Hydrogen Fuel Cell Power System—Development Perspectives for Hybrid Topologies

Abstract

In recent years, the problem of environmental pollution, especially the emission of greenhouse gases, has attracted people’s attention to energy infrastructure. At present, the fuel consumed by transportation mainly comes from fossil energy, and the strong traffic demand has a great impact on the environment and climate. Fuel cell electric vehicles (FCEVs) use hydrogen energy as a clean alternative to fossil fuels, taking into account the dual needs of transportation and environmental protection. However, due to the low power density and high manufacturing cost of hydrogen fuel cells, their combination with other power supplies is necessary to form a hybrid power system that maximizes the utilization of hydrogen energy and prolongs the service life of hydrogen fuel cells. Therefore, the hybrid power system control mode has become a key technology and a current research hotspot. This paper first briefly introduces hydrogen fuel cells, then summarizes the existing hybrid power circuit topology, categorizes the existing technical solutions, and finally looks forward to the future for different scenarios of hydrogen fuel cell hybrid power systems. This paper provides reference and guidance for the future development of renewable hydrogen energy and hydrogen fuel cell hybrid electric vehicles.

1. Introduction

Compared with the level in 2020, the hydrogen consumption in 2050 may increase by six times to about 530 million tons [1]. According to the forecast results of the Hydrogen Roadmap Europe Report, by 2030 fuel cell electric vehicles (FCEVs) will account for 1 per 22 passenger cars and 1 per 15 light commercial vehicles [2]. Japan’s goal is to reach 200,000 FCEVs by 2025 and 800,000 FCEVs by 2030 [2]. Hydrogen fuel cells differ from traditional batteries in that they are not an energy storage device. To be more precise, they receive oxygen and hydrogen and quickly convert them into energy and water [3]. Hydrogen and oxygen are the reactant and oxidant of hydrogen fuel cells, respectively [4]. Hydrogen fuel cells have many advantages compared to traditional batteries:

• They have high efficiency [5] and high energy density [6,7]. The basic energy density of hydrogen is about 2.7 times that of gasoline, diesel, and natural gas, while the theoretical efficiency of hydrogen fuel cells is 83% and the actual efficiency can reach 60–70%. In the literature [8], Zi’ang Xiong et al. proved that the working efficiency of a hydrogen fuel cell-powered forklift is 40.6% higher than that of a lithium battery-powered forklift.
• They are green, pollution-free, and have no exhaust emissions [6,9]. Hydrogen fuel cells only produce water, without producing other greenhouse gas emissions.
• They have low vibration and noise [5], fast hydrogen supply [7], and wide operating temperature range.

Hydrogen fuel cells have great advantages and are bound to become a new widely used power supply mode. However, their disadvantages are low power density and difficulty in self-starting. These defects mean that it is difficult to use them as a sole power source. Thus, a hybrid power supply system dominated by hydrogen fuel cells is an effective way to solve the low power density of hydrogen fuel cells.

2. Modeling of Hydrogen Fuel Cells

At present, hydrogen fuel cells can be divided into five types according to electrolytes [10], including proton exchange membrane fuel cell (PEMFC), alkaline fuel cell (AFC), phosphoric acid fuel cell (PACF), solid oxide fuel cell (SOFC) and molten carbonate fuel cell (MCFC). Compared with other types of hydrogen fuel cells, PEMFC has the advantages of fast starting speed, low operating temperature, high efficiency, low investment, high output value, the most mature development, and great commercial value [6]. The fuel cell used in this paper is PEMFC. A single PEMFC is mainly composed of one proton exchange membrane (which is the most important part), two electrodes, two bipolar plates (80% of the weight of the whole PEMFC), and two gas diffusion layers [11]. The output voltage of the fuel cell is given by the following formula [12,13]:

$$V_{cell}=N_{cell}(E_{Nernst}-{\eta }_{act}-{\eta }_{ohm}-{\eta }_{con})$$

(1)

where $N_{cell}$ represents the number of monomers in series of the hydrogen fuel cell, $E_{Nernst}$ is theoretical output voltage of the fuel cell, ${\eta }_{act}$ represents activation overvoltage, ${\eta }_{ohm}$ represents ohmic overvoltage, and ${\eta }_{con}$ represents concentration overvoltage. The external characteristics of the fuel cell are shown in Figure 1.

