Predicting Charging Times for Mobile Charging Service: Shared Fleet of Electric Vehicles

The mobile charging service is a novel concept that brings a charging station to an electric vehicle. The main component of a mobile charging station is a large battery storing the energy for charging electric vehicles that also need to be recharged. The operation of such a system is costly and needs to be optimised. The outcome of an optimisation is a charging schedule that determines charging sessions for vehicles and mobile stations, trajectories of the van transporting charging stations and shifts for the operating personnel. Charging times constitute one source of uncertainty in such a schedule, therefore accurate prediction models of charging time are relevant. By exploiting data collected during the pilot operation of the Nimbee mobile charging servicing the fleet of shared vehicles, we develop prediction models for the charging time of electric vehicles, as well as, mobile charging stations. By considering three prediction scenarios, we evaluate the performance of several prediction methods, data preparation and feature engineering techniques. Based on the results of experiments, we propose some recommendations on how to optimise the accuracy of such predictions.