Non-intrusive Load Monitoring Method Based on BIC Event Detection and LSTM Network Model

Abstract

A non-intrusive load monitoring method based on Bayesian Information Criterion(BIC) event detection and long short-term memory(LSTM) network model is proposed for the current deep learning-based non-intrusive load monitoring algorithms with high event detection requirements. Firstly, we use the sliding window-based BIC algorithm for event detection, design a unified LSTM network model with low complexity, and finally input the detected events into the LSTM model for recognition. The introduced fast event detection algorithm and the modified LSTM network model improve the accuracy of the algorithm. The overall performance of the proposed algorithm is tested on AMPds dataset and real data. The simulation results show that the above method can effectively improve the accuracy and outperform existing algorithms.