Implementation of high step up converter using RBFN MPPT controller for fuel cell based electric vehicle application

Abstract

Fuel cell-based electric vehicles (EVs) are gaining popularity in the automotive industry due to strict carbon emissions and fuel efficiency regulations. Fuel cells have inherently low voltage characteristics, making it challenging to interface with EV drive systems. This work proposes a unique topology implementing a non-isolated high step-up DC-DC converter to integrate the Proton Exchange Membrane Fuel (PEMF) cell with the EV motor drive. The converter combines a modified Single-Ended Primary Inductor Converter (m-SEPIC) with two winding-coupled inductors, enabling high voltage gain, minimizing conduction losses, and ensuring high efficiency. The output power of a fuel cell is highly dependent on cell temperature and membrane water content, making a maximum power point tracking controller essential for extracting optimal power from the fuel cell stack. The optimum power point of the PEMF cell is identified using a Neural Network (NN)-based Maximum Power Point Tracking (MPPT) technique, employing the Radial Basis Function Network (RBFN) approach compared to the Perturb & Observe (P&O) method. The performance of the RBFN MPPT method is validated under various temperature conditions of the fuel cell system. An extensive analysis and comparison with alternative converters validate the efficacy of the proposed system. The converter is designed for an input voltage of 24 V and boosts it to an output voltage of 240 V, with high static voltage gain, high power density, and minimized total losses. The system is tested and validated through both simulation and real-time implementation using the dSPACE 1104 real-time controller. A 200 W, 3000 rpm, 240 V VSI-based BLDC motor drive integrated with the proposed converter topology is developed as a real-time setup in the laboratory which would confirm the converter’s functionality. Both theoretical analysis and real-time outcomes affirm the suitability of the proposed converter topology for electric vehicle applications.

Introduction

In recent times, a significant challenge has revolved around our increasing dependence on fossil fuels, which is causing considerable harm to the environment. To tackle this pressing issue, people are actively seeking innovative solutions, leading to the exploration of EVs as an alternative. EV adoption has increased recently; however, issues with limited driving range and the availability of charging stations still exist^1–4. In this regard, fuel cell vehicle utilization shows promise as a way to overcome the difficulties faced by electric cars. Fuel cell-powered vehicles have a big advantage over traditional electric vehicles since they don’t need constant external charging. These cells work by using hydrogen and oxygen to drive chemical reactions that generate electricity, with the only waste that these cells create being water. Furthermore, utilizing this water for electrolysis to generate additional hydrogen can enhance overall efficiency. Because of its higher energy density, low noise, and eco-friendliness, PEMF Cells are one of the fuel cell technologies most advantageous for electric vehicles^5–7.

PEMF Cells’ VI characteristics are nonlinear, with a specific operating point yielding maximum output voltage and power⁸. Therefore, employing an MPPT algorithm is important for enhancing PEMF Cell system efficiency. Researchers^9–12 have proposed various MPPT techniques for fuel cell systems in the existing literature. The widely used P&O method for MPPT in fuel cells employs a trial-and-error process, adjusting the operating point based on power changes. Despite its ease, P&O has drawbacks like slow tracking speed and increased oscillations. Reference⁹, introduces a conventional IC-based MPPT controller for PEMF cells, addressing P&O limitations by utilizing power characteristics’ slope for MPP tracking. However, a drawback is the method’s requirement for complex control circuitry.

In Ref¹⁰, an FLC-based MPPT for PEMF cells was proposed to improve the drawbacks of conventional controllers, allowing swift parameter adjustments without a precise mathematical model. However, FLC’s effectiveness depends on chosen membership functions, and fuzzy rules, and faces limitations due to memory requirements. In Ref.¹¹ , a PSO algorithm-based MPPT for PEMF cell, minimizing oscillations at the maximum power point. Yet, initializing duty cycles proves challenging under varying PEMF cell conditions. Conversely, ANN algorithms can identify complex non-linear relationships without a precise model^11,12.

This paper presents a neural network-based MPPT controller for a PEMF Cell-powered electric vehicle, incorporating the RBFN algorithm for optimized power extraction.

