Energy Reviews 3 (2024) 100054

Contents lists available at ScienceDirect

Energy Reviews
journal homepage: www.journals.elsevier.com/energy-reviews

Full Review Article

A systematic review of system modeling and control strategy of proton

exchange membrane fuel cell
Yujie Wang *, Xingliang Yang, Zhengdong Sun, Zonghai Chen
Department of Automation, University of Science and Technology of China, Hefei, Anhui, 230027, PR China

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

Keywords: The proton exchange membrane fuel cell, as a novel energy device, exhibits a wide array of potential applications.
Proton exchange membrane fuel cell This paper offers a comprehensive review and discussion of modeling and control strategies for fuel cell systems.
System modeling and Control, System modeling It commences with a concise introduction to the structure and principles of fuel cells. Subsequently, it outlines
and control
modeling approaches for various fuel cell subsystems, encompassing the fuel cell stack, air supply system,
Water and thermal management
hydrogen supply system, thermal management system, and water management system. Following this, it conducts
a comparative analysis and discussion of prevalent control strategies for the aforementioned subsystems. Lastly,
the paper outlines future research trends and directions in the modeling and control strategies of fuel cells. The
aim of this paper is to provide ideas and inspirations for the design and management of membrane fuel cell
systems from control aspects.

1. Introduction and the anode, and the electrons are also conducted by the external
circuit to the cathode. In the presence of the cathode catalyst [6],
As an emerging energy device, the fuel cell technology has hydrogen ions, electrons, and oxygen are combined to produce water
outstanding advantages such as high efficiency, no pollution, and fast with heat emission [7].
charging, which can meet the energy demand in many fields such as When PEMFC generates electricity, the following chemical reaction
power systems for transportation, aerospace, and marine. Fuel cells can equation occurs internally.
be classified according to the type of electrolyte [1] as phosphate fuel Anode:
cells, molten carbonate fuel cells, alkaline fuel cells, proton exchange
membrane fuel cells (PEMFCs), and solid oxide fuel cells. Among them, H2 → 2Hþ þ2e
the PEMFC [2] can be a potential alternative power system to replace the Cathode:
internal combustion engine. However, the PEMFC system involves
various aspects such as thermodynamics, electrochemistry, and fluid 1
O2 þ2Hþ þ2e → H2 O
dynamics, and the system is affected by various aspects such as tem- 2
perature, humidity, and fuel supply, and these influences are coupled Total reaction formula:
with each other, so the PEMFC system is a nonlinear and complex dy-
namic system [3], which poses a challenge for the design of PEMFC. The 2H2 þ O2 → 2H2 O
following is a brief description of the working principle of the fuel cell.
The chemical reaction equations shown in Fig. 1 show that the
The PEMFC is a chemical power generation device that consists of
chemical product of PEMFC is only water and does not cause pollution to
four main parts: cathode, anode, electrolyte, and external circuit [4].
the environment. However, it is worth noting that the water generated by
When generating electricity, hydrogen and oxygen enter the anode [5]
the PEMFC system needs to be removed in time to avoid flooding the
and cathode of PEMFC respectively. The hydrogen gas loses electrons at
PEMFC. In addition, the chemical reaction exerts heat, which may lead to
the anode and becomes hydrogen ions, which enter the electrolyte so-
high temperature of the PEMFC stack if it is not cooled down in time, thus
lution and cross the proton exchange membrane to the cathode under the
damaging the PEMFC system.
electric field generated by the voltage difference between the cathode
Without a safe and reliable control and management strategy to

* Corresponding author.
E-mail address: [email protected] (Y. Wang).

https://doi.org/10.1016/j.enrev.2023.100054
Received 12 September 2023; Received in revised form 19 October 2023; Accepted 2 November 2023
2772-9702/© 2023 The Authors. Published by Elsevier Ltd on behalf of Shenzhen City Clean Energy Research Institute, Shenzhen University. This is an open access
article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Y. Wang et al. Energy Reviews 3 (2024) 100054

Fig. 1. PEMFC stack reaction principle.

maintain the stable operation of the PEMFC, the PEMFC may face many the paper also outlines research trends and hot topics, with the hope of
problems such as flooding, dehydration, oxygen starvation, etc. These guiding researchers in grasping cutting-edge technologies and identi-
problems may lead to a shortened PEMFC lifetime even damage to the fying research directions. Ultimately, it aspires to make valuable con-
proton exchange embrane [8,9]. tributions to the development of fuel cells. The rest of the paper is
Indeed, there are several pressing issues that need to be addressed in organized as follows: Section 2 gives a brief introduction to the working
the current stage of fuel cell system development. Firstly, within the in- principle and overall structure of the PEMFC. Section 3 introduces the
ternal structure of a fuel cell system, various subsystems coexist, current modeling methods for different fuel cell subsystems including the
including the fuel cell stack, air supply subsystem, hydrogen supply fuel cell stack, air supply subsystem, hydrogen supply subsystem, thermal
subsystem, among others [10]. These subsystems exhibit intricate in- management subsystem, and water management subsystem. Section 4
terdependencies and coupling effects. Effectively characterizing and introduces the current popular control strategies for the PEMFC. Section
managing these interactions represents a pivotal challenge. Secondly, the 5 summarizes the current research progress on PEMFC and looks forward
modeling of fuel cell systems often necessitates a delicate balance be- to future directions.
tween precision and complexity [11]. While accurate models can provide
a more faithful representation of system behavior, they may also entail 2. System architecture of PEMFC
heightened computational burdens [12]. Striking the right balance be-
tween model accuracy and complexity is a critical issue. Furthermore, The PEMFC is a chemical device that uses hydrogen and oxygen to
real-world operation of fuel cell systems introduces dynamic load vari- generate electricity. Compared to the conventional combustion method
ations. Ensuring swift system responses to these load changes for stable of utilizing hydrogen energy, the efficiency of hydrogen energy utiliza-
operation constitutes an urgent concern [13]. With the advancement of tion is significantly increased. The trade-off is high-performance PEMFC
data acquisition and processing capabilities, harnessing big data tech- reactor and its auxiliary systems. The following is a brief description of
nologies for fuel cell system modeling and control to enhance system the overall structure of the PEMFC reactor and its balance of plants. The
performance and stability is now an imperative issue. In practical ap- balance of plants includes: hydrogen supply system, air supply system,
plications, fuel cell system control demands high levels of real-time temperature management system and water management system [14].
responsiveness and reliability. Guaranteeing that control algorithms When the PEMFC is operating normally, the systems coordinate with
can operate stably and reliably in real-time scenarios is a pressing chal- each other to ensure that the PEMFC can provide stable and controllable
lenge. Lastly, fuel cell systems may face various fault scenarios, such as voltage and current to the dynamically changing load.
component failures or sensor malfunctions. Effectively employing fault
diagnosis techniques and implementing fault-tolerant control strategies 2.1. Stack system
to ensure the safe and stable operation of the system under faulty con-
ditions requires immediate attention. The power stack is the core component of PEMFC. The output voltage
This paper provides a comprehensive overview of the latest ad- of a single cell is generally 0.6 V. To meet the actual working re-
vancements in the field of fuel cell system modeling and control by sys- quirements, several cells need to be connected in series to become a
tematically summarizing existing research achievements, including power stack to obtain the required voltage [15]. The chemical reaction
modeling approaches for various subsystems and their corresponding inside the single cell to generate electrical energy is accompanied by the
control strategies. It aims to offer relevant researchers a comprehensive generation of water and the release of heat [16]. How to deal with the
understanding of the current state of research in this field. Additionally, water and heat generated by the chemical reaction is a problem that must

2
Y. Wang et al. Energy Reviews 3 (2024) 100054

be solved for the PEMFC to operate stably and efficiently. reducing the life of the battery. Insufficient water content within the
battery induces dehydration of the proton exchange membrane. This
2.2. Air supply system gives rise to elevated contact and ohmic resistances, subsequently
reducing proton conductivity. This, in turn, leads to heightened heat
The air (oxygen) supply system, mainly includes components such as release and diminished fuel cell efficiency. The amplified heat release
air compressor, inlet humidifier, and pressure-reducing valve [17]. Since may result in localized temperature elevation, further exacerbating
the oxygen involved in the PEMFC anode reaction is obtained directly membrane dryness during the irreversible degradation process. The hu-
from the air, the air needs to be purified. The purified air is compressed midity of the proton exchange membrane is directly related to the per-
into the air compressor so that the air flow rate and pressure can meet the formance of the PEMFC [26], and therefore the humidity of the PEMFC
demand of PEMFC reaction. How to control the air flow rate and pressure needs to be managed.
supplied by the air compressor according to the load change has been a
hot issue studied by scholars [18]. 2.5. Thermal management system
In actual operation, fuel cell system needs to respond quickly to load
changes. However, the air has to pass through a compressor and a The thermal management [27] system, as a subsystem, plays an
transmission pipeline before it reaches the cathode to participate in the important role in maintaining the stable operation of the PEMFC. The
reaction. There is a long transmission delay. If the cathode oxygen cannot energy conversion efficiency of the PEMFC is generally in the range of
be replenished in time, oxygen starvation is likely to occur. To solve this 40%-50% [28,29], and the rest of the energy is generally lost in the form
problem, some scholars have proposed the method of oversupplying of heat. The operating temperature of a fuel cell is approximately 80
oxygen [19] to alleviate the oxygen starvation phenomenon. However, Celsius, which presents more significant challenges in thermal manage-
excessive oxygen supply will increase the power of components thus ment compared to internal combustion engines that typically operate at
reduce the net output power of PEMFC. Therefore, how to design control temperatures around 400 Celsius. Waste heat in PEMFCs arises from
strategies to stabilize the oxygen excess ratio in a suitable range is the various mechanisms, encompassing entropic heating from electro-
focus of scholars' research. chemical reactions, Joule heating due to ohmic resistance, heat associ-
ated with water adsorption and desorption, as well as phase transitions of
2.3. Hydrogen supply system water, such as evaporation and condensation [23,30]. The entropic heat
arises from the entropy change in the electrochemical reactions, which
The hydrogen supply system consists of a hydrogen source, a may account for about 30% of the total waste heat generation, expressed
pressure-reducing valve, a pressure regulating valve, a gas-water sepa- by:
rator, and a hydrogen circulation pump [20]. The hydrogen source is
often some tanks of hydrogen, and the pressure is too high to enter the dU0
qrew ¼ j  T
PEMFC anode directly without pressure reduction. The pressure regu- dT
lating valve needs to adjust the flow and pressure of hydrogen gas sup-
where j is the transfer current density, T temperature, and U0 the equi-
plied to the anode in real-time according to the PEMFC load variation.
librium potential.
Due to the limitations of chemical reaction conversion rate and load
The irreversibility of the electrochemical reactions due to presence of
dynamics, the PEMFC tends to keep the hydrogen on the anode side in
the overpotential in the CLs contributes a major portion of heat genera-
excess during the actual operation. According to the current treatment
tion, which may account for 60% of total heating. In the cathode CL, a
methods for excess hydrogen, the PEMFC hydrogen supply structure can
much larger overpotential η is present due to the sluggish ORR, thus this
be divided into the following types: direct discharge structure, dead-end
type of heating mainly arises in the cathode:
structure, and circulation structure.
qirrev ¼ j  η
2.4. Water management system
This is because many processes including electrochemical reaction
There are three water transport modes in PEMFC [21–23]: kinetics, water condensation and diffusion, evaporation rates, and
transport of reacting gases through porous media are closely related to
(1) Electromigration: the hydrated hydrogen ion H2O þ moves from internal temperature [31]. If not discharged in a timely manner, can lead
the anode to the cathode through the ionomer and PEM with the to high membrane temperature of the PEMFC proton exchange mem-
progress of the electrochemical reaction. brane [32]. Excess air will take away some water and heat, but most of
(2) Diffusion: movement of water under the action of the concentra- the heat needs to be discharged by the cooling system. Currently, for
tion gradient. low-power PEMFC, airflow cooling is convenient and fast, however, for
(3) Hydraulic permeation: the permeation of water driven by the high-power PEMFC cooling fluid is the mainstream cooling means.
capillary pressure.
3. System modeling
Water management [21] is one of the important issues to be consid-
ered for PEMFC. Firstly, PEMFC generates water during operation. Sec- As a nonlinear and strongly coupled industrial system, the output
ondly, in order to obtain higher cell output performance, especially under voltage of PEMFC is affected by various factors such as fuel supply
high temperature operating conditions. It is often necessary to humidify pressure, temperature, operating environment temperature, and its own
the reaction gas input to the PEMFC to provide a good working envi- health status. The description of the behavior of PEMFC during operation
ronment for the fuel cell. The main goal of water management is to has been a hot and difficult problem.
maintain a uniform water distribution and to maintain a certain water
content in the PEM to avoid dehydration and flooding problems. The 3.1. Stack modeling
degree of hydration of the proton exchange membrane determines its
performance and durability [24]. When the water content inside the cell There are various models of PEMFC electric stacks, among which the
is too high, the reaction product water will condense in the pores in the mechanism model, empirical model, and data-driven model are the most
cathode gas diffusion layer and block the flow channel, causing flooding common ones. Among them, the mechanistic model is built entirely
on the cathode side of the PEMFC. Water flooding [25] can even cause based on the mathematical description of the physical relationships
the reverse pole of the battery, producing irreversible losses and greatly within the system, the empirical model describes the system

