Recent Advances in Reinforcement Learning Applications for Building Energy Management: A Mini Review.

Abstract

In 2019, buildings accounted for 55% of the global electricity demand, making them a key contributor to global emissions and a core target for energy efficiency, energy reduction, and policies and measures promoting renewable energy usage. Reinforcement learning (RL) is an agent-based modelling technique that has proven successful in many applications, particularly in artificial intelligence. RL has attracted research attention owing to its utilization in building energy management (BEM) applications. In this work, the latest research advances that utilize this method are investigated and discussed, primarily its usage in modelling complex building energy problems, building energy consumption control, optimization for comfort and cost savings, and the enhancement of demand forecasting algorithms. Furthermore, the combination of RL with other deep learning methods is discussed. As a state-of-the-art technology in smart grid building applications, RL is applied for control purposes and forecasting enhancement.