A Home Energy Management System Optimization Model Based on DNN and RL Adapting to Users’ Uncertain Behaviors

About 35% of the total energy consumption comes from house appliances. Home energy management system (HEMS) is a household optimal system that can improve the efficiency of electricity consumption, increase the consumption of new energy and reduce carbon emissions. At present, the research of HEMS is simulated under the fixed load setting, and the adaptability of the model to the users’ uncertain behaviors is not considered. In this paper, an optimization model of HEMS based on deep neural network (DNN) and reinforcement learning (RL) algorithm is presented. The model aims to minimize the electricity cost and the comfort cost. For the residential case in this paper, the model can reduce 34.2% total cost. The simulation results show that the proposed model can better adapt to the uncertain behaviors of users than the optimization model based on genetic algorithm (GA).