Adaptive Power Management in Fuel Cell Driven Electric Rickshaw Using Neural Network Controlled Interleaved DC-DC Converter

Abstract

Efficient power management in hydrogen-powered light electric vehicles demands precise coordination between the fuel cell energy source, power conversion stage, and motor drive system. This paper investigates the performance of a neural network-based adaptive power management strategy for a Proton Exchange Membrane Fuel Cell (PEMFC) powered electric rickshaw, implemented through a six-phase interleaved DC-DC boost converter controlled by a Function Fitting Neural Network (fitnet) trained using the Levenberg-Marquardt optimisation algorithm. The neural network continuously maps real-time PEMFC terminal voltage and output current to an optimal pulse-width modulation duty cycle, enabling dynamic adaptation of the converter operating point to maximise fuel cell stack utilisation efficiency. The six-phase interleaved architecture is employed to achieve high voltage conversion ratio and substantially attenuate input current ripple, thereby protecting the fuel cell membrane electrode assembly from harmful high-frequency current stress. The converted energy is delivered to a Brushless DC (BLDC) motor through a three-phase voltage source inverter governed by a Hall effect sensor-based six-step commutation controller. Comprehensive simulation studies in MATLAB/Simulink demonstrate that the neural network power management strategy elevates PEMFC stack efficiency from 42.65% to 78.54% relative to a conventional fixed duty cycle approach, while simultaneously improving electromagnetic torque quality and reducing stator current harmonic distortion. The comparative analysis confirms the technical feasibility and performance advantages of the proposed adaptive power management framework for fuel cell electric rickshaw propulsion.

Introduction

The text presents a hydrogen-based electric mobility system designed for three-wheeled e-rickshaws, which are widely used for last-mile transportation in India and other South/Southeast Asian countries. It highlights the limitations of current lead-acid battery systems, such as low energy density, long charging times, and environmental disposal issues, and proposes Proton Exchange Membrane Fuel Cells (PEMFCs) as a cleaner and more efficient alternative. PEMFCs provide zero direct emissions and fast refuelling, but their low and variable output voltage creates a major challenge for powering electric vehicle drivetrains.

To address this, the study proposes a power electronics architecture based on a six-phase interleaved boost converter (IBC) that steps up the fuel cell voltage to a usable level for the motor drive while significantly reducing current ripple. This ripple reduction is important because excessive ripple can damage fuel cell components and reduce lifespan. The six-phase design improves performance compared to traditional three-phase systems by distributing switching stress more evenly and providing better stability.

The system also incorporates an intelligent control strategy using a Function Fitting Neural Network (fitnet) trained with the Levenberg-Marquardt algorithm. This controller dynamically adjusts the converter duty cycle to track the fuel cell’s maximum power point (MPP), ensuring optimal energy extraction under changing operating conditions such as temperature and load variations.

The overall system consists of four main stages: a PEM fuel cell stack, the six-phase boost converter, the neural network controller, and a motor drive system powering a BLDC motor. The fuel cell is modeled using electrochemical equations that account for activation, ohmic, and concentration losses.

In summary, the paper proposes an integrated hydrogen-powered e-rickshaw system that combines advanced power electronics and neural-network-based control to improve efficiency, protect fuel cell health, and enable practical deployment of clean urban transport.

Conclusion

This investigation has demonstrated the performance advantages of a neural network-based adaptive power management strategy, implemented through a six-phase interleaved boost converter, for PEMFC-powered electric rickshaw propulsion. The principal findings are summarised as follows: 1) The six-phase interleaved boost converter achieves a measured voltage conversion gain of 8.39 with a theoretical 83.3% input current ripple attenuation, providing substantially superior membrane protection compared to single-phase and three-phase alternatives. 2) The Function Fitting Neural Network (fitnet, Levenberg-Marquardt, 3×10 neurons) power management controller improves PEMFC stack efficiency from 42.65% to 78.54% (+84.1%) by dynamically tracking the fuel cell maximum power point in response to real-time terminal voltage and current measurements. 3) Neural network control measurably reduces electromagnetic torque distortion and stator current harmonic ripple relative to fixed duty cycle operation, translating to improved ride quality, reduced acoustic noise, lower copper losses, and extended motor lifetime in the e-rickshaw service environment. 4) Hall effect sensor-based six-step commutation with PI speed regulation achieves stable BLDC motor operation across the complete e-rickshaw driving cycle in both operating modes, confirming the architectural independence and modularity of the motor drive system. 5) Future research directions include: experimental hardware validation using a DSP-based real-time controller, integration of a supercapacitor for peak power buffering and regenerative braking energy recovery, and investigation of deep learning architectures for improved MPPT generalisation under degraded PEMFC conditions.

References

[1] Society of Manufacturers of Electric Vehicles (SMEV), “EV Sector Report 2023–24,” New Delhi, India, 2024. [2] N. Mebarki, T. Rekioua, Z. Mokrani, D. Rekioua, and S. Bacha, “PEM fuel cell/battery storage system supplying electric vehicle,” Int. J. Hydrogen Energy, vol. 41, no. 45, pp. 20993–21005, 2016. [3] S. Rafikiran, G. Devadasu, C. H. Basha, P. M. Tom, V. Prashanth, C. Dhanamjayulu, A. Kumbhar, and S. M. Muyeen, “Design and performance analysis of hybrid MPPT controllers for fuel cell fed DC-DC converter systems,” Energy Rep., vol. 9, pp. 4960–4970, 2023. [4] K. J. Reddy and N. Sudhakar, “High voltage gain interleaved boost converter with neural network based MPPT controller for fuel cell based electric vehicle applications,” IEEE Access, vol. 6, pp. 3899–3908, 2018. [5] K. J. Reddy and N. Sudhakar, “ANFIS-MPPT control algorithm for a PEMFC system used in electric vehicle applications,” Int. J. Hydrogen Energy, vol. 44, no. 29, pp. 15355–15369, 2019. [6] Ballard Power Systems, “NEXA Power Module User’s Manual,” MAN5100078, 2003. [7] A. Peer Mohamed and K. R. M. Vijaya Chandrakala, “Implementation of high step-up converter using RBFN MPPT controller for fuel cell based electric vehicle application,” Sci. Rep., vol. 14, p. 29364, Nov. 2024. [8] A. Intidam, N. Ouanjli, A. Mahfoud, M. Al-Dhaifallah, M. El Mahfoud, and R. M. Bajaj, “Development and experimental implementation of optimized PI-ANFIS controller for speed control of brushless DC motor in fuel cell electric vehicles,” Energies, vol. 16, no. 4, p. 1693, 2023. [9] R. Subbulakshmy, R. Palanisamy, S. Alshahrani, and C. Ahamed Saleel, “Implementation of non-isolated high gain interleaved DC-DC converter for fuel cell electric vehicle using ANN-based MPPT controller,” Sustainability, vol. 16, no. 3, p. 1335, Feb. 2024. [10] D. Mane, P. Puranik, and A. Kulkarni, “Design and analysis of electric rickshaw powertrain for Indian urban conditions,” in Proc. IEEE PESGRE, Trivandrum, India, 2022, pp. 1–6. [11] T. Selmi, A. Khadhraoui, and A. Cherif, “Fuel cell-based electric vehicles technologies and challenges,” Environ. Sci. Pollut. Res., vol. 29, pp. 78121–78131, 2022.

Copyright

Copyright © 2026 Anjali N. Mandhare, Kuldeep G. Thakre. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.