Non-intrusive load identification electric quantity decomposition method and system based on priority distribution

Abstract

The invention discloses a non-intrusive load identification electric quantity decomposition method and system based on priority distribution. The method comprises load start-stop edge detection, load start-stop edge classification, first priority load electric quantity decomposition and second priority load electric quantity decomposition. According to high-frequency sampling information collected and calculated by non-intrusive load identification equipment, a maximum likelihood ratio algorithm is adopted to detect a load start-stop edge, and active, reactive and harmonic characteristic information of the edge is obtained; the Euclidean distance of the edge feature data is calculated, and load edge classification is carried out, so that load type identification is realized; and finally, electric quantity decomposition is performed according to the priorities of the two steps. Different electric quantity decomposition strategies are executed based on the priority sequence, the problem that power fluctuation characteristic electric quantity identification of an air conditioner and the like is not accurate is solved, and accurate electric quantity decomposition of multiple electric appliances in the simultaneous working state can be achieved.

Description

technical field [0001] The present invention relates to the technical field of non-intrusive load identification, and more specifically, to a non-intrusive load identification power decomposition method and system based on priority allocation. Background technique [0002] For a long time, the information perception ability of user-side smart energy meters and other measurement equipment is weak, unable to provide accurate fine-grained load data and analysis services, and it is difficult to support two-way information interaction between power companies and residential users. [0003] Non-intrusive load identification technology is a new technology for real-time monitoring of residential power consumption status. It uses high-frequency sampling data from distributed terminals, smart energy meters and other equipment, and applies signal analysis and machine learning algorithms to obtain residential power consumption in a more precise manner. Information, such as the start and...

AI Technical Summary

Problems solved by technology

However, there are few studies on non-intrusive load power decomposition. In the prior art, the Chinese invention patent application (CN111753968A) "Non-intrusive load monitoring smart meter and power decomposition method" adopts sequence-to-point power decomposition with attention mechanism Model, through the self-learning weight fine total power decomposition, and the Chinese invention patent application (CN109840691A) constructed a sub-item power decomposition estimation method based on deep neural network

[0005] Whether it is a deep neural network or a sequence-to-point power decomposition model, the large amount of training sample data necessary for the model is difficult to obtain in practical applications

There are huge differences in the types of electrical appliances in different households, and the update and iteration of the same type of electrical appliances will cause a sharp increase in the number of training samples, otherwise it will lead to a decline in recognition accuracy

At the same time, this type of model must go through a long training and learning process, which is time-consuming, and is not suitable for localized load identification equipment that collects real-time processing and analysis in real time

Images