Forecasting Residential Charging Demand for Public Charging Stations in Urban Areas: A Spatial-Temporal Approach

We present a novel method to forecast public residential charging demand over a longer time horizon at a detailed spatial level. Our method relies on initial EV adoption data, socio-economic attributes, and GIS data, and is applied to the city of Brussels. One-year-ahead predictions are validated against an independent dataset of real- world charging sessions at 253 public charging stations. While the model shows considerable variability in its errors, its performance is comparable to existing models proposed in the literature. We furthermore find that correcting predictions for off-street parking availability does not improve model performance in the current adoption phase.