International Journal of Applied Power Engineering (IJAPE)

Vol. 14, No. 1, March 2025, pp. 244~254

ISSN: 2252-8792, DOI: 10.11591/ijape.v14.i1.pp244-254  244

A comparative analysis of ANFIS and fuzzy controllers for a

dynamic hybrid model

Laoufi Kaltoum1, Youssef Mouloudi2, Abdeldjebar Hazzab1, Abdallah Ben Abdelkader2

1
Laboratory of Command, Analysis and Optimization of Electro-Energetic Systems, University of Bechar, Béchar, Algeria
2
Laboratory of Smart Grids and Renewable Energies (S.G.R.E), Department of Electrical Engineering, Faculty of Technology,
University of Bechar, Béchar, Algeria

Article Info ABSTRACT

Article history: Transitioning from combustion engines to electric motors is essential to
reduce CO₂ emissions and combat climate change. This study presents a
Received Jun 7, 2024 dynamic hybrid model combining a fuel cell and battery for electric vehicles,
Revised Oct 14, 2024 emphasizing simplified parameter extraction from battery datasheets. The
Accepted Oct 23, 2024 model integrates two energy storage systems: batteries for electrochemical
storage and hydrogen for chemical storage, converted into electricity via a
fuel cell stack. This dual approach enables flexible refueling options with
Keywords: electricity or hydrogen. An air compressor in the proton exchange membrane
(PEM) fuel cell stack optimizes performance across varying driving
ANFIS conditions. The research aims to minimize fuel cell consumption and
Battery model enhance energy storage efficiency using Sim Power Systems software. It
Fuzzy employs traditional proportional integral derivative (PID) controllers and
Hybrid fuel cell vehicle advanced optimization techniques, including fuzzy and adaptive neuro-fuzzy
intelligent control inference system (ANFIS), to achieve optimal power distribution between
Proportional integral derivative the fuel cell system (FCS) and the energy secondary source (ESS) for
Sim Power Systems specific road scenarios. The proposed ANFIS-based approach demonstrates
superior control in balancing energy efficiency and driving dynamics,
surpassing both PID and fuzzy logic controllers in key metrics. This
innovative closed-loop control system offers a promising solution for hybrid
electric vehicles, ensuring optimal performance and energy management.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Abdallah Ben Abdelkader
Laboratory of Smart Grids and Renewable Energies (S.G.R.E), Department of Electrical Engineering
Faculty of Technology, University of Bechar
Béchar, Algeria
Email: benabdelkader.abdellah@univ-bechar, [email protected]

1. INTRODUCTION
In light of the growing concern over dwindling fossil fuel reserves and escalating global energy
requirements, hydrogen emerges as a promising energy carrier to supplant traditional fuels in the long run. As
the most abundant element in the universe with non-toxic byproducts, hydrogen presents itself as an
environmentally friendly option, particularly when utilized through a fuel cell hydrogen vehicle (FCHV). The
advent of electric vehicles equipped with energy storage systems (ESS) marks a significant step towards
establishing a fleet of clean transportation methods. Although current battery technologies offer only about
10% of the energy density of conventional fuels [1]-[3], fuel cells (FCs) with their high energy density stand
out as an ideal complementary energy storage solution for clean vehicles [1], [2]. The conversion of
hydrogen (H2) into electricity via fuel cells represents a pivotal electrochemical process, offering a clean
alternative to traditional thermal combustion methods. While fuel cells have been in existence since their

Journal homepage: http://ijape.iaescore.com

Int J Appl Power Eng ISSN: 2252-8792  245

discovery in 1839, their widespread adoption faces several technical and economic obstacles that need to be
addressed before mass commercialization can occur. Challenges such as battery costs, FCS integration, and
hydrogen production and storage need to be overcome to realize the full potential of this technology. This
study focuses on addressing the power selection for the fuel cell system and secondary energy source in a
hybrid FCHV. A dimensioning tool is proposed to ensure compliance with specified dynamic performance
requirements of the vehicle along with an energy criterion, such as hydrogen consumption. The methodology
is exemplified through an application on a prototype hybrid FCHV, considering the unique current/voltage
characteristics of fuel cells and their impact on system performance.The research shifts towards controlling a
fuel cell energy system for residential applications, with a specific emphasis on proton exchange membrane
cells (PEM). Known for their suitability in the automotive and decentralized electric power generation
sectors, PEM cells offer rapid dynamics and operate at relatively low temperatures, simplifying integration
into various systems. This study aims to delve into the physicochemical phenomena within PEMFC cells to
optimize performance and facilitate seamless integration into energy production systems.
Utilizing adaptive neuro-fuzzy inference system (ANFIS), this research endeavors to construct a
versatile model for intricate systems. Subsequently, a dynamic model of a fuel cell energy system for
residential use is outlined, encompassing a cell model, gas reformer model, and power processing unit. The
proposed model includes strategies for controlling cell power, emphasizing the integration of ANFIS and
fuzzy logic for system modeling and control.
The paper is structured as follows: i) Section 1 introduces the PEM fuel cell; ii) Section 2 elaborates
on the polymer membrane fuel cell (PMFC) system; iii) Section 3 discusses hybrid fuel cell vehicles (FCHV);
iv) Section 4 describes electric vehicle systems; v) Section 5 scrutinizes various control methodologies (PID,
fuzzy, ANFIS) through tests on the proposed model; and vi) Section 6 consolidates the simulation results of
the different test cases. In conclusion, this study aims to offer insights into the modeling and control of fuel
cell systems while highlighting the effectiveness of ANFIS and fuzzy logic approaches across diverse
domains, particularly in system identification and control.

