Distributed and Timely Smart Microgrid Management Through Markov Games

Energy delivery within smart microgrids often requires prompt reaction to the system state. In the presence of multiple energy sources, inefficiency may arise due to their lack of coordination. In this paper, we frame the task of efficient energy management as a dynamic program, and we further expand it to the case of multiple agents. We combine this approach with game theory and we leverage the similarity between Markov games concerning information and energy exchanges in networks. This methodological motivation allows us to identify distributed control techniques for efficient energy delivery. Specifically, we highlight how a naive distributed and selfish control of individual nodes may be inefficient from a game theoretic perspective. Yet, a decentralized strategy that combines energy availability and global network cost as shared objectives can significantly improve the outcome, approaching the performance of a centralized resource allocation still in a distributed manner.