Deep Reinforcement Learning-Based Energy Management System for Fuel Cell Hybrid Electric Vehicles Under Rough Pavement Constraints

This paper presents an advanced Deep Reinforcement Learning (DRL) based energy management system (EMS) for fuel cell hybrid electric vehicles (FCHEVs) operating under rough pavement conditions, a common challenge in African countries. Unlike traditional rule-based and optimization-based EMS, the proposed DRL approach dynamically adapts to varying driving conditions, optimizing power distribution while minimizing hydrogen consumption and maintaining battery state-of-charge (SOC) stability. Results show a 27.8^% reduction in hydrogen consumption compared to rule-based EMS and 22.3% compared to optimized EMS, demonstrating superior efficiency and resilience. Additionally, DRL enhances powertrain stability under road irregularities, addressing the energy demand fluctuations and component stress caused by rough terrain. These findings highlight DRL's potential to improve EMS adaptability, efficiency, and durability in real-world driving conditions.