Machine-Learning-Based Home Energy Management Framework via Residents' Feedback

Abstract

This study introduces a smart home energy management (SHEM) framework using an artificial neural network (ANN) approach that incorporates user feedback to gauge preferences regarding cost and comfort. The SHEM framework aims to minimize energy costs by adjusting the operation of home devices according to hourly electricity prices. However, deviations from user preferences can lead to varying levels of dissatisfaction. Residents provide feedback at the end of each day, rating their satisfaction with the energy management system on a scale from 0% (completely dissatisfied) to 100% (completely satisfied). The findings reveal how prioritizing dissatisfaction over cost affects energy management, overall cost, and total dissatisfaction. The ANN-based framework is then tested with two artificial users, demonstrating that the proposed SHEM framework can accurately learn to prioritize dissatisfaction over cost within a few days of operation.