User Behavior Analysis of Substation Area Based on Improved K-Means Quadratic Clustering Algorithm

Abstract

Aiming at the deficiency of low efficiency and accuracy of traditional or single clustering algorithm in user behavior analysis of substation area, an improved dynamic weighted density clustering algorithm combined with k-means (DWDC-k-means) quadratic clustering algorithm was proposed. Firstly, based on the user aggregation degree and the sum of the squared errors (SSE), construct the initial return indicator of clustering and select the optimal clustering k value. Then, dynamic weighted density clustering algorithm is used to select initial clustering centers. Finally, k-means clustering based on the fusion similarity feature of weighted Euclidean distance and differential Pearson correlation distance in the second clustering stage. The algorithm is verified by the measured data set of the smart grid substation area, and the results show that the DWDC-k-means quadratic algorithm proposed in this paper can extract the typical user patterns more accurately than the contrast algorithm, and has a significant improvement in the given clustering performance index.