Demand-Supply-Coupled Charging Station Location Problem for Electric Vehicle Demand Management

Abstract: This paper presents a demand-supply-coupled optimization model for the charging station location problem (CSLP), introducing a learning-based approach to integrate electric vehicle (EV) charging demand management into infrastructure planning. Our model addresses two primary challenges faced in EV charging infrastructure planning. First, conventional CSLP approaches often assume the availability of a known origin-destination (O-D) demand matrix of EV drivers, which is typically difficult for planners to obtain. EV charging demand is also influenced by the availability of charging infrastructure, as drivers may adjust their charging or travel plans based on charging station access. This lack of precise demand data complicates efforts to optimize charging locations effectively. Second, existing CSLP models commonly lack explicit demand measurements, such as the number of daily charging events, which are essential for stakeholders to assess the real-world usage and effectiveness of charging stations.

Our proposed approach consists of two stages. In the first stage, a gradient boosting-based learning model predicts the charging demand for a station using a set of 15 defined features, bypassing the need for an O-D demand matrix. This predictive model provides explicit demand estimations, facilitating more accurate infrastructure planning in scenarios where traditional demand data is unavailable. In the second stage, we introduce a demand-supply-coupled CSLP model designed to optimize the charging usage rates across both existing and potential new stations. To solve this optimization problem, we employ a gradient-based stochastic spatial search algorithm, allowing for efficient identification of optimal station locations.

We validate our model through a case study using six years of charging event data from Kansas City, Missouri. Results demonstrate that our method not only achieves reliable demand predictions but also improves overall charging usage rates by 14%, surpassing two benchmark models. This research offers valuable insights for agencies aiming to strategically expand EV infrastructure by identifying optimal locations for new charging stations based on real-world demand dynamics and usage patterns.