EV Aggregators Use AI to Execute Flexible Grid Services Reliably

⚡ AI Executive Summary

Researchers have developed a machine-learning framework that enables electric vehicle aggregators to reliably commit flexibility services across day-ahead and real-time electricity markets. The system converts theoretical EV flexibility into executable capacity by accounting for traffic patterns, battery state-of-charge dynamics, and real-world activation constraints. The approach uses temporal reinforcement learning to optimize bidding strategies across market layers while managing price uncertainty and reserve activation penalties. This work addresses a critical grid-integration challenge: EV aggregators have long struggled to convert nominal flexibility into reliable grid services because mobility patterns unpredictably alter charging states. By linking traffic forecasts to power-dispatch decisions, the framework reduces the gap between promised and delivered flexibility—a key requirement for grid operators to trust EV participation in ancillary-service and congestion-management programs. The revenue gains demonstrated suggest EV aggregation can move beyond simple energy arbitrage toward multi-service market coordination, strengthening the economic case for demand-side flexibility in renewables-heavy grids.