976 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO.

1, JANUARY/FEBRUARY 2019

Modeling and Control of an Irrigation Station Process

Using Heterogeneous Cuckoo Search Algorithm
and Fuzzy Logic Controller
Jaouher Chrouta , Wael Chakchouk , Abderrahmen Zaafouri, and Mohamed Jemli

Abstract—Water is becoming a precious and very scarce re- the significant increase in agricultural uses. Irrigation has been
source in many countries due to an increasing demand in agricul- the main water consuming activity in the world because it rep-
tural and industrial fields as well as population growth. Therefore,
resents about 80% of the fresh water consumption. Water loss in
we have to optimize the water resources from hydraulic systems,
in order to decrease water losses. In this context, many research irrigation systems is important [1], [2]. To limit water wastage,
studies focused on modeling, identifying, and controlling of hy- this task should be controlled. Before overcoming control prob-
draulic systems, especially for agricultural use. In this paper, we lems, great interest is devoted to modeling and identification
will investigate the problem of identification and control design for of the system. Indeed, we need to optimize the water resources
an irrigation process through an academic irrigation station (IS) by a good water and energy management of the hydraulic sys-
system. Two main objectives will be reached. On the one hand,
we will provide a new optimal Takagi–Sugeno (T–S) fuzzy model tems. In this framework, an irrigation system has been made
of our IS process. The optimal model is obtained through using a by the French company, LEROY SOMMER where practical
fuzzy parameters searching strategy, named intelligent T–S mod- constraints existing in the real irrigation station (IS) have been
eling. This last, is associated with a heterogeneous cuckoo search included to approximate, as well as possible, the behavior of the
(HeCoS) strategy based on the quantum mechanism. On the other real process. Due to highly nonlinear behavior of this industrial
hand, we will devote a lot to the synthesis of an optimal control
law by using fuzzy logic control (FLC) to ensure the global stabil- process, to the nonlinear coupling of different phenomena, the
ity of the closed-loop system. The proposed FLC has two major mathematical solutions of this system are very complex and re-
advantages. First, it does not require a complex process to find quire enormous amounts of computational time. An alternative
a common Lyapunov function for a large number of fuzzy sub- solution is to use system identification algorithms: input–output
systems. Therefore, the designed procedure of the proposed FLC nonlinear models “black-box models,” which have not neces-
is much simpler. Second, by employing the FLC, the closed-loop
system performance can be designed. Finally, and as a result, the sarily to achieve a deep mathematical knowledge of the system
HeCoS strategy is well adopted to find an optimal model for the under study, are identified from data in order to obtain a con-
real processes with high accuracy and strong generalization ability. trol model. See [3]–[5] for an exhaustive discussion. In order to
Experimental results applied to the IS demonstrate the merits of obtain a reliable model for the real irrigation process dynamic,
the proposed FLC.
the Takagi–Sugeno (T–S) fuzzy model [6] recently is widely
Index Terms—Fuzzy logic, heterogeneous cuckoo search used to mathematically represent the real and complex processes
(HeCoS), irrigation station (IS), particle swarm optimization [7]–[9]. The task of building a T–S fuzzy model can be sum-
(PSO), Takagi–Sugeno (T–S) fuzzy model.
marized in two parts: a premise part and a consequent part. The
I. INTRODUCTION aim of the first part is to determine the structure of the model by
extracting the fuzzy rules. This task can be obtained by cluster-
ROM the beginning of civilization, water has been a pre-
F cious and scarce resource in many parts of the globe due to
ing algorithms such as fuzzy C-means[10], fuzzy C-regression
model [9], etc. The aim of the second part is the optimization
Manuscript received February 23, 2018; revised May 30, 2018 and August of system’s parameters. These ones can be calculated by dif-
2, 2018; accepted September 4, 2018. Date of publication September 19, 2018; ferent existing methods such as genetic algorithm (GA) [11],
date of current version December 12, 2018. Paper 2018-IACC-0214.R2, pre- improved gravitational search algorithm (GSA) [12], and parti-
sented at the 2017 International Conference on Advanced Systems and Electric
Technologies, Hammamet, Tunisia, Jan. 14–17, and approved for publication cle swarm optimization (PSO) algorithm [13], [14]. The major
in the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the Industrial Au- weakness of these approaches is that it is hard to guarantee the
tomation and Control Committee of the IEEE Industry Applications Society. best performance of the obtained models due to the separation
(Corresponding author: Jaouher Chrouta.)
J. Chrouta is with the Universite de Tunis Ecole Nationale Superieure between the two parts of building a T–S fuzzy model. Many
d’Ingenieurs de Tunis, Tunis 1008, Tunisia (e-mail:,[email protected]). researches aim at finding a consistent way to get solutions for
W. Chakchouk, A. Zaafouri, and M. Jemli are with the Laboratory of Indus- both the model structures and parameters simultaneously. One
trial Systems Engineering and Renewable Energy Higher National Engineer-
ing School of Tunis, University of Tunis, Tunis 1008, Tunisia (e-mail:, wael. popular and effective method is to put the T–S fuzzy system de-
[email protected]; [email protected]; mohamedjemli2030@gmail. sign into an optimization framework in multidimensional space
com). composed of the model structures and parameters [15]. In par-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. ticular, Du and Zhang [15] presented a new encoding scheme to
Digital Object Identifier 10.1109/TIA.2018.2871392 learn the T–S fuzzy model by using GAs. This algorithm is very

0093-9994 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
CHROUTA et al.: MODELING AND CONTROL OF AN IRRIGATION STATION PROCESS 977