The design of the electrical architecture in an electric vehicle involves various considerations, and the propulsion system voltage is determined by factors such as efficiency, power distribution, and component specifications. Electric vehicles typically use high-voltage propulsion systems to reduce current levels, minimize losses, and improve overall system efficiency. Higher voltage allows for lower current for the same power, which results in reduced resistive losses in the power electronics and motor windings. The chosen voltage is a result of considerations related to efficiency, component specifications, and overall system performance. For example, the Tesla Model S uses a propulsion system voltage of around 400 volts for its battery pack¹³. The earlier models of the Nissan Leaf used a 360 V system to manage the load voltage to suit the requirements of various subsystems within the vehicle¹⁴. The Bolt EV uses a propulsion system voltage of around 350 volts¹⁵. OEMs carefully design the electrical architecture to balance efficiency, component compatibility, safety, and regulatory considerations. The load voltage is often adjusted as needed to meet the specific requirements of different components within the EV.

The design of DC-DC converters for EV applications based on sizing, and equivalent voltage requirements suitable for two/three/four-wheel electric vehicles to handle high voltage and high frequency for improvising the overall efficiency. The common and well-renowned companies like STMicroelectronics-France, ON-Semiconductor-USA, Vishay Intertechnology-USA, and Renesas Electronics-Japan are across the world that are involved in the development of DC-DC converters for EV applications^16–18.

It is not possible to directly link a fuel cell to a vehicle or a load application due to the voltage decrease that occurs as the current rises¹⁹. The isolated converter-based topologies have large leakage inductor problems and utilize high frequencies which impact switching voltage stresses, and weight and reduce the converter efficiency while operating at kW or in high-power applications which include saturation and core losses, high electromagnetic interference, leakage inductance, and a bulky design. In response to these issues, the researchers have opted for a non-isolated high step-up and bidirectional DC-DC converters^20–22. A classical boost converter is utilized, which raises voltage with one duty cycle to match the DC bus voltage²³. However, reaching a unity duty cycle increases voltage stress across power semiconductors, posing challenges for attaining high static voltage gain. Another method involves the utilization of the quadratic boost converter, which, despite its adoption, encounters escalated voltage strain across power switches and diodes situated on the higher-voltage side^24–27. To address the challenge of enhancing voltage static gain, a DC-DC converter utilizing switched capacitors, as described in Refs.^28–30, has been proposed. However, this solution necessitates an additional converter for adaptable voltage regulation. Boost converters interleaved in fuel cell contexts offer benefits such as reducing switch current stress and decreasing input current ripple³¹. In Refs.^32,33, the authors proposed a high voltage gain-based interleaved DC-DC converter that uses three switches for operation, which increases the complexity of controlling the system. Furthermore, the Z source converter, as demonstrated in³⁴ and employing cascaded switched capacitors, delivers a significant voltage increase. The problem with this method is high voltage stress across semiconductor components when the duty cycle is almost zero. Bidirectional DC-DC converters used in energy storage systems have more semiconductor switches, leading to increased conduction losses. To address this issue, an additional snubber circuit is required, which increases the size and cost of the system^35,36.

To achieve enough power conversion and voltage gain, researchers have investigated a variety of isolated converters, as reported in Ref.³⁷. On the other hand, using the SEPIC architecture^38,39, presented a one-switch converter topology with increased static voltage gain that keeps the source’s input current constant. The quadratic three-level topology in Ref.⁴⁰, on the other hand, was under a lot of stress when the duty cycle was 0.6, even though it provided the needed higher voltage ratio for fuel cell applications. In Ref.⁴¹, researchers utilized an HGBDC to increase the output voltage for fuel cell applications. In the past, induction motors usually worked for driving electric vehicles because of their sturdy initial torque, even though their speed regulation abilities were somewhat lacking. However, today, BLDC motors have emerged as a preferred choice in vehicular equipment. This shift is attributed to the enhanced efficiency, extended high-speed operational range, elevated power density, and favourable speed-torque characteristics offered by BLDC motors^42–45.

To overcome the above issues, the proposed method is considered operative and ideal for electric vehicle applications, especially those concerning low nominal voltages and small-capacity fuel cell packets. The schematic diagram of the suggested system with the PEMF Cell, and unique converter powering EV motor can be seen in Fig. 1. Changing the DC-link voltage enables the regulation of the permanent magnet BLDC motor’s speed across a wide range.

Fig. 1.

Representation of PEMF cell powered with suggested converter topology powering EV motor.

This paper introduces a single-switch non-isolated DC–DC converter designed to meet high step-up requirements. The proposed converter employs a modified SEPIC-coupled inductor topology-based Single-Switch Converter. Without comprising the performance, the reduction of inrush current at the input port to the converter is ensured. Furthermore, it reduces voltage stress on the active power switch, enabling the use of lower voltage-rated MOSFETs with low RDS-ON. This reduction in voltage stress contributes to improved conversion efficiency by minimizing conduction and switching losses toward the compact sizing requirement of the proposed system.

