Load Analysis Using Reinforcement Learning for Home Energy Management Systems

This study presents an analysis of electric appliance loads within the context of the Home Energy Management System (HEMS) for identifying non-intrusively the daily consumption routines of the residents. The methodology aims to determine that the periods of use of the devices are the periods of user comfort, as they reflect the daily routine. Applying methods of optimizing electrical consumption without corresponding to this real comfort compromises residents’ satisfaction because changing the usual routine causes discomfort. This view was noted as a gap in the technical literature that focuses on the intersection between energy efficiency and users’ comfort regarding the use of home appliances. Using load analysis and Reinforcement Learning (RL) with multiple agents, the study endeavors to find the ideal times for device usage, aligning them with the residents’ routines. As a result, the study showcases a simulation based on a real residence, employing the proposed load analysis methodology in conjunction with Reinforcement Learning.