AI-based smart home energy management

Abstract

With the increasing request for saving energy and protecting environment, the concept of smart home energy management (SHEM) is developed and applied in our daily life. In this dissertation, the framework of smart home is firstly introduced, and some popular methods of SHEM, e.g., parsimonious random time series, regression or causal models and artificial intelligent based models are reviewed. Then, the methodology on forecasting electricity price is mentioned, such as data collection and pre-processing techniques. Subsequently, a DDPG based model for SHEM is proposed, where the specifical summary and mechanism are mentioned as well. Through the offline training and online testing, three different models of smart home are considered and simulated. The training results denote that the reward increases as the moving of step, while the testing results prove it as well, which means the DDPG based model proposed is effective and accurate.