Abstract: Focusing on the hybrid energy storage system for wide-body truck composed by lithium battery and ultra-ca pacitor,a dedicated energy management strategy is required to significantly suppress the battery operating current,improve the cycling characteristics of the lithium-ion battery,thereby extending the service life of the energy system and reducing the oper ating cost of the wide-body truck.In this paper,an energy management strategy based on the Soft Actor-Critic (SAC) deep re inforcement learning algorithm is designed,and the physical characteristics of the hybrid energy storage system forms the foun dation of reward function design.The proposed strategy is validated using the well-established model,composed by the longitu dinal dynamics model of the pure electric wide-body truck,an equivalent circuit model of the hybrid energy storage system,and the associated models of other components.Simulation results show that,compared with a rule-based control strategy based on frequency separation,the proposed intelligent energy management strategy can effectively mitigate the impact of high current on the lithium-ion battery,reducing the battery peak current by 12.71%.In addition,the terminal voltage of the supercapacitor is effectively controlled,with a minimum voltage of 390 V,which always remains within the allowable operating range.The simula tion results confirmed the feasibility and efficacy of proposed deep reinforcement learning-based energy management strategy for the hybrid energy system of the wide-body mining truck.

Key words: wide-body truck,hybrid energy storage system,energy management strategy,deep reinforcement learning, Soft-Actor-Critic algorithm