention due to the development of advanced learning
techniques that allow automatic decision making and efficient complex problem solving, the application of these approaches to the
tuning of the controller allows better exploration of its parameters search space.

4
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 5. PID controller parameters tuning using computational intelligence approaches.

Computational intelligence approaches include evolutionary computational approaches and swarm intelligence computation such
as Particle Swarm Optimization (PSO), Flower Pollination (FP), Artificial Bee Colony (ABC) [20].
When using CI approaches for tuning the PID controller, a fitness function is calculated using the error between the measured
output variable and the reference. Then, optimization algorithms modify the parameters of the controller depending on the used
optimization algorithm (see Fig. 5).

4.1. Fitness evaluation

The main step in evaluating computational intelligence algorithms is the choice of the objective function that is used to evaluate the
fitness of each agent.
In order to minimize the error signal between the reference and the measured output, the following objective functions are used in
the literature: Mean of the Squared Error (MSE), Integral of Time multiplied by Absolute Error (ITAE), Integral of Absolute Error (IAE),
and Integral of the Squared Error (ISE), and Integral of Time multiplied by the Squared Error (ITSE).
∫τ ∫τ ∫τ
1
MSE = (e(t))2 dt, ITAE = t|e(t)|dt, IAE = |e(t)|dt
t
0 0 0

∫τ ∫τ
2
ISE = (e(t)) dt, ITSE = t(e(t))2 dt
0 0

where e(t) is the error in the time domain.

The objective of the fitness function is to minimize the value of the chosen objective function, the smaller the fitness value of the
corresponding agent the fitter it will be, and vice versa.

5. Fractional order controller design

5.1. Fractional order PID controller

Fractional calculus has three fundamental definitions utilized vastly in the field of control systems, which are Riemann–Liouville
definition, Caputo definition and Grunwald–Letnikov definition.
The fractional PID controller generalizes the classical PID controller which results in higher flexibility due to its two additional
parameters which gives additional specifications and improves the robustness of the controlled system [13].
The general transfer function form of the fractional PID controller is defined by Erenturk [21]:
U(s) ki
H(s) = = kp + α + kd sμ , (α, μ ≥ 0) (17)
E(s) s

5
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 6. Fractional order feedback control of the two-mass wind turbine system.

where kp, ki and kd are the proportional, integral and derivative gains, α is the integration order and μ is the differentiator order.
Generally, the range of α and μ are chosen from 0 to 2 [13].

5.2. Tuning of PIαDμ parameters

The tuning of the controller parameters uses the fractional order system of Eq. (18) as desired closed loop system such that the
system given by Gd(s) satisfies the desired specifications.
1
Gd (s) = ( )m (18)
s
1+ ωu

where the two parameters are given by:

- ωu = ωc the desired unity gains crossover frequency.

[ ( )]
- m = 2 1 − φπm (φm is the phase margin)

By using the design technique proposed in Fergani and Charef [14], the following six successive calculation steps allow the deter­
mination of the five controller parameters. The closed-loop function of Fig. 6 using the obtained controller will make its response
corresponds to the desired reference model given in Eq. (18).
Step 1. Calculation of the variables θi, for 0 ≤ i ≤ 4 from the following equations:
1
θ0 = Gd (ωu ) =
2
(1) m
θ1 = Gd (ωu ) = −
4ωu
m
θ2 = G(2)
d (ωu ) =
4ω2u (19)
2
)
m (m − 4
θ3 = G(3)
d (ωu ) =
8 ω3u
)
(4) 4m (m2 − 2
θ4 = Gd ( ωu ) = −
3 ω4u

Step 2. Calculation of the variables Si, for 0 ≤ i ≤ 4 from the equations

S0 = GIn (ωu )
S1 = G(1)
In (ωu )

S2 = G(2)
In (ωu ) (20)
(3)
S3 = GIn ( ωu )
S4 = G(4)
In ( ωu )

where GIn (s) = Gts(s)

Step 3. Calculation of the terms Go (ωu ), for 0 ≤ i ≤ 4 from equations:

(i)