3
Y. Wang et al. Energy Reviews 3 (2024) 100054

characteristics based on experimental data and empirical formulas, and optimization, and determination of operating conditions for fuel cells.
the data-driven model is modeled using intelligent algorithms to identify Fluid dynamics models can be used to simulate the flow of gases such
the system parameters, as shown in Fig. 2. as hydrogen and oxygen in the channels and pores of fuel cells. By
analyzing gas distribution and pressure drops, the design of channel
3.1.1. Mechanistic model structures can be optimized to ensure uniform gas supply to the reaction
The mechanism model mainly models the material flow and energy area. It is possible to simulate the transport, evaporation, and conden-
changes inside the PEMFC, and establishes precise mathematical re- sation of water in fuel cells. This is crucial for water management and can
lationships between the inputs and outputs of the relevant parameters. prevent water blockage in the channels or affecting the conductivity of
The mechanism model is usually a mathematical model based on suitable the electrolyte membrane. The thermal management of fuel cells is
assumptions using basic conservation laws, such as the fluid dynamics closely related to fluid dynamics, and models can be used to analyze
equations, heat transfer equations, electrochemical reaction equations, thermal phenomena such as conduction, convection, and radiation in fuel
etc. cells, assisting in the design of an effective thermal management system.
By simulating and predicting the electrochemical reaction processes, Additionally, combining the three types of equations mentioned above
it is possible to assess the performance of fuel cells, including indicators can determine the optimal operating conditions, including temperature,
such as voltage output and energy conversion efficiency. This aids in pressure, flow rate, and other parameters, thereby enhancing the effi-
optimizing battery design to achieve higher performance and efficiency. ciency and stability of fuel cells. Usually, multiple mathematical ex-
Through simulating the reaction process, the performance of electrode pressions are established, and the solution requires a joint solution,
and electrolyte materials, including conductivity, activity, and stability, which is more complicated and suitable for the internal description of
can be evaluated, contributing to the development and improvement of PEMFC. The mechanism model can be classified into the one-dimensional
fuel cell materials. Additionally, by simulating electrochemical reaction model, two-dimensional model, and three-dimensional model according
processes, factors that may lead to fuel cell failures can be identified, and to the spatial dimension. Bernardi et al. [33] were the first to develop a
corresponding fault-tolerant control strategies can be designed to one-dimensional mechanistic model of PEMFC. The model only considers
enhance system reliability. Overall, electrochemical reaction models the gas motion and material exchange processes in the direction of the
provide an essential theoretical foundation and tool for the design, proton crossing the proton exchange membrane, the model is relatively
optimization, material selection, and determination of operating condi- straightforward, yet its accuracy is somewhat lacking. Amphlett et al.
tions for fuel cells. [34] used the Stefan-Naxwell equation to describe the material transfer
Thermodynamic models can help evaluate the energy conversion ef- process inside the cell, the Nemst-Planck equation to define the internal
ficiency of fuel cells and understand the performance of fuel cells under resistance of the PEMFC, the Nernst equation to calculate the thermo-
different operating conditions. By simulating the thermodynamic char- dynamic equilibrium electric potential of the PEMFC, and the Tafel
acteristics under different operating parameters, designs can be opti- equation to calculate the activation over-voltage. Further, Mann et al.
mized to enhance efficiency. Thermodynamic models can be used to [35] extended the PEMFC model to a general steady-state model appli-
assess the thermal characteristics of fuel cells under different operating cable to different properties and sizes. Such models describe the material
conditions. This is crucial for designing an effective thermal management exchange and energy transfer between cathodes and anodes with simple
system to maintain the appropriate temperature. Thermodynamic structure and small calculations but ignore the material exchange in
models can be employed to study the thermal coupling effects between other directions, so they do not fully and accurately reflect the parameter
different components, providing a better understanding of the overall changes inside the cell.
thermal characteristics of the fuel cell system. This is crucial for system- Based on the one-dimensional model, a two-dimensional model of the
level integration and optimization. Overall, thermodynamic models offer PEMFC can be developed by considering the variation of physicochem-
an important theoretical foundation and tool for the design, ical parameters perpendicular to the direction of the gas flow. Nguyen
et al. [36] first proposed to develop a two-dimensional model of a
steady-state PEMFC consisting of a channel with two stages. Dannenberg
et al. [37] used the Stefan-Maxwell equation and other mathematical
formulations to construct a mathematical model of mass and energy
transfer over the two-dimensional flow channel inside the PEMFC. Bapat
et al. [38] used a two-dimensional two-phase model to investigate the
anisotropy of the intracell resistivity in relation to the PEMFC operating
current density and the relationship between the cell temperature dis-
tribution. Although the two-dimensional model is more complex than the
one-dimensional model, the overall computational effort is not too large,
and it can reflect the reaction process inside the PEMFC reactor more
realistically than the one-dimensional model, which has been widely
studied by scholars.
The three-dimensional model adds the modeling of the material
transfer process between the flow channel and the diffusion layer to the
two-dimensional model and provides a more detailed and comprehensive
description of the internal working mechanism of the cell, which makes
the calculation results more accurate. Shan et al. [39] established a
three-dimensional dynamic model of PEMFC for the study of water and
heat generated during the operation of PEMFC and its proper manage-
ment. Yong Zhang et al. [40] designed a 3D dot matrix model to simulate
the gas distribution and cooling flow channels, which is not limited to the
single channel scale and takes into account the gas distribution as well as
the cooling flow field effects, etc. Dutta et al. [41] proposed a 3D model
based on s flow channels, which takes into account the effects of ex-
change film thickness, stack voltage, and water flow direction on the
Fig. 2. Typical modeling methods for fuel cell stack. current. The 3D model can reflect the operating characteristics of the

4
Y. Wang et al. Energy Reviews 3 (2024) 100054

PEMFC more realistically, but its complex structure increases the electron transfer number for the reaction. F is the Faraday constant. cin is
computational effort and is not suitable for the design of control the oxidant concentration at the inlet end. cout is the oxidant concentra-
algorithms. tion at the outlet.
The mechanism model can accurately reflect the internal physico- Some empirical models are based entirely on data and empirical
chemical process of PEMFC, and can also show the internal temperature formulas and no longer analyze the internal mechanism of the PEMFC.
distribution. However, its derivation process is complicated, and the Lee et al. [50] developed an empirical model of a PEMFC stack using
model parameters are more and most of them are not easy to obtain, so it numerical simulation of the electrochemical performance. Further,
is suitable for theoretical research but not for engineering applications. Squadrito et al. [51] derived an empirical formula for the PEMFC electric
potential versus current density based on experimental data over the
3.1.2. Empirical model entire current range, thus establishing a mathematical expression for the
The empirical model is characterized by the use of empirical formulas relationship between electric potential and current density. Further, D.
instead of complex formulas, based on the analysis of mechanistic model. Hao et al. [52] analyzed the empirical model of the PEMFC polarization
And it is called semi-empirical model if only some of the formulas are curve and proposed an improved model. In addition, Kim et al. [53]
used empirical formulas. TrucNT et al. [42] developed a semi-empirical proposed an empirical model to fit the PEMFC potential and current
model for the numerical simulation of PEMFC using a partial differen- density data.
tial equation solver, which in turn determined the water content distri- The empirical model usually has fewer parameters, a relatively simple
bution, temperature distribution and permeation of gaseous substances structure, and a much reduced computational effort. So it is often applied
within the membrane. Saygili Y et al. [43] developed a semi-empirical to the design of various system control algorithms.
model of the PEMFC thermal management system, with model parame-
ters identified by a series of dynamic tests. The semi-empirical model 3.1.3. Data-driven model
loses the property of fully reflecting the internal mechanism of the Data-driven modeling approaches do not have explicit equations, but
PEMFC, but its simplified model is significantly less computationally rather black-box models are infinitely close to the test situation obtained
intensive and well-suited for controller design. Here are some widely by learning a large amount of test data through multiple training.
used semi-empirical models. The cell voltage can be obtained by sub- Commonly used data-driven methods include algorithms such as neural
tracting activation, ohmic and concentration overpotentials terms from networks, support vector machines, and fuzzy logic.
the reversible potential [44]: Models based on mechanistic and empirical formulations often
require some reasonable assumptions and approximations to complex
Vcell ¼ Ener  ηact  ηohm  ηconc physicochemical processes. Therefore the computational results based on
The reversible potential is calculated with the Nernst equation: assumptions and approximations to build up a model may not match the
structure of the actual measured data of PEMFC. To develop accurate
"   #
0:5 PEMFC models to obtain PEMFC polarization curves, optimization of
0 ΔS0 RT pO2 pH2
Ener ¼ Ener þ ðT  T0 Þ þ ln PEMFC models using optimization techniques has become an alternative
2F 2F p0 p0
means. Manikandan et al. [54] compared several optimization algo-
rithms, such as simulated annealing, particle swarm optimization, and
with the standard conditions p0 ¼ 1 atm, T0 ¼ 298:15 K, Ener
0
¼ 1:229 V
-1 -1 genetic algorithms, and finally identified HBMO as the best optimization
and ΔS ¼ 165 Jmol K .
0
algorithm. Ma et al. [55] used deep learning neural networks proposed a
The activation overpotential is generally represented by a Butler-
PEMFC aging prediction model, which can be effectively used for the
–Volmer formulation [45]:
PEMFC lifetime prediction problem. Han et al. [56] combined support
     vector machines and empirical formulations to propose a hybrid model
2αF ηact 2ð1  αÞF ηact
i ¼ i0 exp   exp  for predicting the polarization curves of PEMFCs under various operating
RT RT
conditions. Further, Qi Li et al. [57] used an adaptive particle swarm
or the simpler Tafel equation [46], valid for large values of ηact and algorithm to optimize the hyperparameters of the support vector ma-
negligible mass transfer resistance, given by: chine model. Simulation results show that the optimized model can
  accurately predict the dynamic performance of output voltage and
RT i power.
ηact ¼ ln
nF i0 However, it is worth mentioning that the data-driven models are
mostly black-box models, the model construction process requires a large
The exchange current density i0 can be a constant or a function of the
amount of experimental data for training and testing, so the experimental
temperature and the oxygen concentration.
cost is high.
The ohmic overpotential is due to both protonic and electronic re-
sistances across the different layers of the cell and can be written:
3.2. Balance of plant modeling
ηohm ¼ rohm i
The balance of plant (BOP) is the key to ensure the safe and stable
Some scholars [45,47,48] still use the empirical relation for the operation of the PEMFC, different auxiliary systems need to operate in
conductivity of a Nafion 117 membrane obtained back in 1991 by coordination with the reactor system. And different auxiliary systems
Springer et al. [49]: often need to operate in some coordination with each other. Therefore,
   many scholars have studied the coordinated control of auxiliary systems
1 1
σ memb ¼ ð0:005139λ0:00326Þ  exp 1268  and PEMFC reactors. Many scholars have also done a lot of in-depth
303 T
research on the modeling of auxiliary systems. In this paper, the
The concentration overpotential is typically calculated using the modeling can be divided into four major categories according to the
following equation: different auxiliary systems: air supply system model, hydrogen supply
  system model, water management system model and thermal manage-
RT cin ment system model.
EConcentration ¼ ln
nF cout
3.2.1. Modeling of air supply system
Where R is the gas constant, T is the absolute temperature. n is the
The air supply system is one of the key subsystems of PEMFC, and it is a

