Towards More Efficient PEM Fuel Cells Through Advanced Thermal Management: From Mechanisms to Applications

Abstract

Proton membrane exchange fuel cells (PEMFCs) provide an important energy solution to decarbonizing transport sectors and electric systems due to zero carbon emission during the operating process, and how to enhance the system efficiency of PEMFCs is one of the most challengeable issues to hinder the large-scale commercial application of PEMFCs. In recent years, numerous studies have been conducted to explore the feasibility and techno-economic performance of advanced thermal management to promote the efficiency of PEMFC systems. The thermal management of PEMFCs can be implemented from two aspects: one is efficient cooling methods to maintain the PEMFC under proper working temperature range, and the other one is waste heat recovery from PEMFCs to improve the overall system efficiency. Concentrated on these topics, many achievements have been gained by academic and industrial communities, and it is imperative to analyze and conclude these experienced studies from mechanism, technology, and application aspects. Therefore, this review summarized the great advances of thermal management of PEMFCs with efficient cooling and waste heat recovery for the sake of improving the overall efficiency of PEMFC systems, providing guidelines for the future design and optimization of PEMFC systems.

1. Introduction

Proton exchange membrane fuel cells (PEMFCs) have the advantages of low operating temperature and high reliability with quick startup response, which make it a promising candidate for vehicle applications [1,2,3]. PEMFCs are gathering significant attention as a promising solution for the transportation electrification in the Chinese market as shown in Figure 1, due to higher power density and shorter refueling durations compared to conventional electric battery systems [4,5,6]. One of the most important performance indicators for PEMFCs is their system efficiency. From year 2010 to 2020, the measured system efficiency of the worldwide PEMFC passenger vehicle increased from 35% to 60% under a specific testing driving cycle [7,8]. The latest target efficiency of heavy-duty commercial vehicles is set to 68% by 2030 and 72% for the ultimate target by America DOE [9]. Therefore, there is still a long way to improve the thermal efficiency of PEMFCs to propel the commercialization process [10].

Figure 2 shows a typical fuel cell vehicle layout. The core parts of this layout include the fuel cell stack, battery pack, drive motor, thermal and air management system, hydrogen storage system, and electronic control system, which work together to meet the vehicle’s energy requirements and driving performance. A PEMFC system comprises the fuel cell stack and the balance of plant (BOP): the former converts the chemical energy of hydrogen into electricity, while the latter supplies the reactants and ensures the expected reaction boundary conditions for the electrochemical reaction [11,12]. Since it appeared in public, research on improving fuel cell efficiency from stack design and material definition side has never stopped. These works include optimizing catalysts in MEA to improve the electrochemical reactions kinetically [13,14,15], developing a novel membrane with high proton conductivity and low electrical resistance [16], using a metal gas diffusion layer (GDL) with lower contact resistance between the bipolar plate (BP) and GDL [17], water management [18], and designing flow fields to minimize the flow resistance both for reactant gas and cooling systems [12]. Meanwhile, there are operation-oriented solutions to increase the fuel cell system efficiency of PEMFCs, which call for elaborate effort from the vehicle side, including the adjustment of PEMFC operation parameters like temperature [19,20], pressure [21], and reactant stoichiometry ratio [19], the decrease in parasitic power of the air-boosting system from the control side [22], the improvement of water removal methods [23], the modification of system topology by integrating the fuel cell system with energy storage systems [11,24], and the incorporation of a new energy management strategy [25,26]. Though significant improvements in fuel cell performance have been achieved through fundamental and technical solutions over the past decades, cost and reliability remain the challenges for its commercialization [27,28]. Continuous system efficiency increase becomes even more urgent to decrease the fuel cell stack’s product and usage cost. A simulation study carried out by Argonne National Laboratory [29] showed that an aggressive fuel cell system peak efficiency target (from 60% to 68%) could increase fuel economy from 10 to 15% while slightly decreasing cost. The cost decrease is primarily due to the reduction in hydrogen tank cost (8 to 13%).

Compared with the efficiency improvement from the component side, system-level solutions have gotten more and more attention in recent years, which calls for intensive research into the interaction among components (hardware), material property, control strategy, and operation environment optimization at the system implementation level [31,32,33,34]. Around 30~50% of hydrogen’s chemical energy is converted into heat during the electrochemical reaction process inside the PEMFC [35,36]. Effective thermal management is indispensable to control the working temperature in a stable range for the safe and efficient operation of PEMFCs [37]. The thermal management system stands among the quartet of pivotal subsystems that require precise operation and control to optimize fuel cell system performance [38]. The thermal management of PEMFCs includes two aspects. On the one hand, proper cooling methods are incorporated to maintain PEMFCs under the appropriate temperature range for higher working efficiency. On the other hand, recovering waste heat from the exhaust gas and cooling water enhances the overall efficiency of PEMFC systems. This review aims to analyze the progress of the thermal management of PEMFCs related to the improvement of thermal efficiency from mechanisms to applications. The structure of this review is organized as follows: after the background introduction, in Section 2, a brief introduction to the working principles of the PEMFC is presented, along with its associated voltage and efficiency model, which serve as the criteria and basis for the following work. Section 3 summarizes efficiency improvement from cooling methods, with a detailed comparison analysis between liquid-cooling and phase change cooling, where the former is chosen as the leading solution out of consideration of weight, volume, cost, compatibility, and cold start performance. Subsequently, the latest developments in the optimal operating temperature setting and control, temperature uniformity and consistency control, and parasitic power optimization are further stated. Section 4 introduces waste heat recovery from the cooling system, as 80% of the generated heat needs to be dissipated by the cooling system. In addition to thermodynamic cycles, typical topologies and achievements of waste heat recovery methods, such as combined heat, cooling, and power, are reviewed in detail to clarify a clear path. In the Conclusions Section, a technical roadmap for efficiency enhancement and the future development trend are proposed.

2. Working Principle of PEMFC

A fuel cell is a kind of electrochemical device that can convert chemical energy into electric power. As schematically shown in Figure 3, a single PEMFC consists of a proton-conductive membrane in the middle, electrodes and a coated platinum catalyst layer (CL), an MPL, or micro-porous layer (sometimes not used), a gas diffusion layer (GDL), and bipolar plates in both anode and cathode, where the sandwiched structure of membrane, CL and GDL MEA is are called membrane electrode assembly (MEA). The fuel cell completes the following functions: passages for species and heat transport and electronic connection between the bipolar plates and catalyst layers. According to the literature, the oxygen reduction reaction (ORR) at the cathode CL releases the most heat during normal operation [39], so the coolant channel usually lies inside the BP at the cathode side.

