As part of the EU-funded OPEVA project and his PhD research, SoftCPS researcher Tiago Fonseca is developing EnergAIze, a Multi-Agent Reinforcement Learning (MARL) energy management system designed to optimize the integration of Electric Vehicles (EVs) and Renewable Energy Sources (RES) within Renewable Energy Communities (RECs). The system enables prosumers to optimize energy flexibility from assets like EVs (V2G/V1G), heat pumps, and batteries, helping both individuals and communities reduce costs, balance demand, and maximize self-consumption.
Unlike traditional centralized approaches, EnergAIze adopts a decentralized architecture where each household acts as an autonomous agent, making energy decisions based on real-world IoT sensor data. Using a MARL algorithm (MADDPG), prosumers can tailor their energy strategy to minimize costs, prioritize self-consumption, or reduce carbon emissions. Initial tests with real-world data showed reductions in grid energy consumption, lower costs, and improved REC-wide energy sharing, making the system scalable and user-centric.
Evaluated using the CityLearn framework and the EVLearn extension, EnergAIze demonstrated up to 7.24% cost savings at the household level, a 6.45% increase in zero-net energy, and a 9% reduction in peak demand at the community level. Future developments will focus on scaling the system to larger RECs, real-world deployment using edge computing, and integration with platforms like Home Assistant.
For more details, refer to the IEEE SmartGridComm paper and Tiago’s master’s thesis, or request access to the SoftCPS GitHub repository. More information, please contact Tiago Fonseca or Luis Lino Ferreira.