Non-intrusive Load Decomposition Method Based on LSTM Considering Temperature Factor

Non-intrusive load decomposition is a key technology for customer-side energy management. In this paper, a Long Short-Term Memory neural network (LSTM) based method is proposed to load decomposition. We also take into account the impact of nonelectrical factors (temperature) on load decomposition, which is used as the input of LSTM together with the total energy consumption to improve decomposition accuracy. The performance of the proposed method was evaluated using the DRED dataset. The results show that the LSTM model has certain advantages in the energy decomposition of various types of load devices with multistate changing. Meanwhile, the decomposition accuracy of the temperature sensitive loads is improved by considering the temperature factor.