A Deep Deterministic Policy Gradient-Based Energy Management Strategy for Fuel Cell Hybrid Vehicles

A deep deterministic policy gradient (DDPG)-based energy management strategy (EMS) is proposed for fuel cell hybrid vehicles (FCHVs) in this research, which is a deep reinforcement learning (DRL) algorithm and achieves the continuous control in the energy management field in contrast to the general deep Q learning (DQL)-based EMSs. Specifically, the fuel cell system efficiency characteristic is used as the prior knowledge to improve the control effect. In addition, the prioritized experience replay (PER) and other techniques are adopted to improve the computational efficiency of the DDPG. The results demonstrate that the proposed strategy achieves a fast and stable convergence. Additionally, the fuel economy difference between the proposed strategy and the dynamic programming (DP)-based EMS is around 6% under training driving cycles while it is around 3% under a validation driving cycle, which proves the optimality and the adaptability of the proposed strategy.