Charger as Sensor: Revealing Electric Vehicle Spatio-temporal Charging Patterns in Scotland

The spatiotemporally heterogeneous and escalating EV charging demand poses significant challenges to grid stability and infrastructure resilience, necessitating a deeper understanding of charging patterns to support transportation electrification. Apart from sensing charging patterns from limited EV trajectory data or inferring charging activities from SOC-independent trajectory data, geo-tagged chargers emerge as distributed sensors capable of revealing comprehensive charging patterns. This study develops a novel analytical framework that leverages 122,214 real-world charging sessions from 5,834 chargers in Scotland to identify three fundamental charging patterns—duration, frequency and rhythm—through an innovative three-dimensional spatiotemporal data cube approach. By integrating ensemble learning with the SHAP (SHapley Additive exPlanations) framework, we elucidate complex nonlinear interactions among key determinants, including built environment, weather conditions, demographic factors, and charger attributes. Firstly, our findings demonstrate distinct heterogeneity in charging duration and frequency between weekend/weekday and fast/slow charging. Specifically, weekdays exhibit marginally longer charging durations and higher charging frequencies compared to weekends. Fast charging maintains consistent duration, typically completing within one hour regardless of start time. In contrast, slow charging shows substantial temporal variability, with longer durations observed during afternoon and evening periods. Secondly, both fast and slow charging frequencies show long-tail distributions across multi-scale areas, consistent with preferential attachment behavior. Fast-charging frequency also demonstrates greater spatial heterogeneity and higher sensitivity to weather conditions and built environment factors compared to slow-charging systems. Moreover, we also identify four charging rhythm patterns, where nighttime charging is the most dominant followed by daytime, afternoon and morning charging patterns. Built environment emerges as the primary determinant, followed by weather conditions. The study provides critical insights for charging demand management, clean energy integration and informing strategic deployment and maintenance of public EV charging infrastructure.