A Dynamic Pricing System for Electric Vehicle Charging Management Using Reinforcement Learning

Abstract

Multiunit residential building (MURB) residents are an upcoming segment of electric vehicle (EV) owners and potential buyers (around 42% in Europe). Garage-orphaned MURB residents have to mostly rely on public chargers, which currently handle only 5% of the EV charging needs. With EVs becoming more mainstream, public chargers will not be able to match the operations scale without additional deployments. This will not only lead to a demand–supply mismatch in the short term but also impact the growth of EV adoption in the long term. For managing the demand–supply mismatch, dynamic pricing is a widely used control tool, but it is often difficult to make informed pricing decisions when 1) there is variability (both) in demand and supply, 2) users’ spatiotemporal behavior and price elasticity are unknown, and 3) charging preconditions (such as the state of charge) are not freely available. In this article, we present SurCharge, which uses reinforcement learning (RL) to overcome these challenges in dynamic pricing for EV charging. Our approach is evaluated on real-world traffic patterns for Luxembourg by augmenting the Luxembourg Simulation of Urban Mobility traffic scenario simulator with EV charging demand models. The results show that the proposed RL-based SurCharge system delivers a 10%–24% higher revenue margin than other competitive dynamic pricing baselines, without making unrealistic assumptions of prior models and data.