Design and experimental verification of a fuel cell/supercapacitor passive configuration for a light vehicle

Abstract The fuel cell/supercapacitor passive configuration without using any DC/DC converters is promising in automotive applications as it can downsize the fuel cell stack, maintain the peak power capability, improve the system efficiency, and remove the need of additional control. This paper presents the design and characterization of a fuel cell/supercapacitor passive hybrid system for a 60 V light vehicle. A detailed design procedure for the passive hybrid test platform is presented with each component modelled and experimentally verified. The voltage error of the fuel cell and the supercapacitor model in the steady state is within 2% and 3%, respectively. Experimental results also validate the function of the passive configuration under conditions of a step load and a drive cycle. The simulation model of the passive hybrid system matches the measurements when a step load current is applied. The supercapacitor provides the transient current due to its smaller resistance while the fuel cell handles the steady state current, which makes it possible to downsize the fuel cell stack. For the drive cycle examined in this paper, the fuel cell stack can be downsized to one third of the load peak power.


Introduction
The rapid development of the automotive industry has led to a larger number of vehicles worldwide. Internal combustion engine (ICE) vehicles have been very popular and widely used during the past decades. However, there are a few environmental issues associated with the growing usage of ICE vehicles, which include oil depletion and air pollution from greenhouse gas (GHG) emissions, fine particles, and smog. The world crude oil reserves are projected to run out in the next forty years [1], which makes the development of alternative and sustainable energy resources essential.
Transportation electrification is one of the most promising approaches to reduce the impacts of air pollution and fossil fuel consumption. Among others, two types of all-electric vehicles (EVs) are based on battery and fuel cell (FC) technologies [2]. Fuel cell electric vehicles (FCEVs) are efficient, environmentally friendly, and zero-emissions as they only emit water and heat [3][4][5]. In such vehicles, the FC can be used along with a battery or a supercapacitor (SC). This is because the relatively slow dynamic response and the inability to recover regenerative braking energy of the FC make it challenging to meet the varying power demands of the vehicle. Therefore, it is necessary to use a battery, an SC or both together with the FC.
There are three hybridization configurations: fully active, semiactive, and passive depending on the number of DC/DC converters used and the way in which the components are connected to the DC bus [6]. The fully active hybrid system is fully controllable since DC/DC converters are associated with each power source. The semi-active hybrid system with a reduced number of DC/DC converters, can only control the DC-link voltage and the power distribution indirectly. In the passive configuration without any DC/DC converters, power distribution is inherently decided by the characteristics of each device. In the early years of commercial FC electric cars (e.g., Honda FCX [7,8]), FC and SC are hybridized. Later, FC and battery are used in Honda FCX Clarity, Honda Clarity Fuel Cell [9], Toyota FCHV-adv, Toyota Mirai [10,11], and Hyundai NEXO [12] where semi-active or fully active hybrid topologies are employed. In such car models, one DC/DC converter is used for the battery, and the FC is directly connected or connected through one DC/DC converter to the DC bus. The concept of the FC/SC passive hybrid configuration is firstly employed in Honda FCX even though one DC/DC converter is adopted between the FC and the SC. The DC/DC converter is used to pre-charge or electrically disconnect the SC when there is a voltage difference at start-up and regeneration. During the propulsion mode, the converter makes the FC and SC passively connected.
In fully active and semi-active hybrid systems, it is possible to actively control the DC-link voltage and power distribution directly or indirectly. Therefore, advanced power control and energy management strategies are required in such topologies. For example, [13] compares two FC/battery semi-active hybrid topologies and the experimental results show that the topology with a direct battery connection has better fuel economy. This topology is also examined in [14,15] and model predictive control (MPC)-based strategies are used to minimize the hydrogen consumption, FC degradation, and battery state-of-charge (SoC) fluctuations. In [16], convex programming is used to optimize the power-split strategy between the FC and the battery pack. Similarly, FC/SC semi-active hybrid topologies are studied along with various power control methods in [17,18]. A fully active FC/battery hybrid topology is investigated in [19]: the total running cost is optimized based on a MPC framework and the impacts of driving and pricing are also explored. Also, the fully active FC/SC configuration is studied in [20,21] for an electric tram. Moreover, FC/battery/SC active hybrid topologies