Received: 11 June 2019 Revised: 27 October 2019 Accepted: 18 December 2019

DOI: 10.1002/jnm.2717

RESEARCH ARTICLE

Bond graph modelling and control enhancement of an

off-grid hybrid pumping system by frequency optimization

Dhafer Mezghani1 | Abdelkader Mami1 | Genevieve Dauphin-Tanguy2

1
UR-LAPER, UR17ES11, Faculty of
Sciences of Tunis, University of Tunis El
Abstract
Manar, Tunis, Tunisia The climate's dependence requires a complex modelling and other optimization
2
LAGIS UMR CNRS 8146, Ecole Centrale methods for the control of the hybrid system. Moreover, recently, and as for
de Lille, Villeneuve d'Ascq, Cedex, France
both the technological progression of the hardware and software, researchers
Correspondence may more and more optimize problems through the use of enabled platforms.
Dhafer Mezghani, UR-LAPER, The paper aims to discuss this issue. Our hybrid pumping installation contains
UR17ES11, Faculty of Sciences of Tunis,
University of Tunis El Manar, 2092 Tunis,
a photovoltaic generator, a wind source, converters, and an induction motor-
Tunisia. pump group; it has a hybrid and complex character and requires a good model-
Email: [email protected] ling of the energetic point of view before passing the phase of design of control
laws. So we use the graphical approach of modelling called dedicated graph for
this type of system, and this tool showed very interesting results in the level of
modelling, dimensioning, monitoring, and diagnosing. The numerical closed-
loop simulation of the complete model in an appropriate environment allows
us to generate an optimization control in which the appropriate frequency
depends on meteorological conditions. The simulations are made from the cli-
matic conditions measured in a real system using a microcontroller ARDUINO
board. These measurements are used to simulate the adopted control and to
judge on the feasibility of our architecture adopted.

KEYWORDS
hybrid power systems, control systems, optimization, photovoltaic, wind turbine

1 | INTRODUCTION

Water is a vital element for all living beings on earth, and this element exists in an abundant quantity in the planet.
Furthermore, this little quantity of water is not equally distributed between different areas of the world which makes
more than 80% of countries, 40% of population being classified as dry or semi-dry areas and living severe periods of
drought. This can be explained by the fact that precipitation is the primary source of fresh water, and this kind of source
cannot provide the total water demand around the world. So the strategy of rural development in Tunisia must inte-
grate water control as one of its main objectives. However, both for pumping water and for conservation or the
processing of agricultural products, more or less significant amounts of energy are required in proportion to the desired
production volumes. Pumping is an alternative for managing water resources in different fields (surface, groundwater,
desalination, agriculture, or industry uses). Most of conventional pumping stations are powered by fossil energy
sources. The cost of these energies continues to increase. Moreover, the massive use will lead to exhaustion of these
reserves. Besides, they represent real threats to the environment. Hybrid systems are an interesting solution for the elec-
trification of isolated consumers. In particular, the possibility of including renewable energy sources is increasingly

Int J Numer Model El. 2020;1–20. wileyonlinelibrary.com/journal/jnm © 2020 John Wiley & Sons, Ltd 1
2 MEZGHANI ET AL.

appreciated.1-3 Moreover, the complexity of the description of dynamic operation and the classifications of pumping sta-
tions and the multiple parameters involved in these systems have forced researchers to develop simplified models based
on different approaches.4-7
In the search for solutions to energy control applicable to pumping, it seems interesting to know the benefits of a
new pumping technique based on the coupling of two sources of renewable energy, namely, solar energy and wind
energy. Considering their seasonal characteristics, these two energies do not compete but, on the contrary, can value
each other.
So the originality of work is that we propose a hybrid pumping system composed of these two sources of energy with
storage about battery, which consists in the optimal exploitation of the coupling between them, our system is completely
different from the structures adopted in literature.8-17 The frame mainly studied is water pumping. Of course, from the
point of view of renewable energy sources and the control of water optimization, the results appear quite convincing and
interesting, but this does not exclude the fact that the combination of the two sources generates a pretty complex system.
For this, we propose a graphical modelling tool that has shown good results from modelling point of view, structural
analysis, control, identification, and diagnosis. This methodology is called bond graph (BG).18-23 Other modelling
approaches of energetic systems have been developed in the literature, namely, graph theory as24 whose author has pres-
ented the state-space method used for modular modelling of heating, ventilation, and air-conditioning components, and
a dynamic model has been established. This modelling phase by BG is the subject of the first part of this work.
In addition, the control of hybrid source pumping stations is a delicate task. It is responsible for guaranteeing the
requested instructions (related to the field of application as agriculture, the pharmaceutical industry, and water treat-
ment units) indifferent conditions and while maintaining good performance in terms of stability, rapidity, and
precision.
In this context, the second part of this paper is to design a frequency control scheme of the studied system which
takes into account the variations of the climatic conditions. The last part is dedicated to the description of the experi-
mental process and real-time implementation on an embedded target for the acquisition of climatic parameters and the
corresponding optimal frequency computing.
These measurements have given satisfactory results which conform to numerical simulations and ensure optimal
transfer of efficiency from sources to water extraction.
Finally, this paper is concluded with a conclusion and future perspectives of this work.

