Reinforcement Learning Energy Management for Fuel Cell Hybrid Systems: A Review
Qi Li, Xiang Meng, Fei Gao, Guorui Zhang, Weirong Chen, Kaushik Rajashekara
Abstract
Reinforcement learning (RL) is an increasingly popular technique for hybrid system energy management. However, the existing review literature has not emphasized the training environment and reward function setting and has also not sorted out the evolution of RL agents. To fill this gap, this article introduces the principle of an RL-based energy management strategy (EMS), provides literature reviews from both am RL environment and an agent, and finally, offers perspectives for future studies. In this article, the application of an RL-based technique for controlling fuel cell hybrid systems is taken as a case study. Furthermore, this article is also instructive for researchers who are working on other types of hybrid systems.