with two or three DC/DC converters are proposed in [22][23][24][25] and different control strategies are employed to improve the overall performance. Detailed energy management strategies along with future research trends are summarized in [26]. Clearly, active and semi-active hybrid configurations require a robust control strategy and additional DC/DC converters to improve the control performance. However, using DC/DC converters results in additional energy losses and increases the system weight, volume, and cost [27][28][29]. Since no DC/DC converters are used, the passive configuration can mitigate the issues above. Moreover, the passive configuration provides a natural power split depending on the internal resistance of each device. In Fig. 1, an example of a passive hybrid powertrain is shown where no DC/DC converters are used. To ensure that the FC and the battery/SC are properly operating (e.g., their operating voltage ranges are similar and their SoC are suitable), the hybrid configuration needs to be carefully designed [30]. Since the power distribution of each device cannot be actively controlled in the passive configuration, a power control method in an FC/battery passive hybrid system is developed in [31] and the power distribution can be indirectly controlled by the FC working pressure. However, this method makes the FC operating pressure lower than nominal. In [32], the passive configuration of an FC and a Li-ion battery is studied, and the open-circuit voltage (OCV) of the FC is 10 V higher than that of the battery. To avoid overcharging the battery, a MOSFET is used between the FC and the battery to determine the operating mode: FC only or FC/battery hybrid. Usually, the passive hybrid system combining an FC and a battery is promising in unmanned vehicles operating around 24 VDC and with a peak power of up to 1.5 kW. In these cases, it is possible to find FCs and batteries that can operate within a similar voltage range if the battery with a suitable capacity is charged with the maximum current provided by the FC [31][32][33][34].
Due to its wide operating voltage range, SC is well-suited to directly hybridize with FC. Moreover, this combination is suitable for frequent start/stop occasions since SC can provide high transient power with a long cycle life [35]. Therefore, the FC/SC passive configuration can be used to downsize the FC stack, reduce the fuel overconsumption, avoid flooding inside the FC, and improve the FC lifetime [27,36,37]. In [38][39][40], the FC/SC passive configuration is analyzed, sized, and modelled at the cell level. A global system sizing approach for FC and SC is proposed in [41] based on a pre-defined automotive power profile. An FC/SC passive configuration with a 9.5 kW FC stack is developed in [42] and the feasibility under a downscaled drive cycle power profile is experimentally performed.
To obtain a functional passive power split, the sizing of FC and SC in such topology is important. The basic criteria include the DC-link voltage and the energy storage capacity. Moreover, the FC operating current dynamics is proposed in [43] as the third design criterion. This work also develops a lab-scaled FC/SC prototype based on the power profile measured from a Nemo EV. A detailed analysis of an EV with FC/SC passive configuration is performed in [44]; the model of each component is studied, and the power distribution and dynamic characteristics are evaluated at the system level. Due to the unique characteristics of FC and SC, the DC-link voltage is passively determined by the power distribution, which is highly dependent on the driving cycle. Therefore, the energy efficiency of a powertrain system with passive hybridization of FC and SC is analyzed and evaluated based on different drive cycles, and hydrogen consumption and DC-link voltage variation are also demonstrated [45].
Existing studies on the FC/SC passive configuration are mainly focused on component sizing, initial voltage connection, component modelling and powertrain system modelling. However, previous experimental works have been limited to investigating the validation of power consumption on the hybrid configuration while the effects of regenerative power have not been well addressed [42,43]. This paper fills this gap and considers an FC/SC passive configuration for a 60 V light vehicle. The contribution of this work is to present a detailed design procedure for the test platform with each component modelled and experimentally characterized. The FC/SC passive configuration is experimentally validated using a step load current and the Artemis Urban drive cycle. Moreover, the operation of this passive configuration is verified with the regenerative power introduced into the step load current test scenario. The remainder of the paper is organized as follows. Section 2 is dedicated to the system configuration and the development of the test platform for each component. Section 3 presents the models of the FC, SC, and the passive hybrid system as well as the analysis of the system stability. Section 4 shows the experimental results using the step load current and the drive cycle for the hybrid system. Finally, conclusions are drawn in Section 5. Challenges and future trends of FCEVs are presented in Section 6.