It can be seen from Figure 1 that the output voltage curve of PEMFC is divided into three sections: at 0~0.2 A/cm², the activation overvoltage makes the polarization curve drop rapidly; at 0.2~1.0 A/cm², the ohmic overvoltage makes the polarization curve drop gently, and the PEMFC performance is stable; and at 1.0~1.2 A/cm², the polarization curve drops rapidly due to the influence of concentration overvoltage. When the current density is less than 0.8 A/cm², the output voltage is large but the total power is not; when the current density is greater than 1.0 A/cm², the output voltage drops rapidly, resulting in a rapid drop in PEMFC output power, as shown in Figure 1.

Therefore, small or large current density should be avoided when the hydrogen fuel cell is working. However, the power demand of the load is constantly changing, and the method of increasing the rated power of hydrogen fuel cells will greatly increase the cost and space volume of the entire power supply system. The design of hydrogen fuel cells with batteries or supercapacitors can effectively compensate for their low power density. Hybrid power has now become the mainstream scheme of hydrogen fuel cell system applications. When the load power changes, batteries or supercapacitors cooperate with hydrogen fuel cells to play the role of “Peak clipping and valley leveling” and ensure that hydrogen fuel cells operate at their maximum power.

3. Research Status

FCEVs rely on electrical energy to drive the vehicle engine, and this energy mainly comes from the hydrogen fuel cell. Compared with other types of fuel cell, PEMFC is more suitable for FCEVs because it has the advantages of high power density and low corrosion when operating at 60–80 °C [14]. The schematic diagram of an FCEV is shown in Figure 2.

The power system of FCEVs can be separated into three categories: fuel cell + battery (FC + B), fuel cell + battery + supercapacitor (FC + B + SC), and fuel cell + supercapacitor (FC + SC). The FC+B+SC configuration is relatively complex but it also makes the power supply system more controllable. In contrast, FC + B is relatively uncomplicated and its control is relatively simple in most application scenarios. It is currently applicable to most FCEVs. FC + SC removes the battery. Although the external characteristics of the power supply system are not as good as the other two schemes, the hydrogen consumption is the lowest. This is because the loss of the power supply system is reduced after removing the battery.

In this article, the fuel cell module is specified as 180 kW with an optimal power of 100 kW; the battery module is specified as 80 kW, and the output power range of the supercapacitor is 0–50 kW. The power range of the DC transformer is 0–200 kW, and the power range of the three-phase inverter is 0–300 kW. The power demand range of the FCEV during operation is about 50–200 kW. The control method is designed to simulate the real power consumption scenario as much as possible, and the above power range is feasible.

3.1. Topologies of FC + B

As shown in Figure 3, FC + B has four typical topologies [15]. The first is passive topology; fuel cell and battery are directly connected to the load, and there is no DC transformer. The second and the third are semi-active topologies; either the fuel cell or the battery is connected to the load through a DC transformer to simplify the control as much as possible while carrying out power distribution. The fourth is active topology, which is also the most popular topology at present; it can control the power distribution of the fuel cell and the battery at the same time and has great controllability and flexibility. The research status of FC + B is introduced below.

The fuel cell is connected to the DC bus through a DC transformer. The circuit structure of a typical DC transformer is shown in Figure 4. This unidirectional DC/DC transformer can boost the output voltage of the fuel cell and maintain the DC bus voltage.

The battery is connected to the DC bus through a bidirectional DC/DC transformer. The circuit structure of a typical bidirectional transformer is shown in Figure 5. The bidirectional transformer can discharge and charge the battery. The research status of FC + B is introduced below.

In the literature [16], Xiaosong Hu et al. established a model predictive control framework based on Figure 3b from the perspective of cost saving. The authors established a complete model predictive control framework to minimize the total operating cost of FCEVs, including the cost of hydrogen consumption and of the fuel cell itself, as well as the cost caused by fuel cell aging.

The state-of-health (SOH) of the battery is defined by Formula (2):

$$SOH(t)=1-\frac{{\displaystyle {\int }_0^t\left|i(\tau )\right| }d\tau }{2N(c,T_c)Q}$$

(2)

where $i$ is the output current of the battery, $N(c,T_c)$ is the number of battery cycles, and $Q$ is the nominal cell capacity of the battery. Thus, the dynamic discrete $SOH$ model is established, as shown in formula (3):

$$SOH(\mathrm{k}+1)=SOH(\mathrm{k})-\frac{\left|i(k)\right|\Delta t}{2N(c,T_c)Q}$$

(3)

The objective function of model predictive control (MPC) can be expressed as:

$$J_i=C_{h,i}+C_{fcs,i}+C_{bat,i}+D_{soc,i}, i=0,1,2\dots$$

(4)

where $C_{h,i}$ is the hydrogen consumption during the prediction horizon $i$; $C_{fcs,i}$ is the equivalent consumption of fuel cell aging; $C_{bat,i}$ is the equivalent consumption of battery aging; and in order to minimize the aging rate of the battery and extend its service life, parameter $D_{soc,i}$ is introduced. In the literature [16], the optimal control problem is solved by the Sequential Quadratic Programming (SQP) algorithm, which can effectively reduce the total cost.