This work is planned as follows: In “Mathematical modeling of fuel cell system”, and “Controlling technique” represent the modeling of the fuel cell system formation and control scheme techniques. “Design of the novel high step-up converter” refuges the proposed converter design and BLDC motor. “Results and discussion” presents the results and discusses the offered system. Finally, in “Conclusion”, we deliver the conclusion.

Mathematical modeling of fuel cell system

A hydrogen fuel-driven electrochemical apparatus designed to convert hydrogen fuel into electrical energy is commonly referred to as a PEMF cell. The fuel cell feeds fuel and air, which undergo a chemical reaction to transform into electricity and water. An electrolyte and two electrodes, the anode and cathode, make up each fuel cell. The electrolyte separates the hydrogen fuel’s positively and negatively charged ions. The electrolyte helps the cell create electricity at its output when it receives both hydrogen and oxygen. Figure 2 displays the electrical representation of the PEMF cell system. The chemical process shown below produces water and power as byproducts in the fuel cell.

1

Fig. 2.

PEMF cell electrical circuit diagram.

The PEMF cell output voltage expression is given below,

2

In Eq. (2), the variables V_NERNST, V_OHM, V_ACT, and V_CON represent the reversible open circuit voltage, ohmic overvoltage, activation overvoltage, and concentration overvoltage, respectively. The following equations are used to determine each term in Eq. (2):

3

where, T_cell stands for the temperature of the PEMFC cell, while P_H2 and P_O2 refer to the pressures of hydrogen and oxygen, respectively. The ohmic overvoltage, denoted as V_OHM, is computed by using the below equation:

4

where R_M represents the resistance equivalent to electron flow, and R_C stands for proton resistance. It is assumed that the proton resistance remains constant.

5

where, L denotes the membrane’s thickness in centimeters, A stands for the active area of the membrane in square centimeters, and ρM indicates the membrane’s specific resistivity in ohm-centimeters.

The activation voltage, V_ACT, encompasses activation overvoltage and is defined as follows:

6

ℓ₁, ℓ₂, ℓ₃, and ℓ₄ represent the empirical coefficients specific to each cell. Con_O2 refers to oxygen concentration and is calculated as follows:

7

Lastly, the concentration overvoltage, denoted as V_CON, can be determined using the following expression:

8

where R represents the global gas constant, I_max stands for the maximum current density, F indicates Faraday’s constant, and I is the current density. At the fuel cell output, a converter maintains a consistent voltage through the DC link. Comprehensive proposal terms for the 1.26 kW PEMF cell are available, as illustrated in Table 1.

Table 1.

PEMF cell factor specifications.

Factor details	Rating
PMAX (W)	1.26 kW
VMAX (V)	52 A
IMAX (A)	24.23 V
No. of cells	42
Nominal air flow rate	2400 lpm

Controlling technique

The neural network comprises different types of networks, such as RBFN and Multi-Layer Perceptron (MLP). RBFN requires less computation time compared to MLP. Moreover, RBFN also reduces the time needed for model development as it doesn’t require repetition. Therefore, in this proposed approach, RBFN is employed as the MPPT technique for optimizing power from the fuel cell. The RBFN algorithm consists of two stages: supervised and unsupervised methods. The supervised approach trains the weights of the hidden layer, while the unsupervised techniques manage the system parameters.

The RBFN architecture depicted in Fig. 3 contains an input layer, an output layer, and a hidden layer, with interconnected neurons in each layer. The training of RBFN occurs in two stages. Initially, radial basis functions manage the input parameters through an unsupervised training method. After that, the supervised training refines the weights, resembling the backpropagation technique. Within this study, the RBFN controller generates a control duty signal (k) for the highly boosted proposed converter, aiming to regulate the output voltage for the BLDC motor.

Fig. 3.

Structure of RBFN.

Figure 4 illustrates the systematic operation flowchart of the RBFN network for the proposed system. It processes inputs from the PEMFC, involving voltage and current data, through its two neurons constituting the input layer. RBFN is trained based on 1500 data sets of voltage and current with 70% data of the PEMF cell and tested with a 30% split ratio. Based on the maximum voltage and current transition, the appropriate duty cycle will be generated by the Network to trigger appropriately the converter semiconductor switch.

Fig. 4.

Flowchart of RBFN-based MPPT method.

The total input and output of this layer are outlined as follows:

9

10

where i_a¹ represents the input layer, j_a¹ signifies the hidden layer, and net_a¹ denotes the number of nodes denoted by ‘a’ in the input layer. The nodes in the hidden layer of the RBFN network operate based on Gaussian membership functions.