System configuration
For the analysis in this paper, Renault Twizy Urban 80 is selected as the benchmark vehicle and its main specifications [46,47] are listed in Table 1. Twizy is an all-EV and its battery pack is replaced with the FC/SC passive configuration in the case study. The passive hybrid powertrain, also considered in our previous work [2,44,45], consists of an FC stack, a diode, an SC bank, and a propulsion system including a DC/AC inverter, an electric machine, a transmission, and wheels, as shown in Fig. 1. The objective of this paper is to evaluate the FC/SC passive configuration from both experiments and simulations. The propulsion system is considered as the load, which is simulated as a controlled current source for simplicity in this study.
According to vehicle dynamics, the instantaneous wheel power P wheel [48] can be derived as where ρ a is the air density with the value of 1.225 kg/m 3 when the dry air is under 15 • C and standard atmosphere pressure, C d is the aerodynamic drag coefficient, A f is the front area, v is the vehicle speed, C r is the rolling resistance coefficient 1 , m is the vehicle mass, g is the gravitational constant with the value of 9.81 m/s 2 , and α is the slope with zero on a flat road. Once the vehicle speed is known, the wheel power can be calculated. Together with the energy conversion efficiency, η, the wheel power determines the power demand of the FC/SC passive hybrid system. In this study, η is assumed to be 0.9 in the propulsion mode and 0.6 in the regenerative braking mode, respectively.

Design criteria 2.2.1. Specifications of fuel cell
Proton exchange membrane (PEM) FCs are the most attractive candidate for automotive applications. This is mainly because the low operating temperature of this technology enables the fast starting from the idle state to the full power operation mode. The Horizon H-3000 FC stack is used in this study and its specifications are shown in Table 2.

Specifications of supercapacitor
The Maxwell BCAP3000 SCs are selected. Each of them has a rated voltage of 2.7 V and a rated capacitance of 3000 F. The detailed cell parameters can be found in [52]. To make sure that the SC bank is not overcharged, the voltage applied over it should always be within the rated range and the following equations should be met where V SCB is the OCV of the SC bank, V fc is the maximum voltage of the FC stack, V SC is the OCV of the SC cell, and N SC is the number of SC cells. Then, the minimum number of cells needed is 25.33 according to (2). Therefore, 26 cells are connected in series, giving the SC bank a rated voltage of 70.2 V. The total capacitance of the SC bank with seriesconnected SC cells can be calculated as where C i is the capacitance of each individual SC cell. The total energy of the SC bank is where V SCB_max and V SCB_min are the maximum and minimum operating voltage of the SC bank, respectively. The SC bank is connected in parallel with the FC stack, which means that it can only operate within the FC voltage range from 36 V to 68.4 V. Therefore, the total energy of the SC bank is 195 kJ with the total capacitance of around 115 F. The total internal resistance of the SC bank is 7.54 mΩ.

Fuel cell test platform
The FC test platform is shown in Fig. 2. The H-3000 air-breathing FC stack from Horizon Technology is tested, a 10 L HYDROGEN 6.0 hydrogen tank with an initial pressure of 200 bar from Aga Gas AB is used to provide hydrogen to the FC stack, and a two-stage pressure regulator is connected to adjust the hydrogen pressure from the tank. The first stage is to generate a pressure between 0~40 bar, and the second stage is to tune the pressure to around 0.5 bar. During the FC operation, the unreacted hydrogen needs to be purged out from the purge valve, and the output of the purge valve is connected to the outside to avoid the accumulation of hydrogen gas in the lab. The electronic load, EA-9000HP, is connected to provide the dynamic current to the FC stack. The FC controller is used to control the supply valve, purge valve, and air fans as well as to detect the voltage, current, and temperature of the FC stack. The experimental data is collected by   1 Empirical studies show that C r depends on the tire material, design, and working conditions [49]. At low speed levels, C r increases only slightly with the speed while it is proportional to almost the square of the speed at relatively high speed levels [50]. C r is often assumed to be constant to study the vehicle performance and fuel economy.
LabVIEW through serial port communication.