5
Y. Wang et al. Energy Reviews 3 (2024) 100054

strong non-linear system [58]. Accurate and reasonable modeling of the 3.2.2. Modeling of hydrogen supply system
air compressor is the key to achieving accurate control of the oxygen The current hydrogen supply system uses hydrogen bottles for supply
supply. The control goal is to enable the oxygen supply to meet the stack with high purity, and the supply loop is mainly focused on hydrogen
demand. Based on a detailed analysis of the PEMFC system and its auxil- flow, pressure, and humidity control [34]. This has resulted in little
iary systems, Professor Pukrushpan's team [59] at the University of change in hydrogen circuit modeling. Some scholars [70] also considered
Michigan, USA, first established a ninth-order model of the PEMFC system, dual-channel hydrogen supply and thus improved the cell efficiency,
which selected nine variables such as the inlet manifold gas pressure and which was verified in simulation experiments.
the mass and pressure of each reactant in the cathode flow channel as state An early mainstream model of the hydrogen supply system is the one-
variables, and focused on describing the changes of the electric stack and dimensional simple model proposed by Pukrushpan et al. [59] for the
air in the PEMFC. However, the model is too complex to be used for study of hydrogen supply control strategy. For the excess hydrogen re-
controller design. Subsequently, many scholars have simplified the model covery problem, Bao et al. [71] formed an anode loop by an ejector pump
by reducing the order, which is more suitable for designing controllers. and developed a dynamic model of PEMFC capable of characterizing the
Commonly, there is the sixth-order state space model proposed by Seyed hydrogen gas flow, pressure, and humidity. The hydrogen supply sub-
et al. [60] for control and state observer design, and there is also the system is closely linked to the air supply subsystem because of the need to
seventh-order nonlinear model of ordinary differential equations proposed maintain a stable pressure difference between the two sides of the proton
by Kunusch et al. [61], which is focused on air supply system design and exchange membrane. Qinguo et al. [72] developed a one-dimensional
can be used for air supply system modeling. Although the simplified model simulation model of a PEMFC with a proton exchange membrane con-
is already relatively simple and can meet the basic requirements for taining nitrogen and water transport to achieve coordinated control of
designing controllers. However, it is still relatively complex, and subse- the pressure difference between the cathode and anode.
quent researchers have further simplified it on this basis. The structure of the hydrogen supply system is relatively simple, and
In order to model the dynamic behavior of the air supply system, scholars have done less research on the modeling of the hydrogen supply
Danze et al. [62] developed a dynamic model that can satisfy the control system, and there may not be any major changes in the future.
of both the excess oxygen ratio and the cathode pressure. In order to
simplify the model and reduce the computational effort, Liu et al. [63] 3.2.3. Modeling of water management system
improved the air supply system component model in the ninth order In the early years, scholars modeling the water transfer of a single
model using a semi empirical model, including air compressors, cathode PEMFC basically used a macroscopic model based on the first law. The
flow channels, etc. Simulation results utilizing PID control demonstrate basic idea of the model is to solve a series of partial differential equations,
significant improvements in response speed compared to conventional such as the continuity equation, Navier-Stokes equation and some
models, with steady-state time reduced from approximately 4 s to about macroscopic characteristics (mass, momentum, and energy, etc.) of the
2 s. Talj et al. [64] simplified the model to a third-order state model based conservation equations [73].
on some reasonable assumptions and by analyzing the experimental data. The current modeling of PEMFC water management systems is
The simplified third-order model, compared to the fourth-order model, divided into two main categories as shown in Table 1. One category still
incurs an error of no more than 5%. This simplified third-order system follows the idea of numerical modeling of PEMFC and uses a multi-
model is also the classical model preferred by many scholars when dimensional multiphase dynamic model of PEMFC to simulate the dy-
studying the control strategy of air supply systems. Further, Ma et al. [65] namic processes inside PEMFC. JJ. Baschuk et al. [74] proposed a
developed a simplified third-order nonlinear model of the air supply one-dimensional output model of PEMFC considering electrochemical
system, and then, using a perturbation observer to accurately estimate processes in order to analyze the effect of water flooding on the output
the cathode pressure. characteristics of PEMFC, but it did not consider the effect brought by
Nonlinear systems are still more complex compared to linear systems, temperature changes. Hu J et al. [75] has developed a one-dimensional
and it is difficult to design better controllers. Many scholars have model of the main water transfer phenomena in proton exchange mem-
simplified the model by linearizing the nonlinear model. Chen et al. [66] brane fuel cells. The model explicitly considers the water transfer process
proposed a linearized parameter variation (LPV) model based on system in the catalyst layer (membrane water uptake/desorption) and gives a
identification. It is theoretically proved that the LPV system has the same mathematical description of Schroeder's paradox. In addition, in order to
steady state properties and equilibrium point linear models as the iden- solve the two-phase problem in the gas diffusion layer without the use of
tified models. Simulation and experimental results also verify that the complex numerical methods, the assumption of infinite phase transition
LPV model can accurately cover the PEMFC operating conditions from rate is applied and an analytical solution of the two-phase model equa-
startup to shutdown in both dynamic and steady state. Wang et al. [67] tions is derived. Salva J et al. [76] presents a one-dimensional analytical
constructed a linearized model of PEMFC based on the relationship be- model for a specific flow field. It addresses various coupled physical
tween battery current and voltage. Through simulations, they confirmed phenomena, including electrochemistry, as well as the mass and heat
the effectiveness and accuracy of the model. Using the linearized model, transfer of hydrogen, oxygen, and water (in two phases) simultaneously.
they established a predictive control model for air supply to control the The results demonstrate that the values for cell voltage and water content
peroxide ratio. Simulation results indicate that this approach exhibits have a relative error of less than 3.5%. Yin et al. [77] proposed a
superior performance compared to proportional-integral control. one-dimensional mathematical model for water management in order to
In addition, data-driven [68] approaches to model the fuel system
have also been proposed. Deng et al. [69] proposed a data-driven PEMFC
Table 1
model based on a self-encoder. Compared to conventional methods, the
Comparison of various water management modeling methods.
results indicate that, under both simulated and actual data, this method
exhibits the lowest mean absolute error and root mean square error for Method Advantage Disadvantage
the three variables. Importantly, the proposed approach can be readily Numerical modeling  Multi-dimensional  Assumption
extended to other system modeling studies, such as hydrogen supply method  Multi-phase dependent
system models and thermal management system models in fuel cell  Dynamic simulation  Complex
 Reflect physicochemical  Poor applicability
systems. From the above literature summary analysis, it can be seen that behavior
the current modeling method of air supply system gradually transitions Aggregate parameter  Low computational cost  Low accuracy
from simple low-order steady-state modeling to complex high-order dy- modeling method  Simplified partial  Easy loss
namic modeling. But different modeling methods still have their own differential information
 Easy realization
defects.

6
Y. Wang et al. Energy Reviews 3 (2024) 100054

track the saturation distribution of water at the anode. Gerteisen et al. Table 2
[78] developed a one-dimensional, two-phase transient PEMFC model Comparison of thermal management modeling methods.
containing a gas diffusion layer, a proton exchange membrane, and a Method Advantage Disadvantage
cathode catalyst layer. This model forms a suitable basis for identifying
Combined mechanistic and  Reflect temperature  Large computational
the dominant loss mechanisms at different operating points. Marc Prat experimental distribution cost
et al. [48] developed a pore network model that can be used to study the method  High accuracy  Difficulty to design
multiphase transport phenomena in the porous structure of PEMFC. Aggregate parameter  Simple  Low accuracy
Karthik et al. [79] developed a semi-empirical dynamic model of PEMFC modeling method  Less computational cost  Assumptions
 Easy to design dependent
that incorporates the complete transient and dynamic phenomena of
PEMFC liquid flooding and dehydration. Xu S et al. [80], based on the
existing Volume of Fluid (VOF) and Lattice Boltzmann Method (LBM), simulate the temperature distribution characteristics during the opera-
identified and analyzed in detail the main factors affecting droplet dy- tion of the reactor. This model was utilized to investigate the equilibrium
namics. These include the surface wettability of the proton exchange between the stack's distribution characteristics and the power con-
membrane, the initial state of liquid water, the geometry of the channel, sumption of auxiliary components during the system's temperature con-
and the operating conditions. Zhou et al. [81] and [46] proposed a trol process. Subsequently, a thermal management strategy, centered
two-dimensional real-time modeling method, which covers multiple around minimizing auxiliary device power consumption, was put for-
physical domains of fluid and electrochemical characteristics. The ward. Mayyas et al. [91] developed a three-dimensional thermal model
convective properties and diffusion phenomena of the reacting gas are of the PEMFC using the finite difference thermal variation method. The
fully considered. Solsona M et al. [82] proposed a cathode humidifier simulation experiment confirms that the temperature distribution of the
model and derived a multi-input/multi-output nonlinear fourth-order stack predicted by this model aligns closely with the measurements ob-
model. A nonlinear control strategy based on a second-order sliding tained from both infrared detectors and thermocouples. Adzakpa K P
mode was designed and analyzed to demonstrate the validity of the et al. [92] presented a three-dimensional dynamic thermal model of a
model. All of the above models are validated in simulations and have single fuel cell to study the temperature distribution in a fuel cell cooled
some guidance for PEMFC management. However, it is worth noting that with air from the bottom to the top. The model is controlled by a thermal
these models are based on certain assumptions, and the actual operation energy balance and takes into account the inlet gas humidity. Zhang et al.
process may not work due to changes in load dynamics as well as changes [93] simulated the flow characteristics of the air-side. To simulate the
in the external environment. Therefore, the accuracy and practicality of airflow distribution, a three-dimensional simulation model of the fan as
the models still need to be further verified. In addition, these models are well as the heat dissipation was developed. Atyabi et al. [94] built a
still too complex and not suitable for system-level analysis. PEMFC 3D multiphase for active (liquid cooling) and passive (heat pipe)
Another category is the set-total parametric model for system anal- cooling methods numerical thermal management models. Padavu P et al.
ysis. The total parameter model is still based on the premise of material [95] employed a numerical model to simulate a serpentine flow channel
energy conservation, but the spatial material distribution in the system is design with a flow field gradient. This was done to investigate reactant
averaged, and the complex partial differential equations in the traditional distribution, local current density distribution, membrane water content
numerical model are rewritten as ordinary differential equations, which distribution, as well as polarization and power density response. Wei Lin
greatly reduces the computational effort of the model. Xue et al. [83] et al. [96] developed a three-dimensional transient model for evaluating
developed an early system-level total dynamic parameter model using a the effect of PEMFC thermal management on cold start-up performance.
combination of mechanistic and empirical methods, which focuses on In practical control applications [97], high-dimensional models are
describing the transient system behavior. However, based on this simple often difficult to meet the real-time requirements of control algorithms
model, it is not able to effectively characterize the significant transient because of their computational complexity, so many scholars have
dynamic behavior when there is a substantial fluctuation in battery simplified the models so as to establish control-oriented thermal man-
output power. Karnik et al. [84] averaged the water content of the proton agement models. Liso et al. [98] and Saygili Y et al. [43] used the lumped
exchange membrane and proposed a two-volume model for the PEMFC parameter approach to simplify the energy balance equation of the power
electrostack. Considering that the simpler lumped parameter model may pile to a first order differential equation. Zhao et al. [99] developed a
hide some of the key information, some later studies derived a control-oriented thermal management model for PEMFC water-cooled
low-dispersion model based on the lumped parameter model. The systems. The model can effectively simulate the dynamic changes of
low-dispersion models are more sophisticated to provide more detailed PEMFC output voltage. Some other scholars have constructed an aggre-
information about the PEMFC in a certain direction, but at the same time gate parameter model of the PEMFC system to simplify the computation
meet the requirements of system-level applications. For example, McKay and lay the model foundation for designing real-time control strategies.
et al. [85] proposed a two-phase isothermal one-dimensional model for Zhang et al. [100] developed an aggregate parameter model that permits
reactant and water dynamics, which essentially uses the idea of the the evaluation of the effects of operating parameters and parameter in-
two-volume model, but the new model adds a gas diffusion layer and teractions on the thermal performance of a system. Strahl et al. [101]
discretizes the gas diffusion layer into three parts to consider the transfer proposed a lumped parameter model. This model effectively demon-
of two-phase flow. strates the existence of a temperature-dependent maximum net power for
the PEMFC, and thus can be used to design an optimal thermal man-
3.2.4. Modeling of thermal management system agement strategy.
To study the control problem of PEMFC thermal management, it is In addition, for modeling the thermal management system, previous
necessary to establish a model of PEMFC thermal management system scholars often modeled only the cooling fan or the coolant cooling sys-
[86,87]. Most of the models for thermal management are dynamically tem, and few scholars [99] considered the joint action of both.
modeled through a combination of mechanistic and experimental ap-
proaches [88] (see Table 2). 4. Control system
The non-uniformity of air cooling between stack cells results in large
temperature variations between cells. These temperature variations lead The output power of PEMFC is controlled by controlling the fuel
to large voltage differences between cells, which reduces the total power supply pressure and flow rate. The stack temperature and humidity,
of the entire stack [89]. Scholars simulated the temperature distribution which have a direct impact on the safe and stable operation of the fuel
characteristics by building a multidimensional model. YuSet al. [90] cell, also need to be properly controlled. In the following section, two
established a two-dimensional heat transfer model of the reactor to

