Weather-informed optimal scheduling of electric vehicle charging under extreme conditions: A case study from the Scottish islands

Extreme weather can trigger rapid wind cut-outs and price spikes in islanded, wind-dominated distribution networks, calling for weather-aware management of flexible demand. This study presents an integrated framework that couples a data-driven uncertainty layer converting day-ahead wind and load forecasts into calibrated, heteroscedastic, temporally correlated ensembles; a physics-aware scenario reduction that embeds net-injection deviations with bus scaling and ramp weighting before K-means clustering; and an enhanced multi-objective grey wolf optimizer that schedules EV charging or discharging by evaluating voltage quality via AC power-flow results. This model shows that smart charging lowers system-wide voltage dispersion and energy cost relative to baseline operation; under storms, adding vehicle-to-grid further curbs voltage deviation and diesel-backed emissions in the Orkney case. A sensitivity study on EV participation (50%–100%) under behavioral uncertainty shows robust but attenuated gains: the voltage deviation index decreases by about 2% and diesel CO₂ by about 3.7% as participation rises, while EV-side costs scale with the number of responding vehicles. These results suggest that weather-aware, electrically grounded uncertainty aggregation combined with multi-objective metaheuristics can enhance distribution-level resilience and reduce reliance on fossil backup during extreme weather.