Understanding Reinforcement Learning Control in Cyber-Physical Energy Systems

The possibility of modeling a renewable energy system as a Cyber-Physical Energy System (CPES) offers new possibilities in terms of control. More precisely, this document discusses the applicability of Reinforcement Learning (RL) techniques to CPES. By considering a benchmark algorithm, we focus on conceptual and implementation details and on how such details relate to the problem of interest. In this case, we simulate how a RL model can optimize the energy storage control in order to reduce energy costs. The work also discusses the issues that arise in RL models and the possible approaches to these difficulties. Specifically, we propose investigating a better exploitation of the memory mechanism.