6
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

θ0
Go (ωu ) =
1 − θ0
θ1
G(1)
o (ωu ) =
(1 − θ0 )2

θ2 2θ21
G(2)
o (ωu ) = 2
+ (21)
(1 − θ0 ) (1 − θ0 )3

θ3 6θ1 θ2 6θ31
G(3)
o (ωu ) = 2
+ 3
+
(1 − θ0 ) (1 − θ0 ) (1 − θ0 )4

θ4 6θ22 + 8θ1 θ3 36θ21 θ2 24θ41

G(4)
o (ωu ) = 2
+ 3
+ 4
+
(1 − θ0 ) (1 − θ0 ) (1 − θ0 ) (1 − θ0 )5

Step 4. Calculation of the variables Xi, for 0 ≤ i ≤ 4 as follows:

Go (ωu )
X0 =
ωu S0
1 ( )
X1 = G(1)
o (ωu ) − X0 S0 − ωu X0 S1
ωu S0
1 [ (2) ]
X2 = Go (ωu ) − 2X1 S0 − 2 X0 S1 − 2ωu X1 S1 − ωu X0 S2
ωu S0
⎡ ⎤ (22)
(3)
1 ⎢ Go (ωu ) − 3X2 S0 − 6X1 S1 − 3X0 S2 − ωu X0 S3 ⎥
X3 = ⎣ ⎦
ωu S0
− 3ωu X2 S1 − 3ωu X1 S2
⎡ ⎤
(4)
1 ⎢ Go (ωu ) − 4X3 S0 − 12X2 S1 − 12X1 S2 − 4X0 S3 ⎥
X4 = ⎣ ⎦
ωu S0
− ωu X0 S4 − 4ωu X3 S1 − 4ωu X1 S3 − 6ωu X2 S2

Step 5. Calculation of the variables Zi, for 1 ≤ i ≤ 3 using

Z1 = X1 + ωu X2
Z2 = X1 + 3ωu X2 + ω2u X3 (23)
Z3 = X1 + 7ωu X2 + 6ω2u X3 + ω3u X4

Step 6. Calculation of the fractional PIαDμ controller parameters α, μ, Kd, Ki and Kp:

• α as a suitable solution of the second order equation

( 2 ) ( )
Z1 − X1 Z2 α2 + (Z1 Z2 − X1 Z3 ) α + Z22 − Z1 Z3 = 0 (24)

• μ is obtained from
α Z1 + Z2
μ = (25)
α X1 + Z1

• Kd and Ki obtained as follows

μ)
Q2 ω(1−
u
Kd =
μ
(26)
α)
Q1 ω(1+
u
Ki = −
α

• Kp is obtained as follows
Kp = X0 − Ki ω−u α − Kd ωμu (27)

7
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Table 1
Parameters of the two-mass wind turbine system.
Air density ρ = 1.29kg/m3
Turbine radius R = 21.65m
Gearbox ratio ng = 43.165
Turbine inertia Jt = 3.25*105kg.m2
Generator inertia Jg = 34.4kg.m2
Turbine external damping coefficient Kt = 0
Generator external damping coefficient Kg = 0
Shaft stiffness coefficient Kls = 2.691*105Nm/rad
Shaft damping coefficient Bls = 9500Nm/rad/s
Electromagnetic torque coefficients K1 = 0.1082
K2 = 0.2147

Fig. 7. Wind speed profiles: (a) base and step change, (b) variation with mean value of 10 m/s.

Fig. 8. Response of the wind turbine’s speed to base wind profile for computational tuning approaches.

8
 M.L. Frikh et al.
Table 2
Comparative study results for the different controllers obtained with various tuning techniques.
PSO-PID FPA-PID ABC-PID
IAE ISE ITAE ITSE IAE ISE ITAE ITSE IAE ISE ITAE ITSE

kp 10e7 10e7 10e7 9.561e7 10e7 10e7 10e7 10e7 9.389e7 9.9784e7 9.399e7 9.7062e7
ki 10e3 10e7 10e3 9.975e7 10e3 10e7 10e3 10e7 9.2616e6 5.54e7 9.053e5 2.552e7
kd 0 10e2 0 9.907e2 0 10e2 0 10e2 136.7528 273.421 233.379 83.8214
α 1 1 1 1 1 1 1 1 1 1 1 1
μ 1 1 1 1 1 1 1 1 1 1 1 1
Settling time ts (s) 0.0131 0.0127 0.0131 0.0132 0.0131 0.0127 0.0131 0.0127 0.014 0.0129 0.014 0.0134
Overshoot Mp (%) 7.79e-7 0.2056 7.79e-7 0.2228 7.9e-7 0.2056 7.9e-7 0.2056 9.8e-3 0.1205 8.7e-5 0.0488
Rise time tr (s) 7.2e-3 7.1e-3 7.2e-3 7.5e-3 7.2e-3 7.1e-3 7.2e-3 7.1e-3 7.6e-3 7.2e-3 7.6e-3 7.4e-3
Objective function 4.4223 2.176 4.4e-3 6.31e-6 4.4223 2.176 4.4e-3 5.777e-6 5.7352 2.1825 4.8e-3 5.196e-6
9