7
Y. Wang et al. Energy Reviews 3 (2024) 100054

major categories of auxiliary system control strategies are reviewed, coupling of air flow and pressure. The traditional PID control design is
including gas supply system and water thermal management system. simple and the control effect is excellent, but the control accuracy is not
high, and the jitter is easily generated when the load changes drastically.
According this, many scholars have given many methods to optimize the
4.1. Air supply
traditional PID controller so that the controller has higher accuracy and
robustness, such as adaptive PID [110], fuzzy PID [111], neural network
The air supply system of a PEMFC often requires certain control
PID [112], etc. Dongdong [113] proposed a fractional order PID (FOPID)
strategies to maintain the stability of air temperature [102], humidity,
controller based on an unknown input nonlinear observer to regulate the
pressure, flow rate, and other factors to ensure that the PEMFC works
OER and cathode pressure to their desired values at the fastest rate. Such
near the optimal parameters. While the parameters of air temperature
controllers with improved PID algorithms have better accuracy and
and humidity entering the PEMFC vary singularly and easily controlled,
robustness compared to conventional PID controllers. In the traditional
the key to the control of the air supply system is the airflow and pressure
PEMFC air supply control, most of the scholars' researches are to control
control [103]. Due to the large hysteresis of air supply to the air press, it
the peroxide ratio, so as to prevent the occurrence of oxygen starvation
may lead to insufficient oxygen supply during sudden load changes
phenomenon. However, air flow and pressure are coupled with each
[104], i.e., oxygen starvation [105]. In practice, oversupply [106] is
other, and an increase in oxygen flow per unit time is inevitably
usually adopted to alleviate the oxygen starvation phenomenon. How-
accompanied by an increase in pressure. For the strong coupling of pa-
ever, the oversupply ratio of oxygen leads to higher parasitic power and
rameters in the air supply system, Liu et al. [114] proposed a feedforward
reduces the net output power [107]. In addition, the high-power PEMFC
decoupled PID controller to achieve independent control of air supply
air supply pressure and flow rate are strongly coupled, which brings a
flow and air supply pressure. Building upon this, a decoupled fuzzy PID
great challenge to the study of air supply control strategy. Based on the
controller was developed to address challenges related to time-variance,
above analysis, a safe and reliable air supply system management scheme
hysteresis, and random interference. Simulation experiments provided
is the key to the stable and efficient operation of PEMFC [108]. In this
evidence that the decoupled fuzzy PID controller exhibits superior dy-
paper, the control methods are classified into: PID control methods, fuzzy
namic response and stability.
control methods, model prediction algorithms, neural network methods,
Fuzzy control is the application of fuzzy mathematics in the control
and other methods based on the current mainstream air supply control
system, is a kind of nonlinear intelligent control. It is a method of using
strategies, as shown in Fig. 3.
human knowledge to control objects, usually expressed in the form of “if
PID control is widely used in industrial control because of its simple
conditions, then results”, so it is also commonly referred to as language
design and easy implementation, and this control strategy is also suitable
control. Generally used for control object models that cannot be rigor-
for PEMFC air supply system control. Zhao et al. [109] controlled air flow
ously represented mathematically, fuzzy logic methods rely on the
and air pressure separately by two PID controllers, but this experiment
experience and knowledge of experts to effectively control. The air
was conducted separately and independently without considering the

Fig. 3. Typical control methods for air supply system.

8
Y. Wang et al. Energy Reviews 3 (2024) 100054

supply system of PEMFC has time lag, strong coupling, and it is difficult modes to enable its feasibility and effectiveness. The simulation results
to establish an accurate mathematical model, so fuzzy control is very demonstrate superior convergence and robustness when compared to PID
suitable for the control of air supply system. Adisorn et al. [115] used the control, with an approximate 20% reduction in overshoot and a 66%
fuzzy logic controller to control the air supply flow and the experimental decrease in regulation time. This controller consistently achieves rapid
results showed that fuzzy logic control has a better response compared to convergence across a broad spectrum of load currents, even in the pres-
manual control. Kai et al. [116] proposed a feed-forward fuzzy PID ence of various uncertainties, disturbances, and noise. Its feasibility and
control method for air flow control. The accuracy and dynamic perfor- effectiveness have been convincingly validated. Considering the uncer-
mance of the air flow control were improved compared to the conven- tainty of the system parameters during operation and the external per-
tional PID controller and fuzzy PID controller. Fuzzy control is robust and turbations, Liu et al. [123] equated the air supply system as a cascade of
does not rely on a particularly accurate system model, but the step of two systems. And the nonlinear controllers are designed for them sepa-
building a fuzzy rule base requires the designer, who has a lot of design rately. Ma et al. [65] utilized a perturbation observer to accurately esti-
experience. mate the cathode pressure and designed the oxygen excess ratio
The predictive control algorithm utilizes a predictive model to controller by a three-step nonlinear method. However, it is important to
anticipate the deviation between the future output condition and the set note that the observers are solely employed to address the estimation
value. Typically includes dynamic matrix control, model algorithm issue, without taking into account their influence on the overall control
control, internal model control, generalized predictive control, etc. It system. Furthermore, comprehensive experimental tests are necessary to
utilizes predictive models to anticipate the future output of the process. assess real-time performance.
Based on this, a “rolling” optimization strategy is used to calculate the Previous researchers have focused more on intake flow control.
current control inputs. The real-time measured values of the outputs are However, in reality, there is a coupling between cathode gas flow and
quoted for online calibration. It has the ability to handle objects with pressure [124], and changes in flow rate simultaneously cause changes in
time lag with little requirement of mathematical modeling. It is able to pressure, and large pressure fluctuations are equally harmful to the
overcome the time lag of the air supply system [117]. Wang et al. [67] proton exchange membrane. Relatively little research has been done on
used a model predictive controller for air supply based on a model after the joint control of flow and pressure, and there is still very little research
linearization. Based on the linearized model, a model predictive control in this area. Realizing the precise decoupling of the two, considering the
of air supply is developed to control the peroxide ratio. The simulation robust observer with uncertainty, improving the controller accuracy, and
results show the superiority over the proportional-integral control. Duo reducing the interactions between the control objects still need further
et al. [58] designed the corresponding controller using fuzzy generalized analysis and research.
predictive control. The results show that the controller is able to control
the peroxide ratio within the desired range, and the method has a small 4.2. Hydrogen supply
overshoot and regulation time. In addition the computational time
consumed by the method is only 3.4 ms, which is within the acceptable In order to be able to cope with the load changes in real-time [125]
range. Tian et al. [118] proposed a hybrid control scheme based on and to be affected by the chemical reaction rate, the PEMFC often keeps
model predictive control and PID control, and the model predictive the hydrogen on the anode side in excess [126] during the actual oper-
control using neural networks and look-up table methods to reduce the ation. The cyclic structure can make full use of the incompletely reacted
mismatch between the control model and the controlled system. Pre- hydrogen to improve fuel utilization efficiency, and this structure is good
dictive control is somewhat numerous, but it is worth noting that pre- for controlling the pressure difference between the cathode and anode.
dictive control is mostly used for linear systems. For nonlinear models Cheng et al. [71] linearized the nonlinear model and then used a linear
like PEMFC, model predictive control requires some time to be consumed quadratic Gaussian algorithm based on observer feedback for setpoint
upfront to calculate the model, thus forming an approximately linear tracking. And the model predictive control with neural network recog-
model expression. In addition, the model predictive controller has nizer was designed with good robustness. K. Sankarl et al. [127] designed
parameter uncertainty and thus the controller is fragile. a model-based sliding mode controller (SMC). K. Nikiforow [128]
A neural network controller [119] is an intelligent controller that can simultaneously used a pilot anode gas in cycle flow control scheme with
autonomously adjust control variables to optimize the control network three solenoid valves and carried out experiments on a 5 KW PEMFC. Yel
through extensive training, has a simple structure that is easy to design, et al. [129] designed a fuzzy controller to realize the regulation of anode
and does not require consideration of internal logic. Victor et al. [120] hydrogen supply pressure. Yuan et al. [130] proposed an improved PI
proposed a neural controller considering the strong non-linearity of the controller based on fuzzy logic technique, using fuzzy logic to rectify the
system and tested it on a PEMFC hardware-in-the-loop (HIL) simulator. PI parameters to realize the control of hydrogen pressure.
The results show that the controller is highly adaptive and robust. Kim Currently, in the practical application of PEMFC systems, all sub-
et al. [121] designed an approximation-based adaptive controller using a systems are controlled based on measurable external signals. The com-
neural network approximator and adaptive parameter compensation. bination with an online anode observer based on a nonlinear model to
The results demonstrate the effectiveness of the proposed control scheme estimate internal information to control proportional valves or injectors
for load current variations and system uncertainty. The use of a neural and purge valves is an important trend.
network approximator and an adaptive parameter compensation system
are among the first experiments in the field of OER control of a proton 4.3. Water management
exchange membrane PEMFC nonlinear gas supply system. As with the
data model, the neural network controller is a “black box” for the user. PEMFC water management is to balance and optimize the water
Although it is highly fault-tolerant and adaptive, it relies on a large brought inside the battery after humidification [131] of the imported gas
amount of measured data for training. [132] and the water generated by the battery electrochemical reaction
In addition to the aforementioned methods, nonlinear control is [133], both need to exchange membrane to maintain good hydration
frequently employed in managing the gas supply system of PEMFC, [134]. Also need to discharge excessive water inside the battery to ensure
yielding notable achievements. For this strongly nonlinear system, the a higher rate of electrochemical reaction of the battery [135].
dynamics are far more complex than for a linear system. It is obvious that Some scholars have found that the voids in the gas diffusion layer and
a linearized model to portray the PEMFC air supply system cannot the catalyst layer are easily clogged by excess liquid water, and the cross-
describe many of its characteristics. Some scholars have established a sectional area of the flow path is narrowed, resulting in a large hydrogen
nonlinear controller for the air supply system from its nature. Deng et al. pressure drop at the anode [136] (see Table 3). Shi L et al. [137] showed
[122] proposed a nonlinear controller with cascaded adaptive sliding that the balance between water generation and phase change induced

