Research on Vehicle Energy Management Based on Deep Reinforcement Learning Algorithm Under Typical Working Conditions

The drive system of a hybrid electric vehicle is composed of two or more drive systems that can operate simultaneously. In this paper, the collected data is used to construct a typical urban driving condition by using the K-means clustering analysis method. Based on a specific fuel cell hybrid electric vehicle model, mainly according to the concept and characteristics of the deep reinforcement learning algorithm itself, it is proposed to use the Deep Deterministic Policy Gradient algorithm for energy management decision-making. The Equivalent Fuel Consumption Minimization Strategy and the Adaptive Adjustment Adaptive Equivalent Fuel Consumption Minimization Strategy are compared respectively. The simulation experiment results show that the latter reduces the hydrogen consumption of the vehicle operation by about 23%, and at the same time has a significant optimization effect on the maintenance of the optimal battery power.