Non-Intrusive Load Monitoring-based Fuzzy Actor-Critic Reinforcement Learning Home Energy Management

Energy management is economically and environmentally crucial for today’s modern life. This paper presents a home energy management system (HEMS) using fuzzy actorcritic reinforcement learning (FACRL) and non-intrusive load monitoring (NILM), which follows a cyber-physical system (CPS) structure. An LSTM-based neural network predicts future residential consumption using the NILM dataset. The Python-based NILM toolkit disaggregates the whole power into appliancelevel power to obtain the consumption pattern of non-shiftable appliances. Then, the FACRL algorithm schedules the power usage of controllable residential facilities. To the best of our knowledge, it is the first time that the FACRL method has been applied to design a HEMS. The proposed method is applied to manage the consumption pattern of a house in Ottawa according to time-of-use tariffs and the temperature of this city, as well as the comfort level of the user. The results demonstrate low consumption during on-peak hours and the economic benefits of the HEMS.