2. THE POLYMER MEMBRANE FUEL CELL SYSTEM (FCS)

2.1. Working principle
The FCS fuel cell transforms chemical energy (hydrogen and oxygen) into electrical energy. This
electrochemical process represents the reverse electrolysis of water [4]. Electrical energy is generated by a
redox process with oxygen as the oxidant and hydrogen as the reducer. Oxidation occurs at the anode, whereas
reduction transpires at the cathode. Both processes are separated by a membrane functioning as an electrolyte.
The cathode receives gaseous oxygen (or, more broadly, air), whereas the anode is provided with gaseous
hydrogen. Electricity is generated by the two electrons emitted from the hydrogen molecule (Figure 1). The
𝐻 + hydrogen protons traverse the membrane dividing the anode from the cathode, recombining with electrons
and oxygen atoms at the cathode. Ideally, the sole byproducts of the reaction are water and heat. Figure 1
elucidates the concept of the PEM fuel cell.

Figure 1. PEM fuel cell block diagram [4]

2.2. The electrical characteristic of the fuel cell

The role of the electrochemical potential in the redox process. The optimal electrochemical potential
is 1.23 V (standard potential) at standard temperature and pressure conditions (1 atm, 25 °C). In reality, the
open-circuit voltage is somewhat under 1 V [5]. The electrical property of a fuel cell is referred to as the
polarization curve. The cell's voltage is depicted as a function of current density. The membrane's humidity,
operating temperature, and reagent pressure determine it. iPAC (A/cm²) represents the current density.

A comparative analysis of ANFIS and fuzzy controllers for a dynamic hybrid model (Laoufi Kaltoum)
246  ISSN: 2252-8792

𝐼𝑃𝐴𝐶
𝑖𝑃𝐴𝐶 = (1)
𝐴𝑐𝑒𝑙𝑙

With 𝑖𝑃𝐴𝐶 the current of the fuel cell and 𝐴𝑐𝑒𝑙𝑙 the active surface of a membrane.

2.3. Fuel cell system for vehicle

Figure 1 illustrates the four subsystems of a fuel cell vehicle system: the airflow supply system for the
cathode, the hydrogen flow supply system for the anode, the cooling system, and the humidification system.
The airflow supply subsystem comprises a compressor, cooler, water separator, back pressure valve,
connecting pipes, and additional components, facilitating the delivery of essential airflow to the stack
cathode. The hydrogen flow supply subsystem includes a hydrogen tank, a solid valve, a recirculation pump,
a purge valve, and connecting pipes, along with additional components. The purpose of this mechanism is to
ensure that the stack anode is supplied with adequate hydrogen. The air temperature will rise during
compression, necessitating the use of an air cooler to reduce the temperature before reaching the stack.
A humidifier is utilized to improve proton conductivity by adding moisture to the airflow. On the cathode
side of the stack, an electrochemical reaction occurs between oxygen and protons, resulting in the formation
of water and the release of heat. The introduction of coolant into the pile is essential for heat dissipation,
followed by its expulsion into the atmosphere through a radiator to maintain the stack temperature within
acceptable limits. The water separator removes moisture from the exhaust air stream. The back pressure
valve regulates the airflow pressure within the stack to ensure it remains at an optimal level.
A variety of fuel cell models have been developed for different applications [6], [7], with the models
described in [8]-[10] categorized as control-oriented models. The fuel cell system model utilized in this study
incorporates a twin-screw compressor, as cited in references [8], [9], and [11]. A twin-screw compressor
offers multiple benefits when compared to a centrifugal compressor:
- Its nominal rotational speed is approximately an order of magnitude lower than that of a centrifugal
compressor, leading to a more reliable system;
- The flow curve in the compressing map is almost linear at a constant velocity and is less affected by
pressure compared to a centrifugal compressor, hence facilitating system control; and
- A twin-screw compressor does not encounter the "surge" phenomena, hence streamlining the control
system. The compressor map and twin-screw compressor models are derived from [11] and [12]. Figure 2
depicts the twin-screw compressor concept, which is divided into two pieces. The static compressor map
is the initial component that determines the airflow rate and compressor power. The subsequent phase
involves implementing modifications. Due to the disparity between a compressor's operational
environment and its testing conditions, its airflow and energy must be modified in accordance with
thermodynamic principles. The model 𝑃𝑎𝑚𝑏 is tested using inputs of downstream pressure 𝑃𝑠𝑚 , rotational
speed 𝑁𝐶𝑃 , air pressure 𝑃𝑎𝑡𝑚 , pressure 𝑃0 (1.013 bar), and temperature 𝑇 0 (293.15 K). Compressor motor
torque and power are required to ascertain rotational speed. The compressor is connected to the supply
manifold, and its downstream pressure corresponds to the pressure in the supply manifold, as dictated by
the downstream cathode flow field.

Figure 2. The architecture of a fuel cell system for vehicle [13]

Int J Appl Power Eng, Vol. 14, No. 1, March 2025: 244-254
Int J Appl Power Eng ISSN: 2252-8792  247

The compressor's rotation speed may index the compressor power 𝑃𝑐𝑝 from the compressor power
𝑃𝑎𝑚𝑏 . The compressor flow 𝑃𝑎𝑚𝑏 determines the airflow rate 𝑊𝑐𝑝 by using the compression ratio and rotation
speed. The revised air flow rate, 𝑊𝑐𝑝_𝑟𝑒 , is then computed using (2).

