Applications of reinforcement learning for building energy efficiency control: A review. (1st June 2022)

• Record Type: Journal Article
• Authors: Fu, Qiming; Han, Zhicong; Chen, Jianping; Lu, You; Wu, Hongjie; Wang, Yunzhe
• Is Part Of: Journal of building engineering. Volume 50(2022)
• Publication Date: 2022-06-01
• Subjects: Reinforcement learning -- Intelligent buildings -- Energy consumption; Building -- Periodicals; 690.05
• DOI: 10.1016/j.jobe.2022.104165
• Languages: English
• ISSNs: 2352-7102

Abstract

Abstract: The wide variety of smart devices equipped in modern intelligent buildings and the increasing comfort requirements of occupants for the environment make the control of intelligent buildings important and complex. Reinforcement learning, as a class of control techniques in machine learning, has been explored for its potential in the field of intelligent building control. Reinforcement learning methods applied to intelligent buildings can effectively reduce energy consumption. In this paper, we classify reinforcement learning algorithms and analyze the control problems that each algorithm is suitable for solving. In addition, we review the reinforcement learning methods applied to control and manage buildings, outline the problems and future directions of reinforcement learning applications in intelligent buildings, and give our suggestions for researchers who want to use reinforcement learning methods to solve control problems in this field. Highlights: Reinforcement learning has been explored for its potential in the field of building energy efficiency control. We classified reinforcement learning algorithms and analyze control problems that each algorithm may be suitable for. We reviewed studies on the applications of reinforcement learning for building energy efficiency control. The problems and future directions of reinforcement learning applications in intelligent buildings have been outlined. Some suggestions have been presented for using reinforcement learning to solve intelligent building control problems.