Also aimed at the topology of Figure 3b, Chao Jia et al. [17] proposed a new energy management strategy (EMS) based on real-time adaptive model predictive control (AMPC), which optimally distributes the load current of FCEV between the fuel cell and battery in real time.

The goal of the AMPC-based EMS proposed in the literature [17] is to optimize the distribution of load current between the fuel cell and battery so as to achieve an ideal trade-off between the four competing performance indicators. The four performance indicators are: hydrogen loss rate $\Delta m_{H_2}$, fuel cell current change rate $\Delta I_{fc}$, battery power loss $\Delta P_{b,loss}$, and battery SOC $\Delta S_b$. The operation cost function is thus formed.

$$L(x,u,k)={\lambda }_1{m}_{H_2}(k)+{\lambda }_2{I}_{fc}(k)+{\lambda }_3{P}_{b,loss}(k)+{\lambda }_4{S}_b(k)$$

(5)

In each sampling interval of EMS, AMPC solves the constrained optimization problem online to obtain the expected fuel cell current conversion rate $\Delta I_{fc}$ and battery current conversion rate $\Delta I_b$. Its objective function is the sum of the predicted running cost, shown as Formula (6):

$$J={\displaystyle \sum _{l=1}^{N_p}{\Vert Q(k+l|k)(x(k+l|k)-r)\Vert }^2+}{\displaystyle \sum _{v=1}^{N_{u-1}}{\Vert R(k+v|k)·u(k+v|k)\Vert }_2^2}$$

(6)

where ${\Vert \Vert }_2^2$ is the square of the Euclidean norm [18]. Q and R are matrices to reduce errors. The parameters of the prediction model are constantly updated. When the AMPC adapts to the updated parameters, an open-source or commercial solver is used to solve the problem so as to run the AMPC-based EMS quickly.

In the literature [17], the authors built a hardware-in-loop (HIL) test and carried out a comparative study. The proposed EMS is compared with three existing real-time EMSs, including IMPC-based EMS [19], bounded load following strategy (BLFS) [20], and equivalent consumption minimization strategy (ECMS) [20]. The comparative study demonstrates that among the four EMSs, the proposed AMPC-based EMS consumes the least hydrogen energy and has the smallest current change rate, which has the smallest optimization gap compared with the offline benchmark.

Similarly, based on Figure 3b, Moghadari M et al. proposed an EMS based on equivalent consumption minimization [21] and built a power system consisting of four fuel cells and one battery. The goal of the EMS is also to minimize hydrogen consumption. Its objective function is shown in Formula (7):

$$C_{total}(t)=C_{FC1}(t)+C_{FC2}(t)+C_{FC3}(t)+C_{FC4}(t)+C_{batt}(t)\times F(SOC)$$

(7)

where $C_{FCn}(t)$ is the fuel cell hydrogen consumption and $C_{batt}(t)$ represents the hydrogen consumed by battery. $C_{FCn}(t)$ can be expressed by Formula (8). It is a quadratic function in which a, b, and c are fitting coefficients. $P_{batt}(t)$ is the battery power (charge or discharge), shown as Formula (9):

$$C_{FCn}(t)=a{P}_{FCn}^2(t)+b{P}_{FCn}(t)+c$$

(8)

$$C_{batt}(t)=\frac{P_{batt}(t)\alpha }{LHV{\eta }_{MS,ave}}$$

(9)

where ${\eta }_{MS,ave}$ is the average efficiency. The low heat value (LHV) of hydrogen is 119.96 MJ/kg. $\alpha$ is determined by Formula (10):

$$\alpha =\{\begin{array}{l}\frac{1}{{\eta }_{chg,ave}{\eta }_{dis}},P_{batt}\ge 0\\ {\eta }_{chg,ave}{\eta }_{chg},P_{batt}<0\end{array}$$

(10)

where ${\eta }_{chg}$ is battery charging efficiency, ${\eta }_{dis}$ is battery discharging efficiency, ${\eta }_{chg,ave}$ is average battery discharging efficiency, and ${\eta }_{chg,ave}$ is average battery charging efficiency.