2.1. Voltage Model

The working process of a fuel cell is a controlled electrochemical reaction that simultaneously produces electricity, water, and heat. Theoretically, the fuel cell’s efficiency can be influenced by its output voltage.

At the anode, the hydrogen is decomposed into protons and electrons in the catalyst layer of the anode to trigger a hydrogen oxidation reaction (HOR). The electrons are transferred via an external circuit to the cathode which constitutes the output current in fuel cells. At the cathode, oxygen from the air system combined with the protons migrating through the PEM internally combine to trigger an oxygen reduction reaction (ORR). The overall reaction is listed as Equation (1) [40]:

$${\mathrm{H}}_2+{\displaystyle \frac{1}{2}}{\mathrm{O}}_2={\mathrm{H}}_2\mathrm{O}$$

(1)

At standard temperature and pressure condition, with the product water in liquid phase, the total enthalpy change of the above thermodynamic process is $\mathsf{\Delta }\mathrm{H}$: −286 kJ/mol, the Gibbs free-energy change during the process is $\mathsf{\Delta }\mathrm{G}$: −237.3 kJ/mol, so the maximum thermos efficiency ${\mathsf{\epsilon }}_{\mathrm{thermo}}$ of the system is listed as Equation (2) [40]:

$${\mathsf{\epsilon }}_{\mathrm{thermo}}={\displaystyle \frac{\mathsf{\Delta }\mathrm{G}}{\mathsf{\Delta }\mathrm{H}}}=83\%$$

(2)

The reversible voltage ${\mathrm{E}}_{\mathrm{nerst}}$ under ideal temperature and pressure conditions can be calculated as Equation (3) [28,41,42]:

$${\mathrm{E}}_{\mathrm{nerst}}={\displaystyle \frac{\mathsf{\Delta }\mathrm{G}}{\mathrm{n}\mathrm{F}}}=1.229\mathrm{V}$$

(3)

where n is the number of moles of electrons transferred, in this case n2 according to Equation (1), and F is Faraday’s constant, F = 96,845).

The Nernst equation for this reaction under real pressure and activities can be written as Equation (4) [43]:

$${\mathrm{E}}_{\mathrm{rev}}={\mathrm{E}}_{\mathrm{nerst}}-{\displaystyle \frac{\mathrm{R}\mathrm{T}}{2\mathrm{F}}}\ln {\displaystyle \frac{{\alpha }_{{\mathrm{H}}_2\mathrm{O}}}{{\alpha }_{{\mathrm{H}}_2}{\alpha }_{{\mathrm{O}}_2}^{1/2}}}$$

(4)

If the fuel cell is operated below 100 °C, so that liquid water is produced, the activity of water is set to unity ${\mathsf{\alpha }}_{{\mathrm{H}}_2\mathrm{O}}=1$. Activities of hydrogen and oxygen gases are replaced by their unitless partial pressures $({\mathsf{\alpha }}_{{\mathrm{O}}_2}={\mathrm{P}}_{{\mathrm{O}}_2}^{\mathrm{in}},{\mathsf{\alpha }}_{{\mathrm{H}}_2}={\mathrm{P}}_{{\mathrm{H}}_2}^{\mathrm{in}})$. The Nernst equation for this reaction under real temperature, pressure, and activities can be expressed further as Equation (4) [40,43,44]:

$${\mathrm{E}}_{\mathrm{rev}}=1.229-{\displaystyle \frac{\mathsf{\Delta }\mathrm{s}}{2\mathrm{F}}}(\mathrm{T}-298)+{\displaystyle \frac{\mathrm{R}\mathrm{T}}{2\mathrm{F}}}(\ln {\mathrm{P}}_{{\mathrm{H}}_2}^{\mathrm{in}}+{\displaystyle \frac{1}{2}}\ln {\mathrm{P}}_{{\mathrm{O}}_2}^{\mathrm{in}})$$

(5)

The entropy of the reaction $\mathsf{\Delta }\mathrm{s}$ may differ depending on the mixing phase of liquid or gas H₂O in the reaction product, as the coefficient of the temperature effect ranges from $2.298\times {10}^{-4}\mathrm{to}0.910\times {10}^{-3}$ [40,43,44]. For high temperature PEMFCs, the coefficient becomes even lower, so the voltage will become higher [45]. According to Equation (5), reversible output voltage ${\mathrm{E}}_{\mathrm{rev}}$ decreases as the reaction temperature increases.

As depicted in Figure 3, the fuel cell’s output voltage is lower than the reversible voltage due to polarization loss: in the low current density area, activation loss is the dominant one. As the output current density increases, ohmic loss due to ionic and electrical transfer activation loss causes the output voltage to decrease further. In the high current density area, concentration loss becomes dominant.

Referring to the power density distribution as depicted in Figure 3, the highest power density of the fuel cell is located in the medium to high current density area. For real vehicle applications, high power density is desired, which makes the activation loss and ohmic loss unavoidable in real vehicle operation. The fuel cell output is expressed as follows [46]:

$${\mathrm{V}}_{\mathrm{cell}}={\mathrm{E}}_{\mathrm{nerst}}-{\mathsf{\eta }}_{\mathrm{act}}-{\mathsf{\eta }}_{\mathrm{ohm}}-{\mathsf{\eta }}_{\mathrm{con}}$$

(6)

where

• ${\mathsf{\eta }}_{\mathrm{act}}$: activation losses due to reaction kinetics; generally, it could be described by the Butler–Volmer equation or the simpler Tafel equation.
• ${\mathsf{\eta }}_{\mathrm{ohm}}$: ohmic losses from charger transport, e.g., ionic and electronic resistance
• ${\mathsf{\eta }}_{\mathrm{con}}$: concentration losses due to the mass transport of reactants and products inside a fuel cell stack.

As shown in Figure 4, under medium to high load conditions, which represent the typical operating range for FCEVs, the output voltage of a single fuel cell is generally in the range of 0.6~0.7 volts. To meet the specific high drive voltage requirement for vehicle operation, hundreds of fuel cells are electrically connected in series through bipolar plates, which inevitably leads to an increase in cost and space.

