An improved k-means cluster method based on partial correlation coefficient for line-transformer connectivity identification in distribution network

Abstract

Abstract The accurate identification of line-transformer connectivity of the distribution network is very important for load balancing and line loss management. The electric quantity data can be used to identify the relationship, but the recognition accuracy is limited. The essence of relation recognition is unsupervised classification, but as there is no direct center of mass and distance, the traditional mainstream clustering method cannot be applied. Therefore, this paper proposes an improved K-means clustering algorithm based on a partial correlation coefficient. Firstly, spectrum analysis is carried out for the collected power data of the distribution transformer and the 10 kV line, and the high-frequency component of the data is obtained by the high-pass filter. Then, the K-means clustering algorithm improved by the partial correlation coefficient is used to set the electric energy data of each line as the initial centroid. The correlation between the station area and the line is calculated by using the concept of the partial correlation coefficient, and the station area represented by the minimum line distance is assigned to the corresponding line. The results show that this method can improve the accuracy of linear relation identification based on electric quantity data.