Adaptive ensemble ANFIS-based energy management strategy for fuel cell hybrid electric vehicles

This paper proposes a novel ensemble-based Adaptive Neuro-Fuzzy Inference System (ANFIS) strategy for the Energy Management System (EMS) of Fuel Cell Hybrid Electric Vehicles (FCHEVs). Unlike conventional single-controller architectures, the proposed framework integrates multiple ANFIS models trained on diverse driving patterns, dynamically fused via a real-time, State of Charge (SOC)-dependent weighting mechanism. This adaptive approach allows the controller to effectively respond to load fluctuations, thereby enhancing system efficiency and resilience. To ensure robustness, the method is validated through comprehensive MATLAB/Simulink simulations under five distinct driving cycles: WLTP, NEDC, UDDS, Palermo, and Birmingham. Furthermore, the strategy explicitly incorporates semi-empirical degradation profiles for both the Proton Exchange Membrane Fuel Cell (PEMFC) and the lithium-ion battery. Simulation results demonstrate significant improvements in component longevity and charge sustenance. Specifically, the proposed strategy maintained a significantly higher final battery SOC compared to the non-adaptive baseline (improving final SOC from 0.485 to 0.719 under the Birmingham cycle). Most notably, the EMS successfully extends the fuel cell lifespan by approximately 14–24% in urban driving scenarios (Palermo and Birmingham). It achieves more than a twofold increase (107% extension) under WLTP conditions, with only a marginal increase in raw hydrogen consumption. By unifying adaptive learning with health-aware control, this work introduces a scalable EMS framework suitable for real-world operations. To the authors’ knowledge, this is the first study to combine ensemble ANFIS control with SOC-based weighting and integrated degradation modeling, offering a promising pathway toward durable electric mobility.