2.2. Efficiency Model

The final output efficiency of a fuel cell stack is less than the reversible thermodynamic efficiency due to the voltage losses and fuel utilization losses. Real efficiency of a fuel cell can be expressed as follows:

$${\mathsf{\epsilon }}_{\mathrm{stack}}={\mathsf{\epsilon }}_{\mathrm{thermo}}{\mathsf{\epsilon }}_{\mathrm{voltage}}{\mathsf{\epsilon }}_{\mathrm{fuel}}$$

(7)

$${\mathsf{\epsilon }}_{\mathrm{voltage}}={\displaystyle \frac{{\mathrm{V}}_{\mathrm{cell}}}{{\mathrm{E}}_{\mathrm{Nernst}}}}$$

(8)

where ${\mathsf{\epsilon }}_{\mathrm{thermo}}$ is the reversible thermodynamic efficiency of the fuel cell as defined in Equation (2), ${\mathsf{\epsilon }}_{\mathrm{fuel}}$ is the fuel utilization efficiency of the fuel cell, ${\mathsf{\epsilon }}_{\mathrm{voltage}}$ is the voltage efficiency of the fuel cell, ${\mathrm{V}}_{\mathrm{cell}}$ is the real operating voltage of the fuel cell; ${\mathrm{E}}_{\mathrm{Nernst}}$ is the thermodynamically reversible voltage of the fuel cell as defined in Equation (3).

Equation (7) represents the classical way to describe fuel cell efficiency, and is helpful in identifying the contribution for efficiency improvement. For real time vehicle operation, the output of the fuel cell changes frequently even in stable working point. It is feasible to calculate the overall efficiency in terms of energy over the driving cycle with the aid of the integral method.

The fuel cell stack efficiency is defined as the ratio of the electrical power output of the fuel cell stack to the theoretical maximum power that could be produced if all the chemical energy of the fuel were converted into electricity under ideal conditions.

$$\mathrm{E}\mathrm{f}{\mathrm{f}}_{\mathrm{F}\mathrm{C}\mathrm{Stack}}={\displaystyle \frac{{\displaystyle {\int }_{{\mathrm{t}}_1}^{{\mathrm{t}}_2}{\mathrm{V}}_{\mathrm{FCstack}}(\mathrm{t})\times {\mathrm{I}}_{\mathrm{FCstack}}(\mathrm{t})\mathrm{dt}}}{{\displaystyle {\int }_{{\mathrm{t}}_1}^{{\mathrm{t}}_2}{\mathrm{M}}_{{\mathrm{H}}_2}(\mathrm{t})\times {\mathrm{Q}}_{\mathrm{HV}}(\mathrm{t})}}}$$

(9)

where t represents time, ${\mathrm{V}}_{\mathrm{FCstack}}$: fuel cell voltage and current, ${\mathrm{M}}_{{\mathrm{H}}_2}$: hydrogen consumption rate, ${\mathrm{Q}}_{\mathrm{H}\mathrm{V}}$: higher heating value of hydrogen.

Concerning system efficiency, the parasitic power consumed by BOP shall be considered, e.g., thermal management system. Moreover, the system efficiency also includes the recovery of waste energy and it can be expressed as Equation (10) [7,49]:

$$\mathrm{E}\mathrm{f}{\mathrm{f}}_{\mathrm{F}\mathrm{C}\mathrm{S}\mathrm{y}\mathrm{s}\mathrm{t}\mathrm{e}\mathrm{m}}={\displaystyle \frac{{\displaystyle {\int }_{{\mathrm{t}}_1}^{{\mathrm{t}}_2}{\mathrm{V}}_{\mathrm{B}\mathrm{o}\mathrm{o}\mathrm{s}\mathrm{t}}(\mathrm{t}){\mathrm{I}}_{\mathrm{B}\mathrm{o}\mathrm{o}\mathrm{s}\mathrm{t}}(\mathrm{t})\mathrm{d}\mathrm{t}-{\displaystyle {\int }_{{\mathrm{t}}_1}^{{\mathrm{t}}_2}{\mathrm{P}}_{\mathrm{parasitic}}(\mathrm{t})\mathrm{d}\mathrm{t}+{\displaystyle {\int }_{{\mathrm{t}}_1}^{{\mathrm{t}}_2}{\mathrm{P}}_{\mathrm{recover}}(\mathrm{t})\mathrm{d}\mathrm{t}}}}}{{\displaystyle {\int }_{{\mathrm{t}}_1}^{{\mathrm{t}}_2}{\mathrm{M}}_{{\mathrm{H}}_2}(\mathrm{t}){\mathrm{Q}}_{\mathrm{H}\mathrm{V}}(\mathrm{t})}}}$$

(10)

$${\mathrm{P}}_{\mathrm{parasitic}}(\mathrm{t})={\mathrm{P}}_{\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{p}}+{\mathrm{P}}_{\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{r}\mathrm{m}\mathrm{o}}$$

(11)

A comprehensive technology assessment measurement of fuel cell stack efficiency and system efficiency referring to Equations (9) and (10) was conducted by America DOE through a 2016 Toyota Mirai. It was tested at varied steady-state vehicle speeds and loads with in-depth instrumentation to create the efficiency map across the whole power level in a controlled test environment. As shown in Figure 5, the measured peak fuel stack efficiency and system efficiency are 66.0% and 63.7%, respectively. The continuous maximum power is 72 kW, the maximum fuel cell stack output power is 110–114 kW, and the maximum fuel cell system output power is 92 kW with a system efficiency around 40%. As the output power increases, the decrease in system efficiency is significant. For commercial vehicles that typically operate in the medium to high load range, this issue will be particularly prominent with an increase in hydrogen consumption. The problem can be mitigated by shifting the fuel cell load region as far left as possible by increasing the number of cells, but this will lead to a purchase cost increase for the end user. The increase in system efficiency will lead to a reduction in both purchase and usage costs.

3. Efficiency Improvement of PEMFC by Efficient Cooling

Around 30~50% of hydrogen’s chemical energy is converted into heat during the electrochemical reaction process inside the PEMFC [35,36]; the thermal management system is vital to guarantee the fuel cell’s proper work and it will become more prominent when fuel stack operates under high current density (HCD) conditions, as reported by Cai et al. [50], especially for commerical vehicle demanding stable high power. The target of the stack thermal management systems is to control the working temperature in a stable range with minimum temperature deviation throughout the stack and minimum parasitic loss.

