High-precision spatiotemporal prediction of electric vehicle charging demand via data-driven travel behavior modeling

Accurate prediction of the spatiotemporal distribution of electric vehicle (EV) charging demand is essential for charging infrastructure planning and local grid management. However, existing models often rely on coarse spatial zoning, simplified departure-time assumptions, or charging decisions determined mainly by state of charge (SOC), which limits their ability to represent the behavioral mechanisms underlying private EV charging demand. To address these limitations, this study develops a behaviorally grounded, data-driven framework that integrates refined functional zoning, travel-chain generation, destination-constrained charging decisions, and Monte Carlo simulation. Using three years of GPS–SOC trajectory data from 30 privately owned EVs in Jiading District, Shanghai, the framework links 12 functional zone categories with trip-purpose-specific departure-time distributions, empirically derived parking durations, and SOC-based charging feasibility. The proposed model achieves a mean absolute percentage error of 6.8% and a Pearson correlation coefficient of 0.963, outperforming benchmark models in temporal peak alignment, secondary-peak identification, and spatial hotspot localization. The results show that refined functional zoning improves prediction accuracy not simply by increasing spatial resolution, but by better connecting land-use semantics, daily travel behavior, and destination-level charging opportunities. The framework also captures weekday–weekend and seasonal variations while remaining suitable for planning-oriented scenario analysis. These findings demonstrate the value of behaviorally interpretable spatiotemporal modeling for private EV charging-demand prediction and provide practical support for differentiated charging infrastructure deployment in mixed-use urban areas.