Stochastically Stable Synchronous Learning for EV Aggregators Considering Their Collective Age of Information

Abstract

This article exploits potential games to model the integrated demand response/economic dispatch problem and proposes a synchronous log-linear learning (LLL) architecture for generators and electric vehicle (EV) aggregators to simultaneously make real-time decisions. Furthermore, to address the reality that each EV aggregator needs to manage dynamic portfolios of EVs in a geographically spread area and participate on behalf of the most recent demands of its managed EVs, the concept of age of information (AoI) is explicitly modeled and incorporated. It is proven that the proposed framework has guaranteed convergence to a Nash equilibrium that is also a global optimizer, using the stochastic stability theory, perturbed Markov process, and revision trees. Numerical test results on a 15-generator, 15-aggregator benchmark network validate the proposed framework and show that the explicit consideration of AoI improves the dynamic characteristics of the proposed synchronous LLL compared to conventional learning schemes in potential games.