11

12

where and are the standard deviation and mean of the Gaussian function respectively. The output layer contains a single neuron denoted as (j_c), a linear activation function used to generate a control duty signal k. The output neuron computes the sum of incoming signals to produce the output.

13

14

where W_b is the weight associated with connections linking an output and hidden layer. Improving the weights of connections and adjusting learning factors can significantly boost the tracking capabilities of the fuel cell system, as highlighted in reference¹¹. This study utilizes voltage and current data obtained from the PEMF cell system as inputs for the RBFN controller, which then generates the duty cycle k as its resulting output.

Design of the novel high step-up converter

Figure 5 depicts the novel design of the high-efficiency step-up DC-DC converter circuit with increased voltage gain. The model includes Coupled Inductor (CI) L_P and L_S, which are connected with three capacitors (C₁, C₂, C₃), three diodes (D₁, D₂, D₃), output diode D₀ and output capacitor C₀. The purpose of this arrangement is to recover leaked energy from the interconnected inductor, as indicated by the line. N denotes of turns of the windings in the coupled inductor. A coupled inductor is integrated with a modified SEPIC in the novel topology.

Fig. 5.

Circuit diagram of the novel DC–DC high step-up converter topology.

The whole system can be represented as (m-SEPIC+`CI) technology as the converter is working at a low switching frequency (f_s=24 kHz). It achieves an enhanced voltage ratio without relying on a transformer-based isolated converter. Increasing voltage gain and switching frequency is essential for developing fuel cell-powered electric cars. Conventional converters have a high-duty cycle and high output voltage, which may cause reverse recovery difficulties, voltage spikes, and shortened component lifespans, among other problems. The m-SEPIC + CI topology alleviates strain on switches, diodes, and capacitors, ultimately increasing the lifespan and ensuring its suitability for real-world applications. Therefore, this article introduces a high step-up converter topology to enhance voltage gain and overall system efficiency. Figure 6 illustrates the CCM process of the suggested arrangement and its respective operating modes.

Fig. 6.

Various modes of operation of the proposed topology.

Mode 1[t0 < t < t1: The power switch S, diodes D₂ and D₀ are forward biased in this mode as presented in Fig. 6a. The secondary side of the coupled inductor current in I_LS charges both energy intermediate capacitors C_2,C_3, and C₀, the diode D₁, and D₃ are blocked states. Capacitor C₃ charges the output capacitor C₀, and the load through diode D₀.

15

16

17

18

where I_S is the switch current, I_LPmax is the coupled inductor primary side peak current, V_D1 and V_D3 are the diode 1 and 3 voltage, V_C1 and V_C3 are the capacitor 1 and 3 voltage, and I_D2 and I_D0 are the diode 2 and 0 current.

Mode 2[t1 < t < t2: The L_P primary side of the coupled inductor current in I_LP parasitic capacitance of the switch S is displayed in Fig. 6b. The inductor current cannot change abruptly. Now, the diodes D₁, D₂, D_3, and D₀ are in blocked condition. The secondary I_LS inductor current continues to flow through C₁, C_2, and C₃ intermediate capacitors until it becomes zero. The discharge of the output capacitor C₀ provides the energy for the output load.

19

20

21

22

where I_S is the switch current, V_FC is the input voltage, V_D1, V_D2, V_D3, and V_D0 are the diode 1, 2, 3, and 0 voltage, V_C1,V_C2, and V_C3 are the capacitor 1, 2 and 3 voltages.

Mode 3[t2 < t < t3: Figure 6(c) depicts the switch S remaining in the OFF state, indicating that diodes D₁ and D₃ remain forward-biased throughout this timeframe. Through the output diode D₀, the capacitors C₂ and C₃ offer the supply to the load. In this mode, the output diode D₀ remains reversely biased, allowing energy to be transmitted to the load and C₀ to discharge.

23

24

25

26

where V_S is the switch voltage, I_D1 and I_D3 are the diode 1 and 3 current, V_D2, and V_D0 are the diode 3 and 0 voltage.

Figure 7 illustrates the primary theoretical waveforms of the suggested converter with ZVS and ZCS. To streamline the investigation of the proposed converter, ideal characteristics are attributed to capacitors and power semiconductor devices. The converter’s gain, denoted as G, is computed using the winding turns ratio n and the time intervals between these periods in Eq. (27).

27

Fig. 7.

The main theoretical waveform of the proposed topology.