Supercapacitor test platform
The SC bank is tested using the platform shown in Fig. 3. Two DC power supplies, SM30-200, are connected in series to generate a maximum voltage of 60 V. The EA-9000HP is connected in parallel to discharge the SC bank. The oscilloscope is used to collect the experimental data. The SC bank is charged and discharged by a constant current of 50 A. The SC bank is charged to the maximum voltage of about 60 V and discharged to 6 V.

Implementation of passive hybrid configuration
Separate measurements are firstly conducted on the FC stack and the SC bank as shown in Figs. 2 and 3. Then, the FC system, the SC bank, the power supply, and the electronic load are connected to form a passive hybrid system. The schematic of the experimental setup is shown in Fig. 4. The diode, MUR2X100A02, is used to prevent the energy exchange from the SC bank to the FC stack and the power supply. The DC breakers are used to connect or disconnect the separate devices. Since the maximum current of the electronic load is 50 A, the C65N-DC DC breaker with a rated current of 63 A is selected.
To prevent the inrush current when the system starts up, the SC bank is pre-charged at the beginning. The process to run the system is listed as below: Step 1. Connect the cables and make sure that all the components are correctly connected. Also, make sure that the FC system, the SC bank, and the electronic load are electrically disconnected; Step 2. Turn on the power supply and pre-charge the SC bank to at least 50 V. In case of high currents at the system start-up, the SC bank is pre-charged to 60 V in this study;

Fig. 2.
Test platform for fuel cell stack.  Step 3. Turn off the power supply and turn on the FC system; Step 4. Close breaker 1 and breaker 2 to balance the voltages of the FC stack and the SC bank. They should have the same equilibrate voltage of approximately 68 V; Step 5. Close breaker 3 to electrically connect the FC system, the SC bank, and the electronic load. Turn on the electronic load, apply the dynamic load, and run the system; Step 6. After all the tests have been done, reduce the load current to zero, open breaker 1, and turn off the FC system; Step 7. Now only the SC bank is connected to the electronic load. Apply a load current to discharge the SC bank; Step 8. When the SC bank is fully discharged, open breaker 2 and breaker 3, and turn off the electronic load.

Modelling
The FC stack uses the generic FC model provided in the Sim Power Systems (SPS) toolbox of MATLAB/Simulink [53]. The model contains a controlled-voltage source connected in series with an internal resistor, as shown in Fig. 5. The output voltage [54][55][56] is expressed as with V fc is the stack output voltage, E oc is the stack open-circuit voltage, V act is the activation voltage loss of the stack, and V ohmic is the ohmic voltage loss. As shown in Fig. 5, E oc -V act is modelled as the controlled voltage source E. K c is the nominal voltage constant. T is the stack operating temperature, F is the Faraday constant, R is the ideal gas constant, P H2 is the partial pressure of the hydrogen, and P O2 is the partial pressure of the oxygen. N is the number of cells, A denotes the Tafel slope, i fc is the stack output current, i 0 is the exchange current, τ is the time constant of the stack dynamic response, and r fc is the internal resistance of the stack.

