Non-invasive load identification method and system based on self-supervised comparative learning

Abstract

The invention discloses a non-invasive load identification method and system based on self-supervised contrast learning. The method comprises the following steps: data enhancement: extracting multi-label operation features from synthetic data; encoding a Grubrum matrix: converting the power sequence into an image matrix by using Grubrum angle field encoding, and realizing automatic feature extraction; and the self-supervised contrast learning architecture is used for extracting features from the main monitoring channel and the equipment monitoring channel respectively by utilizing a contrast learning architecture comprising two convolutional sub-neural networks, and applying an infoNCE loss function in the architecture so as to optimize parameters, reduce an intra-class distance of a feature space and enlarge an inter-class distance. And finally, determining the label of the test sample according to the average similarity of the samples of each category in the support set extracted from the training set. The validity of the algorithm is verified through two typical NILM common data sets REDD and ECO, the data sets comprise low-frequency power data, and the result shows that the algorithm has high accuracy of identifying the operation of the multi-label electric appliance.

Description

technical field [0001] The invention relates to a non-invasive load identification method, in particular to a non-invasive load identification method and system based on self-supervised contrastive learning. Background technique [0002] In recent years, the proportion of electricity consumption in total energy consumption has been increasing, and the saving of electricity consumption is the key to energy conservation. Several studies have been conducted and it has been found that more electricity can be saved when residents are aware of their real-time consumption of appliances rather than being billed for total electricity consumption each month, thus promoting the development of real-time load monitoring. In addition, demand response (DR) characterized by improved resource efficiency in electricity production requires flexible operation of residential electrical appliances according to the situation in the electricity market, thus relying on real-time consumption informat...

AI Technical Summary

Problems solved by technology

[0009] In view of the deficiencies in the current research on non-intrusive load recognition technology, the purpose of the present invention is to provide a non-invasive load recognition method based on self-supervised contrastive learning to effectively avoid poor feature extraction capabilities and correct multi-label operation recognition in traditional methods. low rate problem

Images