Scalable Optimization of Electric Vehicle Fleets in Virtual Power Plants: Centralized Versus Decentralized Implementation

This paper proposes a decentralized optimization framework for integrating electric vehicle (EV) fleets into Virtual Power Plants (VPPs). The approach relies on the Alternating Direction Method of Multipliers (ADMM) to decompose the global VPP dispatch problem into parallel subproblems, enabling scalable coordination of large EV fleets. To assess performance and optimality, we compare the decentralized scheme against a centralized benchmark that maximizes revenues from energy arbitrage and ancillary services while respecting EV-specific constraints such as state-of-charge limits and departure requirements. Numerical experiments across various fleet sizes (10-4000 EVs) and price profiles (sinusoidal and stochastic) show that while centralized optimization is efficient for small fleets, the decentralized ADMM approach exhibits near-linear scalability and becomes advantageous beyond approximately 4000 EVs. These results underscore the trade-offs between centralized and decentralized VPP management and demonstrate the feasibility of large-scale, real-time EV coordination.