A Review of Recent Advances on Reinforcement Learning for Smart Home Energy Management

Abstract

Smart home energy management is one of the core problems in modern power grids. With the increasing adoption of different types of electric appliances and on-site intermittent renewable energy generation, it has been very challenging to use conventional control techniques for such energy management problems. Reinforcement Learning (RL) has attracted growing research interest recently; it also demonstrates its great potential to enhance smart home performance while addressing some limitations of other advanced control techniques, such as model predictive control. In this paper, we present a review of the recent advances on RL for smart home energy management. The problem of smart home energy management, the background for RL algorithms, and the survey of recent advances on RL for the smart home are presented. However, even though RL-based smart home controls have gained increasing research interest, it is in the beginning research stage. Several questions in this field are still not well-studied and worth further investigation, including data-efficient reinforcement learning, safety concerns, and how to include human behaviors in the loop of making control decisions. In this short survey, we also discuss the challenges and potential opportunities using RL in smart home control.