9
Y. Wang et al. Energy Reviews 3 (2024) 100054

Table 3 operates at maximum efficiency. The optimal efficiency trajectory of the

Comparison of various water management methods. system is obtained under maximum efficiency optimization. Since the
Method Advantage Disadvantage PEMFC system is a nonlinear system with time-varying and uncertain
characteristics, a constrained generalized predictive control is proposed
Channel pressure drop method  Effective judgment  Require
sensors to achieve real-time optimal efficiency trajectory tracking. The proposed
 Good prediction  Noise method can provide better system dynamic response and higher system
performance sensitive efficiency. Y. X. Wang et al. [149] proposed a real-time power optimi-
Aggregate parameter modeling  Accurate judgment  Data zation strategy based on active temperature control. An improved tem-
method dependent
 Fast judgement  Offline
perature perturbation and observation method is designed to obtain the
training optimal temperature reference. And a controller based on the
super-twisting algorithm (STA) is developed to regulate the stack tem-
perature. After determining the optimal temperature, the STA controller
(PCI) flux is the main factor controlling the cell voltage. Song et al. [138] can track the target temperature well with unknown external distur-
identified the boundary of hydrogen pressure change before liquid bances with high robustness. The real system validation results show that
flooding by observing hydrogen pressure during liquid flooding. Then the power is improved by more than 4%. Zhang B et al. proposed the
based on the hydrogen pressure change, water management of PEMFC model predictive control method for controlling the blower in a fuel cell.
was performed before liquid flooding occurred. This method has good Initially, the thermal management model of the fuel cell undergoes
predictability for water flooding. Further, Li et al. [136] proposed a new linearization through Taylor's formula. Subsequently, a closed-loop
method for determining the pressure drop in the cathode flow channel. feedback MPC system is formulated, leveraging the linearized
Some scholars have also suggested using the deviation of the theoretical state-space model. Nevertheless, it is important to highlight that this
voltage from the actual voltage to determine whether a failure occurred algorithm entails significant computational overhead and poses chal-
in the stack. When the deviation is greater than a set threshold, the lenges in its practical implementation within a real-world system. Kwan
battery is considered to be malfunctioning, and then further judgment is et al. [150] proposed a thermodynamic model of a thermoelectric
made by the anode hydrogen pressure drop, and when it is judged to be generator, which was optimized to maintain maximum output power
flooding, certain recovery measures are then taken. using a genetic algorithm approach based on ensuring that it was within
Several authors have investigated the control of gas humidification its operating temperature range. The PEMFC face scenarios of continuous
tracking. Ahmad et al. [139] considered the effect of gas humidification dynamic load changes in real-world application conditions, which can
rate on the operational efficiency of PEMFC. Raman S et al. [133] enabled impair cell performance if the resulting drastic temperature fluctuations
a combination of feedback from PI and PID control algorithms to achieve occur. To address this issue, Tiancai Ma et al. [151] developed a dynamic
fast tracking of gas relative humidity. Ou K et al. [140] developed a model of the air-feed cooling system and designed a corresponding
multiple-input multiple-output (MIMO) fuzzy controller and demon- controller using an integral separation PID control algorithm to reduce
strated its superiority in real-time control of temperature and relative the temperature fluctuation from 7.5 C to 0.4 C. Li Sun et al. [152] used
humidity of a fuel cell. Several scholars have controlled the state of the an open cathode based proton exchange membrane PEMFC using a
PEMFC proton exchange membrane. Liyan et al. [141] proposed a model model and data-driven method to propose a self-anti-disturbance control
prediction controller optimized by a recursive neural network and veri- for hybrid air supply cooling. Since the PEMFC works with fast heat
fied the effectiveness of the method in avoiding fluctuations in cathode generation, the coolant entering the air supply outlet loses its cooling
water concentration. Fu et al. [142] aimed to achieve precise description effect quickly, so there is often a certain temperature difference between
and control of humidity. The internal moisture content was utilized as a the front end and the back end of the cooling system. In order to solve the
feedback signal and a dynamic model was developed by considering phenomenon of temperature gradient existing on the air supply cooling
various auxiliary couplings. A control structure combining feedforward channel. Yuan et al. [153] proposed a bi-directional flow control tem-
and feedback is proposed to keep the internal humidity as close to the perature control method, which effectively reduces the temperature
desired value as possible. Damour et al. [143] proposed a controller gradient.
based on differential flatness, which effectively avoids the phenomena of For the study of coolant cooling control, Jason et al. [154] proposed a
membrane drying and flooding. In addition, some scholars found that the nonlinear perturbation control method to accurately maintain the tem-
low-frequency internal resistance and high-frequency internal resistance perature of PEMFC stack. To address the uncertainty of fuel cell thermal
of PEMFC are sensitive to water flooding and membrane dryness, management, Sun L et al. [155] estimated the uncertain state by real-time
respectively. kurz et al. [144] proposed a system control strategy based input and output data, and then compensated it by active disturbance
on single-cell impedance measurement using the voltage drop caused by rejection control (ADRC). And the guaranteed performance of ADRC
the impedance change to predict water flooding and membrane dryness under uncertainty conditions is proved theoretically. Practical tests show
using this property. that the proposed ADRC method has obvious advantages in both tracking
Whether the above water management strategy can be implemented and regulation performance compared with the traditional PI controller.
depends not only on the merit of the control strategy but also on the Han et al. [156] designed a feedback controller incorporating
accuracy of determining the occurrence of flooding and membrane dry- model-referenced adaptive control for robust temperature control. The
ing in the PEMFC [145]. nonlinear intelligent control strategy, with good fault tolerance for sys-
tem parameter variations, is effective for the control of fuel cell systems
with nonlinear and time-varying temperatures. Jenn-Jiang et al. [157]
4.4. Thermal management
proposed a thermal controller with an intelligent control algorithm to
manage the thermal behavior of the PEMFC.
At present, PEMFC temperature control according to the cooling
In the traditional control strategy, independent controllers are used
method is mainly divided into increased airflow cooling, and liquid
for both the pump and radiator, and some scholars have also proposed a
cooling two methods. For low-power, PEMFC using increased airflow
coordination control problem for the PEMFC where air supply cooling
cooling [146] has the advantages of simple structure and easy control,
and coolant cooling coexist, Li et al. [47] proposed a distributed rein-
but for high-power PEMFCs, using liquid cooling is the mainstream
forcement learning integrated control strategy for solving the coordina-
method [147]. For the study of airflow cooling control, Sangseok Yu et al.
tion control problem of the pump and radiator.
[90] and Yin et al. [148] developed a thermal model and thermal man-
agement system for PEMFC to control the PEMFC temperature so that it

10
Y. Wang et al. Energy Reviews 3 (2024) 100054

5. Summary and outlook Understanding the implications of these assumptions holds significance
for researchers focused on managing water and heat in fuel cells. How-
This section summarizes the future research trends in fuel cell ever, to the best of the authors' knowledge, there has been limited effort
modeling and control strategies, including system modeling, air supply to quantify the influence of each assumption on model accuracy,
system, hydrogen supply system, water management system, and thermal computational efficiency, and the applicable range of each assumption.
management systems, as shown in Fig. 4. Given the intricate nature of the underlying phenomena, conducting a
System modeling: The modeling of PEMFC systems can be classified comprehensive analysis is a challenging endeavor.
into mechanistic, empirical, and data-driven models. The semi-empirical Air supply system: As one of the core subsystems of PEMFC, the
model replaces a part of the overly complex internal mechanism equa- research on the air supply control strategy for the air subsystem has been
tions with empirical formulas, thus greatly reducing the amount of the focus of scholars' attention. In the past, researchers paid more
computation. The fully empirical model replaces all the internal mech- attention to inlet gas flow control to avoid the oxygen starvation phe-
anism equations with empirical formulas, and no longer considers the nomenon. However, in fact, there is a coupling between cathode gas flow
internal reaction mechanism. The data-driven model is a black-box and pressure. Changes in flow rate also cause pressure changes, and large
model that is put together based on a large amount of data training but pressure fluctuations are also harmful to the proton exchange membrane.
requires significant offline computational costs. At present, research on Currently, there are relatively few studies on the joint control of flow and
modeling electrical stacks using traditional methods is relatively well pressure. Achieving precise decoupling of the two, robust observers that
established. Digital twin [158], as an emerging technology, is to make take uncertainty into account, improving controller accuracy, and
full use of data such as physical models, sensor updates, and operation reducing the interactions between control objects all still require further
history. Integrate the simulation process of multi-disciplinary, multi-- analysis and research [160]. In addition, the time lag of fuel cells is a pain
physical quantity, multi-scale, and multi-probability [159]. Complete the point and a difficult problem in the application of fuel cells to realize
mapping in the virtual space so as to reflect the full life cycle process of [161]. Due to its working principle and structure, the time lag cannot be
the corresponding physical equipment. It has not been applied to the completely eliminated using effective control strategies [162], and the
modeling and simulation of fuel cells, which is an important direction for practical applications often use supercapacitors, batteries, and other
future scholars to study. auxiliary means as compensation [163], thus effectively solving the time
The literature on PEM fuel cell modeling presents a plethora of lag problem, so the management of the hybrid power system [164] is also
simplified models tailored for system-level investigations. These models an important direction for future research.
incorporate a diverse array of assumptions and techniques aimed at Hydrogen supply system:Currently, in practical applications of fuel
streamlining calculations by reducing the number of equations. Many of cell systems, the hydrogen system is controlled based on measurable
these assumptions pertain to the behavior of reacting gases within the external signals (e.g. current, pressure, mass flow). The combination with
fuel cell, the various phases of water, and the heat transfer mechanisms. an online anode observer to control based on a non-linear model to

Fig. 4. Summary and outlook.

