A spatiotemporal demand forecasting study of new energy vehicle charging stations

Against the backdrop of Electric Vehicle (EV) proliferation, accurately predicting the spatiotemporal distribution of charging demand is crucial. However, traditional methods face considerable challenges. This study aims to uncover the dynamic patterns of charging demand and construct a predictive framework. Methodologically, the ARIMA time series model is employed to analyze temporal features, Morans I index is used to assess spatial autocorrelation, and K-means clustering identifies spatial patterns. Pearson correlation and spatial regression models are integrated to quantify the influence of geographic dependency and socioeconomic factors. The results reveal a "dual-peak" temporal distribution and a "multi-core" spatial aggregation pattern of charging demand. Traffic flow and regional functional types are identified as key influencing factors.