A Long‐term Energy Management Strategy for Fuel Cell Electric Vehicles Using Reinforcement Learning
Yafu Zhou, Lijian Huang, Xiaoxia Sun, L. H. Li, Jing Lian
Abstract
Abstract The two power sources of a fuel cell electric vehicle (FCEV) are proton electrolyte membrane fuel cell (PEMFC) and Li‐ion battery (LIB). The health status of PEMFC and LIB decreases with the use of FCEV, so the energy management strategy (EMS) needs to give an optimal power distribution based on the health status of power sources throughout the lifetime. However, rule‐based control strategies cannot achieve this. To prolong the service lifetime of two power sources by optimizing power distribution, this article proposes a long‐term energy management strategy (LTEMS) for FCEV, which contains a reinforcement learning module and an improved thermostat controller. By designing a reward function, the reinforcement learning module outputted various LIB state of charge (SOC) boundary which changes with power source attenuation. Based on SOC boundary, the improved thermostat controller will control the fuel cell current under specific driving conditions. Simulation was carried out based on different LIB state of health (SOH) and external temperature, and the simulation results were compared with the data collected from FCEV under rule‐based (RB) strategies. It can be found that the proposed LTEMS can effectively reduce fuel cell and LIB attenuation, and meet the FCEV power demand.