Deep reinforcement learning based energy management strategies for electrified vehicles: Recent advances and perspectives

Electrified vehicles provide an effective solution to address the unfavorable impacts of fossil fuel use in the transportation sector. Energy management strategy (EMS) is the core technology supporting the outstanding performance of electrified vehicles. However, technical bottlenecks in conventional control methods, such as poor real-time performance and limited generalization capability, significantly hinder the development of energy management technology. The recent advances in deep reinforcement learning (DRL) hold enormous potential in addressing relevant problems. To this end, this paper systematically surveys DRL-based EMSs. First, DRL algorithms and useful DRL extensions are briefly reviewed. Next, a comprehensive literature survey of pioneering and representative working is presented. The effectiveness and configuration methods of DRL in energy management issues, as well as a guideline for DRL-based EMSs in different vehicular structures, are appropriately analyzed and summarized. Finally, the main challenges and potential solutions are extracted to enhance further real-world implementation.