Electric Vehicle Charging Navigation Strategy Based on Data Driven and Deep Reinforcement Learning

Abstract

With the popularization and popularization of electric vehicles, the interaction between electric vehicles and power grid and transportation networks is getting deeper and deeper. Considering the multi-data fusion characteristics of dynamic driving behavior and random charging behavior of electric vehicles and the complexity of multi-system modeling, this paper proposes a charging navigation strategy for electric vehicles based on Data-driven mode and deep reinforcement learning. Firstly, the actual operation data of electric vehicles collected by electric vehicle cluster optimization energy storage cloud platform are modeled and mined to obtain the driving, charging data and urban charging station data of electric vehicles. Secondly, the charging navigation model is established by introducing deep reinforcement learning method. The real-time data of electric vehicle-charging station-traffic network is used as the state space of deep Q-network algorithm, and the allocation of charging stations is regarded as the executive action of agent. The optimal action-value function is calculated by establishing the reward mechanism of driving on the way and after arriving at the station, so as to recommend the optimal charging station of return value and plan the driving path for vehicle owners. Finally, a multi-scene simulation example is designed to verify the feasibility and effectiveness of the proposed method