efficient in finding a global optimal solution for the problem,

but it is characterized by a long run time. In [16], a chaotic
search has been embedded in the iteration of basic GSA to find
good model parameters. This algorithm has been successfully
used in system parameters identification [17]. However, the ma-
jor drawback of this algorithm is that it is easily trapped into
local optima. Recently, a new metaheuristic search algorithm,
called cuckoo search (CS) is developed in [18]. It is one of
the latest nature-inspired metaheuristic algorithms [19]. CS is
based on the brood parasitism of some cuckoo species [18].
Although much progress has been achieved on the CS-based
algorithms, significant efforts are required to further improve
its performances. In the work of [20], an improved CS algo-
rithm is performed where heterogeneous CS is implemented to
generate a T–S fuzzy model for nonlinear systems. Motivated
by the efficient performances of such heterogenous CS algo-
rithm in the field of system identification, this paper proposes to
generate a new optimal T–S fuzzy model of the IS process us- Fig. 1. Experimental test bench of the irrigation station.
ing the iTaSuM method. This method is based on the improved
heterogenous cuckoo search algorithm (HeCoS). The HeCoS
algorithm is performed via heterogeneous searching strategy solution; the problem of finding the common Lyapunov function
based on the quantum mechanism. The main advantages of the remains open. In addition, the performance of the closed-loop
iTaSuM method are as follows: fuzzy control system is difficult to be predicted because of its
1) it can approximate an unknown system based on given stability, not performance, that is considered during the design
input–output data; [27]. To overcome this drawback, we propose a new FLC schema
2) it encodes the structures and the parameters of the fuzzy for the T–S system where the system stability can be guaran-
models in a nest vector to simultaneously obtain the opti- teed without resorting to find a common Lyapunov function.
mal structure and parameters. Also, the closed-loop system performance can be considered
Therefore, the experiment results applied to the IS process because the T–S controller generates a closed-loop linear sys-
showed that iTaSuM method generates a good fuzzy model with tem. Indeed, the main purpose of this paper is to treat the tracking
high accuracy. As mentioned above, the second aim of this paper problem of pressure in pipeline and sprinklers. This technique
is to design a fuzzy controller for our irrigation process. Indeed, is used to ensure a high accuracy of the robust control’s perfor-
once the modeling part is completed, a fuzzy controller is syn- mances in the presence of parametric uncertainties, and outputs
thesized in order to optimize the water resources by a good variations in transient and steady states. The T–S fuzzy control
energy management of the hydraulic systems. In this frame- will be experimentally validated through MATLAB-Simulink
work, many research studies dealt with the control problem of environment and the dSpace DS1104 card based on the real-
the IS system via different techniques. In particular, Mejri and time data acquisition control system.
Chaari [21] considered the problem of PI control design applied This paper is organized as follows. After Section I, we provide
to the IS system. However, the major drawback of such con- a description of IS process in Section II. A brief review about
trol low resides in its adaptation to external variations such as the T–S fuzzy model is reported in Section III. In Section IV, we
the extension of drain network. On the other hand, the problem present the HeCoS algorithm and the framework of iTaSuM for
of fuzzy control of IS process is treated in [22]. However, the identification of a T–S fuzzy model. Experimented results are
proposed controller does not ensure the desired speed and sta- presented in Section V in order to validate the performance of
bility. It should be noted that both PI and fuzzy control method data-driven identification of T–S fuzzy models with iTaSuM. In
cited above generally led to a high steady-state error and un- Section VI, the fuzzy control of T–S fuzzy models is developed.
stable tracking performance and requires [23]. In view of these Simulation results and comparative study of fuzzy logic and PI
weaknesses, advanced FLC methods should be developed for controllers are also introduced in this section to demonstrate the
the studied system. In the literature, several versions of FLC various features of the proposed fuzzy logic scheme. The exper-
design have been proposed to control a large class of dynamical imental implementation of fuzzy control to the IS is detailed in
nonlinear systems. Particularly, in [24], Tanaka et al. proposed Section VII. Finally, a conclusion is drawn in Section VIII.
a parallel distributed compensation technique to design a fuzzy
logic controller for T–S fuzzy models. However, the main draw-
back of such technique is the difficulty of finding the common II. DESCRIPTION OF IRRIGATION STATION PROCESS
Lyapunov function for the large number of fuzzy subsystems. The irrigation is a very useful technique used in agriculture,
Several researchers have developed different approaches to find which consists in an artificially brought water from source (lake,
the common Lyapunov function [25], [26]. In fact, these ap- river, . . .) to the plants. The French company LEROY SOMER
proaches involve complex calculations without guaranteeing a provides researchers with an IS by sprinkling (see Fig. 1) in

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
978 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO. 1, JANUARY/FEBRUARY 2019

considered to be a very efficient tool to describe a large class of

nonlinear systems [6].

III. TAKAGI–SUGENO FUZZY MODEL

The T–S fuzzy model of this system can be described by the
following IF-THEN fuzzy rules:
Ri : IF y(k) is Ai0 and . . . and y(k − n) is Ain THEN
y(k + 1) = yi (k + 1)
= ai0 y(k) + ai1 y(k − 1) + · · · + ain y(k − n)
+ bi0 u(k) + bi1 u(k − 1) + · · · + bim u(k − m). (1)
The IF rule function defines the premise part, while the THEN
rule function defines the consequent part of the T–S fuzzy model,
where i = 1, . . . , c (c is the number of fuzzy rules), u and y are
input and output of the plant, respectively, y(k + 1) is the output
of rule i, Ai1 , Ai2 , . . . , Ain are multidimensional antecedents
of the fuzzy sets, ai0 , ai1 , . . . , ain and bi0 , bi1 , . . . , bim are the
coefficients of the linear difference equation of the consequent
part. For simplicity, 1 is rewritten into the following form:
Ri : IF ≺ premise  THEN
y(k + 1) = ai0 y(k) + ai1 y(k − 1) + · · · + ain y(k − n)
+ bi0 u(k) + bi1 u(k − 1) + · · · + bim u(k − m).
(2)
ai ∈ n is the polynomial coefficient that forms the consequent
parameter of the ith rule, xk = [xk 1 , . . . , xk n ]T ∈ n is the
Fig. 2. Pumping station: Block diagram of the hydraulic circuit. input vector of the fuzzy model, Ai1 , Ai2 , . . . , Ain are multidi-
mensional antecedents of the fuzzy sets and yi is the output of
ith fuzzy rule. The global estimated output is calculated by a
weighting of the other outputs of local models according to the
order to reduce the cost of maintaining, improve the standard of following equation:
living of the farmer, and protect the environment. Our system is c
μi .yi (k + 1) 1
made up of following two parts. y(k + 1) = i=1c = c
1) The electrical part: The IS is equipped with two motor i=1 iμ i=1 μi
⎡ c c ⎤
i=1 μi ai1 y(k − 1) + · · ·
pumps with three-phase asynchronous machine, one with i=1 μi ai0 y(k) +
fixed speed and the other with variable speed. Depending ⎢ c c ⎥
on the water requirement, a variable speed drive is made to ⎣ + i=1 μi ain y(k − n) + i=1 μi bi0 u(k) ⎦ (3)
c c
control the flow in the station for a given set pressure. The + i=1 μi bi1 u(k − 1)+ · · · + i=1 μi bin u(k − m)
user communicates with the IS through a control panel
where μi = Ai0 (y(k)) × · · · × Ain (y(k − n)) is the degree of
including all the control buttons and indicators.
membership for rule i, × denotes the “AND” operation, which
2) The hydraulic circuit: The hydraulic circuit includes PVC
is implemented, for example, a triangular or sine norm (typi-
pipes, allowing
 water
 to circulate at a maximum flow rate cally the minimum or product operator). See [6], [29] for an
of about 8 m3 /h under a pressure of 5 bar (see Fig. 2).
exhaustive discussion.
The station is a single input, multioutputs system, where the
input is the pressure and the outputs are the pressure and the flow.
This prototype reproduces the natural irrigation phenomenon IV. HETEROGENEOUS CUCKOO SEARCH ALGORITHM
and takes into account the practical constraints existing in the 1) Cuckoo Search Algorithm: Cuckoo searches are one of
real ISs [28]. It is a complex nonlinear and interconnected sys- the latest nature-inspired metaheuristic algorithms proposed by
tem. The identification of such system based on input–output Yang et al. [18]. It is inspired by the aggressive reproduction
data “black-box” is becoming a promising tool. Therefore, sev- of cuckoo species combining with the behavior of Levy flight.
eral algorithms using the “black-box” technique have been car- The female cuckoo lays her fertilized eggs in nests of other host
ried out recently. Among them is the fuzzy modeling, which birds. In this way, the host birds unwittingly raise her brood.
is widely used to mathematically represent real and complex If a cuckoo’s egg is discovered, the host bird will throw it out
processes. In particular, the well-known T–S fuzzy systems are or abandon her nest and will start her own brood elsewhere.

