Reinforcement Learning based Energy Management for Fuel Cell Hybrid Electric Vehicles
Liang Guo, Zhongliang Li, Rachid Outbib
Abstract
In the paper, a self-learning energy management strategy is proposed for fuel cell hybrid electric vehicles (FCHEV). The studied energy system for FCHEV is composed of fuel cells and lithium batteries. A reinforcement learning (RL) based energy management strategy (EMS) for FCHEV is studied to achieve the power allocation of the two energy sources. The objective is to learn a satisfactory EMS from scratch and only through the interaction of environments. Specifically, Q-Learning, one of the RL methods, is applied to minimize fuel consumption and ensure battery sustainability. Compare with Dynamic Programming (DP), which can reach the best performance of sequential decision problems theoretically, Q-Learning based EMS can achieve results close to DP based EMS. During the process, different objective functions are optimized to be suitable for Q-Learning. Finally, the simulation results with python verify the effectiveness of the method proposed in this paper.