11
Y. Wang et al. Energy Reviews 3 (2024) 100054

estimate the internal information is an important trend. With this feed- References
back control structure, the internal state of the anode side can be well
controlled and the pressure control can be made more meaningful. In [1] J.H. Wee, Which type of fuel cell is more competitive for portable application:
direct methanol fuel cells or direct borohydride fuel cells? J. Power Sources 161
addition, there is an urgent need to find more robust or adaptive control (2006) 1–10.
algorithms to suppress environmental and human disturbances, improve [2] H.A. Gasteiger, S.S. Kocha, B. Sompalli, F.T. Wagner, Activity benchmarks and
the performance of hydrogen subsystems, and significantly reduce requirements for Pt, Pt-alloy, and non-Pt oxygen reduction catalysts for PEMFCs -
ScienceDirect, Appl. Catal., B 56 (2005) 9–35.
development time and cost. [3] S. Zhang, S.B. Beale, U. Reimer, M. Andersson, W. Lehnert, Polymer electrolyte
Water management system: For modeling water management sys- fuel cell modeling - a comparison of two models with different levels of
tems, previous scholars have considered the use of simpler aggregate complexity, Int. J. Hydrogen Energy 45 (2020) 19761–19777.
[4] Z. Wang, Z. Liu, L. Fan, Q. Du, K. Jiao, Application progress of small-scale proton
parameter models, which may hide some of the key information. Sub- exchange membrane fuel cell, Energy Rev. 2 (2023) 100017.
sequent studies have derived low-discretization models based on [5] S. Wang, Y. Wang, Investigation of the through-plane effective oxygen diffusivity
lumped-parameter models. The low-discretization model is more com- in the porous media of PEM fuel cells: effects of the pore size distribution and
water saturation distribution, Int. J. Heat Mass Tran. 98 (2016) 541–549.
plex, but it can provide more detailed information about the PEMFC in a
[6] Y. Wang, T. Liu, H. Sun, W. He, Y. Fan, S. Wang, Investigation of dry ionomer
certain direction, which is an effective method for water management volume fraction in cathode catalyst layer under different relative humilities and
modeling studies. Whether the water management strategy can be real- nonuniform ionomer-gradient distributions for PEM fuel cells, Electrochim. Acta
ized depends not only on the control strategy, but also on whether it can 353 (2020) 136491.
[7] A. Beicha, Modeling and simulation of proton exchange membrane fuel cell
accurately judge the flooding and membrane drying phenomenon of systems, J. Power Sources 205 (2012) 335–339.
PEMFC, so how to make a timely prediction before the flooding and [8] J. Lachaize, S. Caux, M. Fadel, P. Schott, L. Nicod, Pressure, flow and thermal
membrane drying occurs in PEMFC is still the focus of future research. As control of a fuel cell system for electrical rail transport, in: 2004 IEEE
International Symposium on Industrial Electronics, 331, 2004, pp. 335–340.
there are many sensors in the fuel cell water management system, how to [9] A. Goshtasbi, T. Ersal, Degradation-conscious control for enhanced lifetime of
build a physical cyber system to combine the data from many sensors and automotive polymer electrolyte membrane fuel cells, J. Power Sources 457 (2020)
intelligent algorithms to predict and diagnose water management fail- 227996.
[10] Y. Wang, K.S. Chen, J. Mishler, S.C. Cho, X.C. Adroher, A review of polymer
ures in real-time is the focus of future research. Mature cyber-physical electrolyte membrane fuel cells: technology, applications, and needs on
systems compose physical inputs and outputs into a network of inter- fundamental research, Appl. Energy 88 (2011) 981–1007.
acting elements, which closely link sensor networks and intelligent al- [11] J.C. Kurnia, A.P. Sasmito, T. Shamim, Performance evaluation of a PEM fuel cell
stack with variable inlet flows under simulated driving cycle conditions, Appl.
gorithms. Cyber-physical systems are able to improve the connection Energy 206 (2017) 751–764.
between computational and physical elements through intelligent [12] T. Wilberforce, Z. El Hassan, E. Ogungbemi, O. Ijaodola, F. Khatib, A. Durrant,
mechanisms, which improves the adaptability, autonomy, efficiency, J. Thompson, A. Baroutaji, A. Olabi, A comprehensive study of the effect of bipolar
plate (BP) geometry design on the performance of proton exchange membrane
functionality, reliability, and availability of fuel cell water management.
(PEM) fuel cells, Renewable Sustainable Energy Rev. 111 (2019) 236–260.
In addition, there are various kinds of signal-based diagnostic methods, [13] M. Ramezanizadeh, M.A. Nazari, M.H. Ahmadi, L. Chen, A review on the
each with its own characteristics and advantages, and scholars should approaches applied for cooling fuel cells, Int. J. Heat Mass Tran. 139 (2019)
combine these methods to complement their strengths. 517–525.
[14] S. Zhou, L. Fan, G. Zhang, J. Gao, Y. Lu, P. Zhao, C. Wen, L. Shi, Z. Hu, A review on
Thermal management systems: For modeling thermal management proton exchange membrane multi-stack fuel cell systems: architecture,
systems, scholars in the past tended to model only the cooling fan or the performance, and power management, Appl. Energy 310 (2022) 118555.
coolant cooling system, and few scholars have considered the joint role of [15] P.C. Chen, Robust voltage tracking control for proton exchange membrane fuel
cells, Energy Convers. Manag. 65 (2013) 408–419.
the two. In the traditional control strategy, the water pump and radiator [16] W. Zou, D. Froning, Y. Shi, W. Lehnert, A least-squares support vector machine
both use independent controllers, and some scholars have also proposed method for modeling transient voltage in polymer electrolyte fuel cells, Appl.
a coordinated control problem for the coexistence of air supply cooling Energy 271 (2020) 115092.
[17] D. Zhao, L. Xia, H. Dang, Z. Wu, H. Li, Design and control of air supply system for
and coolant cooling in the PEMFC, but further research is still needed. PEMFC UAV based on dynamic decoupling strategy, Energy Convers. Manag. 253
Model uncertainty is another important aspect of thermal management (2022) 115159.
modeling research. Due to the stochastic behavior of physical phenom- [18] H. Chen, Z. Liu, X. Ye, L. Yi, T. Zhang, Air flow and pressure optimization for air
supply in proton exchange membrane fuel cell system, Energy 238 (2021) 121949.
ena, such as the presence and release of supercooled water. Uncertainty [19] W. Linli, S. Zhangyi, H. Mousavi, H. Gu, A novel robust reset controller
also exists due to the difficulty of accurately measuring material prop- implemented through field programmable gate array for oxygen ratio regulation
erties, such as the conductivity and reactivity of the catalyst layer. Future of proton exchange membrane fuel cell, Comput. Electr. Eng. 99 (2022) 107788.
[20] T. Pangaribowo, W.M. Utomo, A.A. Bakar, D.S. Khaerudini, Modeling of control
modeling and management studies should take full account of model
system for hydrogen and oxygen gas flow into PEMFC based on load demand by
uncertainty. It is worth mentioning that fault-tolerant control is a very using fuzzy controller, Int. J. Adv. Trends Comput. Sci. Eng. 1 (2020) 606–611.
effective means to enhance the reliability of PEMFC, and there are fewer [21] Q. Liu, F. Lan, J. Chen, C. Zeng, J. Wang, A review of proton exchange membrane
studies on fault-tolerant control strategies for PEMFC. Through a fuel cell water management: membrane electrode assembly, J. Power Sources 517
(2022) 230723.
reasonable fault-tolerant control strategy, the impact of various faults can [22] H. Pourrahmani, A. Yavarinasab, M. Siavashi, M. Matian, J. Van herle, Progress in
be significantly reduced or even eliminated, thus greatly improving the the proton exchange membrane fuel cells (PEMFCs) water/thermal management:
operational stability and reliability of the PEMFC system, simulation and From theory to the current challenges and real-time fault diagnosis methods,
Energy Rev. 1 (2022) 100002.
industrial applications can further verify the feasibility and reliability of [23] S. Xing, C. Zhao, J. Zou, S. Zaman, Y. Yu, H. Gong, Y. Wang, M. Chen, M. Wang,
its implementation, which is of great practical significance. M. Lin, Recent advances in heat and water management of forced-convection
open-cathode proton exchange membrane fuel cells, Renewable Sustainable
Energy Rev. 165 (2022) 112558.
Declaration of competing interest [24] X.-D. Wang, Y.-L. Wang, Y. Chen, C. Si, A. Su, D.-J. Lee, Proton exchange
membrane fuel cell modeling with diffusion layer-based and sands-based capillary
The authors declare that they have no known competing financial pressure correlations: comparative study, J. Taiwan Inst. Chem. Eng. 45 (2014)
1532–1541.
interests or personal relationships that could have appeared to influence [25] Z. Yang, Q. Du, Z. Jia, C. Yang, K. Jiao, Effects of operating conditions on water
the work reported in this paper. and heat management by a transient multi-dimensional PEMFC system model,
Energy 183 (2019) 462–476.
[26] K. Hong, S. Li, K. Zhu, J. Pang, M. Pan, Effects of relative humidification on
Acknowledgment
durability of membrane electrode assembly of proton exchange membrane fuel
cells, J. Electrochem. Soc. 168 (2021) 064507.
This work was supported by the National Natural Science Foundation [27] X.A. Lu, B. Hc, C. Rz, B. Tw, B. Qz, B. Wx, B. Zt, Thermal analysis and management
of China (Grant No. 62373340). of proton exchange membrane fuel cell stacks for automotive vehicle, Int. J.
Hydrogen Energy 46 (2021) 32665–32675.
[28] X. Jiang, Y. Tao, C. Tang, H. Zhao, W. Li, Simulation of fuel cell power generation
system based on AMESim, IOP Conf. Ser. Earth Environ. Sci. 770 (2021) 012006.

