Distributed reinforcement learning for the management of a smart grid interconnecting independent prosumers

In the context of an eco-responsible production and distribution of electrical energy at
2 the local scale of an urban territory, we consider a smart grid as a system interconnecting different
3 prosumers which all retain their decision-making autonomy and defend their own interests in a
4 comprehensive system which rules, accepted by all, encourage virtuous behavior. In this paper,
5 we present and analyse a model and a management method for smart grids, shared between
6 different kinds of independent actors, which respect their own interests and encourages each to a
7 behavior allowing as much as possible an energy independence of the smart grid from external
8 energy suppliers. We consider here a game theory model, in which each actor of the smart grid is
9 a player, and we investigate distributed machine learning algorithms to allow decision-making
10 leading the game to converge to stable situations, in particular Nash equilibrium. We propose a
11 Linear Reward Inaction algorithm that proves to achieve Nash equilibria most of the time, both for
12 a single time slot and across time, allowing the smart grid to maximize its energy independence
13 from external energy suppliers.