A non-intrusive load monitoring method based on graph signal processing

Through a graph signal processing method, combined with event detection and V-I trajectory feature extraction, load decomposition of high-frequency non-invasive load monitoring is achieved, solving the decomposition performance of the existing methods and improving monitoring accuracy and efficiency.

CN114757284BActive Publication Date: 2025-03-28TIANJIN UNIV
2 Cites 0 Cited by

Patent Information

Application Number
CN202210405700.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-18
Publication Date
2025-03-28
Estimated Expiration
2042-04-18

AI Technical Summary

Technical Problem

The existing high-frequency non-invasive load monitoring methods have shortcomings in load decomposition performance, deep learning methods are complex and have long calculation time, while the separation performance of mathematical and machine learning methods is poor.

Method used

Using a graph signal processing method, load monitoring is realized through the steps of event detection, V-I trajectory feature extraction and graph signal clustering. The specific steps include event detection on high-frequency data, extracting the characteristics of current and voltage, building an undirected graph for graph signal processing, and finally achieving load decomposition through clustering.

Benefits of technology

It improves the accuracy and efficiency of load monitoring, reduces the amount of data usage and calculation, and does not require labeling of training data, and significantly improves the applicability and decomposition results.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present invention relates to a non-invasive load monitoring method based on graph signal processing, which mainly includes event detection, feature extraction based on V-I trajectory and clustering based on graph signal processing; firstly, event detection is performed on high-frequency data; features are extracted from the V-I trajectory composed of the current and voltage of the event; finally, the features of each event are clustered by the method of graph signal processing, so as to realize load monitoring. The present invention applies graph signal processing technology to the field of high-frequency non-invasive load decomposition, and uses the difference in current cycles for event detection. Compared with other high-frequency field methods that use low-frequency power changes to detect events, the amount of data usage and the amount of calculation are reduced. High-frequency current signal features and V-I trajectory features are extracted through high-frequency current signals and V-I trajectory features. The present invention does not need to label training data, and while improving the applicability of the algorithm, it also obtains better power decomposition results and realizes non-invasive load monitoring with higher accuracy.
Need to check novelty before this filing date? Find Prior Art