Dynamic pricing strategy based on day-ahead charging demand forecasts

Abstract

This paper introduces a dynamic pricing model to distribute traffic across electric vehicle charging stations using demand forecasts. Analysis of real-world charging station data collected in Croatia from June 2019 to October 2022 shows that some stations experience heavy usage and long wait times, while others remain underutilized. To address this imbalance, the proposed strategy adjusts prices based on day-ahead demand predictions.

In this study, various time-series forecasting models were compared, and the Seasonal Autoregressive Integrated Moving Average (SARIMA) model demonstrated the best performance based on mean squared error (MSE) and mean absolute error (MAE). Consequently, SARIMA was selected for charging demand forecasting throughout this study.

The proposed dynamic pricing model aims to distribute charging demand more evenly across all stations, improving overall network efficiency. This model enhances station utilization, increases profitability, improves user satisfaction, and reduces waiting times. Compared to the three alternative models, the proposed approach achieves over a 27 % increase in profitability. Additionally, it enables more than 80 % of EVs to charge at their preferred stations, significantly outperforming other models in meeting user preferences. The model also reduces waiting times across the network by over 90 % compared to the second-best approach. Finally, it demonstrates superior load balancing, achieving more than 8 % improvement in mean load distribution variance over the next best method.