2 | BG MODELLING AND SIMULATION OF THE HYBRID PUMPING

INSTALLATION

BG tool (or link graph) is an intermediary between the physical system and the mathematical models that are associ-
ated with it. It provides a unified modelling tool applicable to all physics science fields (electrical, mechanical, and
hydraulic) that will facilitate the study of composed systems. The BG modelling has been used in the applications of
analysis, simulation, computing in control process, and monitoring. It has given a very interesting search results. The
applications of BG are given by Table 1.25,26
The generalized variables power by BG modelling are the effort “e” and the flow “f” where the product Ð expresses
the power transmitted
Ð by the bond. Also, it defines the power generalized variables: the momentum (p = e dt) and the
displacement (q = f dt). These two variables are known as state variables. Knowledge of these variables allows us to
know the dynamic state of the whole system considered.27,28 Besides the description of the system, each of these four
variables has a physical meaning regardless mean while do field. The physical major elements are commonly selected
as effort and flow variables in all the different energetically domains.
The BG elements may be classified as follows:
•Simple passive elements: R, C, and I
•Active elements: sources Se, Sf, and MSe MSf
•Junction elements: 0, 1, TF, GY, MTF, and MGY
For the modelling of the hybrid system, BG technique which is a tool that constitutes an intermediate between the
physical system and its mathematical models (matrix of transfer, state equations, etc) is used. This technique is widely
used to model and simulate dynamic systems. The schematic block of hybrid pumping system is given by Figure 1.
In the case of the hybrid structure studied, we consider the parallel coupling of the two sources (PV and wind)on
the DC bus via protective components. This allows to transform the conditions into DC voltage. The addition of the
MEZGHANI ET AL. 3

TABLE 1 Applications of bond graph

Application Advantages
Modelling *Makes possible the energetic study and makes simpler the building of models for multi-disciplinary systems
*Leads to a systematic writing of mathematical models (linear or nonlinear associated
Analysis *Estimation of the dynamic of the model and identification of the slow and fast variables
*Study of structural properties
Control *Possibility to build a state observer from the model
*Design of control laws from the model
*Design of control laws from simplified models
Identification *No “black box” model
*Identification of unknown parameters, but knowledge of the associated physical phenomena
Monitoring *Graphical heading termination of the “monitorability” conditions and of the number and location of sensors
to make the faults localizable and detectable
Simulation *Specific software (CAMP+ASCL, ARCHER, 20 SIM), a knowledge of the numerical problems which may
happen(algebraic-differential equation, implicit equation) by the means of causality

FIGURE 1 Schematic blocks of HPS

adaptation stages by MPP converters ensure the tension control and the optimization of the quantity of water pumped
in a hydraulic circuit. It is evident that the system Hybrid Pumping System (HPS) is complex, and it requires a model-
ling phase in order to dimension and design adequate control structures.
Our pumping system is hybrid because of the use of converters and multisources of random nature, and its model-
ling is complex (multidisciplinary). Moreover, compared with the classical modelling approach (equations and block
diagrams), the BG present, in all readability, the energy transfer from the sources to the pump. The presence of junc-
tions 0 and 1 allows us to introduce, with flexibility, detectors to follow in real time the power at the input and the out-
put of each system component and to judge on the total efficiency in order to have its maximum value more to the
actions of the controls adopted.
The conversion of the climatic conditions (irradiance Ec, ambient temperature Ta, and junction temperature Tj) into
electricity by the photovoltaic process is a means of solar exploitation. The PV generator is a special energy source that
it is characterized by a nonlinear current-voltage curve. We have a PV field consisting of four serial panels and another
fifth insulated for the measurement of the open circuit voltage.
The behavior of the KANEKA 60 W PV module (Figure 2A) is equivalent to a current source derived by a junction
diode. By neglecting the physical phenomena of photovoltaic cells such as contact resistances and current lost by the
sides of photocells as well as aging cells. The BG model of the equivalent PV panel coupled to a starting capacitance is
given in Figure 2B,C with RD as the nonlinear resistance of the PV diode29 that delivers a maximum power of the order
of 60 W under standard conditions.
For link 3 of the BG model in Figure 2C, the flow variable (current ID) is given by Equation (1).6,7
   
qVpv
I D = I 0 exp −1 , ð1Þ
nKTj
4 MEZGHANI ET AL.

