Dynamic charging management for electric vehicle demand responsive transport

TL;DR

This paper presents a dynamic charging management system for electric ride-hailing fleets that minimizes costs and waiting times by predicting station availability with a hybrid LSTM model within a two-stage optimization framework.

Contribution

It introduces a novel two-stage optimization approach combined with a hybrid LSTM prediction model for efficient electric vehicle charging management under uncertainty.

Findings

Reduced fleet charging waiting times by 48.3%.

Decreased total energy charged by 35.3%.

Improved charging efficiency in a realistic city simulation.

Abstract

With the climate change challenges, transport network companies started to electrify their fleet to reduce CO2 emissions. However, such an ecological transition brings new research challenges for dynamic electric fleet charging management under uncertainty. In this study, we address the dynamic charging scheduling management of shared ride-hailing services with public charging stations. A two-stage charging scheduling optimization approach under a rolling horizon framework is proposed to minimize the overall charging operational costs of the fleet, including vehicles' access times, charging times, and waiting times, by anticipating future public charging station availability. The charging station occupancy prediction is based on a hybrid LSTM (Long short-term memory) network approach and integrated into the proposed online vehicle-charger assignment. The proposed methodology is applied to a realistic simulation study in the city of Dundee, UK. The numerical studies show that the proposed approach can reduce the total charging waiting times of the fleet by 48.3% and the total charged the amount of energy of the fleet by 35.3% compared to a need-based charging reference policy.