Calculation time 156.62 157.12 157.47 164.46 160.87 162.23 161.61 170.28 3.165 3.45 3.39 3.40
FOPID PI
IAE ISE ITAE ITSE IAE ISE ITAE ITSE
kp –1.344e6 4.768e8
ki 4.781e8 3.364e8
kd 3.364e8 0.0
α 0.0553 0
μ –1.023 –1
Settling time ts (s) 2.2e-3 2.8e-3
Overshoot Mp(%) 0.028 0.0318

Computers and Electrical Engineering 95 (2021) 107410

Rise time tr (s) 1.2e-3 1.6e-3
Objective function 1.3e-3 2.6e-4 1.1e-3 2.33e-7 1.3e-3 3.4e-4 9.6e-4 2.551e-7
Calculation time - - - - - - - -
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 9. Response of the wind turbine’s speed to base wind profile for fractional order PID, approximated PI and ABC algorithm.

Table 3
Parameters of the studied FOPID and PID controllers tuning techniques.
Fractional Order PID controller tuning parameters

Crossover frequency ωu = 1000rad/s

Phase margin φm = 85◦

Computational Intelligence approaches parameters

Particle Swarm Optimization (PSO)

Swarm size 50
Iterations 50
C1 1.5
C2 2.0

Artificial Bee Colony (ABC)

Searching agents 50
Iterations 50

Flower Pollination Algorithm (FPA)

Population size 50
Iterations 50
Probability switch p 0.8

6. Results and discussions

The simulation of the two-mass wind turbine system is carried out in Matlab/Simulink environment. The main parameters used for
system simulation are listed in Table 1. To verify the performance and the effectiveness of the proposed controller, the two-mass system
dynamic behavior is simulated under various wind speed profiles namely, base and step change wind (Fig. 7(a)), and gust wind (Fig. 7
(b)). The computational intelligence based controller design approaches has been tested and compared to select the controller pa­
rameters that give the best performance. Then, the selected best controller is compared with the fractional order based controller using
the aforementioned wind profiles.
Fig. 7(a) shows the base and step change in wind speed. The base wind permanently exists in the wind turbine’s operation. In
general, with the time passing, the basic wind speed does not change, so it can be considered invariable and takes the value of 10 m/s.
The step change profile of the wind is characterized by a step-down from 10 to 8 m/s at 0.5 s and a step-up from 8 to 10 m/s at 1 s.
Fig. 7(b) shows gust wind speed, which may be expressed by using a sine function. The described model by gust in Attoui and
Omeiri [22] can be used to verify the robustness of the controller against sudden changes in wind speed.
The computational intelligence based controller design approaches are tested and compared in different wind profiles. In the case
of base wind, the responses of the wind turbine two mass system in different closed loops are given in Fig. 8. Table 2 shows the obtained
controllers parameters after optimization, the obtained performance of the closed loop response in terms of settling time, over-shoot,
rise time, the optimization algorithms performance in terms of different objective functions and calculation time.

10
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 10. Response of the wind turbine’s speed to step changes in wind speed for computational tuning approaches.

Fig. 11. Response of the wind turbine’s speed to step changes in wind speed for fractional order PID, approximated PI and ABC algorithm.