FIGURE 2 Photovoltaic generator. (A) KANEKA PV generator, (B) electrical scheme, (C) reduced BG model. BG, bond graph; PV,
photovoltaic

FIGURE 3 Optimal characteristics of KANEKA K60

   E
I pht = I sc + K isct T j − 25
c
, ð2Þ
1000
 
-Eg
I 0 = I 0n T j 3 exp , ð3Þ
KTj

45-20
Tj = Ec + T a : ð4Þ
800

The simulation of the BG model of our PV module KANEKA 60 W under the 20-sim environment30-33 shows the
power and the current curve as a function of the voltage of the GPV (Figure 3).
We have established a new explicit law allowing to determine the optimal values of the current and of the voltage
These equations are different from the GPV's models which are established in the literature, because they are specific
only to our module and because they are based only on the parameters of the constructor.

   E
I opt = I sc + K isct T j −25
c
−
8 0  1000 1 9
< Vopt -Kvcot × T j − 25 = ð5Þ
k1 I cc exp@ A −1 ,
: U OC k 2 ;

Vopt = 0:76 VCO , ð6Þ

MEZGHANI ET AL. 5

where the coefficients k1 and k2 are determined by experimentation on a KANEKA module by measuring the short-
circuit current Icc and the open voltage VCO for different values of illumination Ec and temperature T j.

   E
I cc = 1:19 + 0:00075 T j − 25
c
−
 1000
1 9
8 0
< -ð-0:28Þ × T j − 25 = ð7Þ
1:19 k 1 exp@ A −1 ,
: 92 k 2 ;

   E
0 = 1:19 + 0:00075 T j − 25
c
−
 1000 1 9
8 0
< VCO -ð-0:28Þ × T j −25 = ð8Þ
1:19 k1 exp@ A−1 :
: 92 k 2 ;

The voltage VCO is obtained by direct measurement of an insulated panel not connected to the hydraulic circuit.
The encapsulation of the sensor during manufacture of the module for the direct measurement of the junction tem-
perature T j in the PV panel is very difficult. So we perform an indirect measurement based on an estimate based on
the measurement of the surface temperature Ts given by

T s = E c ea + bVv + T a : ð9Þ

Based on the simulations carried out under the 20-sim environment for the calculation of the junction temperature T j from
the relationship Equation (4) and the measurements that were made on the surface temperature of our panel TS with embed-
ded board (Figure 4A,B. We have seen a dependence between these two temperatures that it is given by an exponential form:

Ec
T j = T s + ΔT, ð10Þ
1000

witha: Empirically determined coefficient that establish the limit upper of the panel temperature for a low wind speed
and a high solar irradiation (a = −3.4). b: Empirically determined coefficient that establish the limit lower temperature
of the panel with a low wind speed and a high solar irradiation (b = 0.059).
We note that this junction temperature is highly dependent on the ambient temperature and on a quantity of heat
caused by the illumination Ec and the wind Vv at a well-defined altitude. The coefficient ΔT varied according to the sea-
son from the minimum value 3 (month of January) to the maximum value 19 (month of July).
Our AIRX-400 wind turbine (Figure 5A) is made up of a wing and of a shaft coupled to a synchronous generator
with permanent magnet. The conversion to a direct current is ensured by a diode rectifier followed by a buck converter
allowing the tracking of maximum power point of the wind turbine.
The electrical diagram of the association (wing and shaft) is given by Figure 5B.
The diagram of principle is given by the Figure 5C and responds to several parameters by means of the following
equation (Equation 11).

1 ρπR2 Vv 3 Cp ðλ,βÞ
Cturbine = : ð11Þ
2 Ω
According to [6], the power constant depends on the specific velocity and the pitch angle.
 
πðλ + 0:1Þ
Cp ðλ, βÞ = ½0:5 − 0:00167ðβ − 2Þ × sin
18:5 − 0:3ðβ −2Þ ð12Þ
− 0:00184ðλ − 3Þ:

Figure 6A shows the variation of this coefficient Cp in a function of both the specific speed λ and the pitch angle of
the pales β.
6 MEZGHANI ET AL.