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
CHROUTA et al.: MODELING AND CONTROL OF AN IRRIGATION STATION PROCESS 979

In the CS algorithm, each egg of host birds in a nest represents where

a solution, and a cuckoo egg represents a new solution. If a new ⎧
⎪
⎪i : imaginary unit
solution is better than the one in the nest, the worse one will be ⎨ : reduced Planck constant
replaced. To formulate in terms of mathematics, each egg in a (11)
⎪
⎪ m: particles mass
nest represents a solution, and a cuckoo egg represents a new ⎩ 2
∇ : Laplacian.
solution [30].
For the sake of simplicity, we now use the following three In the absence of time, the delta potential is the Schrödinger
idealized rules to describe the CS algorithm [28], [31]. equation for the wave function ψ(γ) of a particle in one-
1) Each cuckoo lays one egg at a time, and dumps it in a dimensional (1D) in a potential v(γ) described by the following
randomly chosen nest. equation:
2) The best nests with high quality of eggs (solutions) will v(γ) = −λδ(γ) (12)
carry over to the next generations.
3) The number of available host nests is fixed, and a host where
can discover an alien egg with a probability pa ∈ [0, 1]. δ the Dirac delta function
In this case, the host bird can either throw the egg away λ intensity of Dirac delta function.
or abandon the nest so as to build a completely new nest
in a new location. Given that each cuckoo has a quantum behavior with its quantum
The new solution is described by the following equation: state formulated by a wave function ψ. |ψ|2 is a probability
density function of the nest position defined by
xi (k + 1) = xi (k) + α ⊕ Levy (4)  
1 2 |γ|
where α  0 is the step size, which should be related to the |ψ(γ)|2 = exp − . (13)
L L
scale of the problem of interest. The product ⊕ means entry-
wise multiplications [30]. The Levy flight essentially provides The solution of |γ| is obtained as
 
a random walk, while the random step length is drawn from a L 1
Levy distribution γ= Ln (14)
2 η
Levy(u) = t−1−β , 0 < β ≤ 2. (5) with
 
The step size generating new nest is different from as defined as |γ|
η = exp −2 . (15)
follows: L
S(k) = α(xi (k) − xj (k)) ⊕ Levy(β). (6) Based on quantum mechanism (QC), the distance between the
ith nest xi and the mean position x̄ of all nest position is
The update of the position of xi is given by the following
equation: L = 2δ |x̄ − x(k)| (16)
xi (k + 1) = xi (k) + ri Si (k) (7) where δ is a control parameter, which controls the difference
between x and x̄ it is defined as follows:
where ri is a random number generated by the uniform distri-
1
n
bution in interval [0, 1]. The CS algorithm employs a discovery
x̄ = xi . (17)
probability pa to replace the nests abandoned by the hosts. Then, n
i=1
the update law is defined as follows:
So, the step size is redefined by the following equation:
xi + r̃∗ if Pa > pa    
x∗ = (8) ∗   1
xi else S (k) = δ x − x̄(k) Ln . (18)
η
where pa is the discovery probability to create a new nest, and
P is a random number in interval [0, 1], while r∗ is the step size Based on the (6), (8), (17), and (18), the heterogeneous search
to generate new nest that is different from that of (6), and it is mechanisms update rules are as follows:
⎧
3 ≺ sr ≤ 1
defined by the following: ⎪
2
⎨ xi (k) + Si (k)
r∗ = rand(xi − xj ) xi , xj ∈ [1, n] . (9) xi (k + 1) = x̄(k) + ζ [x̄i (k) − xi (k)] 13 ≺ sr ≤ 23 (19)
⎪
⎩
2) Quantum Mechanism: The aim of employing quantum xi (k) + ε [g(k) − xi (k)] else.
computing QC in the CS algorithm is that each cuckoo can en- 3) Implementation of iTaSuM Algorithm: This subsection
hance its search ability according to an expected potential well, presents the iTaSuM algorithm (see algorithm 1) [20], [28] and
which can be designed as the best exemplar of all the individuals the flowchart is illustrated in Fig. 13, where the structure and
in the CS. In quantum mechanism, Schrödinger equation is used the parameters of the fuzzy model are presented as a nest. Each
to describe how the quantum state of a physical system changes nest represents a specific T–S fuzzy model. There are two parts
with time [20] in a rule: the premise parameters (structure) and the consequent
∂ −2 2 parameters, cji and σij2
are the premise parameters and αij are
i ψ(γ, t) = ∇ + v(γ, t) ψ(γ, t) (10)
∂t 2m the consequent parameters, where the dimension of ci is the

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
980 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO. 1, JANUARY/FEBRUARY 2019

Fig. 4. Input signal.

Fig. 3. ith rule encoded in a nest vector in HeCoS algorithm [20].

equation:

1 
N
MSE = (yk − ŷk )2 (21)
N
k =1

where N is the number of training samples, and yk and ŷk are

the desired and the fuzzy model outputs, respectively.

