Secondary Frequency Regulation via EV Aggregation-Based Virtual Power Plants: A Mean Field Game Strategy

This paper proposes a scalable mean field game (MFG) strategy for virtual power plants (VPPs) to coordinate large-scale electric vehicles (EVs) in providing secondary frequency regulation services. Firstly, in order to estimate the frequency regulation capability of the geographically dispersed EVs aggregated by the VPP, a joint data- and model-driven rolling evaluation approach based on a temporal multi-graph fusion neural network (TM-GCN) is designed, providing an essential basis for the subsequent optimal bidding and regulation signal decomposition. It is a challenging task to incentivize large-scale heterogenous and self-interested EVs to proactively response to the time-critical frequency regulation command. Therefore, a scalable MFG is formulated for the VPP to rapidly decompose the secondary frequency regulation signals and revenues among numerous EVs, effectively breaking through the curse of dimensionality. The existence and uniqueness of the mean field game equilibrium (MFGE) are validated utilizing the Brouwer fixed-point theorem and the contraction mapping theorem. Moreover, a distributed solution algorithm is developed to search for the MFGE, which exhibits good convergence performance across varying numbers of aggregated EVs. Comparative studies have validated the superiority of the proposed methodology in distributing secondary frequency regulation tasks across large EV fleets while maximizing VPP’s economic benefits.