Demand Forecasting for Electric Vehicle Charging Stations using Multivariate Time-Series Analysis

Explainability improves infrastructure planning: The paper demonstrates that using SHAP (SHapley Additive exPlanations) to identify which factors most influence charging demand—such as weather, day of week, and holidays—enables better decision-making for where and when to build new charging stations, rather than relying on black-box predictions alone.

Precise short-term forecasting addresses real infrastructure gaps: By predicting charging demand at 15-minute intervals for the next day, the approach directly tackles documented problems like long wait times and insufficient infrastructure, suggesting that better forecasting could help station operators allocate resources more efficiently.

Multiple factors shape charging behavior: The research identifies that weather conditions, temporal patterns (day of week and month), and holidays all meaningfully affect EV charging demand, indicating that responsible infrastructure planning must account for these diverse influences rather than assuming uniform demand patterns.

Real-world testing grounds findings in practice: The model was validated using actual data from California State University, Long Beach charging stations, meaning the results reflect genuine charging patterns rather than theoretical scenarios, increasing confidence in the approach's practical applicability.