The closed loop performances show close values between the tested computational intelligence based controller design approaches.
Whereas, the optimization algorithm performances show the superiority of the Artificial Bee Colony (ABC) PID controller design
approach (ABC-PID) both in terms of the ITSE objective function that associates the elapsed time with the error measure and also in
terms of calculation time of the optimization procedure where it is characterized by the fastest controller parameters optimization. In
this work, the calculation time is defined by the elapsed time when the optimization algorithm is executed offline to search for the
optimal parameters.
Next, the ABC-PID is compared with the Fractional order PID (FOPID) controller design approach and the approximated integer
order Proportional Integral (PI) controller. The approximated PI controller is obtained by neglecting the small FOPID α parameter
which is the integration order given in Table 2 and approximating the μ differentiator order to -1 which represents an integer order
integrator.
The comparison results shown in Fig. 9 and Table 2 show the higher performance of the FOPID and the approximated PI controllers
when compared with the ABC-PID controllers in terms of the used objective functions and in the characteristics of the transient
performance of the closed loop response (rise time, settling time, overshoot).
The CI approaches with the parameters given in Table 3 are evaluated in various wind profiles namely: the consequent step changes
in Fig. 10 and gust wind profile in Fig. 12. The responses show the reduced tracking error of the ABC-PID when compared with the

11
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Fig. 12. Response of the wind turbine’s speed to gust wind for computational tuning approaches.

Fig. 13. Response of the wind turbine’s speed to gust wind for fractional order PID, approximated PI and ABC algorithm.

remaining CI approaches.
The ABC-PID controller is evaluated along the FOPID and the approximated PI controllers associated with the two mass system in
the case of series step changes in wind speed shown in Fig. 11 and gust wind in Fig. 13. The obtained results confirm the higher
performance of the FOPID and approximated PI controllers when compared with the ABC-PID based approach in tracking efficiently
the optimal wind turbine speed which will give a better long term cumulative power conversion efficiency. The FOPID has the overall
best performance for the different tested scenarios.
When system parameters uncertainties are considered and changed separately by a ± 50% of their nominal values, the resulting
change in closed loop transient performance is minimal (see Table 4). The FOPID performance remains practically unchanged in the
presence of uncertainties which represents the best result in this robustness test as it can be noted in Table 4.

7. Conclusion

An analytical fractional order PID controller design approach has been applied to control the speed of a two-mass wind turbine
system. The design of an appropriate speed control strategy of the wind turbine alleviates the effects of wind turbulence and high
overshoot and allows better transient performance. Simulation results using various wind profiles show the higher performance of the

12
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

Table 4
Effect of system parameters uncertainties on controller performance.
Controller Settling OvershootMp Rise Settling OvershootMp Rise Settling OvershootMp Rise timetr
timets (s) (%) timetr (s) timets (s) (%) timetr (s) timets (s) (%) (s)
+50% Bls +50% Kls +50% Jt

FOPID 0.0022 0.0285 0.0012 0.0022 0.028 0.0012 0.0033 0.0390 0.0017
PI 0.0027 0.0335 0.0015 0.0027 0.0318 0.0015 0.0040 0.0480 0.0022

ABC-PID 0.0137 0.0547 0.0074 0.0134 0.0488 0.0074 0.0201 0.0715 0.0111

FPA-PID 0.0128 0.2281 0.0072 0.0127 0.2053 0.0071 0.0186 0.3198 0.0107
PSO-PID 0.0134 0.2468 0.0075 0.0132 0.2225 0.0075 0.0194 0.3482 0.0111
-50% Bls -50% Kls -50% Jt
FOPID 0.0022 0.0233 0.0012 0.0022 0.028 0.0012 0.0012 0.015 0.0006

PI 0.0027 0.0269 0.0015 0.0027 0.0319 0.0015 0.0013 0.0170 0.00075

ABC-PID 0.0132 0.0368 0.0074 0.0134 0.0488 0.0074 0.0068 0.0271 0.0056

FPA-PID 0.0125 0.1940 0.0071 0.0127 0.2058 0.0071 0.0065 0.1054 0.0036
PSO-PID 0.0130 0.2150 0.0074 0.0132 0.2230 0.0075 0.0068 0.1135 0.0038

ABC tuning algorithm of the PID controller in terms of convergence time and performance index value when compared with other
computational intelligence tuning techniques. The fractional order design approach gives the best results when compared with the
other studied tuning techniques in terms of robustness and multiple performance evaluation indices.