F I G U R E 4 Surface temperature. (A) Measurements in July,

(B) measurements in October

F I G U R E 5 The studied wind generator. (A) Wind turbine AirX-400,

(B) overview of the wind turbine, (C) electrical diagram of the association (wing
and shaft)
MEZGHANI ET AL. 7

F I G U R E 6 Power characteristics of AirX-400. (A) Power

coefficient, (B) optimal power coefficient of AirX-400

We have observed that the pitch angle influences directly on the aerodynamic efficiency of the wind turbine; a 2 varia-
tion of the angle results in a 10% decrease in Cp. For the case of our AIR-X wind turbine, we have set the pitch angle β at 2 .
Figure 9 illustrates the evolution of the Cp coefficient of our Air X-400 wind turbine for a lot of specific speed values.
Also, for a pitch angle β = 2 , a polynomial interpolation of order 3 applied on the curve Cp of our wind turbine
AIRX-400 gives the following Equation (13):

Cp ðλÞ = C3 λ3 + C2 λ2 + C1 λ + C0 : ð13Þ

The BG model of the wing and shaft association is shown in Figure 6B.
The model of the generator dedicated to the method of BG resembles to a harmonic representation. The three arms
of the diagram in Figure 7B represent the phases of the modeled generator. Each arm consists of four types of elements:
two passive elements of type R, I, and then two junctions of type GY. These elements represent respectively the value of
the resistance of the windings, the cyclic inductance, the gyrator of variable ratio r given by an external device, and the
equi-flux junction allowing an association equivalent to a series connection of these elements [6].
The generator SGPM is coupled to an uncontrolled rectifier, the electrical diagram of which is given in Figure 7C.
It is possible to represent our rectifier by an average hop graph model represented by three transformers whose
gains vary according to the conductions of the diodes. We have associated with this rectifier an LC circuit to reach the
continuity of the current.
The bus current Idc is given by the logic control presented by Equation (14).
   
1 1
I dc = ð2S1 −S2 − S3 Þ i1 + ð2S2 − S1 − S3 Þ i2
3 3
  ð14Þ
1
+ ð2S3 −S1 − S2 Þ i3 :
3

In the model of the Figure 7D, we have added a Ca capacity to solve the causality problem. The filter circuit is
coupled to a buck, Figure 8A, for the search of the maximum power point.
8 MEZGHANI ET AL.

F I G U R E 7 The studied wind generator. (A) Bond graph model

of the wing and shaft, (B) bond graph model of SGPM, (C) electrical
model of rectifier, (D) bond graph model of the rectifier with filter
circuit

The average BG model of this buck is given by Figure 8B. The development of the average BG model of the buck
converter is widely treated in our work.6,7
In order to model all the elements that constitute a wind generator and referring to the BG models of each wind tur-
bine element, we have obtained the word BG model of the wind turbine that is represented in Figure 9.
The numerical simulations performed on our wind turbine alone (Figure 17) show the evolution of its power as a
function of the mechanical speed for various values of wind speed Vv. The evolution of the Pturbine power as a function
of the mechanical rotational velocity Ω (Figure 17) allows us to determine the characteristic of the maximum power of

F I G U R E 8 Buck converter models. (A) Electrical model of the buck

converter, (B) bond graph model of the buck converter
MEZGHANI ET AL. 9

F I G U R E 9 AirX-400 wind
turbine modelling. (A) Word
bond graph model of the wind
turbine, (B) electrical power of
wind turbine AirX-400,
(C) control scheme

the turbine that will be exploited for the search of the maximum power point MPPT corresponding to a maximum value
of the coefficient (Cpmax = 0.49) and a specific speed (λopt = 9). So it is highly desirable that the wind system be operated
in such a way that it operates on this point.
So we have derived a new law of the maximum power MPPT extracted from the wind turbine AIRX-400 given by
 
λopt :Vv
Pmax = K = K1 Vv 2 with K1 = 2:52: ð15Þ
R

The wind turbine must produce maximum power by making the best using way of wind energy; the black curve cor-
responds to the point of operation at maximum power MPPT. Indeed, at nominal speed and at a wind speed of 12.5 m/
s, the power of the turbine is maximum which is about 395 W for a rotation speed of 2000 rpm (200 rd/s). The control
of this system based on two functions, the maximum power extraction algorithm (MPPT) and the control of the DC bus
at the output of the buck by an internal MPPT control. For our application, the wind turbine is coupled to a battery
pack, where the internal controller MPPT uses hysteresis. This stops the wind turbine in silent control mode as soon as
the battery pack is fully charged or for turbulent wind values. The wind turbine resumes power generation when the
battery voltage falls just under the full load of around 24 V. Figure 18 illustrates the synoptic diagram of the internal
controller of the wind turbine. Based on the description of the manufacturer of the AIRX-400 turbine, the control of
Figure 18 is based on the limitation of the current generated by the three-phase rectifier to limit the power in the wind
system, a such as limit value of 550 W which corresponds to a current of the order Imax = 22.5 A for a wind speed of
14.5 m/s.
For the HPS architecture of the Figure 1, the matching stages 1 and 2 are respectively a buck converter and a boost
converter. We show their average BG models in a cyclic relationship as shown in Figure 10A,B.23
The coupling of the two sources in parallel is ensured by two power diodes to a battery pack which is charged at a
maximum voltage of 24 V. Figure 10C illustrates the BG model of the DC bus by means of protection by power diodes.
The diodes are modeled by transformer elements modulated by conduction signals SWD1 and SWD2 and internal
resistors R of low values so as not to influence the voltage value (Vbuck ffi Vhyb ffi 24 V).
A DC/AC converter is the voltage inverter shown in the Figure 11A. It is controlled by a symmetrical regular sam-
pling pulse width modulation strategy32 enabling, for each switching period, to give a variable average supply voltage
across the motor terminals17-23.
10 MEZGHANI ET AL.

