Spatio-temporal forecasting model for EV charging demands

The rapid expansion of electric vehicles (EVs) presents formidable challenges for power systems, especially regarding the scaling of EV charging infrastructure to meet the growing demands of EV fleets. These demands are influenced by complex interdependencies between spatial and temporal factors, such as transport work, weather conditions, traffic density, route and charging infrastructure, leading to imprecise charging demand predictions by existing models that do not fully address all factors. This study tackles this problem by introducing an innovative predictive model, named Weather Traffic Routes and Chargers (WeTRaC), which offers high-precision forecasts of spatio-temporal charging demands for EV fleets of e-buses and e-trucks, managing various operational conditions and ensuring efficient use of charging infrastructure. The model combines graph neural networks with detailed physics-based vehicle simulations using real-world inputs collected from cities around the world to produce state-of-charge (SOC) predictions. By pinpointing critical areas and peak times for charging demand, the model can optimise the placement of charging stations, thereby preventing grid overload and facilitating a green transition. It significantly accelerates prediction times with only a minimal trade-off in accuracy, as demonstrated in our simulated studies, making it an ideal tool for analysing vehicle fleet charging demand.