F(SOC) is a function whose role is to make the battery SOC as close to the expected value as possible. The value of $\beta$ determines the rate of current change. The function is shown in Formula (11):

$$F(SOC)=1-\frac{2\beta (SOC(t)-0.5(SO{C}_{\max }+SO{C}_{\min }))}{SO{C}_{\max }+SO{C}_{\min }}$$

(11)

In the literature [21], the authors compared the performance of ECMS with the strategy of dynamic programming (DP) algorithm and obtained the optimal EMS. The results show that ECMS can achieve an effect close to DP.

Compared with the topology of Figure 3b, the topology of Figure 3c rarely appears in practical applications. This is because in the whole power supply system, the hydrogen fuel cell is the main energy output component and the battery plays an auxiliary role. The number of electrical components used in the circuit topology of Figure 3c and Figure 3b is the same and the control complexity is similar, but the actual effect is not as good as that of Figure 3b. Therefore, there is less research on the topology of Figure 3c.

Compared with the topology of Figure 3b, the hydrogen fuel cell and battery in Figure 3d are respectively connected to the bus through a DC/DC converter, and their control complexity is higher than that in Figure 3b. Therefore, researchers prefer to use a fuzzy logic control method. Fuzzy logic is a science based on multivalued logic and uses the method of fuzzy sets to study fuzzy thinking, language forms, and laws. Fuzzy logic control is better at controlling systems that are complex or difficult to accurately describe.

In the literature [22], the Ziwei Yang team proposed a new EMS based on a fuzzy logic control strategy on the basis of Figure 3d. The SOC of the lithium battery is calculated by an ampere hour integration method, as shown in Formulas (12)–(15):

$$U=U_{oc}-IR$$

(12)

$$U_{oc}=f(SOC)$$

(13)

$$R=\Delta U/I$$

(14)

$$SOC=SO{C}_0-{\displaystyle \int I}dt$$

(15)

This strategy determines the fuel cell output according to the SOC of the cell and the load power demand. The IF-THEN rule is used to distribute the power between fuel cell and battery pack so that the fuel cell can respond quickly, followed by the load variation; this is also to achieve energy distribution optimization between fuel cell and battery pack. At the same time, the battery can provide or absorb dynamic energy to ensure unit operation time efficiency of the system.

In the example, the number of fuzzy subsets of SOC is 3, the number of fuzzy subsets of load power is 5, and the number of fuzzy subsets of output power of the fuel cell system is 5. Fifteen fuzzy control rules are formulated. The weighted average method is used for deblurring.

In the literature [22], the author has proved that the proposed fuzzy logic control strategy can better adapt to the situation of load change and has strong robustness, and the experiment verified the correctness of the EMS to avoid overcharge and overdischarge of the battery pack in the whole work cycle. The fuzzy logic control strategy can extend the service life of the fuel cells.

Also based on the fuzzy control theory, Yigeng Huangfu et al. proposed an optimal fuzzy logic control EMS based on condition recognition [23] on the basis of Figure 3d (i.e., to identify driving conditions) and selected the parameters of the corresponding fuzzy logic control system from the preset parameter database. However, the premise is to use a genetic algorithm to optimize the database offline.

A new learning vector quantization (LVQ) neural network, which improved upon the Self-Organizing Feature Map (SOFM), is proposed in this literature. This method avoids the lack of classification information in SOFM. LVQ network is widely used in pattern recognition because of its simple structure and low requirement for input vectors.

Figure 6 shows the main structure of the proposed optimization strategy, which applies to the topology of Figure 3d. After the sensor collects data, the data is sent to the condition identifier. The control center calculates the data, selects the appropriate optimization parameters from the database, and sets them as controller parameters.

Comparative experiments have been carried out in this literature. Compared with the fuzzy logic control system with fixed parameters, the proposed optimization strategy can effectively reduce the fluctuation of hydrogen consumption and reduce the output power of the fuel cell, extending its service life.

