Energy trading in energy communities through load management using deep reinforcement learning

Energy Communities (ECs) play an essential role in the active engagement of end-users and the deployment of Renewable Energy Sources (RES), contributing to net-zero carbon emissions. In ECs, several entities/homes join to exchange a surplus or demand reduction during peak hours with the grid. In this paper, the end-user offers flexibility by adjusting the load demand by following different load-shaping patterns and avoiding sharing adjustments during critical hours. Therefore, load reshaping is proposed in ECs to provide a surplus to the aggregator and further share the surplus with the grid to overcome the uncertainty and intermittency caused by the growing number of RES and inadequate power system flexibility. Within this framework, this paper proposes a load management aggregation strategy based on Proximal Policy Optimization Deep Reinforcement Learning (PPO-DRL), designed to coordinate end-user demand to meet grid requirements, satisfy constraint limits, and enhance end-user economic benefits. In addition, the proposed approach is validated by comparing several algorithms to assess individual entity/home contributions, meet the overall cost minimization objective, and enable charging of the Electric Vehicles (EVs) while prioritizing user comfort. The results' effectiveness is validated over the fixed pricing mechanism for real-time management of EVs charging with a minimum Mean Absolute Error (MAE) of 0.0560 W. • An energy community trading model with end-user participation. • Deep Reinforcement Learning Framework for intraday trading with different load reshaping techniques. • The proposed solution enables dependence on the resource locally to manage and optimally charge EVs. • The end-user privacy is managed through aggregator management for flexibility control without sharing user data.