3.1. Demonstrations of Cooling Methods in PEMFC

There are various cooling methods investigated for the thermal management of PEMFCs, including air cooling, liquid cooling, and phase change cooling [51]. Given the limitations inherent in air cooling concerning their efficacy and spaces, air-cooling methods are not deemed suitable for high-power PEMFCs in vehicular applications [51]. The liquid-cooling method is the most commonly used cooling method in PEMFCs, while phase change cooling indicates great potential due to the increasing power density of PEMFCs. Phase-change-cooling methods include various modes such as the evaporation, boiling, and cooling of liquid working fluids and melting/solidifying phase change materials. In some research, heat pipe cooling is also treated as phase change cooling [52]. The flow-boiling cooling process is characterized by the heated liquid rapidly transforming into vapor throughout its volume, and the saturation temperature is approximately 10~20 K lower than the optimum PEM stack operating temperature [53].

For stationary applications, the size and weight of a cooling system may not be a critical issue; however, for vehicle applications, where space and weight are always challenges, a comprehensive evaluation must be carried out during the early phase to balance the space, weight, cost, and heat rejection ability of the cooling system as well as its parasitic power. The liquid-cooling method has high power cooling ability, flexible controlled cooling ability, and it is also efficient in heat recovery. Liquid cooling is a suitable cooling method for FCEVs [54]. When the fuel cell stack power exceeds 5 kW, liquid cooling shows its advances, with lower parasitic power than air cooling and higher cooling capacity than phase change cooling [55]. For a cold start, a critical function in vehicular applications in low-temperature environments; by circulating the cooling fluids through the fuel cell stack, the ice formation within the PEMFC can be mitigated, thereby facilitating a smoother startup process [56,57]. Liquid cooling shows its uniqueness in the convenience of integrating another cooling system in FCEVs. Due to their maturity, liquid-cooling methods have been highly considered by industrial companies, especially from the component level, to achieve cooling performance. Honda applied a wave flow channel separator inside the fuel stack to enhance cooling efficiency for FCX clarity; the uniform cooling of the generating surfaces achieved by this method has enabled the number of cooling layers required for each cell to be reduced by 1/2 [8]. To maximize the performance of the Mirai fuel cell stack, precise temperature control over the vehicle operation load range is realized with the aid of a water control valve. The Toyota Mirai has a cooling water control valve (electromagnetic three-way valve). The cooling water control valve is specifically designed to distribute the cooling water amount between the radiator and bypass circuits [58].

Lots of research on phase change cooling has been carried out by numerical simulation. Fly et al. [52] made a comparison study between system layout requirement for evaporative and typical liquid cooling. Due to higher heat transfer coefficients in the condensing radiator during phase change, the radiator frontal area required for an evaporative-cooled system was 27% less than a conventional liquid-cooled system. Also, pressurizing the coolant channel facilitates the boiling temperature regulation, as validated via numerical simulation by Choi et al. [59,60]. For phase change cooling, the design of cooling devices has also received sustained attention and research, e.g., flow-boiling-type cooling plates with multi-microchannels; and heat pipe in the shape of ultra-thin vapor chambers are proven to be effective in temperature uniformity control [53,61]. HFE-7100, with a boiling temperature of 61 °C, is a recognized two-phase working fluid for low-temperature PEMFC cooling with low temperature deviation [60,62]. There are few papers reporting on experimental studies on two-phase cooling for fuel cells, apart from Garrity et al. [53,63].

Cooling methods with phase change materials (PCMs) have emerged as a potential technology due to the high latent heat of PCM. Guo et al. [64] invented a heat-peak regulator via a thermal accumulator filled with phase change material that separately exchanges heat with the fuel cell coolant and the air conditioner refrigerant. At a circumstance temperature of 38 °C, the maximum temperature of the fuel cell stack running can be reduced from 89 °C to 83 °C under a defined cycle, which shows its potential to reduce the parasitic power in the cooling system. PCM-based phase change cooling can also be combined with other cooling methods. Chen et al. [65] proposed a delayed liquid-cooling strategy in a hybrid battery thermal management system (TMS) combined by liquid cooling and phase change materials (PCMs), which reduces cooling power consumption by 33.3% without sacrificing the cooling performance. James et al. [66] studied the potential of a hybrid PCM/liquid-cooling system fuel-cell-powered heavy-duty truck where 80% of heat rejection is dissipated by the cooling fan into the air, and the remaining 20% is dissipated by PCMs. The parasitic power of the fan is only 35% compared to that of 100% cooling by liquid flow. The required phase change material weighs up to 79 kg with a volume of up to 97 L when the phase change material is set to carnauba wax, which has the lowest mass–cost combination. PCM-based thermal management systems incrementally affect the weight of battery modules and their cost is extremely high. In addition, most PCMs possess very low thermal conductivity, which restricts the cooling ability used in thermal management systems.

Fan et al. [67] prove that when waste heat recovery from Meg watts stationary PEMFCs is combined with cooling technique selection, liquid cooling is more suitable for places where winter temperatures are above zero because it saves cost, while phase change cooling presents better adaptivity in colder zones. As summarized in

Table 1. Comparison between typical cooling strategies for PEMFCs in vehicle scenario.

Cooling Strategy	Techniques	Advantages	Disadvantages/Challenges
Liquid cooling	-Cooling channels integrated in bipolar plates	-Strong cooling capability -Flexible control of cooling capability -Cost and space	-Radiator size -Cooling fan size -Coolant degradation -Large parasitic power
Phase change cooling	-Through flow-boiling	-Elimination of coolant pump -Simplified system -Cooling homogeneity	-Development of suitable working media -Cold start -Cost, space, and weight -Instant high cooling power -Engineering experience

, for mobile applications, the liquid-cooling method is suitable for cost, weight, volume, compatibility, and cold start performance.

3.2. Control Strategies Used in Cooling Systems of PEMFC

Among all the cooling methods, the liquid-cooling method is widely adopted in commercial PEMFCs. As depicted in Figure 6, a typical liquid-cooling system used in PEMFC vehicles comprises a radiator, a reservoir, a cooling pump, an electric radiator fan, and a bypass valve or electric three-way valve [58,68]. The former two constitute the main sources of parasitic power in cooling systems. When the temperature of the outlet coolant is much lower than the upper limit value, the cooling fluid can directly return to the stack back, bypassing the circuit of the radiator, thus saving the parasitic power consumption of the circulating pump through a three-way thermostat valve [68]. A coolant control valve driven by a stepping motor can be applied instead to distribute the coolant flow circulated by a coolant pump or radiator. The coolant pump and radiator fan are the two indispensable heat-dissipating control devices. To enhance the cooling ability, the successful implementation of a cooling system also relies on proper control strategies. Significant research gaps still exist to further advance the integration of coupled TMS schemes with decoupled temperature control strategies and algorithms into high-power liquid-cooling PEMFC stack, especially for FCEVs, which are frequently subjected to transient high-power thermal load variation, high durability requirements, and cost constraints [69,70].