For the topology of Figure 3d, there are other control methods besides the above fuzzy control methods. Giuseppe Graber et al. proposed a theory [24] for the topological structure of Figure 3d. In this literature, the author considered the power distribution between two energy sources and proposed a joint algorithm. This method can improve the utilization rate of hydrogen energy and prolong the life expectancy of fuel cell and battery.

$$\{\begin{array}{l}\Delta I_{disch}^{MAX}(t+dT)=\frac{V_1(t)-V_{\min }/\lambda }{R_{lumped}}\\ \Delta I_{ch}^{MAX}(t+dT)=\frac{\lambda ·V_{\max }-V_1(t)}{R_{lumped}}\end{array}$$

(16)

The maximum current change rate of lithium battery charging and discharging is shown in Formula (16), where $R_{lumped}$ is the magnitude of the fuel cell or battery concerning reference current during an interval, $V_1(t)$ is the terminal voltage at time $t$, and $V_{\min }$ and $V_{\max }$ are the minimum and maximum voltages of the hydrogen fuel cell, respectively. The role of coefficient $\lambda$ is to reduce the working area of the battery to increase its life, and its value is between 0 and 1.

The control strategy distributes the current change of hydrogen fuel cell and battery according to the change of load current demand. Formula (17) is the distribution rule where $\Delta I_{PT}$ is the change value of load current and $a$ and $b$ are the parameters given by the algorithm.

$$\{\begin{array}{l}\Delta I_{FC}=\frac{1}{(1+a)}(a·\Delta I_{PT}+b)\\ \Delta I_{Batt}=\frac{\Delta I_{PT}-b}{(1+a)}\end{array}$$

(17)

In this literature, the authors estimated the life $L{S}_{FC}$ of the hydrogen fuel cell by using the empirical degradation model shown by Formula (18):

$$L{S}_{FC}=\frac{\Delta P}{{\displaystyle {\sum }_{j=1}^M(c_2{I}_{FC}^2(t_j-t_{j-1})+c_1{I}_{FC}^{}(t_j-t_{j-1})+c_0)}}$$

(18)

where $\Delta P$ represents the attenuation value of fuel cell from the beginning to the end of its life. The denominator represents the fuel cell performance decay rate. A quadratic function can express the relationship between the fuel cell decay rate and the fuel cell current, where $c_2,c_1,c_0$ are fitting coefficients. Therefore, the lifetime of the fuel cell can be estimated.

The lifetime of the battery $L{S}_{Batt}$ is estimated in Formula (19):

$$L{S}_{Batt}=\frac{N_{LC}·C_{Batt}}{{(I_{equ}/L{S}_{Batt})}^{\alpha }}, I_{equ}={\displaystyle {\sum }_{j=1}^M{I}_{Batt}(t_j-t_{j-1})}$$

(19)

where $N_{LC}$ is the number of charging and discharging times and its value is given by the manufacturer. $I_{Batt}$ is the battery current, and $\alpha$ is the Peukert coefficient.

The method proposed in the literature needs to obtain detailed electrical parameters of fuel cell and battery. However, in general, manufacturers do not provide complete datasets, which is an obstacle to the application of this method.

On the basis of Figure 3d, the Taehyung Kim team proposed a new topology consisting of two DC buses by distinguishing load types [25], as shown in Figure 7.

The purpose is to reduce the number of elements in the circuit (one switch and one inductor) while adding an additional voltage bus. This method can reduce the cost of the power supply system and is applicable to the scenario where the weight and volume of the hydrogen fuel cell system are highly required.

The proposed control structure regulates the battery current and two bus voltages simultaneously. In the control circuit, the capacitor current command is generated by two PI regulators. In addition, the battery controller generates a battery voltage command. These three values are used to calculate the current command of the main inductor and the duty cycles of the three switches (S1~S3).

3.2. Topologies of FC + B + SC

There are many topologies of FC + B + SC. This paper lists four typical topologies, as shown in Figure 8. The service life of the supercapacitor is much longer than that of fuel cell and battery. It can absorb excess electric energy or quickly supplement insufficient electric energy. In the FC + B + SC topology, the fuel cell is the main power source and is connected to the DC bus through the DC transformer. According to whether battery and supercapacitor are controlled, the four topologies are divided into passive, semi-active, and active. The passive topology does not need to design an EMS. When the load changes, battery and supercapacitor distribute power according to the internal resistance, but this will reduce the utilization of supercapacitor. There are two forms of semi-active topology. The topology shown in Figure 4b is widely used at present. The supercapacitor is boosted by a DC transformer, which can significantly improve the utilization of the supercapacitor. In addition, a bidirectional converter can also change the amplitude frequency characteristics of the system and improve the responsiveness of the load. The active topology uses four DC transformers, so the energy control algorithm is complex and the system cost is high.