The Eq. (27) mentioned above, assuming a duty cycle of k = 0.6 and a CI winding turns ratio of n = 1, results in a voltage gain of G = 10. Equations (28)–(30) are employed to determine the capacitor voltage.

28

29

30

The capacitors are assumed to have the same value, meaning C₁ = C₂ = C₃, and the variation in capacitor voltage is determined using Eq. (31).

31

The determination of the coupled inductor value is achieved through the utilization of Eqs. (32) and (33).

32

33

where V_FC is the PEMF cell voltage, Δi_L is the ripple current, and n is the no. of turns. The resonant frequency (f_r) of the proposed system is calculated using the Eq. (34).

34

Equation (35) is employed to compute the ripple voltage across the capacitor, considering a minimum 15% change in capacitor voltage, denoted as ∆V_Cap.

35

The output capacitor C₀ is determined based on the change in ripple voltage (∆V_Cap), calculated using Eq. (36).

36

The converter output voltage is denoted as V₀, and defined by Eq. (37).

37

The stress on the switch voltage is computed utilizing Eq. (38).

38

Equations (39)–(42) determines the stress on the diode’s voltage and input current.

39

40

41

42

The voltage strain through both the capacitor and diode is equivalent to half of the total V₀ output voltage. Equation (43) provides the total efficiency of the designed converter.

43

The transfer function expression based on the input and output voltage of the proposed converter topology is represented in Eq. (44).

44

The frequency response features of the proposed converter using the Bode Plot are illustrated in Fig. 8, depicting the relationship between input and output voltages. From Fig. 8, it’s clear that the bode plot of the proposed system shows a phase margin of 0.1° with a slope of -20dB/dec while the gain margin is 38.04 dB/dec. Therefore, the system is stable and possesses a fast response. The bode plot indicates the stability of the proposed converter, affirming its suitability for closed-loop operation.

Fig. 8.

Bode plot for the proposed converter topology.

Figure 9 compares the static gains of several high step-up converter topologies along with the suggested architecture. The CCM of the converter confirms its strong voltage gain, achieving 10.24 at a minimum duty cycle of 0.6 and a lower 24 kHz switching frequency. Similarly, the converter shows a voltage gain of 20, highlighting its strong voltage gain capabilities, with the same switching frequency and standard duty cycle. It is also clear that, for certain duty cycles, the suggested converter performs better than other converters regularly.

Fig. 9.

Comparison of converters static voltage gain.

Table 2 compares the components used in the suggested converter with those in other converter designs in terms of no. of components, voltage gain, maximum voltage stress across active and passive components, power density and efficiency. Six different kinds of converters, all operating at a f_S of 50 kHz, are chosen for comparison from references^{26,30,31,38,46,47}.

Table 2.

Parameter comparison of various converters.

Topology	In26 ref.	In31 ref.	In30 ref.	In38 ref.	In46 ref.	In47 ref.	Proposed converter
Techniques	CI	CI + L	3 Winding-CI	2-CI	CI	3 Winding-CI	CI + mSEPIC
No. of MOSFETs	1	1	1	2	1	1	1
No. of capacitors	5	4	5	3	4	3	4
No. of diodes	6	4	6	4	6	3	4
Voltage gain
Frequency in kHz	50	50	50	10	50	50	24
Voltage stress of switch
Voltage stress of diode
Input voltage VIN or VFC	24 V	24–30 V	25 V	30 V	24 V	43 V	24 V
Output voltage VOUT	400 V	200 V	500 V	380 V	359 V	300 V	240 V
Power output POUT	175 W	150 W	200 W	1.2 kW	600 W	200 W	200 W
Efficiency η	91.41%	90.81%	95%	92.4%	94.7%	95%	95.25%
Power density	Medium	Medium	High	Medium	High	High	High

In comparison to other converters, the suggested converter stands out for having a larger voltage gain, high power density, and a lower switching frequency and voltage. Furthermore, under full load circumstances, the efficacy of the suggested converter is 95.25%, which exceeds that of conventional converter topologies, and it runs at a low switching voltage. Power density is pivotal in comparing DC–DC converters, with higher power density reducing overall size, weight, and volume. Achieving this involves considering overall system losses, including passive and active elements, along with PCB layout. The semiconductor devices used in the prototype determine power density, categorized as high, medium, or low. Circuit specifications in Table II, based on switching frequency, mass, and volume, indicate a high-power density for the proposed topology.