𝑇𝑎𝑚𝑏 𝑃0
𝑊𝑐𝑝_𝑟𝑒 = × 𝑊𝑐𝑝 (2)
𝑇0 𝑃𝑎𝑚𝑏

With the formulae given in [9], the air temperature exiting the compressor 𝑇𝑐𝑝 is determined.
The hydrogen fuel consumption of PEMFC is determined by (2) [14]. Load power variations and
repetitive start-stop cycles represent two significant challenges that negatively impact the performance and
longevity of fuel cell systems [15]. The identified issues result in substantial complications, such as membrane
drying and water flooding, which may cause irreversible damage to the channel and gas diffusion layers [16].
𝑡 𝑉𝐹𝐶 (𝑡)𝑖𝐹𝐶 (𝑡)
𝑀𝐻2 = ∫0 𝑑𝑡 (3)
𝜂𝑑𝑐 (𝑡)𝜌𝐻2

In the computation model, the reference value of the air flow 𝑊𝑐𝑝∗ is calculated as in [17].

𝑎𝑡𝑚
∗ 𝑀𝑉 ∅𝑎𝑡𝑚 𝑃𝑆𝑎𝑡 1 𝑛𝐼𝑠𝑡
𝑊𝑐𝑝 = [1 + 𝑎𝑡𝑚 ]
𝑎𝑡𝑚 ) 𝑦 𝜆𝑏𝑒𝑠𝑡 𝑀𝑂2 (4)
𝑀𝑎 (𝑃𝑎𝑡𝑚 −∅𝑎𝑡𝑚 𝑃𝑆𝑎𝑡 𝑂2 4𝐹

𝑎𝑡𝑚
Where ∅𝑎𝑡𝑚 is air relative humidity; 𝑃𝑆𝑎𝑡 is the vapor saturation pressure with relative humidity ∅𝑎𝑡𝑚 ; 𝑃𝑎𝑡𝑚
is atmospheric pressure; 𝑀𝑉 is the molar mass of vapor; 𝑀𝑎𝑎𝑡𝑚 a is the air molar mass with relative humidity
∅𝑎𝑡𝑚 ; 𝑀𝑂2 is oxygen molar mass; 𝑛 is the number of single fuel cells; F is Faraday’s constant; and 𝑦𝑂2 is
𝑀𝑂2
oxygen mass fraction in the air, calculated by: 𝑦𝑂2 = 0.21 𝑎𝑡𝑚 .
𝑀𝑎
To model the concentration of oxygen in the cathode, we first define a parameter called the oxygen
excess ratio 𝜆𝑂2 .
𝑊𝑂2_𝑖𝑛
𝜆𝑂2 = (8)
𝑊𝑂2_𝑟𝑐𝑡

Where 𝑊𝑂2_𝑖𝑛 is the flow of oxygen into the cathode and 𝑊𝑂2_𝑟𝑐𝑡 is the mass of oxygen reacted in the cathode;
low values of 𝜆𝑂2 indicate low oxygen concentration in the cathode or oxygen starvation. The rate of oxygen
reacted 𝑊𝑂2_𝑟𝑐𝑡 depends on the current drawn from the stack 𝐼𝑠𝑡 . The mass flow rate out of the supply
manifold 𝑊𝑠𝑚 , depends on the downstream (cathode) pressure.
𝑛𝐼𝑠𝑡
𝑊𝑂2_𝑟𝑐𝑡 = 𝑀𝑂2 (9)
4𝐹

1
𝑊𝑂2_𝑖𝑛 = 𝑦𝑂2 𝑊𝑠𝑚 (10)
1+∅𝑎𝑡𝑚

3. HYBRID FUEL CELL VEHICLE (FCHV)

The vehicle's electrical power system, as seen in Figure 3, comprises the electrical machine, the
secondary power source, and the fuel cell system. To integrate these components into one or several buses, it is
essential to establish an electrical architecture. The electric machine is methodically outfitted with a converter
to facilitate torque management. The converter is reversible in power to give the necessary energy during the
traction phases and to recover the energy provided by the electric machine during the vehicle braking phases.
At the power source level, a primary function of the power converter(s) is to regulate the power
distribution between the PAC system and the auxiliary energy source. Furthermore, its local control often
enables the regulation of current, voltage, and/or power transferred to safeguard the linked devices [18], [19].
Incorporate a variable into the power allocation process: Hybridization enables the distribution of power
demand between the fuel cell system and the auxiliary energy source. This enables the modification of the
operational parameters of the FCS system to optimize performance in higher yield areas through effective
management strategies, consequently decreasing hydrogen consumption. Within the powertrain of a hybrid
fuel cell vehicle, multiple operational modes can be delineated. The operational modes of traction, braking,
and halting generate distinct energy flows.

A comparative analysis of ANFIS and fuzzy controllers for a dynamic hybrid model (Laoufi Kaltoum)
248  ISSN: 2252-8792

Figure 3. Example of hybrid fuel cell vehicle architecture

3.1. The secondary source of energy

The main feature of the secondary energy source is its ability to reverse the power generation process.
The secondary source may be replenished through kinetic energy recovery or by utilizing the fuel cell.
A battery functions as an electrochemical energy converter, enabling the chemical storage of energy. The
primary technologies utilized in hybrid automobiles include lead-acid batteries, nickel metal-hydride (Ni-MH),
and lithium-ion (Li-ion) [20]. Nickel-metal hydride (Ni-MH) technology is widely recognized for its
advantageous characteristics in terms of capacity, lifespan, and economic efficiency. Li-ion technology
demonstrates high specific power (W/kg) and specific capacity (Wh/kg); however, improvements are
necessary regarding cost, operational safety, service life, and low-temperature performance. The primary
technology exhibits a low specific capacity due to the significant weight of the batteries; however, it remains a
robust and economical choice that benefits from ongoing advancements [21], [22].