F I G U R E 1 0 Average BG models of coupling elements. (A) Buck

converter, (B) boost converter, (C) DC voltage continuous Bus. BG, bond graph

FIGURE 11 Voltage inverter. (A) Schematic diagram of the tension

inverter, (B) averaged BG model of inverter associated to motor. BG, bond
graph

Referring to [23], the model BG obtained in the Figure 11B remains valid for both the average and switching
models.
MEZGHANI ET AL. 11

F I G U R E 1 2 Induction motor models. (A) Park model,

(B) Thevenin model, (C) BG model. BG, bond graph

FIGURE 13 BG model of centrifugal pump and hydraulic circuit. BG, bond

graph

We indicate by ρi the cyclic ratio of each arm of inverter (i = A, B, C). This inverter is coupled to an induction
motor-pump, his Thevenin electrical schemes is given along the axis D, and deduced from the model of Park, it is repre-
sented by Figure 12A,B. The BG model correspondent is given by the Figure 12C [23].
In the BG model in Figure 12C, the elements MSe take the following values: (MSe1 = ωsΦqs, MSe2 = −ωs Φds,
MSe3 = −ωs Φqs, MSe4 = ωs Φds, MSe5 = Rr Lr Φds,MSe = Lr Φqs ). For our application, the used pump is a centrifugal
Rr

type, and the two tanks are communicating, Equation (16) characteristic of the hydraulic network which related the
flow Q and the mechanical speed Ωm is given by the first law of similarity.
12 MEZGHANI ET AL.

FIGURE 14 BG model of the studied hybrid system. (A) BG

model in closed loop, (B) BG model with controllers. BG, bond graph

F I G U R E 1 5 Adaptation stages control schemes. (A) Adaptation stage

1, (B) adaptation stage 2

0 1
 
B 2 ðb2 − ψÞ C Ωnom
Ωm = @ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA Q = Q: ð16Þ
−b1 − b1 2 −4 b0 2 ðb2 −ψÞ Qnom

In addition, the motor-pump is characterized by its following aerodynamic Equation (17).

dΩm
Cem = C2 Ωm 2 + C1 Ωm + J : ð17Þ
dt

For thus, the BG model of hydraulic circuit is given by Figure 13. The element R: Rnlp contains the parameters C2
and C1.
MEZGHANI ET AL. 13

3 | F REQUENC Y CONTR OL OF HYBR ID PUMPING SYS TEM

The obtained model is then realized in 20-sim software environment. A closed-loop V/f control system, applied to an
induction motor, is fed by a PWM three-phase voltage inverter. For a wind speed (between 3.1 and 14.2 m/s), the hybrid
system functions on the sun wire. And since the weather conditions (EC and Ta) are variable, then it is necessary to
adapt the operation point of load to the maximum power provided by the hybrid source through the frequency of the
inverter which varies depends on the climatic conditions in order according to get an optimum frequency control. For
thus, it makes as possible to ensure an optimum efficiency of the structure. The simulation scheme is given by the
Figure 14.
The hybrid source is formed by coupling the PGV with the wind turbine to the DC bus through the DC/DC con-
verters. Its word BG model is given by Figure 15.
In Figure 15, the commands applied to the buck and boost converters are given respectively in Figure 15A,B.
The digital simulation of the structure in closed loop shows that the stator frequency fS varies in function of the vari-
ables EC and Tj, a polynomial interpolation in a high order enables us to define two frequencies fsE and fsT given by
Equation (5), and finally, the value of stator frequency applied to the machine is calculated by the algorithm of
Figure 15.
The adopted MPP method for the hybrid source (wind turbine and GPV) generates the stator frequency fs. It
depends on the insulation Ec and the temperature Ta as shown in Figure 16. It is given by the following Equation (18).

fs = f 0 + ðsignðΔE Þ:ΔfE Þ:CE + ðsignðΔT Þ:ΔfT Þ:CT: ð18Þ

In this equation, the variations DfE and DfT are given by the system Equation (19).

