Non-intrusive load decomposition method based on seq2seq

Abstract

The invention provides a non-intrusive load decomposition method based on seq2seq. The non-intrusive load decomposition method comprises the following steps: 1, designing a seq2seq model; 2, function extraction, wherein Conv1D is used to carry out convolution and pooling on a power sequence on a one-dimensional scale, and power features are extracted by relying on a plurality of convolution kernels with the same weight; 3 load identification based on LSTM (Long Short Term Memory); and 4, carrying out seq2seqBCL load decomposition. The invention provides a non-invasive load decomposition algorithm (seq2seq Based on CNN and LSTM, seq2seqBCL) based on seq2seq based on combination of a convolutional neural network (CNN) and a long short-term memory network (LSTM) in order to solve the problem that an existing non-invasive load decomposition method is relatively low in decomposition accuracy under a low-frequency sampling condition (1Hz and below), the deep learning model takes a power time sequence as the input of a network, and performs feature extraction through a CNN (Convolutional Neural Network). In consideration of the time sequence of electric power data, the LSTM layer is added to carry out electric appliance identification, and compared with a seq2seq model in NILMTK, the number of network layers is reduced, and the network structure is simplified.

Description

technical field [0001] The invention belongs to the field of non-invasive load detection and relates to a non-invasive load decomposition method based on seq2seq. Background technique [0002] The development of Non-intrusive Load Monitoring (NILM) can be roughly divided into three stages: the proposal stage, the machine learning stage, and the deep learning stage. Non-invasive load monitoring was first proposed by Professor Hart in the 1980s, and in 1992, Professor Hart first proposed a non-invasive load monitoring system. It was not until 2008 that scholars proposed a method based on integer programming. Kolter et al. used the FHMM model for non-intrusive load decomposition in 2011. After testing on the REDD dataset, this method achieved the best monitoring performance at that time. In 2014, the Non-Intrusive Load Monitoring Toolkit (NILTK) was released. This toolkit is an open source tool specifically designed to compare energy decomposition algorithms in a repeatable m...

AI Technical Summary

Problems solved by technology

The NILM load identification algorithm has relatively high requirements on the quality of the data set. Some scholars have conducted research using ENERTALK and found that when the sampling frequency of the data set is lower than 1-3 Hz, the performance of NILM will be greatly affected, and the decomposition accuracy is low.

Images