4. CONTROL STRATEGIES
4.1. Adaptive neuro-fuzzy inference system (ANFIS)
The ANFIS model exemplifies a sophisticated method for enhancing Takagi Sugeno fuzzy inference
systems, which are recognized as the most widely utilized framework in practical applications. The
architecture utilizes multilayer networks, formed by cells that perform designated duties while conforming to
established criteria [18], [19]. To exemplify the fundamental idea of ANFIS, we examine a fuzzy inference
system with two inputs (x, y), and a singular output z, as seen in the schematic in Figure 4. This framework
effectively combines neural network flexibility with fuzzy logic accuracy, providing a robust instrument for
intricate system modeling and optimization.
The parameters of the ANFIS system are classified as premise membership function parameters and
consequent parameters. The learning process employs a hybrid methodology: the gradient descent algorithm
optimizes the premise parameters, while the least-squares approach accurately computes the consequent
parameters. This cooperative learning method enhances the model's accuracy and adaptability. The operational
framework of the ANFIS system may be outlined as follows:

Rule 𝑖: if 𝑥 is 𝐴𝑖 ; and 𝑦 is 𝐵𝑖 ; then 𝑧𝑖 = 𝑝𝑖 𝑥 + 𝑞𝑖 𝑦 + 𝑟𝑖

where 𝐴𝑖 and 𝐵𝑖 represent the fuzzy sets associated with the function node. As in Figure 4, the ANFIS is
composed of five layers.
˗ Layer 1: Each node i in this layer uses a membership function given by: o1i = μAi (x). Where 𝑥 is the input
for node 𝑖 (same for input 𝑦, 𝑜𝑖1 = 𝜇𝐵𝑖 (𝑦)). 𝜇𝐴𝑖 represents the membership function of 𝐴𝑖 and it specifies
the degree of membership with which 𝑥 satisfies. Generally, 𝜇𝐴𝑖 (𝑥) is chosen as a Gaussian function.

(𝑥− 𝑐𝑖 )
𝜇𝐴𝑖 (𝑥) = 𝑒𝑥𝑝 [− ] (11)
𝜎𝑖

With 𝑐𝑖 , 𝜎𝑖 are parameters that refer to the premise parameters. The values of these last changes according
to various exhibitions of the function of belonging.
˗ Layer 2: The outputs of this layer are the weights of the rules, which are obtained by a simple
multiplication of the entries in each cell. multiplication of the entries in each cell.

𝑜𝑖2 = 𝜇𝐴𝑖 (𝑥) × 𝜇𝐵𝑖 (𝑦) (12)

Int J Appl Power Eng, Vol. 14, No. 1, March 2025: 244-254
Int J Appl Power Eng ISSN: 2252-8792  249

˗ Layer 3: This layer corresponds to the normalization of rule weights by the relationship.
wi
o3i = ̅̅̅
wi = , i = 1,2 (13)
w1 +w2

˗ Layer 4: Each node 𝑖 in this layer is a node that is calculated as (14).

𝑜𝑖4 = ̅̅̅
𝑤𝑖 𝑧𝑖 = ̅̅̅(𝑝
𝑤𝑖 𝑖 𝑥 + 𝑞𝑖 𝑦 + 𝑟𝑖 ) (14)

Where ̅̅̅
𝑤𝑖 are the outputs of layer 3 and (𝑝𝑖 , 𝑞𝑖 , 𝑟𝑖 ) are the consequent parameters of the output function.
˗ Layer 5: The single node in this layer sums all the input signals and returns, as output, the approximate
value of the desired function expressed by (15).
∑2
𝑖=1 𝑤𝑖 𝑧𝑖
𝑜𝑖5 = ∑𝑖 ̅̅̅
𝑤𝑖 𝑧𝑖 = ∑2
(15)
𝑖=1 𝑤𝑖

ANFIS design integrates input parameters and generates subsequent parameters, represented as a
linear combination in the output.
𝑤1 𝑤2
𝑧= 𝑧 + 𝑧 = ̅𝑤 𝑤2 2 = (𝑤
̅̅1̅𝑧1 + ̅̅̅̅𝑧 ̅̅̅1̅𝑥)𝑝1 + (𝑤
̅̅̅1̅𝑦)𝑞1 + (𝑤
̅̅̅1̅)𝑟1 +
𝑤1 + 𝑤2 1 𝑤1 + 𝑤2 2
(𝑤
̅̅̅̅𝑥)𝑝
2 2 + (𝑤
̅̅̅̅𝑦)𝑞
2 2 + (𝑤
̅̅̅̅)𝑟
2 2 (16)

With 𝑝𝑖 , 𝑞𝑖 , 𝑎𝑛𝑑 𝑟𝑖 are linear consequential parameters.

The gradient descent method optimizes the premise parameters, while the consequent parameters are
determined using the least-squares technique. This approach ensures the optimal identification of the
resultant parameters, provided the premise parameters stay unchanged. A balance between computational
complexity and performance results dictates the selection of these tactics. Consequently, the hybrid learning
approach outperforms solitary gradient descent in efficacy and facilitates the segmentation of the parameter
set for optimized computation [23].