V. EXPERIMENTAL RESULTS
This section is meant to validate the performance of the pro-
posed iTaSuM method on the IS process and to compare the
performance of the proposed HeCoS algorithm with other three
existing algorithms. These algorithms include PSO algorithm
[32], [33], which is widely used to optimize the parameters of
the linear regression models, many researchers have focused
on this algorithm [13], [34]–[36]. However, it suffers from pre-
mature convergence, and is trapped easily into local optima,
which will significantly affect the model accuracy [7], [37].
The MsPSO which is a heterogeneous multiswarm PSO al-
gorithm used to generate an optimal T–S fuzzy model, which
is called optimizing fuzzy model (OptiFel). The OptiFel en-
codes the system structures and the parameters into particles to
find the optimal solution simultaneously. In the MsPSO algo-
rithm, the whole population is grouped into four subswarms and
heterogeneous search strategies are employed to accomplish the
optimization task [37]. The GAMsPSO is an improved version
of the MsPSO algorithm with an adaptive inertia weight based
on gray relational analysis used in order to obtain an optimal T–
S fuzzy models [14]. The parameter settings of each algorithm
are listed as follows:
PSO: w = 0.729; c1 = c2 = 0.729.
MsPSO: w = 0.5−0.9t
tm ; c1 = c2 = 1.4949.
same as that of input variable ui , and σij  0. Hence, a T–S
GAMsPSO∗ : w(t) = fw (gt† );c1 = c2 = 1.4949.
fuzzy model can be obtained if nest vector representations are
†: gt Stands for the gray relational of tth iteration.
finally determined by the HeCoS algorithm discussed earlier. ∗
: In GAMsPSO, cm in = 1, 5; cm ax = cfinal = 2, 5
For a dataset with N inputs, we can calculate Fij (uji ) according
HeCoS: Pa = 0.15, α = 1.3, β = 1.3 and δ = 0.9, and the
to the Gaussian function defined by [20]
number of agents popsize = 40.
 2
 For all algorithms, authors used the following parameters: the
(uj − cj ) maximum iteration number is set to 50 and the number of agents
Fij (uji ) = exp − i 2 i . (20)
σij popsize is set to 40.
To identify the sprinkler station, dSpace card (T M S320F 240)
In order to minimize the error between the output of fuzzy is used for taking the real measures of the pressure and the flow
model and the output data, we used the mean square error outputs which are represented, respectively, in Figs. 5 and 6,
(MSE) as the fitness of each individual defined by the following while the input signal is represented in Fig. 4.

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
CHROUTA et al.: MODELING AND CONTROL OF AN IRRIGATION STATION PROCESS 981

Fig. 5. Real curve of the pressure evolution.

Fig. 6. Real curve evolution of the flow. Fig. 7. (a) Real value and the pressure output of all algorithms. (b) Error
between the real value and the algorithms output.

This dataset originally consists of 25 000 input–output ob-

servations [u(k), y1 (k), y2 (k)]. At each sampling time k, u(k)
is the input pressure of the IS, while y1 (k) and y2 (k) are, re-
spectively, the pressure and flow outputs of the IS process. The
simulation was conducted to predict y1 (k) and y2 (k) are based
on u(k), y1 (k − 1), y2 (k − 2). In order to take all mentioned
issues into account, two experimental cases have been simu-
lated. In the first case, all the 25 000 data pairs are used as
training data. The rule number was set to 3 to build the T–S
fuzzy model by iTaSuM. The performance results of our fuzzy
model on whole data set are shown in Figs. 7 and 8. Indeed,
Fig. 7(a) depicts the real and the predicted output of the pres-
sure variable and Fig. 7(b) draws the corresponding error signal.
Furthermore, Fig. 8(a) depicts the real and the predicted output
of the flow variable and Fig. 8(b) shows the corresponding error
signal. According to the results presented in Figs. 7 and 8, we
can conclude that there
are significant differences between HeCoS and the other al-
gorithms (PSO, MsPSO, and GAMsPSO). The detailed per-
formance comparisons among different models are shown in
Table I. The MSE value of iTaSuM is 0.0026 and 0.0019, which
are, respectively, referred to the pressure and the flow outputs,
showing the performance of the proposed approach in achieving
model accuracy. Compared with the other methods, it is clear
that the identification accuracy of our approach is quite high.
In the second case, the whole data set has been divided in two
parts such that the first 12 500 input–output pairs were taken Fig. 8. (a) Real value and the flow output of all algorithms. (b) Error between
as training data, while the last ones are considered as testing the real value and the algorithms outputs.

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
982 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO. 1, JANUARY/FEBRUARY 2019

TABLE I
COMPARISON RESULTS OF THE IRRIGATION STATION PROCESS (CASE 1). BOLD VALUES ARE THE BEST RESULTS

Fig. 9. Training data. (a) Real value and the pressure output of all algorithms. Fig. 10. Testing data. (a) Real value and the pressure output of all algorithms.
(b) Error between the real value and the algorithms outputs. (b) Error between the real value and the algorithms outputs.

data. The training data set is used to build the model, and testing algorithms based on the MSE and variance accounting for
data set is to demonstrate the robustness of the model. Also, (VAF) tests [14]. According to the results of the MSE and
the fuzzy model is constructed by three rules. In this case, the VAF tests, it is clear that the identification accuracy of HeCoS
comparisons between the real process and the obtained model algorithm is quite high than other methods. The parameters
for training and testing data are depicted in Figs. 9 and 10 for the values, obtained with optimal fuzzy model of the IS process
pressure output, and Figs. 11 and 12 for flow output. It can be using HeCoS algorithm after the learning processes, listed
easily seen that the fuzzy model outputs are in good accordance in Table IV. The experimental results presented in all fig-
with the real process outputs. For comparison, Figs. 9(b), 10(b), ures show that iTaSuM approach based on HeCoS algorithm
11(b), and 12(b) illustrate the error between the real value and is well adapted to find the optimal fuzzy model of the hy-
the others outputs obtained from different algorithms. It shows draulic process with high accuracy and a strong generalization
that the fuzzy model generated by HeCoS matches well with the ability.
real system and better than others algorithms. It should be men-
tioned that in this case, a training MSE value of 0.0031 and a VI. FUZZY LOGIC CONTROL DESIGN
testing MSE value of 0.0047 have been achieved for the pressure
A. Proposed FLC
output. For the flow variable, the training MSE value is 0.0022
and the testing MSE value is 0.0034. These results of MSE show On combining the T–S fuzzy plant model and the FLC,the
the performance of the HeCoS algorithm in achieving model number of fuzzy subsystems generated becomes too large [37].
accuracy. Tables II and III present a comparative study be- This is because the increasing number of fuzzy subsystems can
tween the HeCoS algorithm and PSO, MsPSO, and gray MsPSO be explained by the introduction of the FLC (each control rule

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
CHROUTA et al.: MODELING AND CONTROL OF AN IRRIGATION STATION PROCESS 983

