Recent advances in data mining and machine learning for enhanced building energy management

Abstract

Due to the recent advancements in the Internet of Things and data science techniques, a wide range of studies have investigated the use of data mining (DM) and machine learning (ML) algorithms to enhance building energy management (BEM). However, different classes of DM and ML algorithms feature different mechanisms and capabilities, resulting in their distinct roles and performance in BEM. Appropriate integration of different categories of DM and ML algorithms in BEM is essential to promote their wide application and provide guidance for new topic areas. This study presents a literature review of the use of DM and ML techniques in key areas of BEM, including building performance evaluation, energy usage prediction, and demand flexibility optimization. The review categorizes DM and ML techniques into three main categories, including supervised DM, unsupervised DM, and reinforcement learning (RL). Unsupervised techniques are primarily used for building energy performance assessment, while supervised techniques are mainly employed for building performance benchmarking and energy usage prediction. RL has been utilized for optimal building control to improve efficiency, demand flexibility, and indoor thermal comfort. The strengths, shortcomings, and integration of these methods in terms of their applications in BEM are discussed, along with some suggestions for future research in this field.