Machine Learning Predictive Models in Smart Home Energy Management: Progress and Challenges

Abstract

: Home Energy Management Systems (HEMS) is becoming an essential part of the low-carbon economy and smart cities due to the global energy crisis and climate change issues. Conventional Home Energy Management Systems have significant difficulties in dealing with the complexity and heterogeneity of energy data, which hinders their practicality and wide adoption. In this comprehensive paper, the survey and review the recent progress in machine learning techniques for smart home energy management applications, highlighting key developments. The systematically categorize and thoroughly analyze the performance of various machine learning models across multiple critical tasks: energy consumption prediction, user behavior analysis, demand response optimization, and Heating, Ventilation, and Air Conditioning (HVAC) control. The findings show that deep learning methods and reinforcement learning approaches consistently achieve better prediction accuracy and enhanced adaptivity; however, persistent issues exist, such as data heterogeneity, the weak generalization ability of models across different settings, and difficulties in practical real-world deployment and implementation. To address these challenges, the strongly suggest that future research should concentrate on cross-task joint optimization strategies, multi-source data fusion techniques, and long-term adaptive model frameworks to significantly facilitate practical, intelligent energy management in homes.