Figure 4. Architecture of ANFIS controller

5. THE SIMULATION RESULTS AND DISCUSSION

Figure 5 presents a the vehicle's power requirements dynamic hybrid model combining a fuel cell and
battery for electric vehicles to create a multi-domain simulation of a fuel cell vehicle (FCV) drivetrain.
Nevertheless, a commonly employed fuel cell vehicle (FCV) powertrain is exemplified by the Honda
Clarity [24]. The electric motor is energized by a fuel cell and a battery to function the fuel cell vehicle (FCV).
The electric motor, battery, fuel cell, and DC/DC converter are the four components of the FCV electrical
subsystem. Where the electric motor is a 100-kW internal permanent magnet synchronous machine
functioning at a supply voltage of 288 VDC. This motor incorporates integrated magnets and contains eight
poles. Flux vector control is essential to achieve a maximum motor speed of 12,500 rpm.
Parameters Sim Power Systems the fuel cell stack has 400 cells, delivering a peak output of 90 kW.
A supplementary source that provides more power when required is a 13.9 Ah, 288 VDC, 25 kW lithium-ion
battery [24], [25]. We desire clarity in Figure 6 to illustrate the multiple working modes of the FCV during a
whole cycle, namely: acceleration, cruising, recharging the battery during acceleration, and regenerative
braking. The simulated operation time in acceleration mode is roughly one minute. The FCV accelerates from
0 km/h to around 90 km/h in 12 seconds, thereafter decelerating to 80 km/h in 16 seconds. This exceptional
outcome is achieved by maintaining the accelerator pedal at 70% for the initial 4 seconds, followed by 25% for
the subsequent 4 seconds after complete release, and then at 85% when the pedal is re-engaged. Pressure is
applied for 4 seconds, then a brake setting of -70% is set until the end of the exercise.

A comparative analysis of ANFIS and fuzzy controllers for a dynamic hybrid model (Laoufi Kaltoum)
250  ISSN: 2252-8792

Figure 5. Multi-domain simulation of an FCV power train based on Sim Power Systems and Sim Driveline [24]

Figure 6. Specify driving road topology

The following illustrates what occurs when the FCV moves: initially, at time t = 0 seconds, the fuel
cell vehicle (FCV) is at rest, and the driver engages the accelerator pedal to 70% of its capacity, relying on
the battery to power the engine until the fuel cell is activated. Starting at t = 0.725 seconds, the vehicle's
speed gradually increases from 0 km/h to 44 km/h. However, the fuel cell is unable to reach the reference
power because of its large time constant, requiring ongoing battery assistance. At t = 4 seconds, when the
accelerator pedal is released to 25%, the fuel cell is unable to promptly decrease its power output, resulting in
the battery supplying the required torque to sustain a constant speed of 55 km/h. Figure 7 shows the change
in values representing the speed of the car from take-off to arrival, as the graphs show the dynamic response
of the fuel cell and battery system. Figure 7(a) illustrates the fuel cell output power lagging behind demand,
while Figure 7(b) shows the vehicle speed stabilizing at 55 km/h. Figure 7(c) presents a gradual depletion of
the battery state of charge, and Figure 7(d) demonstrates the torque output, highlighting how the battery
supports the fuel cell during changes in accelerator input.
At t = 6 seconds, following a duration of two minutes, the power output of the fuel cell aligns with
the reference power, rendering the battery unnecessary. At t = 8 seconds, the driver accelerates by pressing the
accelerator pedal to 85%, causing the fuel cell system to function at a speed of 86 km/h, with the battery
providing supplementary power. At t = 8.05 seconds, the combined output from the fuel cell and battery is
insufficient to meet the required demand, leading to a torque mismatch, attributed to the fuel cell's response
time. At t = 8.45 seconds, the observed torque aligns with the reference value due to the increase in fuel cell
power, resulting in a decrease in battery power to 9 kW. Figure 8 shows the motor speed. Figure 8(a)
demonstrates the fuel cell power achieving the reference level, whereas Figure 8(b) depicts the stabilization of
vehicle speed at 86 km/h. Figure 8(c) illustrates the reduction in battery power as it supports the fuel cell,
while Figure 8(d) emphasizes the torque output following the reference.

Int J Appl Power Eng, Vol. 14, No. 1, March 2025: 244-254
Int J Appl Power Eng ISSN: 2252-8792  251

Car Speed Drive Torque

90 250

Ref Cr
80 200 PI
44.6
Fuzzy
70 44.4 150 Anfis
44.2
86
Speed (Km/h)

60 100
4.05 4.1 4.15

Cr (Nm)
50 85 50
200
(a)
40 84 (b) 0
80
12 12.5 13 13.5
75 150
70
30 -50 8.2 8.4
65
4.05 4.1 4.15 4.2 50
20 PI -100 40
30
Fuzzy 20
10 -150
Anfis 10
12 12.2 12.4 12.6 12.8
0 -200
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time in sec Time in sec
Flow Rate Air Flow Rate Fuel
2500 500
475
PI 470 PI
450 465
1000 Fuzzy 460 Fuzzy
400 455 Anfis
2000 800
Anfis

Flow Rate Fuel (lpm)

Flow Rate Air (lpm)

4 4.1 4.2 9 10
160 350 250
140 2100
120 200
1500 100
80 2000 300
60 4 4.1 4.2
0.35 0.4 0.45 0.5 0.55 250
(c) 8 9
(d)
1000 200

150
220
500 100 10 200
1000
8 180
50 160
500 6
12 12.5 12.2 12.4 12.6
0 0.12 0.14 0.16
0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time in sec Time in sec

Figure 7. The change in values represents the speed of the car from take-off to arrival:
(a) car speed, (b) drive torque, (c) flow rate air, and (d) flow rate fuel

Stack Consumption Air Stack Consumption Fuel

2500 900
PI PI
Stack Consumption Fuel (lpm)
Stack Consumption Air (lpm)

