DDPG-guided Adaptive Equivalent Consumption Minimization Strategy for a Range-Extended Electric Vehicle

To improve the optimality without significantly increasing the computational burden, this paper proposes a DDPG-guided adaptive equivalent consumption minimization strategy (A-ECMS) for a range-extended electric vehicle. DDPG agent is trained to find the optimal equivalent factor sequence by trial and error and ECMS is used to achieve power distribution based on numerical optimization. In real-time applications, the actor network outputs the equivalent factors based on the vehicle states. This architecture can combine the exploration capability of DDPG with the stable and efficient optimization capability of ECMS. Simulation results show that the equivalent fuel consumption of DDPG-guided A-ECMS is reduced by 7.1% compared with that of PI-based A-ECMS.