Uncited link

Eqs. (1), (11), (12), (19)–(27)

CRediT authorship contribution statement

Mohamed Lamine Frikh: Conceptualization, Writing – original draft. Fatma Soltani: Conceptualization, Writing – original draft.
Nadia Bensiali: Conceptualization, Formal analysis, Writing – original draft. Nadir Boutasseta: Conceptualization. Nadir Fergani:
Conceptualization, Writing – original draft.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.
The authors declare the following financial interests/personal relationships which may be considered as potential competing
interests:

References

[1] Mérida J, Aguilar LT, Dávila J. Analysis and synthesis of sliding mode control for large scale variable speed wind turbine for power optimization. Renew Energy
2014;71:715–28. Nov.
[2] Hong CM, Chen CH, Tu CS. Maximum power point tracking-based control algorithm for PMSG wind generation system without mechanical sensors. Energy
Convers Manag 2013;69:58–67. May.
[3] Flah A, Khan IA, Agarwal A, Sbita L, Simoes MG. Field-oriented control strategy for double-stator single-rotor and double-rotor single-stator permanent magnet
machine: design and operation. Comput Electr Eng Mar. 2021;90:106953.
[4] Boukhezzar B, Siguerdidjane H. Comparison between linear and nonlinear control strategies for variable speed wind turbines. Control Eng Pract Dec. 2010;18
(12):1357–68.
[5] Eberhart R, Kennedy J. Particle swarm optimization. In: Proceedings of the IEEE international conference on neural networks. 4; 1995. p. 1942–8.
[6] Gaing ZL. A particle swarm optimization approach for optimum design of PID controller in AVR system. IEEE Trans Energy Convers 2004;19(2):384–91.
[7] Karaboga D, Basturk B. A powerful and efficient algorithm for numerical function optimization: artificial bee colony (ABC) algorithm. J Glob Optim 2007;39(3):
459–71. Oct.
[8] Gozde H, Taplamacioglu MC. Comparative performance analysis of artificial bee colony algorithm for automatic voltage regulator (AVR) system. J Frankl Inst
2011;348(8):1927–46. Oct.
[9] Dash P, Saikia LC, Sinha N. Flower pollination algorithm optimized PI-PD cascade controller in automatic generation control of a multi-area power system. Int J
Electr Power Energy Syst 2016;82:19–28. Nov.
[10] Oustaloup A. La commande CRONE : commande robuste d’ordre non entier. Paris: Hermès; 1991.
[11] Podlubny I. Fractional-order systems and PIλDμ-controllers. IEEE Trans Autom Control 1999;44(1):208–14. Jan.
[12] Bettayeb M, Mansouri R. Fractional IMC-PID-filter controllers design for non integer order systems. J Process Control 2014;24(4):261–71. Apr.
[13] Shah P, Agashe S. Review of fractional PID controller. Mechatronics 2016;38:29–41. Sep.
[14] Fergani N, Charef A. Process step response based fractional PIλDμ controller parameters tuning for desired closed loop response. Int J Syst Sci 2016;47(3):
521–32. Feb.

13
M.L. Frikh et al. Computers and Electrical Engineering 95 (2021) 107410

[15] Boukhezzar B, Siguerdidjane H. Nonlinear control of a variable-speed wind turbine using a two-mass model. IEEE Trans Energy Convers 2011;26(1):149–62.
Mar.
[16] Ghasemi S, Tabesh A, Askari-Marnani J. Application of fractional calculus theory to robust controller design for wind turbine generators. IEEE Trans Energy
Convers 2014;29(3):780–7. Sep.
[17] Liu J, Gao Y, Geng S, Wu L. Nonlinear control of variable speed wind turbines via fuzzy techniques. IEEE Access 2017;5:27–34.
[18] Novak P, Ekelund T, Jovik I, Schmidtbauer B. Modeling and control of variable-speed wind-turbine drive-system dynamics. IEEE Control Syst 1995;15(4):28–38.
Aug.
[19] Boukhezzar B, Siguerdidjane H, Hand MM. Nonlinear control of variable-speed wind turbines for generator torque limiting and power optimization. J Sol Energy
Eng 2006;128(4):516–30. Aug.
[20] Engelbrecht AP. Computational intelligence: an introduction. 2nd Ed. Chichester, UK: John Wiley & Sons, Ltd; 2007.
[21] Erenturk K. Fractional-order PIλDμ and active disturbance rejection control of nonlinear two-mass drive system. IEEE Trans Ind Electron 2013;60(9):3806–13.
Sep.
[22] Attoui I, Omeiri A. Modeling, control and fault diagnosis of an isolated wind energy conversion system with a self-excited induction generator subject to
electrical faults. Energy Convers Manag 2014;82:11–26. Jun.

14