What makes demand wait? Modeling cross-day flexibility in electric vehicle charging

This study proposes an interpretable framework for understanding the dynamics of daily electric vehicle (EV) charging demand using three years of residential charging data from Denmark. By aligning daily boundaries with charging demand peaks, we uncover behavioral patterns driven by price signals, routines, seasonality, and weather. A key insight is that EV users react more consistently to relative price changes, especially changes in nighttime prices relative to the previous day, than to absolute values. Demand also shows strong routine and seasonal influences, with nighttime prices dominating in winter and midday prices becoming more influential in summer. Comparing machine-learning algorithms (Random Forest, Catboost) with a Linear Regression model, we demonstrate that the latter provides the best balance between performance and interpretability, indicating distinct weakly patterns of aggregated charging behavior. Overall, the proposed framework offers a structured approach to uncovering EV demand dynamics, providing system operators and flexibility providers with clear, actionable insights.