TABLE II
COMPARISON RESULTS OF THE IRRIGATION STATION PROCESS (CASE 2). BOLD VALUES ARE THE BEST RESULTS

TABLE III
COMPARISON RESULTS OF THE IRRIGATION STATION PROCESS USING VAF TEST VALIDATION.
BOLD VALUES ARE THE BEST RESULTS

Fig. 11. Training data. (a) Real value and flow output of all algorithms. Fig. 12. Testing data. (a) Real value and flow output of all algorithms.
(b) Error between the real value and the algorithms outputs. (b) Error between the real value and the algorithms outputs.

of the FLC is simultaneously applied to all the model rules of Control Rule i: IF premise i .
the T–S fuzzy model). Therefore, the performance of the closed-
loop system becomes more difficult to design and analyze. To THEN num [u(k)] = ui (k)
overcome these drawbacks, a new FLC is proposed. This FLC = adi0 y(k) + adi1 y(k − 1) + · · · + adin y(k − n)
has two consequents in each rule. They are a numerator part
and a denominator part. While the number of fuzzy subsystems + bdi1 u(k − 1) + · · · + bdim u (k − m) + R
involved remains the same, the overall closed-loop T–S fuzzy den [u(k)] = bi0 (22)
model-based system can be linearized as a transfer function with
arbitrary designed coefficients. This FLC of the following form where num [u(k)] is the numerator of u(k), den [u(k)] is
[38] is proposed. the denominator of u(k), adij = acj − aij for all j = 0, 1,

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
984 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO. 1, JANUARY/FEBRUARY 2019

Fig. 13. Flowchart of iTaSuM for the identification of the T–S fuzzy model.

TABLE IV
PARAMETERS IN EACH RULES OF iTaSuM OF FLOW (TOP) AND PRESSURE (BOTTOM) OUTPUTS

2, . . . , n, bdil = bcl − bil for all l = 1, 2, . . . , m, R is the ref-

erence input of the closed-loop system, acj and bcl are some 
c 
c
+ ··· + μi adin y(k − n) + μi bdi1 u(k − 1) + · · ·
coefficients independent of i to be designed later. i=1 i=1
The main difference of the proposed FLC of (22) from the
conventional FLC [24]–[26] is that there are two consequents in 
c 
c
+ μi bdim u (k − m) + μi bi1 u(k − 1) + · · ·
the consequent part of the rules: a numerator and a denominator.
i=1 i=1
From (22), the inferred output of the proposed FLC is as follows: 

c 
c
c + μi bim u (k − m) + μi R . (24)
i= 1 μ i u i (k ) c
num [ui (k)] c
μi i=1 μi ui (k)
i=1 i=1
u(k) = = c i= 1 =  c . (23)
den [ui (k)] i = 1 μi bi 0
 c i=1 μi bi0
i = 1 μi It can be seen that no double summation term appears in the
closed-loop system of (24) when the proposed FLC of (22) is
Substituting this control signal into the T–S fuzzy plant model used. Hence, the number of fuzzy subsystems remains to be c.
of 3, the closed-loop fuzzy control system becomes Furthermore, from (24) and (22)
c 
1  
c
1 
c
y(k +1) = c μi ai0 y(k) + μi ai1 y (k −1) + · · · y(k + 1) = c μi (ai0 + adi0 ) y(k)
i=1 μi i=1 i=1 i=1 μi i=1


c 
c 
c 
c
+ μi ain y(k − n) + μi adi0 y(k) + μi adi1 y (k −1) + μi (ai1 + adi1 ) y(k − 1) + · · ·
i=1 i=1 i=1 i=1

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
CHROUTA et al.: MODELING AND CONTROL OF AN IRRIGATION STATION PROCESS 985

Fig. 14. Block scheme of the controlled pumping station based on fuzzy logic.


c 
c
+ μi (ain + adin ) y(k − n) + μi (bi1 + bdi1 ) u(k − 1)
i=1 i=1


r
+··· + μi (bim + bdim ) u (k − m) + R
i=1

1 
c 
c
= c μi ac0 y(k) + μi ac1 y(k − 1) + · · ·
i=1 μi i=1 i=1
 c c
+ μi acn y(k − n) + μi bc1 u(k − 1) + · · ·
i=1 i=1


c Fig. 15. Decision surface of fuzzy controller.
+ μi bcm u (k − m) + R.
i=1
TABLE V
Since acj and bcl are independent of i INFERENCE MATRIX OF THE FUZZY CONTROLLER

1 c c
y(k + 1) =  c ac0 y(k) μi + ac1 y(k − 1) μi
μi i=1 i=1
i=1


c 
c
+ · · · + acn y(k − n) μi + bc1 u(k − 1) μi + · · ·
i=1 i=1 the performance of the closed-loop system can be designed, and
 the system stability can be guaranteed without resorting to find

c
a common Lyapunov function. To use the fuzzy controller, we
+bcm u (k − m) μi + R
i=1
must program through the tool FUZZY OF MATLAB. Entries
are chosen as Gaussian form (bell) and converted to linguistic
= ac0 y(k) + ac1 y(k − 1) + · · · + acn y(k − n) variables using three fuzzy levels for each submodel: Z, P ,
+ bc1 u(k − 1) + · · · + bcm u (k − m) + R. and N . The Output variable have five membership functions
as follow, NB: Negative big, NS: Negative small, Z: Zero, PB:
Furthermore, if we choose bdil such that bcl = 0 for all l = Positive big, PS: Positive small. The variation law of fuzzy
1, 2, . . . , m, the overall closed-loop system becomes controller is shown in Fig. 15. Thus, the FLC inference method
y(k + 1) = ac0 y(k) + ac1 y(k −1) + · · · + acn y(k − n) + R. requires nine fuzzy rules presented in Table V. The pumping
(25) station controlled by the fuzzy logic shown in Fig. 14.
Hence, the closed-loop system is reduced to a linear system of The rules can be written in the following way:
(25) by applying the FLC of (22). Since (25) is a characteristic f (ε is A) and (ε̇ is B) then Ucf = Si (ε, ε̇) . (26)
equation of the linear system,the coefficients ac0 , acl , . . . , acn
govern the location of the closed-loop poles. By designing the Using the method max–min as a inference engine and the gravity
poles of the closed-loop system, the coefficients can be found, center method for the defuzzification, the output of the fuzzy

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
986 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO. 1, JANUARY/FEBRUARY 2019

Fig. 17. Evolution of the pressure with the fuzzy controller.

Fig. 18. Tracking error with the fuzzy controller.

Fig. 19. Evolution of the flow with the fuzzy controller.

smooth law of order what provides us an energy saving on the

output of the fuzzy controller.
Then, Fig. 16 describes the algorithm of the irrigation buckled
system controlled with fuzzy logic steps.