The supercapacitor is connected to the DC bus through a 2-Quadrant DC/DC buck-boost transformer. The circuit structure of a typical buck-boost transformer is shown in Figure 9. The transformer is controlled by two complementary pulse width modulation (PWM) signals, which can realize the discharge and charging of the supercapacitor. The transformers connected to fuel cell and battery have been described in Section 3.1 and will not be repeated here. The research status of FC+B+SC is introduced below.

For the topology in Figure 8b, Huan Li et al. first designed the ECMS without considering the power source voltage reduction [26] and then compared the proposed ECMS with the existing rule-based control strategy (RBCS) and other strategies. Finally, an online adaptive equivalent power consumption minimization strategy (AECMS) is proposed by introducing the change of power source performance to ensure the continuity of battery charging and extend the service life of fuel cells.

The AECMS optimization formula is shown in Formula (20):

$$f_w(t)=K_{FC}{m}_{FC}(t)+K_{BA}{m}_{BA}(t)+K_{SC}{m}_{SC}(t)$$

(20)

where $f_w$ represents equal hydrogen consumption of fuel cell, battery, and capacitor, $m_{FC}$ is hydrogen consumption of the fuel cell, $m_{BA}$ is equivalent hydrogen consumption of the battery, and $m_{SC}$ is equivalent hydrogen consumption of the supercapacitor. $K_{BA}$ and $K_{SC}$ are coefficients. Their roles are to limit the SOC of battery and supercapacitor. $K_{FC}$ is the fuel cell coefficient. The three coefficients are calculated as follows:

$$K_{FC}=\{\begin{array}{l}{(1-2\times \frac{\eta -{\eta }_{opt}}{{\eta }_{\max }-{\eta }_{\min }})}^2,\eta \ge 0.4\\ {(1-2\times \frac{\eta -{\eta }_{opt}}{{\eta }_{\max }-{\eta }_{\min }})}^4,\eta <0.4\end{array}$$

(21)

$$K_{BA}=\{\begin{array}{l}{(1-2\times \frac{s-S_{bat,\mathrm{int}}}{B_{bat,\max }-B_{bat,\min }})}^4,S_{bat,\min }\le s\le S_{bat,\max }\\ {(1-2\times \frac{s-S_{bat,\mathrm{int}}}{S_{bat,\max }-S_{bat,\min }})}^{20},s<S_{bat,\min },s>S_{bat,\max }\end{array}$$

(22)

$$K_{SC}=S_{eff}\times S_{peak}$$

(23)

$$S_{eff}=\{\begin{array}{l}{(1-2\times \frac{as+b-S_{opt}}{S_{\max }-S_{\min }})}^2,S_{\min }\le s\le S_{\max }\\ {(1-2\times \frac{as+b-S_{opt}}{S_{\max }-S_{\min }})}^{20},s<S_{\min },s>S_{\max }\end{array}$$

(24)

$$S_{peak}=\{\begin{array}{l}1 0\le I_{load}\le 30\\ -0.01\times I_{load}+1 I_{load}<0 or I_{load}>30\end{array}$$

(25)

Formula (21) calculates the $K_{FC}$ with the current efficiency $\eta$ and the optimal efficiency ${\eta }_{opt}$ of the fuel cell, taking into account the maximum efficiency ${\eta }_{\max }$ and the minimum efficiency ${\eta }_{\min }$. Similarly, Formula (22) calculates the $K_{BA}$ with the current battery SOC $s$ and the battery initial SOC $S_{bat,\mathrm{int}}$, taking into account the maximum battery SOC $S_{bat,\max }$ and the minimum battery SOC $S_{bat,\min }$. In Formula (23), supercapacitor coefficient $K_{SC}$ is the product of peak power coefficient $S_{peak}$ and efficiency coefficient $S_{eff}$. $S_{eff}$ plays a role in limiting the SOC range of the supercapacitor. Formula (24) calculates the $S_{eff}$ with the current supercapacitor SOC $s$ and the optimal SOC $S_{opt}$ of the supercapacitor, taking into account the maximum SOC $S_{\max }$ and the minimum SOC $S_{\min }$. $a$ and $b$ are SOC conversion coefficients which convert the SOC of the supercapacitor into the SOC of the battery. Formula (25) calculates the $S_{peak}$ using the load current $I_{load}$.

