Toward intelligent energy management for fuel cell hybrid electric vehicles: Advances in topologies, perception, decision, and connected systems

Driven by the intensifying pressures of the global energy crisis and environmental protection mandates, fuel cell hybrid electric vehicles (FCHEVs) have emerged as a promising solution for the decarbonization of heavy-duty applications. However, the sluggish dynamic response of fuel cells and the heterogeneity of multi-source powertrains present significant challenges to system durability and the hydrogen economy. To address these issues, this paper presents a systematic and critical review of energy management strategies, structured around a novel four-layer framework comprising physical, perception, decision, and system layers. First, the mapping relationship between powertrain topologies and control freedom is elucidated as the foundational physical layer. Second, perception techniques, including micro-scale velocity prediction and macro-scale driving pattern recognition, are synthesized. Third, a comprehensive taxonomy of decision-making algorithms is provided. The trade-offs between optimality and real-time capability are critically evaluated across rule-based, optimization-based, and learning-based strategies. While online optimization achieves up to 29% energy savings, specific focus is given to deep reinforcement learning algorithms such as deep Q-networks and proximal policy optimization that bridge global optimality and real-time execution. Fourth, the integration of vehicle-to-everything connectivity is discussed to extend the control horizon from single-vehicle optimization to cooperative eco-driving. Finally, visionary paradigms, including high-fidelity digital twins, foundation model-based generative decision-making, embodied perception, and neuro-symbolic safety guardrails, are proposed to provide valuable guidelines for next-generation intelligent FCHEVs.