Assessing the impact of electric vehicles on power grids using sparse mobility data: A GIS- and machine learning-based approach

The electrification of light-duty transport introduces new and variable demand patterns, challenging distribution network planning and operation. As Electric Vehicle (EV) adoption grows, grid operators need better tools to address this novel demand and formulate decisions on investments and infrastructure upgrades. Yet, limited and low-quality mobility data complicates efforts to assess EV-induced grid stress. This study presents a novel bottom-up spatiotemporal energy model that uses low-resolution mobility data to evaluate electric mobility’s impact on distribution networks. By integrating Geographic Information Systems (GIS) and machine learning (ML), the model refines EV selection routines based on journey and driver features. It maps mobility patterns onto real power system elements, using tailored metrics to assess network stress. Using real-world data and implemented on the network twin of a Regional Distribution System Operator (DSO) in northern Italy, the model provides a dataset of geo-referenced EV arrivals with a one-hour resolution, comparing twelve future 2030 scenarios with the current one. The approach offers clear, actionable insights for grid operators and urban planners, bridging the gap between energy and transport infrastructure. The study finds that in the future scenarios, a significant impact is limited geographically to some areas of the Region’s capital city and less than 1% of grid elements report values over the operational levels in the most cumbersome scenario. Moreover, this e-mobility-induced stress is confined to the peak traffic hours of the day.