The effectiveness of the proposed ECMS was verified by experiments in the literature. The experimental results show that the maximum efficiency point of ECMS is changed when the battery and fuel cell degrade to a certain extent. The aging models of fuel cells and batteries are established in the literature, and the online prediction method is designed.

The high cost of hydrogen fuel cells is an important factor which hinders the development of hydrogen fuel cells. An important reason for fuel cell degradation is the start/stop cycle. At 55 °C, the service life of an existing PEMFCS is 1100 h [27]. Therefore, one of the ways to prolong the service life of hydrogen fuel cells is to reduce their working time. Similarly, based on the topology of Figure 8b, Alessandro Ferrara et al. proposed a fixed setpoint strategy [28]. Compared with the charge balance strategy, this strategy can better coordinate the hydrogen consumption and start/stop cycle. When the vehicle has a short journey, the battery and supercapacitor can meet the vehicle operation without starting the fuel cell to prolong its life. The high manufacturing cost of hydrogen fuel cells is an important factor that hinders the development of hydrogen fuel cells. The research shows that the fuel cell operates very stably, the performance degradation caused by dynamic load can be ignored, and the only important reason for fuel cell aging is the number of start/stop cycles.

In the literature [28], the authors compare three control methods on the basis of considering the startup/shutdown times of hydrogen fuel cells: Fixed-Setpoint EMS, Charge-Balancing EMS, and Predictive EMS, as shown in Figure 10, Figure 11 and Figure 12.

The Fixed-Setpoint EMS keeps the fuel cell operating at maximum efficiency to minimize hydrogen consumption. On the one hand, the EMS needs to turn off the fuel cell periodically to avoid overcharging the battery; on the other hand, EMS should minimize the number of fuel cell start/stop times and extend its service life.

The Charge-Balancing EMS adjusts the operating point of the fuel cell according to the battery SOC, as shown in Figure 11. The setpoint is updated through the map-based control logic [29]. The fuel cell operates at maximum efficiency if the charge is in the appropriate range. When the SOC of the battery is high or low, the output power of the fuel cell will decrease or increase to avoid overcharging or discharging the battery.

The main idea of Predictive EMS is that if the battery can complete the task independently, the EMS will not start the fuel cell. If not, the EMS controls the fuel cell to work at the most effective point. This strategy is not an online EMS and can only be used as a reference.

In the literature [28], the authors compared the performance of the three strategies in two scenarios of mixed driving and urban driving. The results show that the simulation results of Fixed-Setpoint EMS are better than that of Charge-Balancing EMS.

In the above two studies, the authors gave their respective control schemes for the topology of Figure 8b. Compared with Figure 8b,c focuses more on the control of the battery, while the supercapacitor plays an auxiliary role.

For the topology of Figure 8c, Jianlin Wang et al. proposed a real-time Predictive EMS in the literature [30]. In order to predict the power demand of the vehicle, the author designed a short-term memory (LSTM) neural network speed predictor which can predict the future speed of the vehicle. The LSTM neural network allows information to be maintained and transferred, which means that the EMS can predict future information through historical information.

The EMS uses a wavelet transform algorithm to limit the current change rate of fuel cell and battery and adopts a rule-based strategy to control the SOC of supercapacitor and battery within the normal range so as to ensure the dynamic performance of the vehicle under future driving conditions.

The wavelet transform method has a strong ability to analyze and capture the transient response of power demand signals. This method decomposes the predicted power demand of the vehicle into high-frequency and low-frequency components. Figure 13 is the flowchart of three-level wavelet transform decomposition.

Here, $H$ represents a high-pass filter and $L$ represents a low-pass filter. After signal decomposition, the response of the high-frequency signal is realized by the supercapacitor and the response of the low-frequency signal is completed by hydrogen fuel cell and battery, which can effectively reduce the current change rate of hydrogen fuel cell and battery and prolong the working life of the hydrogen fuel cell system.

3.3. Topologies of FC + SC

The charging and discharging rate of the supercapacitor is very fast but its energy density is very low. Therefore, the topology of fuel cell + supercapacitor (FC + SC) is seldom used. FC + SC is applicable to scenarios with a small load fluctuation range but high fluctuation frequency. SC is responsible for providing instantaneous power when the load fluctuates and is also responsible for the startup of FC. The topology is shown in Figure 14.

Mince Li et al. compared the topology schemes of FC + B and FC + SC and implemented the EMS based on the dynamic planning method [31]. The simulation shows that there is no topology with the lowest hydrogen consumption and the slowest aging among the four topologies. Figure 14a is recommended when only hydrogen consumption is involved. When it is necessary to consider both energy consumption and aging, the topologies in Figure 3b and Figure 14b are recommended.

