Intelligent Energy Management for Hybrid Electric Vehicles Using Soft Actor-Critic with Mixed Samplings

This paper presents a deep reinforcement learning (DRL)-based energy management strategy for hybrid electric vehicles (HEVs), which achieves multi-objective co-optimization of fuel efficiency, emission and battery health in the framework of an improved soft actor-critic (ISAC) algorithm. To address the limitations of traditional energy management strategies that are difficult to remain adaptability and robustness in variable driving environments, the ISAC algorithm employs a mixed sampling mechanism that emphasizes the importance of recent experience while assigning sample priorities according to temporal difference errors, which not only reduces the computational complexity but also facilitates the convergence speed of learning. Simulation results verify that ISAC-based energy management strategy outperforms standard SAC in terms of convergence speed, fuel economy and battery health.