The BLDC motor for an EV is connected to the proposed topology. In variable-speed drive applications, BLDC motors are becoming more and more common as an alternative to synchronous and induction motors. Their silent operation, small construction, great power density, and high-power factor are the reasons behind this. In a BLDC motor, permanent magnets are used in the rotor, while the stator structure is like that of an induction machine. Therefore, fine-tuning the rotor’s location is essential to controlling the motor’s speed. With each sensor spanning a 60-degree gap, the hall sensors produce hall signals dependent on the location of the rotor. Then, using a decoder circuit, these produced Hall signals are transformed into switching signals for the VSI.

Results and discussion

Simulations under typical operating circumstances of the PEMFC system were executed using MATLAB/Simulink. To assess the accuracy and effectiveness of the proposed MPPT controller with the suggested topology is performed in both simulation and hardware. Table I contains the PEMF Cell system’s parameters. The dynamic reaction of the PEMFC system is examined by observing abrupt shifts in temperature: starting at 320 K for 0 to 0.1 s, and then rising to 350 K for 0.1 to 0.2 s. Figure 10, illustrates these temperature variations across distinct time intervals.

Fig. 10.

Varying temperature of PEMF cell.

Figure 11, displays the waveforms depicting the voltage output, and power of the PEMF cell resultant to the mentioned temperatures. During specific time intervals, the PEMF cell produces 1070 W of power from 0 to 0.1 s, and peaks at 1220 W between 0.1 and 0.2 s. Figure 12 depicts the DC link voltage and power acquired via the P&O-based MPPT method, resulting in power outputs of 790 W at 320 K and 940 W at 350 K. Furthermore, the figure exhibits the DC link voltage output and power employing the unique RBFN method, achieving 910 W at 320 K and 1020 W at 350 K. The importance of the suggested MPPT technique for the fuel cell is assessed against the P&O-based MPPT controller. As seen in Fig. 12, it becomes manifest that the proposed technique yields an increased DC link voltage matched to the P&O.

Fig. 11.

PEMF cell voltage and power output at different temperature conditions.

Fig. 12.

DC link voltage and power output under varying temperature conditions.

Table 3 shows a comparison of the PEMF Cell DC link voltage at different temperatures using P&O and RBFN MPPT controllers. In contrast to the P&O, the suggested RBFN controller produces a DC link power of 1020 W to reach a maximum power point within 0.12s. Figure 13 is a graph that shows how the BLDC motor starts up and stays stable at different temperatures related to the PEMF cell. This illustration showcases motor factors such as electromagnetic torque, back EMF, and speed during dynamic temperature fluctuations within the fuel cell system. The BLDC motor torque waveform shows a slight changeover in the varying temperature conditions and becomes stable. The electromagnetic torque measures 2.2 N-m and 2.4 N-m corresponding to the specified cell temperatures. The back EMF waveform remains unchanged even as speed conditions vary. During these temperature variations, the BLDC motor operates at specific speeds: 2570 rpm from 0 to 0.1 s with operating converter power 910 W, and 3100 rpm from 0.1 to 0.2 s with 1020 W.

Table 3.

Parameters details of the proposed converter topology.

Parameters	P&O with PEMFC		RBFN with PEMFC
PEMFC temperature (K)	320	350	320	350
Period (s)	0–0.1	0.1–0.2	0–0.1	0.1–0.2
Duration required to reach maximum power point (s)	0.038	0.138	0.025	0.12
DC link power (W)	790	940	910	1020
DC link voltage (V)	180	210	210	250

Fig. 13.

Characteristics of BLDC motor for FC-based RBFN MPPT technique.

The analysis of the suggested topology considers the 370 K temperature-based constraint. Figure 14 highlights the voltage and current of the switch and diode 1 in the proposed converter. The converter operates under soft-switching conditions, ensuring ZVS and ZCS, as shown in Fig. 14. Additionally, the proposed converter produces half of the output voltage as the switching voltage (~ 150 V), which improves the system’s efficiency. Figure 15, displays the voltage and current waveforms of the diodes in the suggested topology, clearly indicating a ZCS and ZVS operation to that depicted in Fig. 6. Both the diode D₂ and D₃ voltage values are 210 V, and the diode D₀ voltage is 100 V. The capacitor voltage waveforms are represented in Fig. 16, detailing that the number of capacitors C₁ and C₂ is equivalent to that of C₃. From Fig. 16, it is detected that the value of capacitor C₁ is 110 V, the C₂ voltage is about 102 V, and the C₃ capacitor voltage is 230 V. The number of capacitors C₂ and C₃ voltages are equivalent to the proposed converter output voltage. Figure 17 shows a real-time model of the proposed converter established and verified in a lab setting to confirm the viability of the theoretical analysis. Figure 18a depicts the proposed converter topology hardware setup. Figure 18b represents the elements put together on a PCB board measuring around 72 ×  52 mm². The converter has a volume of 109 cm³ (assuming a switch height of 29.2 mm), and it possesses a power density of 1.829 W/cm³. However, the configuration prioritizes credible operational results overachieving the highest volume density. To validate the converter’s operation, the experimental setup indicates the proposed converter which operates equivalently under a scaled-down model as per the designed theoretical converter claimed.

