Energy management strategy for hybrid electric vehicles based on double Q-learning

This paper presents an energy management strategy (EMS) using double Q-learning to reduce fuel consumption for hybrid electric vehicle (HEV). The ultimate goal of the EMS is to make the engine work in a high-efficiency zone by reasonably distributing mechanical energy from the engine and electrical energy from the power battery during the driving process of the vehicle, so that the vehicle can achieve optimal performance and achieve the purpose of reducing fuel consumption and emissions. Double Q-learning is a kind of reinforcement learning algorithms, which can avoid the maximization bias generated in Q-learning, so that the EMS can achieve better control effect. This paper simulates and compares the strategies including double Q-learning, rule-based, and Q-learning. The results demonstrate that the presented strategy can availably improve fuel economy and maintain the stability of SOC.