Reinforcement learning based smart home energy management

Abstract

With the rapid economic development and population growth, modern society has an increased energy demand. With limited non-renewable capacity and renewable energy technologies that have not yet been promoted on a large scale, the large demand for energy consumption presents a shortage of supply. Saving energy and achieving efficient use of energy is a critical task. At the same time, the energy consumption of residential buildings is an essential part of the total energy consumption, and to achieve adequate control of energy consumption, we can start by controlling the energy consumption of residential buildings, which is relatively easy to achieve in reality. This thesis discusses how to perform energy control for smart homes, which is divided into three main parts: 1) Prediction of future electricity price based on LSTM algorithm. A more accurate electricity price is the basis for subsequent energy control of home appliances. 2) Modeling of various household appliances and energy management algorithms with different operational characteristics in smart homes. 3) The modeling of the smart home optimization problem and the development of a home energy management algorithm based on reinforcement learning. Minimize the electricity costs for consumers while considering their comfort level.