Total travel costs minimization strategy of a dual-stack fuel cell logistics truck enhanced with artificial potential field and deep reinforcement learning

Abstract To fulfill the increasing power level of fuel cell, a self-adaptive energy management strategy (EMS) with considerations of the efficiency and health of dual-stack fuel cell (DFC) and the total traveling costs for a logistics truck is proposed. The virtual attractive/repulsive forces generated by artificial potential field (APF) functions are applied to DFC and battery system as performance regulator in order to guarantee the efficiency of DFC and the maintenance of SOC. Deep reinforcement learning algorithm, namely deep deterministic policy gradient (DDPG), is leveraged to automatically adjust the virtual force exerted to APF functions in order to assist the power allocation between various energy sources. In comparison to identical power allocation via equivalent hydrogen consumption minimization strategy, APF function generated uneven power distribution of DFC by prohibiting high/low current and frequently start/stop operations of single fuel cell, especially under charge depletion stage. Meanwhile, DDPG-tuner is effective to soften the interaction effect between DFC and battery while meeting the multi-objectives of the EMS. The proposed EMS in cooperation of APF function and DDPG tuner is expected to cope with the dynamic price fluctuation of various energy sources and beneficial to reduce total travel costs as well as extend the DFC's longevity.