Smart Home Energy Management Using Non-Intrusive Load Monitoring Integrated With Deep Reinforcement Learning

This paper presents a Deep Reinforcement Learning (DRL)-based Home Energy Management System (HEMS) that integrates Non-Intrusive Load Monitoring (NILM) (hereafter referred to as NDRL-HEMS), to realize a self-adaptive system that updates itself in response to changing user consumption patterns, a feature lacking in traditional HEMS. In the proposed system, NILM is employed for accurate load disaggregation, extracting appliance-level profiles used to train a DRL agent. This agent is trained with objective to maximize household self-sufficiency, and it strategically schedules shiftable loads to align appliance operations with solar generation and dynamic electricity pricing. A one-year simulation compares the conventional HEMS with the proposed NDRL-HEMS, showing a 10% increase in self-sufficiency and a reduction in electricity costs by 200 €. The grid-level analysis further indicates that widespread adoption of NDRL-HEMS can also alleviate grid congestion.