Long Short-Term Memory Based Refined Load Prediction Utilizing Non Intrusive Load Monitoring

This paper proposes a refined load prediction method (RLPM) based on the Non-Intrusive Load Monitoring (NILM) and the Long Short-Term Memory Network (LSTM). Specifically, the aggregated load data obtained from electricity meter is decomposed to different load components using NILM. On this basis, a behavior pattern-based household appliances classification method is established, which divides the individual household load data into the fixed load and the fluctuating load. Then LSTM is established separately for the fixed load series and fluctuating load series to realize the load prediction. Finally, the load prediction results for individual household and group households can both be obtained. The experimental results show that the proposed refined load prediction method (RLPM) can improve the accuracy of the load prediction results for both individual household and group households, with a significant application value.