Keywords
daily load data; CRITIC weighting method; singular value decomposition; weighted FCM clustering; clustering center selection
Abstract
Daily load data clustering is an important method for analyzing and extracting users ’ electricity consumption characteristics.The indicator weights of the dimensionality-reduced sampling data for clustering will affect the clustering results.Therefore, a daily load data clustering method based on the integration of the CRITIC weighting singular value decomposition （SVD） dimensionality reduction method （C-SVD） with an improved weighted fuzzy C-means （FCM） algorithm is proposed.Meanwhile, to solve the problem that traditional FCM is susceptible to the initial clustering centers, an adaptive initial clustering center determination method called density-distance center selection （DDCS） is proposed.Firstly, SVD is adopted to perform dimensionality reduction on the load data.Then, the CRITIC weighting method is used to configure weights for the dimensionality-reduced indicators.Then, the DDCS method is utilized to determine the initial clustering centers.Finally, the weighted FCM algorithm is applied to cluster the load data.Simulation examples show that, compared with traditional methods, the proposed method has strong robustness and can significantly improve the accuracy of load data clustering results.
DOI
10.19781/j.issn.1673-9140.2025.05.009
First Page
90
Last Page
97
Recommended Citation
WAN, Jinwei; XING, Jie; SHAN, Yinghao; JIN, Beiyu; and HOU, Meiqian
(2025)
"Improved FCM load clustering method based on C-SVD dimensionality reduction,"
Journal of Electric Power Science and Technology: Vol. 40:
Iss.
5, Article 9.
DOI: 10.19781/j.issn.1673-9140.2025.05.009
Available at:
https://jepst.researchcommons.org/journal/vol40/iss5/9