4. Summary and Outlook

In this paper, the control circuit topology of hydrogen fuel cells is introduced in detail, and the research status and development level over the last three years are reviewed. Different topologies have different electrical characteristics and are suitable for different application scenarios. FC + B + SC is applicable to scenarios where the load dynamic range changes greatly and the power supply response is required to be fast; FC + B is applicable to scenarios with little dynamic load change but high requirements for power supply stability; and FC + SC is suitable for scenarios with small load dynamic range changes. The topology of Figure 8a has the fastest response speed. This is because the supercapacitor and battery are directly connected to the DC bus, which can quickly provide a large amount of power shortage. However, this topology is rarely used because the battery and supercapacitor are directly connected to the DC bus without control, which cannot ensure that the hydrogen fuel cell operates at a high efficiency. The topology in Figure 14a has the lowest hydrogen consumption because the number of electrical components is the minimum and the circuit loss is the minimum. However, this topology is also rarely used because the hydrogen fuel cell in this topology ages very quickly. The comparison summary of the four commonly used topologies is shown in

Table 1. Control schemes of several common topologies.

Topological Structures	the Most Frequently Used Control Methods	Inputs	Outputs	Advantages
Figure 3b	Realtime adaptive model predictive control (AMPC)-based energy management strategy	Hydrogen fuel cell current Hydrogen fuel cell current Change rate Battery SOC Battery current	Hydrogen fuel cell reference current	Reduce hydrogen consumption, reduce current fluctuation of fuel cell, and extend service life of fuel cell
Figure 3d	Energy management strategy based on fuzzy logic control strategy	Load current Battery SOC	The output power Hydrogen fuel cell reference current	The system has strong robustness, and extend its service life
Figure 8b	Online adaptive equivalent power consumption minimization strategy (AECMS)	Battery SOC Supercapacitor SOC Load current	Hydrogen fuel cell reference current Supercapacitor reference current	Minimize hydrogen consumption; three power sources fully reach their potential
Figure 8c	Realtime predictive energy management strategy	Battery SOC Supercapacitor SOC Load current The historic velocity	Hydrogen fuel cell reference current Battery reference current	Reduce the power frequency of the fuel cell and battery; keep the SOC of supercapacitor in reasonable range

.

In the future, with the continuous promotion of hydrogen fuel cells, their application scenarios will continue to expand. In order to meet the needs of different loads, more topological structures will be extended. Rules-based, optimization-based and intelligence-based EMS are gradually integrated to complement each other’s strengths. The problem of control accuracy may be solved by artificial intelligence online control in the future.

The current EMS of hydrogen fuel cell hybrid power systems can effectively improve the power, economy, and durability of FCEVs. Based on the current progress, from the perspective of being more efficient, energy-saving, intelligent, and environment-friendly, the EMS of FCEV power systems needs to be further studied in the following respects:

• Widening the energy source of hybrid power. The characteristics of fuel cells are soft, and the battery management system with multiple energy sources has become the development trend for the future. However, further research on the optimization of the combination of different energy sources is needed.
• Under the general environment of information intelligence, the Internet of Vehicles will be gradually popularized. Multi-time scale condition prediction based on big data and traffic information fusion can assist the hydrogen fuel cell system to optimize the EMS online.
• The control mode with higher control accuracy needs to be solved by “artificial intelligence” and “big data analysis”. Monitoring the vehicle operation data and the driving habits of drivers in the road section, adjusting the control strategy in the vehicle through artificial intelligence data processing, and switching different control methods to achieve optimal control are important factors.
• While vigorously promoting and developing the hydrogen fuel cell hybrid power system and EMS, the production and operation costs of hydrogen fuel cells have become a major problem that needs to be solved urgently. High cost is an important reason that hinders the promotion of hydrogen fuel cells. The development of hydrogen fuel cells is affected by the overall development level of the hydrogen energy industry chain, so the research and development cycle is long and the situation is complex.

This paper provides reference and guidance for the future development of the circuit topology of renewable hydrogen energy and hydrogen fuel cell hybrid electric vehicles. The development of hydrogen fuel cells is affected by the overall development level of the hydrogen energy industry chain, so the research and development cycle is long and the situation is complex. More teams need to join the research field of hydrogen fuel cell hybrid power, increase investment in this field, and jointly develop hydrogen fuel cell hybrid power technology.