Load curve integrated spectrum clustering algorithm considering double-scale similarity

Abstract

The invention discloses a load curve integrated spectrum clustering method considering double-scale similarity, and the method comprises the steps: the load form change similarity is calculated through the cosine distance of a load difference vector, and a double-scale similarity measurement mode for measuring the load similarity degree is constructed; through integrated clustering, clustering performance improvement and effective combination of two measurement modes are realized, spectral clustering is used as a basis clustering model generation algorithm, different similarity measurement modes are selected, different clustering cluster numbers and random operation are set to ensure the diversity of base clusters, a weighted consistency matrix and spectral clustering are used as a clustering integration strategy, in the clustering integration process, a Thevenberg index DBI or a new index MDBI is adopted as a clustering evaluation index, the reciprocal of the DBI or the MDBI is adopted as a weight self-adaptive setting basis to calculate a consistency matrix, and then final integrated clustering division is achieved through spectral clustering. The method has excellent clusteringeffectiveness and robustness, and can avoid the defect that a single spectral clustering algorithm needs parameter optimization for different data sets.

Description

technical field [0001] The invention relates to a load curve integrated spectrum clustering algorithm considering double-scale similarity, and belongs to the field of load characteristic analysis of power systems. Background technique [0002] Under the background of urban energy Internet, the construction and improvement of electricity collection information system and dispatching, transportation inspection, and marketing business systems have promoted the rapid accumulation of electric power data resources. Valuable information such as energy consumption characteristics is hidden in power data, which needs to be mined by applying data analysis technology. As an unsupervised learning technology, clustering is suitable for the classification of unlabeled load curves, providing power companies with classification results based on differences in load characteristics, helping power companies to accurately grasp the laws of user energy consumption behaviors, and providing power ...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem summarized in the background technology, the present invention provides a load curve integrated spectr...

Images