B. Simulation Results of Fuzzy Controller

Fig. 16. Flowchart of a regulation cycle with the fuzzy controller.
The reference signal is rich. We find at the beginning a died
zone, then a ramp, and a series of steps signals with different
controller can be written in the following form: amplitude.
  The pressure output tracking using the fuzzy controller and
Si (ε, ε̇) = min μiA (ε) , μiB (ε̇) (27) the output tracking error are shown, respectively, in Figs. 17
Ucf G = max (Si (ε, ε̇)) . (28) and 18. It can be seen that good tracking performance has been
achieved by the proposed fuzzy model (see Fig. 21). Figs. 19
The discourse universe of output currency in five fields. The and 20 show the evolution of the flow output using the fuzzy
decision surface of fuzzy controller, Fig. 15, reflects a probable controller with and without saturation.

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
CHROUTA et al.: MODELING AND CONTROL OF AN IRRIGATION STATION PROCESS 987

Fig. 20. Evolution of the flow using the fuzzy controller with saturation.
Fig. 22. Evolution of the pressure for the two types of regulators.

Fig. 21. Evolution of the control law. Fig. 23. Evolution of flow for two types of regulators.

TABLE VI
COMPARATIVE TABLE BETWEEN PI AND FUZZY CONTROLLERS,
RELATING TO THE PRESSURE
comparison with the PI controller. Time response of the variation
of the water flow according to time is given in Fig. 23, which
clearly shows that the flow under our controller is less than that
obtained by the PI controller. Accordingly, the FLC saves more
water and then more energy than the PI controller, which is
the main objective of this paper. Therefore, it can be concluded
that the proposed method provides less-conservative results and
has better advantages in comparison with the existing ones.
TABLE VII Moreover, the suggested controller works well when applied to
COMPARATIVE TABLE BETWEEN PI AND FUZZY CONTROLLERS, the IS process. Experimental implementation will be presented
RELATING TO FLOW
in the next section to validate the theoretical results.

VII. EXPERIMENTAL IMPLEMENTATION

To validate the performances of the proposed control law, an
experimental study is presented in this section. In this context,
the operational algorithm of pumps and system regulation are
The flow response in Fig. 19 of IS system controlled by fuzzy
shown in Fig. 16. The scheme used for the experimental setup
controller exceeds the maximum flow (8 m3 /h) accepted by the
is shown in Fig. 24.
practical system, which forces us to add a saturation to flow
The experimentation has been carried out using MATLAB-
output.
Simulink and dSPACE DS1104 real-time controller board. This
board contains a Motorola Power PC 603e model that operates at
C. Comparative Study the speed of 250 MHz and a DSP (T M S320F 240 − 20 MHz).
In this subsection, the performance and effectiveness of the The developed experimental test bench of the electrical con-
developed controller are examined in terms of accuracy and trol cabinet is shown in Fig. 24. This cabinet is composed by
speed. The comparative analysis between the suggested FLC different blocks: A3, A4, and A5, in yellow color, represent, re-
and PI controllers is presented in Tables VI and VII. As can be spectively: flow sensors, pressure before, and after a pressure
seen, the proposed controller provides better results than the PI drop valve. The block A2 in green color is the electrical regula-
controller. Fig. 22 displays the response of the pressure of the IS tor pressure card. We can see the speed variator in the red block
process under the two methods. From this figure, it turns out that A1. Finally, two asynchronous motors drive the pumps. The first
the proposed method has better behavior and fattest response in one has a power of 1.1 kW and runs at a fixed speed. The second

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
988 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO. 1, JANUARY/FEBRUARY 2019

Fig. 25. Experimental results for pressure output.

Fig. 26. Experimental results for flow output.

Fig. 27. Control law.

Fig. 24. Experimental testing configuration.

has a power of 1.5 kW and a variable speed. Both the motors

are located inside the blue rectangle.
Table VIII summarizes the characteristics of the two moto-
pumps.
The practical process involves the following steps.
1) We measure the pressure (bar) from block A5 of the
station.
2) We send this information to the controller board of our Fig. 28. Tracking error.
dSpace through ADC 16 bits input.
3) The PI controller parameters are loaded in real time using Experimental results are shown in Figs. 25–28. Fig. 25 shows
MATLAB-Simulink software. the pressure output tracking (in green) and Fig. 28 gives tracking
The proposed fuzzy control set point is calculated in dSpace error curves. From these figures, it can be seen that satisfactory
(closed-loop system with PI controller). The DS1104 sends the tracking performance has been achieved. Fig. 27 presents the
appropriate control signal to our irrigation by mean of DAC corresponding control curve. The tracking errors fluctuate in
connectors. the transient period, but after some time, the tracking errors

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
CHROUTA et al.: MODELING AND CONTROL OF AN IRRIGATION STATION PROCESS 989