Characterization
The model parameters are obtained from the tested polarization curve of the FC stack. Fig. 6 (a) compares the measured and simulated polarization curves. The stack voltage decreases gradually with the increase of the current while the stack power increases with the current. The simulation results show a good match with the experimental data with a voltage error within 2%. When step load currents are applied to the FC stack, the simulated and measured voltage profiles are shown in Fig. 6 (b). It can be observed that when the load current changes from 40 A to 50 A, the FC voltage first drops to 40.9 V instantaneously and then increases slightly to 41.5 V. When the load current changes from 50 A to 40 A, the voltage has an overshoot and then declines slightly to 44 V. Therefore, the voltage has an instantaneous change and there is a time delay for the voltage to reach the steady-state, which suggests that the FC stack behaves as a first-order system when the step load current is applied. The time delay is closely related to the temperature and the air supply subsystem. The air supply subsystem is simulated by a first-order inertial link and the temperature is considered as a constant.
The relationship between the voltage and the current of the FC stack can be simulated as a linear function. If the FC stack does not operate at the OCV but with a higher current, the polarization curve can be linearized to provide the linear model of the studied FC stack, which can be described as where k fc is the linear coefficient shown in Fig. 7. k fc is estimated to be 0.2388 Ω when the FC stack works in the ohmic region where the current is above 10 A.

Modelling
The SC bank is represented by a generic Stern model [57,58] shown in Fig. 8, which is also available in the SPS toolbox of MATLAB/Simulink [59]. The output voltage can be obtained as follows where Q SCB is the total electric charge, i SCB is the current of the SC bank, and r SCB is the internal resistance. Considering the self-discharge phenomenon of the SC bank, the electric charge when i SCB = 0 is modified as follows

Characterization
The self-discharge test is shown in Fig. 9 (a), where the SC bank is charged to 59.2 V by 50 A and then the charging current reduces to 0 A within several minutes. After this, the power supply is turned off and the SC bank voltage drops to 55.6 V as the charge redistributes. The measured and simulated dynamic characteristics are shown in Fig. 9 (b). The smallest error is obtained when C SCB = 105.85 F and r SCB = 0.0225 Ω. Due to the degradation of the SC cells and the resistances of the connectors and cables, the real capacitance is 10 F lower than the calculated value, and the real resistance is a bit higher than the calculated value in Section 2.2.2.
The obtained voltage error is less than 3% when the voltage is above 16 V and the voltage error is a little bit higher when the voltage is below 16 V. This is because the low voltage is more difficult to model. In the meantime, the measurement error should also be considered.

System stability
In the FC/SC passive configuration, the FC stack, the SC bank, and the load are connected in parallel. Moreover, a diode is placed between the FC stack and the SC bank. Based on the component models above, the signal block diagram from the load current to the FC current is shown in Fig. 10 (a) and the signal flow graph is derived in Fig. 10 (b). Mason's gain formula depicted in Fig. 10 (c) is used to determine the transfer function: where T k is the path gain of the k th forward path from the input to the output, Δ k is the T k path cofactor, and Δ is the determinant of the graph. There are two forward paths (T 1 and T 2 ) from i load to i fc and three individual loops (L 1 , L 2 , and L 3 ). T 1 and T 2 touch all the individual loops and there are no non-touching loops. Therefore, (9) can be written as Since r SCB << k fc , the transfer function is a low-pass filter and the cutoff frequency is 1/(2πC SCB (r SCB +k fc )) by ignoring the diode resistance, which is 5.8 mHz with the model parameters determined above.
Due to the degradation of FC and SC, the coefficients of the transfer function shown in (11), which are influenced by the utilization rate,    have a certain degree of uncertainty. This means that the precise values of C SCB , r SCB and k fc cannot be determined readily. Empirical data shows that the SC performance degrades when its capacitance decreases and ESR increases [60]. The FC end of life is defined with respect to 10% of its voltage drop at the nominal current, which means that its internal resistance increases due to degradation [61].
Suppose the ranges of C SCB , r SCB and k fc are given as where δ i > 0(i = 1, 2, 3) and δ i is constants for a given time interval. Therefore, by ignoring the diode resistance, the actual controlled object (Σ 0 , ΔΣ) can be described as The nominal system Σ 0 is described as and the set of uncertainties can be described as a parameter vector set: Therefore, (13) can be written as with ΔG(s) = G(s) − G 0 (s). According to (14), the nominal system Σ 0 has a negative pole of − 1/(C 0 (r 0 +k 0 )) located in the left half of the complex plane, which means that Σ 0 is stable. Also, the uncertainty system ΔΣ is stable. Therefore, the actual object (Σ 0 , ΔΣ) is stable and not influenced by parameter variations.

