Non-intrusive Load Monitoring Method Based on Transfer Learning and Sequence-to-point Model

Abstract

Non-intrusive load monitoring (NILM) or load disaggregation refers to the task of estimating the appliance power consumption given the aggregate power consumption readings. The sequence-to-sequence model with neural networks is widely used in the NILM, but the data processing of this method is complicated and the accuracy of load disaggregation is low. The sequence-to-point model improves the accuracy of load disaggregation by changing the data processing method and the neural network structure. The UK-DALE dataset is used to compare the load disaggregation effects of seq2seq and seq2point models. In addition, the problem of insufficient training data is solved by a model training method based on transfer learning. The effectiveness of this training method is verified by collecting a small amount of data from private dataset.