12
Y. Wang et al. Energy Reviews 3 (2024) 100054

[29] L. Xing, W. Xiang, R. Zhu, Z. Tu, Modeling and thermal management of proton [59] J.T. Pukrushpan, Control of Fuel Cell Power Systems : Principles, Modeling,
exchange membrane fuel cell for fuel cell/battery hybrid automotive vehicle, Int. Analysis, and Feedback Design, Control of Fuel Cell Power Systems : Principles,
J. Hydrogen Energy 47 (2022) 1888–1900. Modeling, Analysis, and Feedback Design, 2004.
[30] Q. Chen, G. Zhang, X. Zhang, C. Sun, K. Jiao, Y. Wang, Thermal management of [60] S.M. Rakhtala, A.R. Noei, R. Ghaderi, E. Usai, Design of finite-time high-order
polymer electrolyte membrane fuel cells: a review of cooling methods, material sliding mode state observer: a practical insight to PEM fuel cell system, J. Process
properties, and durability, Appl. Energy 286 (2021) 116496. Control 24 (2014) 203–224.
[31] W.-M. Yan, F. Chen, H.-Y. Wu, C.-Y. Soong, H.-S. Chu, Analysis of thermal and [61] C. Kunusch, P.F. Puleston, M.A. Mayosky, A.P. Husar, Control-oriented modeling
water management with temperature-dependent diffusion effects in membrane of and experimental validation of a PEMFC generation system, IEEE Trans. Energy
proton exchange membrane fuel cells, J. Power Sources 129 (2004) 127–137. Convers. 26 (2011) 851–861.
[32] H. Chang, F. Cai, X. Yu, C. Duan, S.H. Chan, Z. Tu, Experimental study on the [62] M.A. Danzer, J. Wilhelm, H. Aschemann, E.P. Hofer, Model-based control of
thermal management of an open-cathode air-cooled proton exchange membrane cathode pressure and oxygen excess ratio of a PEM fuel cell system, J. Power
fuel cell stack with ultra-thin metal bipolar plates, Energy 263 (2023) 125724. Sources 176 (2008) 515–522.
[33] D.M. Bernardi, M.W. Verbrugge, Mathematical model of a gas diffusion electrode [63] Z. Liu, L. Li, Y. Ding, H. Deng, W. Chen, Modeling and control of an air supply
bonded to a polymer electrolyte, AlChE J 37 (2010) 1151–1163. system for a heavy duty PEMFC engine, Int. J. Hydrogen Energy 41 (2016)
[34] A. Jc, M. Rf, P. Ba, R. Pr, H. Tj, B. Rm, Performance modeling of the ballard mark 16230–16239.
IV solid polymer electrolyte fuel cell .1. Mechanistic model development, [64] R.J. Talj, D. Hissel, R. Ortega, M. Becherif, M. Hilairet, Experimental validation of
J. Electrochem. Soc. 142 (1995) 142. a PEM fuel cell reduced order model and a moto-compressor higher order sliding
[35] R.F. Mann, J.C. Amphlett, M. Hooper, Development and application of a mode control, IEEE Trans. Ind. Electron. 57 (2010) 1906–1913.
generalised steady-state electrochemical model for a PEM fuel cell, J. Power [65] M. Yan, Z.B. Fan, B. Jga, C. Hong, C. Ts, Oxygen excess ratio control of PEM fuel
Sources 86 (2000) 173–180. cells using observer-based nonlinear triple-step controller, Int. J. Hydrogen Energy
[36] T.V. Nguyen, A water and heat management model for proton-exchange- 45 (2020) 29705–29717.
membrane fuel cells, J. Electrochem. Soc. 140 (1993) 2178. [66] F.X. Chen, J.R. Jiao, S.G. Liu, Y. Yu, S.C. Xu, Control-oriented LPV modeling for
[37] K. Dannenberg, P. Ekdunge, G. Lindbergh, Mathematical model of the PEMFC, the air supply system of proton exchange membrane fuel cells, Fuel Cell. 18
J. Appl. Electrochem. 30 (2000) 1377–1387. (2018) 433–440.
[38] C.J. Bapat, S.T. Thynell, Effect of anisotropic electrical resistivity of gas diffusion [67] Y.X. Wang, Y.B. Kim, Real-time control for air excess ratio of a PEM fuel cell
layers (GDLs) on current density and temperature distribution in a Polymer system, IEEE ASME Trans. Mechatron. 19 (2014) 852–861.
Electrolyte Membrane (PEM) fuel cell, J. Power Sources 185 (2008) 428–432. [68] Y. Li, P. Pei, Z. Ma, P. Ren, H. Huang, Method for system parameter identification
[39] Y. Shan, S.Y. Choe, A high dynamic PEM fuel cell model with temperature effects, and controller parameter tuning for super-twisting sliding mode control in proton
J. Power Sources 145 (2005) 30–39. exchange membrane fuel cell system, Energy Convers. Manag. 243 (2021)
[40] Z. Yong, H. Shirong, J. Xiaohui, Y. Yuntao, X. Mu, Y. Xi, Characteristics of proton 114370.
exchange membrane fuel cell considering “dot matrix” gas distribution zones and [69] Z. Deng, Q. Chen, L. Zhang, K. Zhou, H. Liu, Data-driven reconstruction of
waveform staggered flow field with cooling channels, Energy Convers. Manag. interpretable model for air supply system of proton exchange membrane fuel cell,
267 (2022) 115881. Appl. Energy 299 (2021) 117266.
[41] S. Dutta, S. Shimpalee, J. Zee, Three-dimensional numerical simulation of straight [70] X. Bai, L. Luo, B. Huang, Q. Jian, Z. Cheng, Performance improvement of proton
channel PEM fuel cells, J. Appl. Electrochem. 30 (2000) 135–146. exchange membrane fuel cell stack by dual-path hydrogen supply, Energy 246
[42] N.T. Truc, S. Ito, K. Fushinobu, Numerical and experimental investigation on the (2022) 123297.
reactant gas crossover in a PEM fuel cell, Int. J. Heat Mass Tran. 127 (2018) [71] O. Cheng, Modeling and control of air stream and hydrogen flow with
447–456. recirculation in a PEM fuel cell system—I. Control-oriented modeling, Int. J.
[43] Y. Saygili, I. Eroglu, S. Kincal, Model based temperature controller development Hydrogen Energy 558 (2006) 232585.
for water cooled PEM fuel cell systems, Int. J. Hydrogen Energy 40 (2015) [72] Q. Zhang, Z. Tong, S. Tong, Z. Cheng, H. Lund, M.J. Kaiser, Modeling and dynamic
615–622. performance research on proton exchange membrane fuel cell system with
[44] V. Kannan, H. Xue, K.A. Raman, J. Chen, A. Fisher, E. Birgersson, Quantifying hydrogen cycle and dead-ended anode, Energy 218 (2021) 119476.
operating uncertainties of a PEMFC–Monte Carlo-machine learning based [73] P.H.A. N
obrega, A review of physics-based low-temperature proton-exchange
approach, Renew. Energy 158 (2020) 343–359. membrane fuel cell models for system-level water and thermal management
[45] G.A. Futter, P. Gazdzicki, K.A. Friedrich, A. Latz, T. Jahnke, Physical modeling of studies, J. Power Sources 558 (2023) 232585.
polymer-electrolyte membrane fuel cells: understanding water management and [74] J.J. Baschuk, X. Li, Modelling of polymer electrolyte membrane fuel cells with
impedance spectra, J. Power Sources 391 (2018) 148–161. variable degrees of water flooding, J. Power Sources 86 (2000) 181–196.
[46] Z. Yang, Q. Du, Z. Jia, C. Yang, J. Xuan, K. Jiao, A comprehensive proton exchange [75] J. Hu, J. Li, L. Xu, F. Huang, M. Ouyang, Analytical calculation and evaluation of
membrane fuel cell system model integrating various auxiliary subsystems, Appl. water transport through a proton exchange membrane fuel cell based on a one-
Energy 256 (2019) 113959. dimensional model, Energy 111 (2016) 869–883.
[47] J. Li, Y. Li, T. Yu, Distributed deep reinforcement learning-based multi-objective [76] J.A. Salva, A. Iranzo, F. Rosa, E. Tapia, Validation of cell voltage and water content
integrated heat management method for water-cooling proton exchange in a PEM (polymer electrolyte membrane) fuel cell model using neutron imaging
membrane fuel cell, Case Stud. Therm. Eng. 27 (2021) 101284. for different operating conditions, Energy 101 (2016) 100–112.
[48] M.E. Hannach, M. Prat, J. Pauchet, Pore network model of the cathode catalyst [77] K.M. Yin, H.K. Hsuen, Mathematical model of proton exchange membrane fuel cell
layer of proton exchange membrane fuel cells: analysis of water management and with consideration of water management, Fuel Cell. 13 (2013) 1213–1225.
electrical performance, Int. J. Hydrogen Energy 37 (2012) 18996–19006. [78] D. Gerteisen, T. Heilmann, C. Ziegler, Modeling the phenomena of dehydration
[49] T.E. Springer, T. Zawodzinski, S. Gottesfeld, Polymer electrolyte fuel cell model, and flooding of a polymer electrolyte membrane fuel cell, J. Power Sources 187
J. Electrochem. Soc. 138 (1991) 2334. (2009) 165–181.
[50] J.H. Lee, T.R. Lalk, A.J. Appleby, Modeling electrochemical performance in large [79] K. Murugesan, V. Senniappan, Investigation of water management dynamics on
scale proton exchange membrane fuel cell stacks, J. Power Sources 70 (1998) the performance of a ballard-mark-V proton exchange membrane fuel cell stack
258–268. system, Int. J. Electrochem. Sci. 8 (2013) 7885–7904.
[51] G. Squadrito, G. Maggio, E. Passalacqua, F. Lufrano, A. Patti, An empirical [80] S. Xu, P. Liao, D. Yang, Z. Li, B. Li, P. Ming, X. Zhou, Liquid water transport in gas
equation for polymer electrolyte fuel cell (PEFC) behaviour, J. Appl. Electrochem. flow channels of PEMFCs: a review on numerical simulations and visualization
29 (1999) 1449–1455. experiments, Int. J. Hydrogen Energy 48 (2023) 10118–10143.
[52] D. Hao, J. Shen, Y. Hou, Z. Yi, W. Hong, An improved empirical fuel cell [81] D. Zhou, T.T. Nguyen, E. Breaz, D. Zhao, S. Cl
enet, F. Gao, Global parameters
polarization curve model based on review analysis, Int. J. Chem. Eng. 2016 (2016) sensitivity analysis and development of a two-dimensional real-time model of
1–10. proton-exchange-membrane fuel cells, Energy Convers. Manag. 162 (2018)
[53] Junbom Kim, Modeling of proton exchange membrane fuel cell performance with 276–292.
an empirical equation, J. Electrochem. Soc. 142 (1995) 2670–2674. [82] M. Solsona, C. Kunusch, C. Ocampo-Martinez, Control-oriented model of a
[54] T. Manikandan, S. Ramalingam, A review of optimization algorithms for the membrane humidifier for fuel cell applications, Energy Convers. Manag. 137
modeling of proton exchange membrane fuel cell, Journal of Renewable 8 (2016) (2017) 121–129.
1103–1112. [83] X. Xue, J. Tang, A. Smirnova, R. England, N. Sammes, System level lumped-
[55] R. Ma, T. Yang, E. Breaz, Z. Li, P. Briois, F. Gao, Data-driven proton exchange parameter dynamic modeling of PEM fuel cell, J. Power Sources 133 (2004)
membrane fuel cell degradation predication through deep learning method, Appl. 188–204.
Energy 231 (2018) 102–115. [84] A.Y. Karnik, A.G. Stefanopoulou, J. Sun, Water equilibria and management using a
[56] I.S. Han, C.B. Chung, A hybrid model combining a support vector machine with an two-volume model of a polymer electrolyte fuel cell, J. Power Sources 164 (2007)
empirical equation for predicting polarization curves of PEM fuel cells, Int. J. 590–605.
Hydrogen Energy 42 (2017) 7023–7028. [85] D.A. Mckay, J.B. Siegel, W. Ott, A.G. Stefanopoulou, Parameterization and
[57] L. Qi, W. Chen, Z. Liu, G. Ai, H. Jin, Nonlinear multivariable modeling of prediction of temporal fuel cell voltage behavior during flooding and drying
locomotive proton exchange membrane fuel cell system, Int. J. Hydrogen Energy conditions, J. Power Sources 178 (2008) 207–222.
39 (2014) 13777–13786. [86] J. Xu, C. Zhang, R. Fan, H. Bao, Y. Wang, S. Huang, C.S. Chin, C. Li, Modelling and
[58] D. Yang, R. Pan, Y. Wang, Z. Chen, Modeling and control of PEMFC air supply control of vehicle integrated thermal management system of PEM fuel cell vehicle,
system based on T-S fuzzy theory and predictive control, Energy 188 (2019) Energy 199 (2020) 117495.
116078. [87] A. De las Heras, F. Vivas, F. Segura, M. Redondo, J. Andújar, Air-cooled fuel cells:
keys to design and build the oxidant/cooling system, Renew. Energy 125 (2018)
1–20.

13
Y. Wang et al. Energy Reviews 3 (2024) 100054

[88] A.J. Headley, D. Chen, Critical control volume sizing for improved transient [117] H. Hu, K. Ou, W.-W. Yuan, Fused multi-model predictive control with adaptive
thermal modeling of PEM fuel cells, Int. J. Hydrogen Energy 40 (2015) compensation for proton exchange membrane fuel cell air supply system, Energy
7762–7768. 284 (2023) 128459.
[89] S. Yu, Y. Lee, K. Ahn, Two dimensional numerical model for thermal management [118] Y. Tian, Q. Zou, J. Liu, J. Han, Novel hybrid control scheme of a proton exchange
of proton exchange membrane fuel cell with large active area, Transactions of the membrane fuel cell air supply system, Energy Technol. 10 (2022) 2100906.
Korean Society of Mechanical Engineers - B 32 (2008) 359–366. [119] U. Javaid, A. Mehmood, J. Iqbal, A.A. Uppal, Neural network and URED observer
[90] S. Yu, D. Jung, Thermal management strategy for a proton exchange membrane based fast terminal integral sliding mode control for energy efficient polymer
fuel cell system with a large active cell area, Renew. Energy 33 (2008) electrolyte membrane fuel cell used in vehicular technologies, Energy 269 (2023)
2540–2548. 126717.
[91] Abdel Raouf, Mayyas Dilip, Ramani Arunachala, M. Kannan, Keng Hsu, Cooling [120] V.M. Sanchez, R. Barbosa, L.G. Arriaga, J.M. Ramirez, Real time control of air feed
strategy for effective automotive power trains: 3D thermal modeling and multi- system in a PEM fuel cell by means of an adaptive neural-network, Int. J.
faceted approach for integrating thermoelectric modules into proton exchange Hydrogen Energy 39 (2014) 16750–16762.
membrane fuel cell stack, Int. J. Hydrogen Energy 39 (2014) 17327–17335. [121] B.M. Kim, H.C. Yun, S.J. Yoo, Adaptive control of proton exchange membrane fuel
[92] K. Adzakpa, J. Ramousse, Y. Dub
e, H. Akremi, K. Agbossou, M. Dostie, A. Poulin, cell air supply systems with asymmetric oxygen excess ratio constraints, IEEE
M. Fournier, Transient air cooling thermal modeling of a PEM fuel cell, J. Power Access 8 (2019) 5537–5549.
Sources 179 (2008) 164–176. [122] H. Deng, Q. Li, Y. Cui, Y. Zhu, W. Chen, Nonlinear controller design based on
[93] Q. Zhang, L. Xu, J. Li, M. Ouyang, Performance prediction of proton exchange cascade adaptive sliding mode control for PEM fuel cell air supply systems, Int. J.
membrane fuel cell engine thermal management system using ID and 3D Hydrogen Energy 44 (2019) 19357–19369.
integrating numerical simulation, Int. J. Hydrogen Energy 43 (2018) 1736–1748. [123] Z. Liu, J. Chen, H. Chen, C. Yan, Air supply regulation for PEMFC systems based on
[94] A. Seyed, A. Ebrahim, U. Chinonyelum, Comparison of active and passive cooling uncertainty and disturbance estimation, Int. J. Hydrogen Energy 43 (2018)
of proton exchange membrane fuel cell using a multiphase model, Energy 11559–11567.
Convers. Manag. 268 (2022) 115970. [124] H. Chen, Z. Liu, X. Ye, L. Yi, S. Xu, T. Zhang, Air flow and pressure optimization for
[95] P. Padavu, P.K. Koorata, S. Kattimani, Model based evaluation of water air supply in proton exchange membrane fuel cell system, Energy 238 (2022)
management and membrane hydration in polymer electrolyte fuel cell with 121949.
reactant flow-field gradients, Int. J. Heat Mass Tran. 214 (2023) 124460. [125] T. Zeng, C. Zhang, A. Zhou, Q. Wu, C. Deng, S.H. Chan, J. Chen, A.M. Foley,
[96] L. Wei, Z. Liao, A.M. Dafalla, Z. Suo, F. Jiang, Effects of endplate assembly on cold Enhancing reactant mass transfer inside fuel cells to improve dynamic
start performance of proton exchange membrane fuel cell stacks, Energy Technol. performance via intelligent hydrogen pressure control, Energy 230 (2021)
9 (2021) 2100092. 120620.
[97] C. Yan, J. Chen, H. Liu, H. Lu, Model-based fault tolerant control for the thermal [126] M. Siva, S. Meshram, D. Patle, U. Gara, Model based control strategies to control
management of PEMFC systems, IEEE Trans. Ind. Electron. 67 (2020) 2875–2884. voltage of proton exchange membrane fuel cell, Chem. Prod. Process Model. 16
[98] V. Liso, M.P. Nielsen, S.K. Koer, H.H. Mortensen, Thermal modeling and (2020) 69–85.
temperature control of a PEM fuel cell system for forklift applications, Int. J. [127] K. Sankar, A.K. Jana, Nonlinear multivariable sliding mode control of a reversible
Hydrogen Energy 39 (2014) 8410–8420. PEM fuel cell integrated system, Energy Convers. Manag. 171 (2018) 541–565.
[99] X. Zhao, Y. Li, Z. Liu, Q. Li, W. Chen, Thermal management system modeling of a [128] K. Nikiforow, P. Koski, J. Ihonen, Discrete ejector control solution design,
water-cooled proton exchange membrane fuel cell, Int. J. Hydrogen Energy 40 characterization, and verification in a 5kW PEMFC system, Int. J. Hydrogen
(2015) 3048–3056. Energy 42 (2017) 16760–16772.
[100] Y. Zhang, M. Ouyang, Q. Lu, J. Luo, X. Li, A model predicting performance of [129] X. Ye, T. Zhang, H. Chen, J. Cao, J. Chen, Fuzzy control of hydrogen pressure in
proton exchange membrane fuel cell stack thermal systems, Appl. Therm. Eng. 24 fuel cell system, Int. J. Hydrogen Energy 44 (2019) 8460–8466.
(2004) 501–513. [130] H. Yuan, H. Dai, W. Wu, J. Xie, X. Wei, A fuzzy logic PI control with feedforward
[101] S. Strahl, R. Costa-Castell