Δf E = f sE0 −f sE forΔT = 0
, ð19Þ
Δf T = f sT 0 −f sT forΔE = 0

F I G U R E 1 6 Experimental process.
(A) Calibration of LDR, (B) data
acquisition (Ec and TS) with ARDUINO,
(C) KALEAS wind anemometer,
(D) pumping station
14 MEZGHANI ET AL.

where frequencies fsE0 and fsE (resp. fsT0 and fsT) are respectively the frequencies to E0 (resp. T0) and Ec (resp. Ta) are
calculated by polynomial interpolation, and they are given by the system of Equation (20).
(
f sE = 10 −6 Ec 3 + 0:194Ec + 0:167 for Tp = Ct
ð20Þ
f sT = −10 − 4 Tp 2 − 0:054Tp + 53:76 for Ec = Ct

At the STC (25 C, 1000 W/m2), the corresponding reference frequency f0 (49 Hz).

4 | P E R F O R M A N C E RE S U L T S A N D DI S C U S S I O N

In order to validate the BG model of the pumping hybrid system and in order to simulate the real behavior of hybrid
system, we must have all experimental measurement results and all the acquisition of the various sizes (Figure 17A).
The illustrated block measurement, in Figure 17A, is a database of the metrological parameters EC and TS measured
respectively by an LDR and an LM35 sensor. The CPU of microcontroller ARDUINO, shown in Figure 17B, uses its
data for the calculation of the corresponding optimal frequency according to the law of Equation 18. In order to reach
this, we have to go through a calibration step for the photo-resistor to determine the equivalent in W/m2 from the mea-
sured voltage.

F I G U R E 1 7 Experimental measurements. (A) Irradiance measurements during experimental months (Ec), (B) surface temperature Ts,
(C) wind speed measurements during experimental months (Vv), (D) frequency according junction temperature measurements during
experimental months (fs), (E) photovoltaic voltage measurements during experimental months (Vpv)
MEZGHANI ET AL. 15

For the temperature's sensor, whenever the temperature increases by 1 C, the output voltage automatically
increases by 10 mV.
The average wind speed at the entrance of the blades of the wind turbine AIRX-400 is measured using a KALEAS
anemometer (Figure 17C).
The real pumping system is given by Figure 17D. This inverter (element 1) fed a motor-pump EBARA (element 2).
The electrical specifications of our induction machine are described by Table A2. The system includes different sensors:
pressure (element 7), flow (element 5), and level (element 6). The pipe (element 3) which is connected by a valve assem-
bly (element 4). For this installation, one PV panel is described in Table A2, so we combine four PV panels in series
delivering enough power for pumping.
In the 20-sim environment, we have simulated the full BG model, and we have used the measurements recorded on
the climatic conditions (Ec, Ts, and Vv) in a database for 4 months (Figure 17A-C). We observe that the photovoltaic
voltage Vpv is around 240 V for strong winds and high temperatures and 260 V for low winds and intense radiation.
This voltage Vpv decreases as a function of the surface temperature of the photovoltaic field (Figure 17E). The generated
frequency by the optimized law increases exponentially as a function of illumination and surface temperature
(Figure 17D).
In order to highlight the advantage of the hybrid pumping structure compared to the photovoltaic structure that is
the subject of several works.6 We find that the behavior of the two structures during the month of January is similar
since the illumination is very low and the wind does not exceed 4 m/s (Figure 18A,B).
Since the head of the hydraulic system is inversely proportional to the flow, during the seasonal averages (October
and April), a slight increase in the flow (Figure 19A,B) of the hybrid structure is observed (Figure 20A,B), which affects
the total efficiency of the structure by an average of 15% (Figure 21A,B).

F I G U R E 1 8 Experimental results during January. (A) Flow

measurements, (B) head measurements, (C) efficiency
16 MEZGHANI ET AL.

During the summer, the irradiance is intense; the photovoltaic structure records a pumped water quantity in the
order of 28 m3 per day. This quantity may be increased for instances or the illumination will decrease provided that
there is an intense wind that reaches 9 m/s (32 km/h) as shown in Figure 22A. This increase records the value 48 m3
per day, which shows advantage of the hybrid pumping structure compared to the PV pumping. This type of structure
ensures maximum management of the quantity of pumped water even if the efficiency does not exceed 15%
(Figure 22C).
Experimental tests carried out for typical months of the year show that our pumping structure powered by multi-
sources (photovoltaic-wind-battery), not connected to the electricity grid, presents an optimization of energy. This is
ensured by a maximum power point tracking system (MPPT) for both random energy sources (photovoltaic and wind).
This optimization is based on PI regulators and limiters. The battery regulator is not the subject of this study. We note
that the amount of water pumped is largely exceeds to that obtained by a photovoltaic pumping structure which allows
us to preserve the quantity not used by the consumer.

F I G U R E 1 9 Experimental results of pumped water flow. (A) in

October, (B) in April

FIGURE 20 Experimental results of head. (A) in October, (B) in

April
MEZGHANI ET AL. 17

FIGURE 21 Obtained efficiency. (A) in October, (B) in April

F I G U R E 2 2 Experimental results during July. (A) Flow

measurements, (B) head measurements, (C) efficiency
18 MEZGHANI ET AL.

