Multi-agent Reinforcement Learning Based User-Centric Demand Response with Non-intrusive Load Monitoring

This research proposes a multi-agent reinforcement learning framework as a home energy management algorithm that focuses on user needs and preferences as well. The proposed method aims to secure the smart grid from power outages due to overloading. The system predicts appliance-level load demand for the following day using non-intrusive load monitoring (NILM) and four neural network-based supervised learning methods to pick the more accurate forecasting method. The Python-based NILM toolkit is utilized to analyze disaggregation methods on the forecasted demand to obtain appliance-level energy consumption. The user feedback and time-based price values are employed to optimize appliance scheduling. The simulation results of each stage of the algorithm are presented. The results demonstrate a 15% reduction in the electricity cost.