3.2.1. Optimum Control of PEMFC Temperature

The optimal operating temperature is largely dictated by the mitigation of electrode “flooding”, PEM hydration, and material degradation [71]. Under low-temperature conditions, water vapor condenses into liquid water, which can accumulate within the porous structure and impede the diffusion of reactants to the catalytic sites, thereby increasing the transport resistance and leading to the concentration polarization of the electrodes [69,72]. Under high operation temperature, though the fuel cell reversible voltage shall decrease theoretically as Equation (4) expressed, the measured FC efficiency increases due to the enhanced electrode kinetics for the electrochemical reaction, increased membrane proton conductivity, and improved mass transfer of the reactants [19,73]. However, a too high temperature will cause the membrane to dry out and reduce the surface area of the catalyst, which will in turn weaken the stack performance [74]. The effect of temperature on the efficiency of PEMFCs is demonstrated by Desantes et al. [75] through experiment, as shown in Figure 7. In this figure, it can be seen that with the increase in the current load, the temperature influence increases to 3% (absolute value) at the peak load. Zhiani. et al. [73] compared the action performance through a comparison test between low temperature (55 °C) and pressure (34 kPa) and high temperature (75 °C) and pressure (172 kPa) and demonstrated the effect of temperature on the activation of PEMFCs and consequently, the efficiency output.

There are various approaches to achieve proper working temperature of PEMFCs by cooling system integrated with the controller, including Linear Quadratic Integral (LQI) controllers, adaptive Linear Quadratic Regulator (LQR) controllers, fuzzy logic controllers optimized by genetic algorithms, and fuzzy PID control strategies. Sun et al. [76] proposed a multi-model predictive control method based on an adaptive switching strategy and multi-object optimization; compared to traditional PID control, model predictive control, the multi-mode predictive control model is proven to be more robust in precise temperature within a deviation of 1 K. Hu et al. [77] made a comparison among different temperature-controlling strategies on hydrogen consumption including constant temperature setting, rule-based temperature setting, and optimal temperature-tracking control. The optimal temperature-tracking control can save around 25% of hydrogen consumption under CHTC-HT cycle, a China-specific driving cycle for heavy-duty commercial vehicles with GVW > 55,000 kg. To save the vast experiment load for optimizing the operation temperature for each working load in vehicle driving, Tang et al. [78] developed the automatic controller calibration tool, which was integrated with the dynamic model of fuel cell temperature behavior based on metaheuristic optimization algorithms and CSO-SVR model. Compared to steady working conditions, the effect of temperature becomes even more remarkable under dynamic conditions.

3.2.2. Temperature Uniform Control of PEMFC

The highly concentrated electrochemical reaction location, sandwich-like assembly, local overheating, and non-uniform temperature distribution significantly contribute to the attenuation and deterioration of stack performance [69,79]. As shown in Figure 8I, Zhang et al. [80] acquired temperature distribution in the anode/cathode flow fields of a liquid-cooled fuel cell stack with five fuel cells by three-dimensional multiphase non-isothermal modeling. Huang et al. [81] developed a thermal characteristic test platform for hydrogen fuel cells wherein the infrared temperature for individual cells is measured for a 110 kW fuel cell stack. The test results show that the maximum surface temperature difference among the fuel cells can be up to 20 K, as shown in Figure 8II. These results demonstrate the importance of improving the temperature uniformity of PEMFC.

Rojas [72] presents a control-oriented modeling methodology for a liquid-cooled PEMFC. The model can predict temperature variation across the stack only via the voltage of individual cells. Thus, real-time temperature distribution inside the fuel cell stack is available without additional temperature-measuring devices. Cheng et al. [82] developed a model-based thermal management system of a 30 kW fuel cell for a hybrid city bus model. Due to significant time delay of the real thermal system, a thermal management controller consisting of nonlinear feedforward and Linear Quadratic Regulator (LQR) state feedback was implemented to avoid the large temperature fluctuation for precise temperature control. For better fuel cell temperature distribution control, Liu et al. [70] improved the fuel cell temperature distribution by regulating the input and output coolant water temperature in a decoupling way. A controllable electric pump was used to circulate the coolant water, and a radiator with controllable electric fan was used to release the excess heat into the atmosphere. The experimental results show that the temperature control deviation range is within 0.2 °C even under the dynamic load conditions.

3.2.3. Reduction of Parasitic Power by Optimal Control

According to a DOE technical report [66,83], for a Class 8 heavy-duty fuel cell with 313 kW electric power output, the parasitic power of the radiator fan is expected to be 27 kW in 2025, and the value shall decrease to 22 kW in 2030. The parasitic power of the radiator fan is expected to undergo a 5% reduction between technology years. The classic proportional and integral (PI) controller is not able to adapt to dynamic changes in the TMS’s running parameters due to varied thermal loads or external disturbance, so the parasitic power of the TMS cannot be optimized [84]. Subjected to higher current density output with unpredictable disturbances, more and more demand has been raised toward simultaneously improving system dynamic response and lessening parasitic power. It is an inspiring trend to combine the advantages of robust non-PID control and PID control for temperature-decoupled management [68].

Chen et al. [71] tried to minimize the parasitic power via optimization of the cooling channel and development of an innovative cooling strategy in parallel. Chang et al. [85] attempted to decrease the parasitic power of the thermal management system with precise fan speed control strategy. Han et al. [86] studied four different control strategies to optimize the parasitic power of a 75 kW automotive PEMFC where the control strategy for the coolant pump and radiator fan are varied. As shown in Figure 9, different cooling system control strategies shall be varied according to the working load level. For low current density output, a pump working with state feedback control and a fan working under PI control show the lowest parasitic power. For high-load driving conditions, constant speed control performs better. An excellent temperature controller integrated into TMS was recognized to satisfy these attributes as follows: optimal tracking of the set-point temperature, minimizing parasitic losses, and simplifying algorithm implemented by favorable hardware platform [69].