5 | C ON C L U S I ON

We presented an atypical BG model of a pumping hybrid installation, then we established an optimum V/f control in
function of the climatic conditions. Finally, some measurements were carried out on the experimental device, and they
allowed us to validate the adopted control and check the studied operation reliability of hybrid chains. The simulations
carried out on the basis of the meteorological measurements show that the hybrid system has good energy efficiency
through a pumping system coupled directly to a PV generator with an increase in the quantity of pumped water. This
shows the performance of our architect adopted for hybrid pumping compared with other works. This control can be
improved by using other commands such as fuzzy logic or sliding mode.

ORCID
Dhafer Mezghani https://orcid.org/0000-0002-3637-9177

R EF E RE N C E S
1. Bianchini A, Magnelli N, Ferrara G, Carnevale EA, Ferrari L. Optimization of a PV-wind-diesel hybrid system for a remote stand-alone
application. Energy Proc. 2015;81:133-145.
2. Maleki A, Pourfayaz F. Optimization of grid independent diesel-based hybrid system for power generation using improved particle
swarm optimization algorithm. Int J of Smart Elec Enginee. 2015;4:29-36.
3. Cherif H, Belhadj J. Energy output estimation of hybrid wind-photovoltaic power system using statistical distributions. J of Elec Sys.
2014;10:117-132.
4. Brian DV, Byron AN. Analysis of off-grid hybrid wind turbine/solar PV water pumping systems. Sol Energy. 2012;86:1197-1207.
5. Roberto Valer L, Melendez TA, Cristina Fedrizzi M, Zilles R. Variable-speed drives in photovoltaic pumping systems for irrigation in
Brazil. Sus Energy Tech and Assess. 2016;15:20-26.
6. Mezghani D, Othmani H, Sassi F, Mami A. A new optimum frequency controller of hybrid pumping system bond graph modelling-
simulation and practice with ARDUINO board, international journal of advanced computer science and applications. Int J Adv Comput
Sci Appl (IJACSA). 2017;8(1):78-87.
7. Mezghani D, Othmani H, Mami A. Bond graph modeling and robust control of a photovoltaic generator that powered an induction
motor pump via SEPIC converter. Int Trans Electr Energ Syst. 2018;29(3):1-21.
8. Li R, Wul B, Li X, Zhoul F, Li Y. Design of wind-solar and pumped-storage hybrid power supply system, 3rd IEEE International Confer-
ence on Computer Science and Information. Dent Tech. 2010;5:402-405.
9. Kumar A, Kochhar E, Upamanyu K. Photo voltaic and wind energy hybrid sourced voltage based indirect vector controlled drive for
water pumping system. 2015 IEEE International Conference on Electrical, Computer and Communication Technologies (ICECCT). 2015;
1-5.
10. Ma R, Li X, Wu Z, Zhang Q. Multi-objective optimal scheduling of power system considering the coordinated operation of photovoltaic-
wind- pumping system storage hybrid power, 5th International Conference on Electric Utility Deregulation and Restructuring and
Power Technologies; 2015.
11. Olorunfemi O. Analysis of current source induction motor drive fed from photovoltaic energy source. IEEE T on Energy Conv. 1991;6:
99-106.
12. Ahmad GE, Hussein HMS, El-Ghetan HH. Theoretical analysis and experimental verification of PV modules. Renew Energy. 2003;28:
1195-1168.
13. Nichita C, Luca D. Large band simulation of the wind speed for real time wind turbine simulator. IEEE T Energy. 2002;17:523-529.
14. Bratcu AI, Munteanu I, Bacha S, Picault D, Raison B. Cascaded DC–DC converter photovoltaic systems: power optimization. IEEE T Ind
Electron. 2011;58:403-411.
15. Saied M, Hanafy A, Elgabaly MA, Sharaf A. Optimal design parameters for a PV array coupled to a dc motor via dc-dc transformer. IEEE
T on Energy Conver. 1991;6:258-264.
16. Dongbing Z. Designing a SEPIC Converter. Application Note: National Semiconductor (Texas Instrument); 2013.
17. Hmidet A, Hasnaoui Z, Dhifaoui R. Digital control of MPPT structures for water pumping systems. 7th IET International Conference on
Power Electronics, Machines and Drives, Manchester, UK. 2014; 1-7.
18. Dauphin-Tanguy G. Les Bond Graphs. Paris: Hermès edition; 2000.
19. Roboam X, Astier S, Foch H, Fontès G, Gandanegara G, Piquet H, Saisset R, Sareni B, Turpin C. Graphes de liens causaux pour systèmes
à énergie renouvelable (partie 2). France: Techniques de L'ingénieur, 2007.
20. Zeddini MA, Pusca R, Sakly A, Mimouni MF. PSO-based MPPT control of wind-driven self-excited induction generator for pumping sys-
tem. Renew Energy. 2016;95:162-177.
21. Umesh B, Umanand L. Bond graph model of doubly fed three phase induction motor using the Axis rotator element for frame transfor-
mation. Sim Mod Prac and Theo. 2008;16:1704-1712.
22. Mezghani D, Cabani I, Ellouze M, Mami A. Linearizing control of a photovoltaic structure and stability by lyapunov directly on bond
graph. J of Elec Sys. 2007;4:181-192.
MEZGHANI ET AL. 19