Fig. 14.

Switch and diode 1: voltage and current waveforms of the proposed system.

Fig. 15.

Diodes voltage waveforms of the proposed system.

Fig. 16.

Capacitors voltage waveforms of the proposed system.

Fig. 17.

Real-time setup of the proposed system.

Fig. 18.

(a) The hardware setup of the proposed converter topology, and (b) the PCB of the proposed converter topology.

Table 4, highlights a full breakdown of the element specifications utilized in the proposed converter prototype. The waveform confirms the working principle even if the actual voltage (246 V). The converter output voltage V₀ and PEMF cell voltage V_FC (~ V_IN) waveforms of the proposed converter are offered in Fig. 19a. The input and output currents of the proposed system are 8.2 A and 0.8 A. The input voltage is 24 V, while the stepped-up output voltage across the developed converter measures around 246 V.

Table 4.

Specifications of the developed converter.

Converter parameters	Designed values
VFC fuel cell stack voltage (V)	24
Converter voltage, V0 (V)	240
Power, POUT (W)	200
fs ( kHz)	24
C1, C2, C3 capacitors (µF)	220
Capacitor COUT (µF)	440
Duty cycle	0.66
Voltage gain	10 at n = 1

Fig. 19.

(a) Input and output (voltage and current) waveforms, (b) switch voltage and current waveforms, (c) diode 2 voltage and current waveforms, (d) diode 3 voltage and current waveforms, (e) capacitor 2 voltage and current waveforms, and (f) capacitor 3 voltage and current waveform.

The voltage and current waveforms of the proposed converter switch are shown in Fig. 19b. These waveforms confirm that the switch operates under soft-switching conditions, with ZVS during both turn-on and turn-off operations. The observed switching voltage is 152 V, which is half the output voltage, and the current is 14.8 A. The voltage and current waveforms of diode D₂ are presented in Fig. 19c, showing a voltage of 210 V and a current of 10 A across D₂, along with ZCS operation. Similarly, Fig. 19d shows the voltage and current waveforms of diode D₃, with values of 198 V and 12 A. The waveforms for diode D₃ also clearly demonstrate ZVS operation. From these observations, the switch and diodes’ voltage and current waveforms indicate that the proposed converter performs soft-switching with low voltage stress, which increases the system’s efficiency. Figure 19e,f show the voltages and currents across the different intermediate capacitors. The values of the voltage across the capacitor waveforms V_C2 and V_C3 are 93 V and 178 V, respectively. The testing environment and measurement equipment affect the desired low ripple (< 5%) in the DC output voltage. It indicates that the converter’s output voltage stays constant at a duty cycle of 0.66 when comparing the results of the simulation with the experiment, as shown in Figs. 12 and 19a. The input voltage of 24.5 V is increased to 10 times its initial value, or 246 V at 0.1–0.2s, in the simulation results shown in Fig. 12. Experiments confirm that the results of the simulation match the observations, as shown in Fig. 19a. Figure 20a,b, illustrate the proposed converter efficiency concerning output power and power losses. The power level ranges from 100 to 1200 W.

Fig. 20.

(a) Efficiency comparison of the different converters at full load, and (b) comparison of the apparatus power losses at full load.

The proposed topology operates with a lower switching voltage than the conventional methods. Notably, the maximum efficacy of the proposed topology reaches approximately 95.25% under specific parameters: k = 0.66, N = 1, and fs = 25 kHz. Under full load conditions, the measured efficiency of the proposed method is 93.5%. Under 500 W ranges, the proposed converter operates at a maximum efficiency of 96%. Figure 20b, it becomes evident that the proposed converter exhibits low tolerance at the measured power level compared to the Coupled Inductor – Voltage Multiplier Cell (CI-VMC) topology⁴⁸. The predicted total loss variation of the proposed converter and CI-VMC is 6.78 W. Comparatively, the proposed converter demonstrates lower losses than the traditional converter, comprising a diode conduction loss of 6.1 W, and a coupled inductor winding loss of 3.82 W. The net capacitance loss of the proposed CI-VMC is 1.7 W. The overall power loss of the suggested converter is 7.9 W is less than the CI-VMC with 10.5 W. Table 5, presents the cost assessment of the proposed converter for power ratings of 200 W.