3.2.4. New Fundamental Strategy

As this review concentrates on the effect of thermal-managed systems, especially control-related, further effective energy management strategies such as external energy maximization strategy (EEMS) and the equivalent consumption minimization strategy (ECMS) will not be discussed further. However, the importance of this method will not be impacted. Different from the gradient descent approach applied in EEMS and ECMS, Quan et al. [26] have demonstrated an optimization method based on a direction prediction optimal foraging algorithm (OFA/DP), which has the advantages of high optimization capability and simple parameter definition; 38.62% minimization in hydrogen consumption has been achieved. From a vehicle engineering perspective, since energy management directly affects vehicle drivetrain control, the cutting-in phase of the entire project requires comprehensive and systematic consideration. It is observed that existing energy management strategies without integration of fuel cell or battery temperature dimension are more or less incapable of performing very well [87,88]. The issue of balancing efficient thermal control and high-efficiency output is an imperative matter for advanced TMS [89]. Kandidayeni et al. [90] studied the integration of temperature dimension into the energy management system, and the hydrogen consumption was found to be 5.3% lower with a bounded load following the strategy control algorithm. For the same reason, this review suggests integrating the design of the thermal system in terms of hardware and control strategy simultaneously at the earliest stage. Artificial intelligence (AI) has started demonstrating its potential in revolutionizing thermal system controls, offering significant improvements in efficiency, precision, and adaptability. Tian et al. [91] identified the optimal temperature for maximizing PEM fuel cell performance by integrating an artificial neural network (ANN) with the genetic algorithm (GA). As shown in Figure 10, the predicted current–voltage curve under different operation temperatures and varied current densities with AI method shows good consistency with the 3D simulation. A total of 1500 data generated from a 3D-validated multi-physics model of PEM fuel cells were used to train, validate, and test the ANN model. Each identification case took less than 1 s for a standard node, compared with about 20 h for predicting a single I-V curve using the 3D multi-physics simulation. After the ANN was properly trained, it was shown to predict fuel cell performance with an error of less than 2.5% and 1.5% under low and high current densities, respectively. Integrating artificial intelligence (AI) in thermal management systems involves using advanced algorithms and machine learning techniques to monitor, predict, and optimize system performance based on real-time data, especially as the computational capabilities of intelligent vehicles today keep increasing.

Kharat et al. [92] demonstrated that a reinforced learning agent (RL) exhibits fewer temperature fluctuations, better efficiency, lower fuel usage, and better cooling than PID and MPC baselines. A primary benefit of the AI method is the order of magnitude speedup in training and inference compared to conventional physics-based modeling. Sarani et al. [93] leveraged response surface methodology and artificial neural networks for predictive modeling and optimization of PEMFC performance under operating conditions. Both RSM and ANN models demonstrated high predictive accuracy, with R2 values of 98.66% and 99.11%, respectively, due to their ability to capture the nonlinear relationships between input parameters and system performance. Intelligent, data-driven control approaches will enable improved understanding and behavior of the object, leading to better performance and reliability [92]. Besides AI’s unique intuition in fuel cell components and material definition, artificial intelligence multi-object optimization (AI-MOO) [94,95] has shown its potential to enhance fuel cell efficiency from the thermal and energy management sides. Effective thermal control requires a good design of the cooling system.

4. Efficiency Improvement of PEMFC by Waste Heat Recovery

Owing to the narrow operating temperature range, low-grade heat recovery presents more significant challenges and is less viable compared to the recovery of high- and medium-grade heat [74]. As illustrated in Figure 11, waste heat from fuel cells can be reclaimed directly through heat utilization and indirectly through thermal cycles and thermoelectric conversion. Additionally, combined heat and power (CHP) subsystems and combined cooling and power (CCP) subsystems play pivotal roles. The conversion of heat into electricity via thermos generators constitutes a waste heat recovery (WHR) method that has gained significant attention.

4.1. Overall Energy Distribution in PEMFC

Chen and Wang et al. [71] revealed that the reactant gas flows in the gas flow channel (GFC) can only remove 5% water heat production in a fuel cell stack under the standard reaction stoichiometric ratio. A similar conclusion is that the heat taken away by the exhaust reactant in PEMFC is only about 3~10% [71,77,81,89,96] due to its low exhaust temperature. In few cases, a higher percentage of heat taken away by exhaust reactants is reported, which may depend on individual models and simulation conditions [97]. As depicted in Figure 12, 50% of the hydrogen high heating value (HHV) is converted into gross electric power and around 45% of hydrogen energy is converted into heat in a PEMFC [81]. Focusing on the heat created, around 11% is applied for vaporizing the cooling water, while around 80% of the heat produced is dissipated by the cooling system which becomes the primary waste heat source.

In China, low-grade/low-enthalpy sources below 200 °C cover 63% of the total amount of industry waste heat, and 70% of geothermal heat source temperatures are below 150 °C; low-grade WHR has broad promotion prospects [98]. The application of the cooling method significantly impacts the decision of WHR, as summarized by Nguyen et al. [54]. Usually there are two ways to utilize waste thermal energy below 200 °C: direct thermal utilization and conversion of thermal power (electricity), which will be further described in detail.

4.2. Waste Heat Recovery Methods Investigated in PEMFC

Due to the widely existing waste heat, various waste heat recovery technologies are investigated in PEMFCs, such as the Organic Rankine Cycle, and thermoelectric generators. This section will introduce and discuss these demonstrations in detail, providing guidelines for future waste heat recovery technology implementation in PEMFCs.

4.2.1. Organic Rankine Cycle

Organic Rankine Cycle (ORC) is a kind of Rankine cycle characterized by using organic fluid as the work medium; ORC can be applied to harvest waste energy from low-temperature heat resources like solar energy, geothermal energy, waste heat in internal combustion engine (ICE) coolant circuits, and exhaust gas and PEMFC [98,99,100]. Most of the ORC-related research being championed is modeling-based and does not present a real-life scenario of the fuel cell system investigated in the transport sector [28]. ORC is an essential technology for the low-grade heat recovery from PEMFC due to its relatively high thermal efficiency. Park et al. [101] summarize the current state-of-the-art experimental ORC system performance based on 200 scientific works according to specific selection criteria covering the waste temperature range. As shown in Figure 13, the overall heat to electrical power conversion efficiency was around 44% of the Carnot cycle efficiency. Taking the working temperature of 80 °C, the environment temperature of 20 °C, and the Carnot cycle efficiency defining the ceiling of ORC, which is around 17%, the average thermal efficiency of ORC is about 7.4%.

Representative research results on waste heat recovery with ORC are summarized in Table 2, where ORC is combined with other waste heat solutions, e.g., regenerator [102], recuperator, and metal hybrid for hydrogen storage [103], which brings around a 5% efficiency increase. In 2016, He et al. [104] reported a thermal efficiency of 4.73% achieved with a heat pump combined with ORC for WHR from the fuel cell of 49.8 kW. The ORC efficiency listed in Table 2 shows a comparable average value to that in Figure 13, and the low working temperature interval of ORC may be a possible reason. Therefore, the optimization of the ORC scenario is important to further enhance the thermal efficiency of the ORC and overall efficiency of PEMFC systems, such as combining regenerator [102] or recuperator [103]. In the meantime, experiment results carried by vehicle manufacturers such as Groupe PSA in France and Hino on engine coolant waste heat recovery show a promising value for vehicle application [100,105].