Step load evaluation
To evaluate the performance of the FC/SC passive hybrid configuration, the FC stack and the SC bank are connected in parallel following the procedure mentioned in the second part, and the experiment is conducted using step load currents. Note that the FC stack, the SC bank, the power supply, and the electronic load are connected in parallel and they share the same voltage theoretically. Therefore, only the load voltage is measured. The test results are shown in Fig. 11 (a).
When the positive load current is applied during 20 ~ 354 s, the FC and SC together provide the current to the electronic load. The SC bank responds instantly and supplies almost all the current and power due to its smaller internal resistance. Since the SC bank supplies almost the load power, its state of charge decreases, which ultimately results in a decrease in its voltage. The system voltage is always below 60 V, which means that the FC stack does not operate at the OCV because of the long stabilization time of approximately 120 s, as shown in the figure. This can protect the FC stack against gas starvation and degradation. From the polarization curve of the FC stack, it can be observed that its current increases when its voltage decreases. In this way, the FC current increases gradually to the steady-state due to the reduced system voltage and the FC does not suffer from any current or power transients. When the negative load current is applied during 356 ~ 400 s, the SC bank is charged by this load current and the voltage of the SC bank increases. This results in the decreased FC current until a new operation equilibrium is reached. In summary, the SC bank responds to the transient current and power instantaneously while the FC stack takes over the steady-state current and power. The current distributions obtained from the equivalent circuit model and the measurements are compared in Fig. 11 (b). In the steady state, the FC currents obtained from the equivalent circuit model have error within 4% compared to the measurements, while the errors are slightly higher during transients. This is because the resistance of the FC stack changes with the operating conditions, such as temperature, pressure, and flow rate. The resistance of the SC bank also changes with the temperature.

Drive cycle evaluation
To further evaluate the passive hybrid system, the Artemis Urban drive cycle is used to calculate the power profile for the light vehicle characterized in Table 1. The speed profile and the reference power profile considering a reduced power scale of the calculated power for the FC/SC passive configuration are shown in Fig. 12 (a). A simulation is conducted to evaluate the feasibility of the FC/SC passive configuration with the load power as the input. Before the simulation starts, the initial voltage of the SC bank is set to be 65 V to equilibrate the FC stack and the SC bank. A breakdown of the current, voltage, and power distribution for the FC stack and the SC bank can be seen in Fig. 12 (b). During the drive cycle, the FC and the SC reach a new equilibrium and the voltage is passively decided. The FC follows the load current and power. However, it does not experience the high-frequency load. This is because SC handles the high-frequency current and power.
The power histograms of the load, FC, and SC are shown in Fig. 12 (c) and "Pct." represents the percentage of the specific power in the total power range. It is clear that the FC peak power is only about one third of the load peak power. This makes it possible to downsize the FC stack, i. e., for this simulation the FC stack only needs a peak power of one third of the load peak power. Also, the SC evens out the power from the FC and the FC power becomes smoother, which can reduce the fuel  overconsumption avoid flooding inside the FC, and improve the FC lifetime [27,36,37]. The FC average power almost equals the load average power since the SC average power is only 0.04 kW. For an ideal SC, its average power should be zero. In reality, for a lossy SC, its average power is positive. The non-zero SC average power results in an energy difference of about 0.01 kWh between the start and the end of this simulation.
The experimental results of the FC/SC passive configuration under a downscaled Artemis Urban drive cycle current are shown in Fig. 13. As shown in Fig. 13 (a), when the load current changes frequently, the SC bank reacts to this high-frequency change while the current and power of the FC change gradually due to its relatively higher internal resistance. The maximum FC current/power change rate is 1.6 A/s and 82.4 W/s, which is only 2.3% and 2.7% of the nominal current/power per second, respectively. Therefore, the FC degradation is reduced in this passive hybrid configuration. The current distributions between simulation and experiment with the same initial conditions are compared in Fig. 13 (b) and the simulation shows good consistent with the experiment. The histograms in Fig. 13 (c) shows that the FC supplies almost the average power of the load while the SC provides nearly zero average power. This results in an energy difference of 0.005 kWh during this drive cycle test. Since the SC handles most of the peak power, the FC can be downsized. In fact, the FC peak power is only about one third of the load peak power.