o, Model-based analysis for the thermal management of compensation for hydrogen pressure in vehicular fuel cell system, Int. J. Hydrogen
open-cathode proton exchange membrane fuel cell systems concerning efficiency Energy 46 (2020) 5714–5728.
and stability, J. Process Control 47 (2016) 201–212. [131] H.N. Vu, D.T. Le Tri, H.L. Nguyen, Y. Kim, S. Yu, Multifunctional bypass valve for
[102] B. Yang, J. Li, Y. Li, Z. Guo, K. Zeng, H. Shu, P. Cao, Y. Ren, A critical survey of water management and surge protection in a proton-exchange membrane fuel cell
proton exchange membrane fuel cell system control: summaries, advances, and supply-air system, Energy 278 (2023) 127696.
perspectives, Int. J. Hydrogen Energy 47 (2022) 9986–10020. [132] W. Wan, Y. Yang, Y. Li, C. Xie, J. Song, Z. Deng, J. Tan, R. Zhang, Operating
[103] S. Sharma, G. Babu, A new control strategy for a higher order proton exchange conditions combination analysis method of optimal water management state for
membrane fuel cell system, Int. J. Hydrogen Energy 45 (2020). PEM fuel cell, Green Energy Intell. Transport. 2 (2023) 100105.
[104] Y. Wang, M. Li, J. Gao, H. Chu, H. Chen, Fault-tolerant control through dynamic [133] S. Raman, S. Swaminathan, S. Bhardwaj, H.K. Tanneru, B. Bullecks,
surface triple-step approach for proton exchange membrane fuel cell air supply R. Rengaswamy, Rapid humidity regulation by mixing of dry and humid gases
systems, Int. J. Hydrogen Energy 47 (2022) 1804–1819. with feedback control for PEM fuel cells, Int. J. Hydrogen Energy 44 (2019)
[105] M. Karimi, Voltage control of PEMFC using A new controller basedon 389–407.
reinforcement learning, International Journal of Information and Electronics [134] H. Fu, J. Shen, L. Sun, K.Y. Lee, In-depth characteristic analysis and wide range
Engineering 2 (2012) 752–756. optimal operation of fuel cell using multi-model predictive control, Energy 234
[106] P. Wang, Y. Ma, J. Li, Y. Gao, Y. Zhang, D. Ma, A novel control algorithm of the air (2021) 121226.
supply subsystem: based on dynamic modeling of proton exchange membrane fuel [135] A.Y. Karnik, J. Sun, A.G. Stefanopoulou, J.H. Buckland, Humidity and pressure
cell, Processes 10 (2022) 1499. regulation in a PEM fuel cell using a gain-scheduled static feedback controller,
[107] G.B. Kareem, B. Ganesan, Robust analytical proportional-integral-derivative IEEE Trans. Control Syst. Technol. 17 (2008) 283–297.
tuning rules for regulation of air pressure in supply manifold of proton exchange [136] A. Yuehua Li, B. Pucheng Pei A, C. Ziyao Wu A, A. Huachi Xu, A. Dongfang Chen,
membrane fuel cell, Asia Pac. J. Chem. Eng. 16 (2020) e2569. A. Shangwei Huang, Novel approach to determine cathode two-phase-flow
[108] R. Pan, D. Yang, Y. Wang, Z. Chen, Performance degradation prediction of proton pressure drop of proton exchange membrane fuel cell and its application on water
exchange membrane fuel cell using a hybrid prognostic approach, Int. J. Hydrogen management, Appl. Energy 190 (2017) 713–724.
Energy 45 (2020) 30994–31008. [137] L. Shi, B.P. Setzler, Y. Yan, Understanding the Ebalance for water management in
[109] D. Zhao, D. Kr€ ahenbühl, B. Blunier, C. Zwyssig, M. Dou, A. Miraoui, Design and hydroxide exchange membrane fuel cells, J. Power Sources 536 (2022) 231514.
control of an ultra high speed turbo compressor for the Air Management of fuel [138] M. Song, P. Pei, H. Zha, H. Xu, Water management of proton exchange membrane
cell systems, in: 2012 IEEE Industry Applications Society Annual Meeting, 2012, fuel cell based on control of hydrogen pressure drop, J. Power Sources 267 (2014)
pp. 1–6. 655–663.
[110] J. Li, T. Yu, A new adaptive controller based on distributed deep reinforcement [139] A. Haddad, R. Bouyekhf, A.E. Moudni, Dynamic modeling and water management
learning for PEMFC air supply system, Energy Rep. 7 (2021) 1267–1279. in proton exchange membrane fuel cell, Int. J. Hydrogen Energy 33 (2008)
[111] Z. Baroud, M. Benmiloud, A. Benalia, C. Ocampo-Martinez, Novel hybrid fuzzy- 6239–6252.
PID control scheme for air supply in PEM fuel-cell-based systems, Int. J. Hydrogen [140] K. Ou, W.-W. Yuan, M. Choi, S. Yang, Y.-B. Kim, Performance increase for an open-
Energy 42 (2017) 10435–10447. cathode PEM fuel cell with humidity and temperature control, Int. J. Hydrogen
[112] B. Kai Ou A, B. Ya Xiong Wang, A. Zhen Zhe Li, A. Yun De Shen, A. Dong Ji Xuan, Energy 42 (2017) 29852–29862.
Feedforward fuzzy-PID control for air flow regulation of PEM fuel cell system, Int. [141] L. Zhang, P. Mu, S. Quan, Model predictive control of water management in
J. Hydrogen Energy 40 (2015) 11686–11695. PEMFC, J. Power Sources 180 (2008) 322–329.
[113] D. Zhao, F. Li, M.A. Rui, G. Zhao, Y. Huangfu, An unknown input nonlinear [142] H. Fu, J. Shen, L. Sun, K.Y. Lee, Fuel cell humidity modeling and control using
observer based fractional order PID control of fuel cell air supply system, IEEE cathode internal water content, Int. J. Hydrogen Energy 46 (2021) 9905–9917.
Trans. Ind. Appl. 56 (2020) 5523–5532. [143] C. Damour, M. Benne, B. Grondin-Perez, J.P. Chabriat, B.G. Pollet, A novel non-
[114] L.A. Zhao, A. Hc, P.B. Lian, A. Xy, A. Sx, Z.A. Tong, Feedforward-decoupled closed- linear model-based control strategy to improve PEMFC water management – the
loop fuzzy proportion-integral-derivative control of air supply system of proton flatness-based approach, Int. J. Hydrogen Energy 40 (2015) 2371–2376.
exchange membrane fuel cell, Energy 240 (2021) 122490. [144] T. Kurz, A. Hakenjos, J. Krmer, M. Zedda, C. Agert, An impedance-based predictive
[115] A. Thomya, Y. Khunatorn, Design of control system of hydrogen and oxygen flow control strategy for the state-of-health of PEM fuel cell stacks, J. Power Sources
rate for proton exchange membrane fuel cell using fuzzy logic controller, Energy 180 (2008) 742–747.
Proc. 9 (2011) 186–197. [145] B. Xie, G. Zhang, Y. Jiang, R. Wang, X. Sheng, F. Xi, Z. Zhao, W. Chen, Y. Zhu,
[116] K. Ou, Y.-X. Wang, Z.-Z. Li, Y.-D. Shen, D.-J. Xuan, Feedforward fuzzy-PID control Y. Wang, “3Dþ 1D” modeling approach toward large-scale PEM fuel cell
for air flow regulation of PEM fuel cell system, Int. J. Hydrogen Energy 40 (2015) simulation and partitioned optimization study on flow field, eTransportation 6
11686–11695. (2020) 100090.

14
Y. Wang et al. Energy Reviews 3 (2024) 100054

[146] T. Zeng, C. Zhang, Z. Huang, M. Li, S.H. Chan, Q. Li, X. Wu, Experimental [156] J. Han, S. Yu, S. Yi, Advanced thermal management of automotive fuel cells using
investigation on the mechanism of variable fan speed control in Open cathode a model reference adaptive control algorithm, Int. J. Hydrogen Energy 42 (2016)
PEM fuel cell, Int. J. Hydrogen Energy 44 (2019) 24017–24027. 4328–4341.
[147] L. Huang, J. Chen, Z. Liu, M. Becherif, Adaptive thermal control for PEMFC [157] Jenn-Jiang Hwang, Thermal control and performance assessment of a proton
systems with guaranteed performance, Int. J. Hydrogen Energy 43 (2018) exchanger membrane fuel cell generator, Appl. Energy 108 (2013) 184–193.
11550–11558. [158] Y. Wang, X. Kang, Z. Chen, A survey of digital twin techniques in smart
[148] L. Yin, Q. Li, T. Wang, L. Liu, W. Chen, Real-time thermal Management of Open- manufacturing and management of energy applications, Green Energy and
Cathode PEMFC system based on maximum efficiency control strategy, Asian J. Intelligent Transportation 1 (2022) 100014.
Control 21 (2019) 1796–1810. [159] Y. Zhang, C. Huang, H. Huang, J. Wu, Multiple learning neural network algorithm
[149] Y.-X. Wang, Q. Chen, J. Zhang, H. He, Real-time power optimization for an air- for parameter estimation of proton exchange membrane fuel cell models, Green
coolant proton exchange membrane fuel cell based on active temperature control, Energy and Intelligent Transportation 2 (2023) 100040.
Energy 220 (2021) 119497. [160] M. Li, Y. Wang, P. Yu, Z. Sun, Z. Chen, Online adaptive energy management
[150] T.H. Kwan, X. Wu, Q. Yao, Multi-objective genetic optimization of the strategy for fuel cell hybrid vehicles based on improved cluster and regression
thermoelectric system for thermal management of proton exchange membrane learner, Energy Convers. Manag. 292 (2023) 117388.
fuel cells, Applied Energy -Barking Then Oxford- 217 (2018) 314–327. [161] Y. Wang, Z. Sun, X. Li, X. Yang, Z. Chen, A comparative study of power allocation
[151] T. Ma, W. Lin, Y. Yang, M. Cong, Q. Zhou, Research on control algorithm of proton strategies used in fuel cell and ultracapacitor hybrid systems, Energy 189 (2019)
exchange membrane fuel cell cooling system, Energies 12 (2019) 3692. 116142.
[152] L. Sun, G. Li, Q.S. Hua, Y. Jin, A hybrid paradigm combining model-based and [162] L. Wang, M. Li, Y. Wang, Z. Chen, Energy management strategy and optimal sizing
data-driven methods for fuel cell stack cooling control, Renew. Energy 147 (2020) for hybrid energy storage systems using an evolutionary algorithm, IEEE Trans.
1642–1652. Intell. Transport. Syst. 23 (2021) 14283–14293.
[153] W.W. Yuan, K. Ou, Y.B. Kim, Thermal management for an air coolant system of a [163] Y. Wang, Z. Sun, Z. Chen, Energy management strategy for battery/
proton exchange membrane fuel cell using heat distribution optimization, Appl. supercapacitor/fuel cell hybrid source vehicles based on finite state machine,
Therm. Eng. 167 (2019) 114715. Appl. Energy 254 (2019) 113707.
[154] J.R. Kolodziej, Thermal dynamic modeling and nonlinear control of a proton [164] L. Wang, M. Li, Z. Chen, An energy management strategy for hybrid energy storage
exchange membrane fuel cell stack, J. Fuel Cell Sci. Technol. 4 (2007) 255–260. systems coordinate with state of thermal and power, Control Eng. Pract. 122
[155] L. Sun, Y. Jin, F. You, Active disturbance rejection temperature control of open- (2022) 105122.
cathode proton exchange membrane fuel cell, Appl. Energy 261 (2020) 114381.

15