800
Fuzzy 350 Anfis
2000 1000
Anfis 700 300 Fuzzy
800
600 250
1500
600 7.98 8 8.02 8.04 8.06 8.08
8 8.058.18.15 500

1000 400
(a) (b)
300
500
200 480 450
1100 460
1000 100 440 400
0 1000 420
0 400 350
900 800 380
4 4.1 4.2 12 12.5 4 4.1 4.2 12.2 12.4 12.6
-500 -100
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time in sec Time in sec
4 Power Fuel Cell System (FCS) x 10
4 Power battery
x 10 5
10 4
4 x 10
x 10 PI PI
4
x 10 2.8
5.8
4.4 Fuzzy 4 Fuzzy
4.2 2.7
8 5.6 4 Anfis x 10
4
Anfis
5.4 3.8 1.3 2.6
3.6 3 1.2 8 8.2 8.4
5.2
8 8.028.048.068.08 1.1
6 5 1
4.8 3000 2 0.9
Power (W)

Power (W)

2800
4 4.1 4.2 0.7 0.8
2600
4 1 4
(c) 4
2400 (d) -2
x 10
x 10 2200
-2.2
7 0
14.81515.2 -2.4
2 -2.6
4
4000 6 x 10 -2.8
-1
2000 12 12.5 13
5 -2.2
0
0
4 -2 -2.4
0.68 0.7 0.720.740.76
12 12.5
4.1 4.2 4.3
-2 -3
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time in sec Time in sec

Figure 8. The motor speed shows (a) stack consumption air, (b) stack consumption fuel,
(c) power fuel cell system, and (d) power battery
A comparative analysis of ANFIS and fuzzy controllers for a dynamic hybrid model (Laoufi Kaltoum)
252  ISSN: 2252-8792

At t = 10.9 seconds, the battery's state of charge (SOC) decreases to 39.99%, requiring recharging due
to its fall below the 40% threshold. The fuel cell manages the distribution of energy between the battery and
the engine, enabling the battery to draw a maximum of -24 kW. In this phase, the battery is recharged while
the fuel cell is in operation; however, the necessary torque is not achieved, resulting in a constant motor speed
of 86 km/h until t = 12.8 seconds. Figure 9 shows the power motor of the fuel cell system (FCS). Figure 9(a)
illustrates a decline in state of charge (SOC) below 40%, whereas Figure 9(b) demonstrates a constant vehicle
speed. Figure 9(c) depicts the negative battery power during the recharging process, while Figure 9(d)
demonstrates the torque output that falls short of meeting the required demand.
At t = 12 seconds, the driver modifies the accelerator pedal to -70%, triggering regenerative braking
in which the motor functions as a generator. The kinetic energy of the FCV is converted into electrical energy
stored in the battery; however, the required torque of -140 Nm cannot be attained because the battery's
maximum absorption capacity is limited to 25 kW. The power output of the fuel cell diminishes as a result of
its response time, leading to a gradual reduction in vehicle speed from 75 km/h to a complete stop. By t = 15
seconds, the fuel cell power attains approximately 2.38 kW, indicating the minimum output. Figure 8(c)
illustrates the energy absorption by the battery, Figure 9(a) demonstrates the impact of regenerative braking on
power output, Figure 9(b) presents the reduction in vehicle speed, and Figures 8(d) and 9(d) emphasizes the
decrease in fuel cell power during this process.

Fuel Cell System (FCS) Voltage Fuel Cell System (FCS) current
450 350 130
120 PI
110
400 300 100 Fuzzy
90
Anfis
350 8 8.05 8.1
250 1.5
1
340 0.5
300 0
Voltage (V)

Current (A)

338 200 -0.5

345 0.71 0.72 0.73
250 336
334 340
150
(a) 200
332 335
330
(b)
4 4.2 325 100
150 419
358 12.2 12.4 12.6 12.8
418.5 356
354 50 160
100 160
418 352 PI
350 140
50 417.5 348 Fuzzy 0 140 120
346 12.2 12.4 12.6
0.7 0.72
Anfis 4 4.1 4.2
8 8.05 8.1 8.15
0 -50
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time in sec Time in sec
x 10
4 Power Motor State of charge
12 4
40.8
x 10 PI 40.116 40
7.2 40.7 40.114
10 7 Fuzzy 40.112
6.8 39.995
6.6 40.11
6.4
Anfis 40.6 40.108
8 6.2 40.106 39.99
3.95 4 4.05 9.8 10 10.2 10.4
8.028.048.068.088.1 40.5 40.186
6
Power (W)

40.184
SOC %

40.4
4 40.182
(c) (d) 40.3 40.18
4
2 4 x 10 7.95 8 8.05
x 10
3 4 40.2
0
2 40.1 PI
2.5
-2 0 Fuzzy
40
4.05 4.1 4.15 4.2 12 12.5 13 Anfis
-4
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Time in sec Time in sec

Figure 9: The power motor of (a) FCS voltage, (b) FCS current, (c) power motor, and (d) state of charge

6. CONCLUSION
This paper presents a control-oriented model for proton exchange membrane (PEM) fuel cells,
incorporating cathode and anode mass flow transients with a membrane hydration model. The validation of the
proposed model required comprehensive testing, resulting in findings that exhibited a high degree of accuracy.
Employing this control-centric paradigm, an adaptive neuro-fuzzy inference system (ANFIS) is employed to
regulate increased oxygen levels, preventing oxygen deficits and hydrogen flow disruptions during sudden
variations in current load inside the fuel cell system.
ANFIS exhibits superiority over conventional PID controllers by amalgamating the benefits of fuzzy
logic controllers with those of PID controllers. Comparative simulations demonstrate the efficacy of ANFIS in

Int J Appl Power Eng, Vol. 14, No. 1, March 2025: 244-254
Int J Appl Power Eng ISSN: 2252-8792  253

managing hydrogen and airflow in PEM fuel cells compared to traditional PID controllers. Due to the
sensitivity of fuzzy logic control to system features and load variations, PID parameters are carefully
calibrated to enhance system control through rule inference, hence ensuring clarity and precision in
control objectives.
Given the critical role of air and hydrogen flow in system performance, ANFIS is an indispensable
tool for enhancing control precision. The integration of ANFIS significantly improves control methodologies,
leading to superior system performance. In the future, the incorporation of photovoltaic panels in electric
vehicles, augmented by supercapacitors for increased electric charge, alongside a sophisticated energy
management system employing artificial intelligence, presents the potential for enhanced vehicle efficiency
and operational efficacy.

