Hybrid-priority model predictive control for large-scale EV charging under transformer capacity constraints and incomplete information

Purpose This study addresses the scheduling of large-scale electric vehicle (EV) charging under strict transformer capacity constraints, fixed electricity pricing and incomplete battery information. It aims to optimally balance emergency task prioritization, peak load shifting and user fairness in highly stochastic residential environments. Design/methodology/approach A closed-loop Hybrid-Priority Model Predictive Control framework is proposed for deployment at the community aggregator level. To tackle state of charge unobservability, a data-driven estimation module using historical charging profiles and safety buffers is integrated. The receding-horizon optimization incorporates a multi-dimensional dynamic priority index that synthesizes departure urgency, remaining energy demand, emergency weighting and fairness evolution. Subsequently, a three-stage power allocation algorithm comprising a high-priority guarantee, demand-based distribution and fragmented capacity filling precisely maps feasible power commands to individual chargers. Findings Evaluated under dynamic base load fluctuations and fully randomized initial vehicle states, simulation results with a heterogeneous fleet of 300 EVs show the strategy strictly prevents transformer overloading. It achieves a 96.5% completion rate for highly urgent emergency vehicles and 99.5% overall. It effectively shifts flexible non-urgent loads to late-night off-peak periods, demonstrating strong robustness against estimation errors and system uncertainties. Originality/value This work offers a highly robust, non-price-driven resource allocation solution for legacy residential communities. It effectively mitigates system uncertainties through a receding-horizon feedback mechanism, enabling the optimal coordination of heterogeneous fleets without relying on dynamic price signals or perfect state observability.