Data-driven on-demand energy supplement planning for electric vehicles considering multi-charging/swapping services

Abstract

Electric vehicles (EVs) have been experiencing steady growth in many countries in recent years. Given the increasing transportation electrification, it is urgent to establish an efficient on-demand energy supplement system for EVs. In this paper, we present a data-driven two-stage charging/swapping service scheme, where the EV owners can select multi-services, including fast charging at the fast-charging station (FCS), slow charging at the charging post (CP), and battery swapping at the battery-swapping station (BSS). In the first stage, a service recommendation is provided according to the proposed hybrid recommendation algorithm based on the collaborative filtering (CF) algorithm. In the second stage, the on-demand energy supplement orders are dispatched to the swapping/charging infrastructure. To ensure the long-term revenue of the energy supplement system, we formulate the Markov Decision Processes (MDPs) of different types of charging/swapping infrastructures. Then, deep reinforcement learning (DRL) and mixed-integer linear programming (MILP) are jointly used to solve the large-scale sequential decision-making problem. The proposed methodologies are numerically verified in case studies. According to the simulation results, compared with the state-of-art, our methods can better relieve the burden of the power sectors and shows better performance in daily revenue, answer rate, and queue length at FCS.