REFERENCES
[1] H. F. Gharibeh, A. S. Yazdankhah, and M. R. Azizian, “Energy management of fuel cell electric vehicles based on working
condition identification of energy storage systems, vehicle driving performance, and dynamic power factor,” Journal of Energy
Storage, vol. 31, p. 101760, Oct. 2020, doi: 10.1016/j.est.2020.101760.
[2] A. U. Rahman, I. Ahmad, and A. S. Malik, “Variable structure-based control of fuel cell-supercapacitor-battery based hybrid
electric vehicle,” Journal of Energy Storage, vol. 29, p. 101365, Jun. 2020, doi: 10.1016/j.est.2020.101365.
[3] H. F. Gharibeh, A. S. Yazdankhah, M. Reza Azizian, M. Farrokhifar, and A. Alahyari, “Impact assessment of installed
photovoltaic on fuel cell hybrid electric vehicle considering fast identification of power changes,” in 2020 International Youth
Conference on Radio Electronics, Electrical and Power Engineering (REEPE), IEEE, Mar. 2020, pp. 1–6, doi:
10.1109/REEPE49198.2020.9059152.
[4] A. D. J. Larminie, Fuel Cell Systems Explained, 2nd ed. Hoboken, NJ, USA: John Wiley & Sons, 2003.
[5] P. H. Rodtaz, “Dynamics of the polymer electrolyte fuel cell: experiments and model-based analysis,” Ph.D. dissertation, Swiss
Federal Institute of Technology, Zurich, Switzerland, 2003.
[6] B. Gou, W. Na, and B. Diong, Fuel cells: Modeling, control, and applications, Boca Raton, FL, USA: CRC Press, 2010.
[7] Y.-P. Yang, F.-C. Wang, H.-P. Chang, Y.-W. Ma, and B.-J. Weng, “Low power proton exchange membrane fuel cell system
identification and adaptive control,” Journal of Power Sources, vol. 164, no. 2, pp. 761–771, Feb. 2007, doi:
10.1016/j.jpowsour.2006.11.038.
[8] J. T. Pukrushpan, A. G. Stefanopoulou, and H. Peng, “Control of fuel cell breathing,” IEEE Control Systems, vol. 24, no. 2, pp.
30–46, Apr. 2004, doi: 10.1109/MCS.2004.1275430.
[9] J. T. Pukrushpan, H. Peng, and A. G. Stefanopoulou, “Simulation and analysis of transient fuel cell system performance based on
a dynamic reactant flow model,” in ASME International Mechanical Engineering Congress and Exposition, Jan. 2002, pp. 637–
648, doi: 10.1115/IMECE2002-32051.
[10] W. Garcia-Gabin, F. Dorado, and C. Bordons, “Real-time implementation of a sliding mode controller for air supply on a PEM
fuel cell,” Journal of Process Control, vol. 20, no. 3, pp. 325–336, Mar. 2010, doi: 10.1016/j.jprocont.2009.11.006.
[11] A. Guo, W. Chen, Q. Li, Z. Liu, and H. Que, “Air flow control based on optimal oxygen excess ratio in fuel cells for vehicles,”
Journal of Modern Transportation, vol. 21, no. 2, pp. 79–85, Jun. 2013, doi: 10.1007/s40534-013-0009-8.
[12] T. C. Ma, “Research on the control of fuel cells engine system analysing and modeling,” Ph.D. dissertation, Tongji University,
Shanghai, China, 2007.
[13] D. Zhou, A. Al-Durra, F. Gao, A. Ravey, I. Matraji, and M. G. Simões, “Online energy management strategy of fuel cell hybrid
electric vehicles based on data fusion approach,” Journal of Power Sources, vol. 366, pp. 278–291, Oct. 2017, doi:
10.1016/j.jpowsour.2017.08.107.
[14] R. Touileb and A. Abbou, “Evaluation of energy management system of a hybrid energy source in EV,” International Journal of
Power Electronics and Drive Systems (IJPEDS), vol. 14, no. 4, pp. 1911–1918, 2023, doi: 10.11591/ijpeds.v14.i4.pp1911-1918.
[15] L. Xu, C. D. Mueller, J. Li, M. Ouyang, and Z. Hu, “Multi-objective component sizing based on optimal energy management
strategy of fuel cell electric vehicles,” Applied Energy, vol. 157, pp. 664–674, Nov. 2015, doi: 10.1016/j.apenergy.2015.02.017.
[16] P. Thounthong and P. Sethakul, “Analysis of a Fuel Starvation Phenomenon of a PEM Fuel Cell,” in 2007 Power Conversion
Conference - Nagoya, IEEE, Apr. 2007, pp. 731–738, doi: 10.1109/PCCON.2007.373048.
[17] R. Saïsset, C. Turpin, S. Astier, and J. M. Blaquière, “Electricity generation starting from a fuel cell hybridised with a storage
device,” in IEEE Vehicle Power and Propulsion Conference, Paris, 2004.
[18] H. Otmane, Y. Mouloudi, and B. Mokhtar, “Comparative analysis of cascaded fuzzy-PI controllers based-MPPT and perturb and
observe MPPT in a grid-connected PV system operating under different weather and loading conditions,” International Journal of
Power Electronics and Drive Systems (IJPEDS), vol. 10, no. 4, pp. 1986–1994, 2019, doi: 10.11591/ijpeds.v10.i4.pp1986-1994.
[19] R. Sakthivelsamy and K. Subramaniyan, “Modelling and performance analysis of free body dynamics of electric vehicles,”
International Journal of Power Electronics and Drive Systems (IJPEDS), vol. 15, no. 1, pp. 1–7, Mar. 2024, doi:
10.11591/ijpeds.v15.i1.pp1-7.
[20] M. Nizam and M. R. A. Putra, “Analysis of lithium-ion indirect liquid cooling battery thermal management system with high
discharge rate,” International Journal of Power Electronics and Drive Systems (IJPEDS), vol. 14, no. 3, pp. 1414–1420, Sep.
2023, doi: 10.11591/ijpeds.v14.i3.pp1414-1420.
[21] R. Touileb and A. Abbou, “Simulation and optimization of hydrogen consumption in a fuel cell/battery hybrid vehicle,”
International Journal of Power Electronics and Drive Systems (IJPEDS), vol. 14, no. 2, pp. 662–672, Jun. 2023, doi:
10.11591/ijpeds.v14.i2.pp662-672.
[22] T. Sutikno and S. Padmanaban, “Integrated motor drive for vehicle electrification: a step toward a more sustainable and efficient
transportation system,” International Journal of Power Electronics and Drive Systems (IJPEDS), vol. 14, no. 2, pp. 649–652, Jun.
2023, doi: 10.11591/ijpeds.v14.i2.pp649-652.
[23] W. Zhou, L. Yang, Y. Cai, and T. Ying, “Dynamic programming for new energy vehicles based on their work modes Part II: Fuel
cell electric vehicles,” Journal of Power Sources, vol. 407, pp. 92–104, Dec. 2018, doi: 10.1016/j.jpowsour.2018.10.048.