23. Mezghanni D, Andoulsi R, Mami A, Dauphin-Tanguy G. Bond graph modelling of a photovoltaic system feeding an induction motor-
pump. Sim Mod Prac and Theo. 2007;15:1224-1238.
24. Ye Y, Jiongde C, Jingmei F, Shaofan W. Modular modeling of air-conditioning system with state-space method and graph theory. Inter-
national Journal of Refrigeration. 2019;99(3):9-23.
25. Azmani A, Dauphin-Tanguy G. Program for computer aided modelling and analysis. Bond Graph for Engineers. Netherlands: Elsevier Sci-
ence Pub; 1992:263-278.
26. Ould BB. Bond graph approach as analysis tool in thermofuid model library conception. J of the Frank Inst. 2003;340:1-23.
27. Karnopp DC, Margolis DL, Rosenberg RC. System Dynamics, Mod and Sim of Mecha Sys. Fourth ed. New York: John Wiley & Sons Inc;
2005.
28. Sanchez R, Dauphin-Tanguy G, Guillaud X, Colas F. Bond graph based control of a three phase inverter with LC filter connection to pas-
sive and active loads. Sim Mod Prac and Theo. 2010;18:1185-1198.
29. Houssem CHAOUALI, Hichem OTHMANI, Dhafer MEZGHANI, Abdelkader MAMI. Improving the control strategy of a standalone PV
pumping system by fuzzy logic technique. International Journal of Advanced Computer Science and Applications. 2017;8(3):170-175.
30. Broenink JF. 20-Sim software for hierarchical bond-graph/block-diagram models. Sim Mod Prac and Theo. 1999;7:481-492.
31. Susmethaa Siri R, Keerthana C, Arthi A, Senthil Rani S. Hybrid water pumping control system for irrigation using Arduino. Int J of
Enginee Res & Techno. 2015;4:859-863.
32. Othmani H, Mezghani D, Ahmed B, Mami A. New approach of incremental conductance algorithm for maximum power point tracking
based on fuzzy logic. International Journal of Grid and Distributed Computing (IJGDC). 2016;9(7):121-132.
33. http://www.20sim.com/

How to cite this article: Mezghani D, Mami A, Dauphin-Tanguy G. Bond graph modelling and control
enhancement of an off-grid hybrid pumping system by frequency optimization. Int J Numer Model El. 2020;1–20.
https://doi.org/10.1002/jnm.2717

A P P EN D I X

TABLE A1 Characteristics of
Specification Value
AIRX-400
Start-up wind speed, m/s 3.13
Output voltage, V 24
Rated power, W 400
Turbine controller: μP
Rated wind speed, m/s 12.5
Rotor diameter, m 1.15
Coeff of turbine
c3 = −210−4, c2 = −2.810−3, c1 = 9.410−2, c0 = 10−4

TABLE A2 Specifications of EBARA

Electrical Data Value

Nominal output power, W 370
Nominal elect power, W 550
Max flow rate, L/min 35.6
Max head, m 7
Statoric resistor, Ω 24.6
Rotoric resistor, Ω 16.1
Mutual self, H 1.46
Rotoric self, H 1.48
Statoric self, H 1.49
Coeff of pump C1 = 1.7510-3 kg.m .s , C2 = 7.510-6Kg.m .s ,J = 6.510-3 Kg.m−1, b0 = 4.5210−4 min2.m.tr−2, b1 = −1.96610−3 m.min2.
−2 −1 −4 −2

tr−1.L−1, b2 = −0.012 min2.m.L−2, Ψ = 4.0816 10−3 min2.m.L−2

20 MEZGHANI ET AL.

TABLE A3 Characteristics of
Electrical Data Value
KANEKA 60
Nominal output Pmpp, W 60
Nominal Voltage Umpp 67
Nominal current Impp, A 0.9
Open circuit voltage Uoc, V 92
Short circuit current Isc, A 1.19

Kisct: Temp coeff of Isc, %/ C 0.075

Kvcot: Temp coeff of Uoc, mV/ C −280

Temp coeff output, %/ C −0.23