Zhao et al. [106] put forward a hybrid power system that utilizes ORC to recover waste heat from PEMFC, as depicted in Figure 14I. The findings demonstrate that the total electric efficiency of the hybrid power system can be increased by approximately 5% with the assistance of the ORC-based waste heat recovery system. This study also emphasized the effects of fuel flow rate, operating pressure, turbine inlet pressure, and backpressure on the efficiency of the thermodynamic process. He et al. [104] compared the performance of two ORC-based systems for recovering waste heat from PEMFCs. One is a standalone ORC system, while the other is an ORC and heat pump combined system, as shown in Figure 14II. The results reported a thermal efficiency of 4.73% for the ORC and heat pump combined system, and it was proved that the ORC and heat pump combined system is more feasible for the cooling of PEMFC. Azad et al. [107] analyzed the performance of two-stage ORC (STORC) and PEMFC integrated system, as shown in Figure 14III; the results indicate that operating a STORC with wasted heat from the PEMFC could improve overall efficiency by 1.9%. Liu et al. [108] proposed a new thermal management concept in which the organic material is applied both for phase change cooling and waste heat recovery; the system layout is shown in Figure 14IV. By using a specially designed cooling plate in the cooling circuit to generate a higher vapor content in the working fluid, more power is provided to drive the turbine, resulting in an ORC thermal efficiency of 7%.

Table 2. Typical waste heat recovery by ORC in PEMFCs.

Overheating/Recuperator	Waste Heat Source	Research Type	Organic Fluid	ORC Eff.	Overall Eff.	Improved Efficiency (Absolute Value)	Ref.
/	1007 kW Fuel cell	simulation	R123 R245ca R245fa	10.94% 10.70% 10.59%	/	5.24% 5.13% 5.08%	[106]
With transcritical CO2 cycle and cold energy of liquefied natural gas	1047 kW electric power output with 1190 kW heat	simulation	/	/	72%	33%	[109]
ORC	49.8 kW fuel cell	simulation	R245fa	4.03%	/	/	[104]
Heat pump and ORC	49.8 kW fuel cell		Water for HP R123 for ORC	4.73%	/	/
Recuperator + Metal Hybrid for Hydrogen storage	1180 kW fuel cell	simulation	R123	6.52%	44.3%	2.3%	[103]

Table 2. Typical waste heat recovery by ORC in PEMFCs.

Overheating/Recuperator	Waste Heat Source	Research Type	Organic Fluid	ORC Eff.	Overall Eff.	Improved Efficiency (Absolute Value)	Ref.
/	1007 kW Fuel cell	simulation	R123 R245ca R245fa	10.94% 10.70% 10.59%	/	5.24% 5.13% 5.08%	[106]
With transcritical CO2 cycle and cold energy of liquefied natural gas	1047 kW electric power output with 1190 kW heat	simulation	/	/	72%	33%	[109]
ORC	49.8 kW fuel cell	simulation	R245fa	4.03%	/	/	[104]
Heat pump and ORC	49.8 kW fuel cell		Water for HP R123 for ORC	4.73%	/	/
Recuperator + Metal Hybrid for Hydrogen storage	1180 kW fuel cell	simulation	R123	6.52%	44.3%	2.3%	[103]

In the combined system, there are numerous heat transfer processes occurring between the heat source and the working fluids. The exergy destroyed in the heat exchangers (evaporator and condenser) amounts to 74% of the overall exergy loss while the share of evaporator is higher. The below parameters must be comprehensively considered: pinch point temperature difference (PPTD) of the evaporator and condenser, heat exchanger effectiveness, pump and turbine efficiency, installation, environment protection, and running costs [110,111].

Comprehensive energetic and exergetic analyses were conducted by Sadeghi et al. [112] for an ORC system with the aid of AI. Employing the binary zeotropic mixture of R717/water as its working fluid, the optimized cycle system exhibits high thermal efficiency and more than 4% improvement of thermal and exergy efficiency compared to the conventional ORC cycle.

4.2.2. Other Thermodynamic Cycles

Except for ORC, other thermodynamic cycles also show the potential to recover waste heat from PEMFC and enhance the overall system efficiency, such as the CO₂ cycle and Kalina cycle. Ahmadi et al. [109] proposed a hybrid WHR system consisting of a transcritical carbon dioxide cycle and a liquefied natural gas to reduce the condensing temperature in a PEMFC. With the proposed system, the generated power of the system increases by 39%. The Kalina cycle can be used in low temperature applications for cogeneration with different power cycles to recover waste heat from gas turbine, diesel engine, geothermal heat, and solar energy as well [113]. It uses a zeotropic mixture of water and ammonia as working fluid to maximize the power output from the turbine and reduce the power input in the pumps [114]. Due to the advantage of heat utilization with ammonia–water mixture, the Kalina cycle usually shows higher efficiency than ORC with the same boundary condition [115]. The Kalina cycle offers many advantages compared to ORC, including superior performance, higher flexibility, and reduced heat transfer temperature difference between its working fluid and heat source. The Kalina cycle is recommended for further investigation for waste heat recovery in FCEV. Taking liquid nature gas (LNG) as heat sink, Wang et al. [116] improved the turbine output power by lowering the turbine back pressure. However, the working pressure level is higher in the Kalina cycle than that in the ORC cycle, which calls for high level engineering integration, and the risk of two-phase working fluid for ORC at the expander’s outlet is lower for the ORC cycle [117].

Wang et al. [118] proposed to classify the waste heat source into convext type, straight type, and concave type according to the heat profile’s Temperature–Enthalpy (T-H) diagram. For convex-type waste heat source, the thermal match between the organic working fluid’s evaporation curve and heat source is best, so ORC is the best selection, while for the straight and concave type of waste heat, the Kalina cycle is more suitable. The work can provide a reference to choose a suitable technology to recover low temperature waste heat for power generation in the process industry.