Conclusions
This paper deals with the design and characterization of a passive FC/SC configuration for a 60 V light vehicle. The detailed test procedure is developed. The FC and SC models are experimentally characterized and verified. Simulations results using these models match the experimental data well. The dynamic response of the FC stack is characterized as a combination of the resistive characteristics and a first-order system. The behavior of the SC bank is simulated by resistors and capacitors connected in series. It turns out that the internal resistance of the SC bank is around 10 times smaller than that of the FC stack.
The feasibility of the FC/SC passive configuration is experimentally verified when a step load current is applied: the SC bank provides the transient current and power while the FC stack takes over the steadystate current and power. The FC/SC passive configuration allows the FC to have a long stabilization time and to avoid the OCV operation, which can reduce the FC degradation and improve the FC lifetime. The FC/SC passive configuration is further simulated and experimentally verified under the Artemis Urban drive cycle. Results show that the lowpass filter effect makes it possible to downsize the FC stack, which only needs a peak power of one third of the load peak power. The SC evens out the current and power from the FC to prevent the FC from experiencing large currents and power variations.
Future work related to this paper will be focused on several aspects. First, a four-quadrant electronic load needs to be introduced to validate the performance of the FC/SC passive system under various drive cycles with regenerative braking. Second, properly size of the FC stack and the SC bank should be designed to optimize the cut-off frequency. Third, to achieve a better controllability, active hybrid topologies may be required while passive hybrid configurations would be better for applications where the travel route and operation speed are pre-defined.

Outlook and future trends
In the past few decades, FCEVs have made significant progress in terms of environmental adaptability, driving range, and performance. FCEVs have achieved -40 • C cold start. Heat dissipation at high temperatures is no longer an issue. The driving range has been extended due to the development of the hydrogen tank. In today's passenger car, the hydrogen tank volume has increased to 60 L, the pressure has increased to 700 bar, and the mass ratio of hydrogen and tank has increased to 5.7%. With improvements in the power density of FC, power converter, and electric machine, the size and weight of FCEVs are also reduced. However, there are still some challenges for the development of FCEVs. The durability of FCEVs has not met the DoE target due to the relatively short lifetime of the FC stack. Hydrogen refueling infrastructure needs to be expanded. In fact, only 432 stations have been established by the end of 2019 although more stations are expected to be established in the future. One of the biggest concerns is the relatively high cost of FCEVs (e. g., 6000 USD for a Mirai). Therefore, lifetime enhancement of the FC stack, establishment of hydrogen stations and cost reduction of FCEVs are some of the main challenges.
As reported in this paper, great efforts have been made to optimize different hybridization topologies between FC, battery, and SC as well as power control and energy management strategies. On the other hand, further improvements in the performance of these power devices are still required. In fact, li-ion battery/SC hybrid device is being developed to fully take advantage of battery and SC. With the hybridization of FC and Li-ion SC, the power density and energy density of the FCEV powertrain can be improved, which will significantly enhance the driving range and the dynamic response.

Funding
This work was supported by the Swedish Electromobility Center and the Swedish Energy Agency.

Author statement
Q.X. and Y.L. conducted the investigation and the conceptualization, Q.X. and S.L participated in the experiment and validation. Q.X. wrote the original draft and S.L. and Y.L. participated in the review & editing.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.