Table 5.

Cost evaluation of the proposed converter.

Components	No. of items	Parameters	USD (200 W)
Switch, S	1	IXFB110N60P3	$15.52 (with gate driver)
Diodes (D1, D2, D3 and D0 )	4	VS—15EWX06FN-M3	$10.71 (for 4 numbers)
Coupled inductor (with core and wire)	2	205 µH (A—core 2.1*3.2 cm2)	$18.37 (for 2 numbers)
Capacitors (C1, C2, and C3 )	3	2.08 µF/250 V	$5.28 (for 3 numbers)
Output capacitor (C0)	1	140 µF/600 V	$ 3.52

The inductor’s core losses, including the coupled inductor losses, are negligible when compared to other losses; hence, they are disregarded in the efficiency calculation. Consequently, the efficacy of the developed converter remains consistently greater than 95% across various power levels, rendering it a promising choice for electric vehicle applications. The analysis distinctly indicates that as the power rating increases, the novel-designed and developed converter becomes more cost-effective with a high step-up gain of 10, proving its efficiency of 95.2%, with a high-power density of 1.829 W/cm³ produced total losses of 7.9 W. Specifically, the converter considered for a power rating of 200 W underwent laboratory setup, testing, and validation of the results. Therefore; the proposed converter suits its credibility to be utilized for low and medium-powered EV applications.

Conclusion

This manuscript proposes a non-isolated, high-gain modified SEPIC-based CI converter designed for fuel cell electric vehicle systems. The proposed converter topology is designed to step up the PEMFC voltage from 24 V to 240 V, making it suitable for BLDC drive-based EV systems. The converter operates with a voltage gain of 10 and a low switching voltage stress of 150 V with soft switching. The steady-state investigations based on CCM operation prove the stability of the proposed converter. The suggested high-gain DC converter performs better compared to other converter topologies. Two MPPT techniques, namely P&O and RBFN, were implemented with the proposed converter, making it particularly well-suited for use in electric vehicle applications. These methods were analyzed, and results were obtained under two different PEMFC system temperature conditions. The outcomes proved that the RBFN-based PEMFC MPPT methods operate at maximum powers of 910 W and 1020 W, outperforming the P&O method. A real-time prototype of the proposed converter was developed, and its performance was validated under laboratory conditions with a VSI-based BLDC EV drive system. Experimental results confirm that the proposed converter achieves a high step-up voltage gain of 10 with an efficiency of 95.2%, with a high-power density of 1.829 W/cm³, total losses of 7.9 W making it cost-effective with compact design of size 72 × 52 mm² suitable for low and medium-powered EV applications.

Abbreviations

ANN

Artificial neural network

BLDC

Brushless DC

CCM

Continuous conduction mode

CI

Coupled inductor

EVs

Electric vehicles

EMF

Electromagnetic Force

FLC

Fuzzy logic controller

HGBDC

High gain boost DC–DC converter

IC

Incremental conductance

m-SEPIC

Modified single-ended primary inductor converter

MPPT

Maximum power point tracking

OEMs

Original equipment manufacturers

P&O

Perturb and observe

PCB

Printed circuit board

PEMFC

Proton exchange membrane fuel cell

PSO

Particle swarm optimization

RBFN

Radial basis function network

SEPIC

Single-end primary inductor converter

VI

Voltage-current

VSI

Voltage source inverter

ZCS

Zero current switching

ZVS

Zero voltage switching

V_FC

Fuel cell voltage

I_FC

Fuel cell current

k

Duty cycle

G

Gain

D

Diodes

S

Switch

V₀

Output voltage

I₀

Output current

C

Capacitor

L

Inductance

Author contributions

A.P.M.—Conception, design of the work; analysis; K.R.M.—analysis, interpretation of data; drafting the work. S.S.—conception, design of the work; U.S.—Analysis, interpretation of data; D.A.—Reviewing and supervision. All authors reviewed the manuscript.

Acknowledgements

The authors would like to thank for the work which was supported by the research grants [SEED-2022-CE-95] and [13354-psu-2023-PSNU-R-3-1-EI-); Research, Development, and Innovation Authority (RDIA), Prince Sultan University; Saudi Arabia and Amrita Vishwa Vidyapeetham, Coimbatore, Tamil Nadu supported by the Amrita Seed grant [ASG2022107]. The authors would like to acknowledge the support of Prince Sultan University for paying the Article Processing Charges (APC) of this publication.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.