A comparative analysis of ANFIS and fuzzy controllers for a dynamic hybrid model (Laoufi Kaltoum)
254  ISSN: 2252-8792

[24] P. Mercier, “Fuel Cell Vehicle (FCV) power train,” MATLAB Central File Exchange, 2024, [Online]. Available:
https://www.mathworks.com/matlabcentral/fileexchange/33309-fuel-cell-vehicle-fcv-power-train.
[25] O. Tremblay and L.-A. Dessaint, “Experimental validation of a battery dynamic model for EV applications,” World Electric
Vehicle Journal, vol. 3, no. 2, pp. 289–298, Jun. 2009, doi: 10.3390/wevj3020289.

BIOGRAPHIES OF AUTHORS

Laoufi Keltoum obtained the L.Sc. degree in electrical engineering from the
University of Oran, Algeria, in 1995, and the M.S. degree in 2004 from the University of Oran,
Algeria, currently a doctoral student of the Faculty of Technology, University from Béchar,
Algeria. She has been teaching at Bechar University since 2004. Her main research activities
focus on FACTS technology, control of electrical systems, and optimal energy distribution and
stability improvement using power electronics in the framework of constraint compensation
and energy optimization. She can be contacted at email: [email protected] or
[email protected].

Youssef Mouloudi received the state engineer degree in electrical engineering in

2006 from the University of Bechar and the M.S. degree in 2009 from Bechar University,
Algeria. He is currently a Ph.D. student of the Faculty Technology, University of Bechar,
Algeria. He is working at SONELGAZ "Algerian Electrical Society" since 1995. His areas of
interest are FACTS systems, power filters, applications of power electronics, and stability
improvement. Correspondence address: Department Science and Technology "Electrical
Engineering", Faculty of Technology, Bechar University, BP 417, 08000 Bechar, Algeria. He
can be contacted at email: [email protected] or [email protected].

Abdeldjebar Hazzab secured a state engineering degree in electrical engineering

from the University of Sciences and Technology of Oran (USTO), Algeria, in 1995. He
furthered his academic pursuits at the Electrical Engineering Institute of USTO, obtaining a
Master of Science degree in 1999, followed by a doctoral degree (Ph.D.) from the same
institution in 2006. Currently, he serves as a professor of electrical engineering at the
University of Bechar, Algeria, where he serves as the Director of the Research Laboratory of
Control Analysis and Optimization of Electro-Energetic Systems. Additionally, he is an
adjacent Professor at École de Technologie Supérieure (ETS), Université de Montréal, Canada,
and a researcher at the energy intelligence research and innovation center - Cégep de Sept-Îles.
His research endeavors encompass the domains of power electronics, electric drive control,
and artificial intelligence, with a particular emphasis on their practical applications. He can be
contacted at email: [email protected] or [email protected].

Abdallah Ben Abdelkader received the L.Sc. degree in electrical engineering

from Bechar University, Bechar, Algeria, in 2006 and M.Sc. degree in electrical engineering
from University of Sciences and Technology of Oran (USTO) Algeria, in 2013, respectively,
where he has been working toward the doctoral degree in the Department of Electrical
Engineering, 2013. He is currently a research member of Smart Grids and Renewable Energies
Lab, Department of Electrical Engineering, Tahri Mohammed University Bechar, Algeria.
His main research activity is focused on FACTS technology, power system control, and
optimal power dispatch. Faculty of Technology, Bechar University, BP 417, 08000
Bechar, Algeria. He can be contacted at email: [email protected] or
[email protected].

Int J Appl Power Eng, Vol. 14, No. 1, March 2025: 244-254