4.2.3. Combined Heat/Cooling and Power Based on PEMFC

The utilization of waste heat for cockpit heating [119,120], preheating of reactants [121], heating to facilitate water evaporation [122], and hydrogen release from metal hydride tanks [123] already demonstrate their advancements in the overall system’s power-saving capabilities, e.g., Quan et al. [120] has demonstrated 50.8% waste heat utilization after a systematic improvement via crucial parameter research. PEMFCs can also work as the main power in distributed combined heat/cooling and power (CHP/CCP) systems, which can provide electricity and reuse waste heat during electricity production with high energy efficiency. As shown in Figure 15, the PEMFC-based combined heat and power (PEMFC-CHP) system shows its advantage to provide electricity and heat to specific users, achieving a sustainable reduction in carbon emissions [124]. The influence of CHP/CCP systems on the enhancement of overall efficiency, particularly in low environmental temperature operations, is evident and can provide direct guidance for the design of Organic Rankine Cycle (ORC) systems [67]. AI is helpful in combined heat and power cycle for configuration optimization: as Mehdi et al. researched in their report, it is shown via multi-objective optimization of the proposed cycle that the exergy efficiency of the cycle increases by 31% and the cost of electricity generation decreases by 18% [125].

4.2.4. Thermoelectric Generators

The thermoelectric generator is an important low-grade heat recovery technology due to its simple structure, no moving parts, environment friendliness, extremely low noise [126]. For application, as early as 2009, BMW studied exhaust waste heat recovery in an ICE power BMW 530I; the conversion efficiency of TEGs measured is around 1~2.5% for typical driving cycles, showing a distinct reliance on waste heat source [127,128]. A fuel consumption of 3% reduction is reported from the Honda team through simulation and test in a series hybrid vehicle [129]. A TEG operates at approximately 20% of the Carnot efficiency over a wide temperature range, as summarized by [130]. Orr et al. [131] proposed to apply TEG and heat pump in combination to enhance the heat recovery efficiency and create a completely solid state and passive WHR system. On the contrary, Hasani et al. [132] concluded that using thermoelectric coolers can be a suitable solution for recovering waste heat from a PEM fuel cell. In their experiment, a heater is applied to simulate the waste heat produced as that from a 5 kW PEMFC; a specially designed waste heat recovery system can recover about 10% of waste heat transferred through cooling and the thermoelectric conversion efficiency is 0.35%.

4.2.5. Metal Hydrides

Nasri et al. [123] introduced the method of WHR for FCEVs with thermochemical energy storage metal hydrides. Using metal hydrides to store the waste heat of the powertrain components during normal operation and releasing it for the startup at very low ambient temperatures, the simulated range increases from 152 km to 178 km at −20 °C ambient temperature. A main barrier to the practical use of metal hydrides is the prohibitively high temperatures and pressures necessary for reversible operation [54,127]. Most metal hydrides cannot store much hydrogen, and such hydrides have slow kinetics and do not release hydrogen at low temperature [133]. This kind of waste heat recovery is not recommended for combination with liquid cooling [54].

Different from cooling system definition in PEMFCs, there is not a clear technical path for waste heat recovery for FECVs, even though combined heat and power as well as combined chilling and power proved a possible solution for vehicle application within narrow temperature range with one choice. Thermodynamic cycles can recover waste heat, showing that better environment compatibility though extensive research on matching and feasibility studies are urgently needed. Due to the enthusiasm shown by the automotive industry and academia in the field of waste heat utilization, especially in the utilization of individual cooling water waste heat, it is optimistic that with the increasing demand for system efficiency and the reduction in parasitic power in thermal management systems, research and development in waste heat utilization will be reignited. It is reasonable to expect more integrated cooling system and waste heat recovery from bottom-level hardware and configuration design [134].

5. Conclusions

The original motivation of this review is to summarize the latest achievements in thermal management of PEMFCs for improvement of thermal efficiency, specifically focusing on working mechanism and engineering application from the perspective of system integration and optimization. The first item is related to thermal management system itself to maintain the PEMFC under proper working temperature range, which means the pursuit of efficiency enhancement through the optimization of working temperatures, solutions aimed at temperature uniformity, and strategies for minimizing parasitic power consumption of the cooling system. The other one is waste heat recovery from PEMFCs to improve the overall system efficiency, including Organic Rankine Cycle, Kalina cycle, and thermoelectric generators, etc. Focusing specifically on the fuel cell stack, we anticipate an increment of approximately 5% in system efficiency through integrated thermal management system including efficient cooling and waste heat recovery, and this goal can be achieved with the aid of optimum operation temperature and control, parasitic power reduction, and waste heat recovery. The main conclusions and findings are drawn as follows:

(1)

Optimum operation temperature and control: Achieving a 2–3% increase in efficiency by maintaining the fuel cell stack at its optimal operating temperature. This is crucial for ensuring stable operation and maximizing performance, as the efficiency of the stack is greatly influenced by its temperature.

(2)

Parasitic power reduction: A further 1–2% efficiency gain can be realized by minimizing the parasitic power loss associated with auxiliary systems such as fans and water pumps and actuator valves.

(3)

Waste heat recovery: An additional 2–3% improvement in efficiency is feasible through the integration of an Organic Rankine Cycle (ORC) system for waste heat recovery. It is worth noting that the efficiency of the ORC system is bound by the Carnot cycle; its efficiency will further increase under cold conditions when the condenser working temperature becomes lower.

(4)

The provided information shall be valuable to guide the system layout and control strategies for ITMS with ORC at the early design stage.

Systematic research on the integration of the cooling system and waste heat recovery system shall be a focus in the recent future to study the overall thermal system behavior and parasitic power reduction for system efficiency improvement. With the marketization of fuel cell commercial vehicles and the increasing demand for high current density (HCD) fuel cells, the urgency of this research is further strengthened. Moreover, the integration of AI and enhanced computing capabilities is progressively enabling the engineering realization of comprehensive models, multi-objective optimization, and precise control systems that were previously unattainable. With the comprehensive embrace of this function by the industry, more in-depth and specific research on a certain function will increase explosively as AI is boosting intelligent vehicle research from every area.

Though there are some direct efficient methods to increase the fuel cell system efficiency through energy management, e.g., EEMS and direction prediction optimal foraging algorithm (OFA/DP)-based ECMS, this review concentrates on the thermal management side out of concern of integrated system design. This review mainly concentrates on the efficiency improvement side; performance degradation caused by non-standard operation conditions which may hinder the marketing of FCEVs shall also be considered in developing a new fuel cell system [135]. In addition to that, hybrid PEMFCs and battery are becoming a popular powertrain for unmanned underwater vehicles (UUVs), which also call for efficient thermal management system and high efficiency [136].