Current signal classification algorithm

Abstract

The invention discloses a current signal classification algorithm, which comprises the following steps of: dividing a current signal waveform into different sections according to physical characteristics of the current signal waveform, respectively extracting the shapes, statistics and harmonic characteristics of the sections, and calculating the similarity between section pairs. The current signal classification algorithm also comprises the following steps: segmenting a current signal, separating a standby current and an overshoot current, and only leaving a working current; extracting shapedistribution characteristics of the working current section; extracting statistical characteristics of the working current section; extracting harmonic characteristics of the working current section;calculating the similarity between the segment pairs; and searching and analyzing a maximum clique set in the similarity graph by adopting a maximum clique search algorithm to obtain a class which isautomatically classified. According to the current signal classification algorithm, the current signals generated by different electric appliances in different working states can be quickly and accurately classified, so that subsequent processing is facilitated.

Description

technical field [0001] The invention belongs to the field of non-invasive load monitoring, and relates to a method for extracting characteristics of current signals, specifically a classification algorithm for current signals. Background technique [0002] In recent years, with the continuous development of smart grid, the lean and intelligent analysis of electrical signals has attracted more and more attention. For example, non-intrusive load monitoring (NILM) technology uses feature extraction and machine learning algorithms to analyze pooled voltage and current and monitor appliance usage without the need for sub-meters. Many feature extraction methods have been proposed for non-intrusive load monitoring, such as wavelet features, voltage-current traces, current harmonics, active / reactive power, and automatically learned deep features, etc. In the paper [Non-Intrusive Load Monitoring Using Semi-Supervised Machine Learning and Wavelet Design], signal features are extracte...

AI Technical Summary

Problems solved by technology

Active / reactive power characteristics are difficult to distinguish between appliances with similar power

Deep learning methods require a large amount of labeled data for supervised or semi-supervised learning, and labeled data are often expensive and scarce in actual scenarios, which greatly limits their development.

Images