under FLC controller are almost the same. Fig. 26 illustrates REFERENCES
the evolution curve of the flow output. Its stability is obtained [1] R. R. Perez, V. F. Batlle, and L. S. Rodriguez, “Robust system identification
at fact to the interconnection between pressure and flow in the of an irrigation main canal,” Adv. Water Resour., vol. 30, no. 8, pp. 1785–
structure of our real process. The curve of flow has a relation 1796, 2007.
[2] R. R. Perez, J. P. Gonzalez, B. P. Reyes, and S. P. Pereira, “Distributed
with the energy absorbed by the pump units. control under centralized intelligent supervision in the güira de melena
The experimental results of the FLC control strategy confirm irrigation system,” Ingenieria Hidraulica en Mexico, vol. 18, no. 2, pp. 53–
the results obtained in simulations. From Figs. 25 and 26, it is 68, 2003.
[3] J. Sjöberg et al., “Nonlinear black-box modeling in system identifica-
clear that the proposed FLC design strategy provides a good tion: A unified overview,” Automatica, vol. 31, no. 12, pp. 1691–1724,
agreement and performance qualities for the IS process. Con- 1995.
sequently, we can note that the ratio speed/power for both the [4] A. Juditsky et al., “Nonlinear black-box models in system identification:
Mathematical foundations,” Automatica, vol. 31, no. 12, pp. 1725–1750,
outputs (pressure and flow) of our system with fuzzy controller 1995.
are satisfactory. [5] M. Mohammadzaheri, S. Grainger, and M. Bazghaleh, “A comparative
study on the use of black box modelling for piezoelectric actuators,” Int.
J. Adv. Manuf. Technol., vol. 63, no. 9–12, pp. 1247–1255, 2012.
VIII. CONCLUSION [6] T. Takagi and M. Sugeno, “Fuzzy identification of systems and its ap-
plications to modeling and control,” IEEE Trans. Syst., Man, Cybern.,
In this study, we investigated an optimal T–S fuzzy model of vol. SMC-15, no. 1, pp. 116–132, Jan./Feb. 1985.
an IS by using iTaSuM. In order to identify the structure and [7] J. Chrouta, A. Zaafouri, and M. Jemli, “System identification using gray-
parameters of the T–S fuzzy system, we used HeCoS strategy based adaptive heterogeneous multi-swarm PSO algorithm: Application
to an irrigation station,” J. Circuits, Syst. Comput., vol. 27, no. 4, 2018,
based on the quantum mechanism for the framework optimiza- Art. no. 1850059.
tion. The main benefit of this approach (iTaSuM) is to obtain [8] J. Chrouta, A. Zaafouri, and M. Jemli, “Modelling and identification of
simultaneously all the model’s structures and parameters. Fur- an irrigation station using hybrid possibilistic c-means and fuzzy particle
swarm optimisation,” Int. J. Comput. Appl. Technol., vol. 55, no. 1, pp. 30–
thermore, this method reduces the inconsistence in the tradi- 38, 2017.
tional two-stage identification process where the structures and [9] A. J. Telmoudi, M. Soltani, L. Chaouech, and A. Chaari, “Parameter
the parameters of T–S fuzzy models are encoded into a nest vec- estimation of nonlinear systems using a robust possibilistic c-regression
model algorithm,” in Proc. Inst. Mech. Eng., Part I: J. Syst. Control Eng.,
tor in order to guarantee that the obtained fuzzy models will have 2018, pp. 1362–1381.
good performance. Then, we developed an optimal control law [10] J. C. Bezdek, Eds., “Objective function clustering,” in Pattern Recognition
by using FLC based on the dynamic models of the IS process With Fuzzy Objective Function Algorithms. Boston, MA, USA: Springer,
1981, pp. 43–93.
obtained by (iTaSuM). The FLC control technique was com- [11] C. Li, J. Zhou, B. Fu, P. Kou, and J. Xiao, “T–S fuzzy model identification
pared with a classical PI control strategy in order to prove the with a gravitational search-based hyperplane clustering algorithm,” IEEE
robustness and the effectiveness of the proposed control tech- Trans. Fuzzy Syst., vol. 20, no. 2, pp. 305–317, Apr. 2012.
[12] X. Ding, Z. Xu, M. Liu, and J. Wu, “T–S model identification based on
nique. Consequently, this comparative study shows that the FLC silhouette index and improved gravitational search algorithm,” in Proc.
controller offers some advantages, such as the energy of control 33rd Chin. Control Conf., 2014, pp. 4423–4428.
effort is minimized and the system controller is characterized [13] Z.-H. Zhan, J. Zhang, Y. Li, and H. S.-H. Chung, “Adaptive particle swarm
optimization,” IEEE Trans. Syst., Man, Cybern., Part B (Cybern.), vol. 39,
by faster response and zero-steady-state error. no. 6, pp. 1362–1381, Dec. 2009.
Finally, the simulations and experimental results showed that [14] J. Chrouta, A. Zaafouri, and M. Jemli, “An improved heterogeneous
iTaSuM can generate a good T–S fuzzy model with high accu- multi-swarm PSO algorithm to generate an optimal T–S fuzzy model
of a hydraulic process,” Trans. Inst. Meas. Control, vol. 40, no. 6,
racy and strong generalization ability in IS dataset. The exper- pp. 2039–2053, 2017.
imental results demonstrate that the FLC controller can ensure [15] H. Du and N. Zhang, “Application of evolving Takagi–Sugeno fuzzy
the desired performances desired irrigation process. Therefore, model to nonlinear system identification,” Appl. Soft Comput., vol. 8,
no. 1, pp. 676–686, 2008.
the proposed method allows a good water management and a [16] C. Li, J. Zhou, J. Xiao, and H. Xiao, “Hydraulic turbine governing system
minimization of energy consumption. identification using T–S fuzzy model optimized by chaotic gravitational
search algorithm,” Eng. Appl. Artif. Intell., vol. 26, no. 9, pp. 2073–2082,
2013.
APPENDIX [17] C. Li and J. Zhou, “Parameters identification of hydraulic turbine gov-
erning system using improved gravitational search algorithm,” Energy
TABLE VIII
Convers. Manage., vol. 52, no. 1, pp. 374–381, 2011.
PUMPS PARAMETERS
[18] X.-S. Yang and S. Deb, “Cuckoo search via Levy flights,” in
Proc. World Congr. Nature and Biologically Inspired Comput., 2009,
pp. 210–214.
[19] D. Chitara, K. R. Niazi, and N. Gupta, “Cuckoo search optimiza-
tion algorithm for designing of multimachine power system sta-
bilizer,” IEEE Trans. Ind. Appl., vol. 54, no. 4, pp. 3056–3065,
Jul./Aug. 2018.
[20] X. Ding, Z. Xu, N. J. Cheung, and X. Liu, “Parameter estimation of
Takagi–Sugeno fuzzy system using heterogeneous cuckoo search algo-
rithm,” Neurocomputing, vol. 151, pp. 1332–1342, 2015.
[21] A. Z. M. R. Mejri and A. Chaari, “Control of a station of irrigation by
In Table VIII, N n represents the rated speed, cos(ϕ) is the sprinkling,” in Proc. Int. Conf. Control, Eng. Inf. Technol., Sousse, Tunisia,
power factor, In is the rated current, Hmax represents the maxi- Mar. 2013, pp. 122–127.
[22] M. R. Mejri, A. Zaafouri, and A. Chaari, “Hybrid control of a station of
mum height of pumps, and Qm ax is the maximum flow generated irrigation by sprinkling,” Int. J. Eng. Innovative Technol., vol. 3, no. 1,
by pumps. pp. 9–17, 2013.

