A Reinforcement Learning approach to the management of Renewable Energy Communities

Leonardo Guiducci, Giulia Palma, Marta Stentati, Antonio Rizzo, Simone Paoletti

Abstract: Optimal management of renewable energy is an important pillar of environmental sustainability, as it maximizes the use of clean and renewable resources. This article considers the optimal management of a renewable energy community that receives incentives for virtual self-consumption. This incentive scheme has been adopted in the Italian energy framework since 2020. The optimization problem maximizes the social welfare of the community, which includes the incentive together with the exploitation of renewable energy sources. A key role in such a problem is played by the battery energy storage system (BESS), which is crucial in balancing supply and demand. We propose a novel Reinforcement Learning-based BESS controller, aiming at maximizing the community social welfare by acting in real time and relying only on data available at the current time-step. Through different simulations in several scenarios, we demonstrate the effectiveness of our approach and its ability to outperform a state-of-the-art rule-based controller. Moreover, we assess the proposed approach by comparing its performance with that of the actual, though ideal, optimal control policy based on an oracle providing perfect knowledge of future data.