City-Scale Intelligent Scheduling of EV Charging and Vehicle-to-Grid Under Renewable Variability

Rapid electrification of road transport and growing shares of variable renewable generation are pushing urban low-voltage feeders toward their operating limits. Uncoordinated electric vehicle (EV) charging can create transformer overloads, voltage violations, and unfair delays, while most existing smart charging schemes either ignore distribution network constraints or treat fairness and risk in an ad hoc way. This paper proposes a city-scale hierarchical scheduling framework that coordinates EV charging and vehicle-to-grid (V2G) services under renewable variability. In the upper layer, a LinDistFlow-based optimal power flow computes feeder-constrained power envelopes and shadow prices over a rolling horizon, capturing transformer and voltage limits under photovoltaic (PV) uncertainty. In the lower layer, each station solves a queue-aware receding-horizon optimization that allocates charging/V2G set points across plugs using α-fair and lexicographic objectives, with conditional value-at-risk (CVaR) constraints on waiting times and state-of-charge (SoC) shortfalls. A digital twin of a medium-sized city with 24 stations (238 plugs) on five feeders and PV shares between 25% and 55% is used for evaluation. Compared with uncoordinated charging and myopic baselines, the proposed scheduler reduces feeder peak loading and PV curtailment while improving user experience and equity: average waits and 90% CVaR of waits are lowered, the Gini coefficient of waiting times drops (e.g., from 0.31 to 0.22), and SoC shortfalls are significantly reduced, all while respecting voltage limits. Each receding-horizon step executes in under 30 s on commodity hardware, indicating that the framework is practical for real-time deployment in city-scale smart charging platforms.