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.
990 IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 55, NO. 1, JANUARY/FEBRUARY 2019

[23] W. Chakchouk, J. Chrouta, C. B. Regaya, A. Zaafouri, and A. Sellami, Wael Chakchouk received the master’s degree in
“Discrete-time adaptive backstepping control: Application to pumping automatics and industrial informatics and the Ph.D.
station,” in Proc. Inst. Mech. Eng., Part I: J. Syst. Control Eng., vol. 232, degree in electrical engineering from the Higher Na-
no. 6, pp. 683–694, 2017. tional Engineering School of Tunis (ENSIT), Uni-
[24] K. Tanaka, T. Ikeda, and H. O. Wang, “Robust stabilization of a class versity of Tunis, Tunis, Tunisia, in 2012 and 2018,
of uncertain nonlinear systems via fuzzy control: quadratic stabilizability, respectively.
h/sup/spl infin//control theory, and linear matrix inequalities,” IEEE Trans. He is currently a Research Member with the
Fuzzy Syst., vol. 4, no. 1, pp. 1–13, 1996. Laboratory of Industrial Systems Engineering
[25] K. Tanaka and M. Sano, “A robust stabilization problem of fuzzy control and Renewable Energy, ENSIT, Tunis. His re-
systems and its application to backing up control of a truck-trailer,” IEEE search interests include identification, control the-
Trans. Fuzzy Syst., vol. 2, no. 2, pp. 119–134, May 1994. ory and real-time control of nonlinear and complex
[26] H. O. Wang, K. Tanaka, and M. F. Griffin, “An approach to fuzzy control systems.
of nonlinear systems: Stability and design issues,” IEEE Trans. Fuzzy
Syst., vol. 4, no. 1, pp. 14–23, Feb. 1996.
[27] L.-K. Wong, F. F. Leung, and P. K.-S. Tam, “Design of fuzzy logic con-
trollers for Takagi–Sugeno fuzzy model based system with guaranteed
performance,” Int. J. Approx. Reasoning, vol. 30, no. 1, pp. 41–55, 2002.
[28] J. Chrouta, A. Zaafouri, and M. Jemli, “Engineering optimisation by het-
erogeneous cuckoo search algorithm: Application to an irrigation station,”
in Proc. Int. Conf. Adv. Syst. Elect. Technol., 2017, pp. 285–290.
[29] L. Wong, F. H. Leung, and P. K. Tam, “Stability design of T–S model
based fuzzy systems,” in Proc. 6th IEEE Int. Conf. Fuzzy Syst., 1997,
vol. 1, pp. 83–86. Abderrahmen Zaafouri was born in 1967, in Sidi
[30] X.-S. Yang and S. Deb, “Engineering optimisation by cuckoo search,” Int. Bouzid, Tunisia. He received the B.Sc. degree in
J. Math. Modelling Numer. Optim., vol. 1, no. 4, pp. 330–343, 2010. electrical engineering and the master’s degree in
[31] N. J. Cheung, X.-M. Ding, and H.-B. Shen, “A nonhomogeneous cuckoo automatic control from the High Normal School
search algorithm based on quantum mechanism for real parameter opti- of Technical Education of Tunis (ENSET), Tunis,
mization,” IEEE Trans. Cybern., vol. 47, no. 2, pp. 391–402, Feb. 2017. Tunisia, in 1993 and 1995, respectively, and the Ph.D.
[32] A. Nickabadi, M. M. Ebadzadeh, and R. Safabakhsh, “A novel particle degree from the National Engineering School of
swarm optimization algorithm with adaptive inertia weight,” Appl. Soft Tunis, Tunis, in 2000.
Comput., vol. 11, no. 4, pp. 3658–3670, 2011. Since 2016, he has been a Professor with the
[33] Y. Shi and R. Eberhart, “A modified particle swarm optimizer,” in Proc. Higher National Engineering School of Tunis, Tu-
IEEE Int. Conf. Evol. Comput. Proc., IEEE World Congr. Comput. Intell., nis. He is currently a Member with the Laboratory
1998, pp. 69–73. of Industrial Systems Engineering and Renewable Energy, École Supérieure
[34] M.-S. Leu and M.-F. Yeh, “Grey particle swarm optimization,” Appl. Soft des Sciences et Techniques de Tunis, Tunis. His research interests include ap-
Comput., vol. 12, no. 9, pp. 2985–2996, 2012. proaches to robust control of uncertain systems, modeling and identification,
[35] J. Xuan and H. Ming, “An improved quantum particle swarm optimization electrical machines, sensorless vector control of ac motor drives, and advanced
and its application,” in Proc. 6th Int. Conf. Comput. Sci. Netw. Technol., digital motion control.
2017, pp. 28–31.
[36] J. Qiao, H. Zhou, and C. Yang, “Bare-bones multiobjective particle swarm
optimization based on parallel cell balanceable fitness estimation,” IEEE
Access, vol. 6, pp. 32493–32506, 2018.
[37] N. J. Cheung, X.-M. Ding, and H.-B. Shen, “OptiFel: A convergent het-
erogeneous particle swarm optimization algorithm for Takagi–Sugeno
fuzzy modeling,” IEEE Trans. Fuzzy Syst., vol. 22, no. 4, pp. 919–933,
Aug. 2014.
[38] K. Tanaka and M. Sugeno, “Stability analysis and design of fuzzy control
systems,” Fuzzy Sets Syst., vol. 45, no. 2, pp. 135–156, 1992.
[39] L. Wong, F. Leung, and P. Tam, “Fuzzy model based control system
with state feedback,” in Proc. 4th Int. Conf. Control, Autom., Robot. Vis., Mohamed Jemli was born in Nasr’Allah, Tunisia, on
Singapore, 1996, vol. 3, pp. 1914–1918. November 2, 1960. He received the B.S. and D.E.A.
degrees from the École Supérieure des Sciences et
Techniques de Tunis (ESSTT), Tunis, Tunisia, in
1985 and 1993, respectively, the Ph.D. degree from
Jaouher Chrouta was born on May 10, 1987, in
the Ecole Nationale dIngenieurs de Tunis, Tunis, in
Gafsa, Tunisia. He received the B.S. and M.S. degrees 2000, and the HDR degree from the ESSTT, Tunis,
in electrical engineering from the High School of Sci-
in 2011, all in electrical engineering.
ences and Techniques of Tunis, University of Tunis,
From 1998 to 2001, he was an Aggregate Teacher
Tunis, Tunisia, in 2010 and 2012, respectively, and
with the Institut Superieur des Etudes Technologiques
the Ph.D. degree in electrical engineering from the de Rades. In 2001, he has been an Assistant Professor
Higher National Engineering School of Tunis (EN-
with the ESSTT. Since 2012, he has been a Professor with the Ecole Nationale
SIT), University of Tunis, Tunis, in 2018.
Superieure dIngnieurs de Tunis (ENSIT). He has authored or coauthored more
He is currently a Research Member with the Lab-
than 100 papers published in international conference proceedings and technical
oratory of Industrial Systems Engineering and Re-
journals in the area, and also holds many patents. His research interests include
newable Energy, ENSIT, Tunis. His research interests
electrical machines, sensorless vector control of ac motor drives, advanced dig-
include nonlinear systems analysis, modeling and identification, control and
ital motion control, renewable energy, and power quality.
optimization.

Authorized licensed use limited to: ANNA UNIVERSITY. Downloaded on September 03,2024 at 09:27:38 UTC from IEEE Xplore. Restrictions apply.