Clustering algorithm of power load curves in distribution network based on analysis of matrix characteristic roots

In order to solve the problem of determining the clustering number of power load curves, this paper presented a clustering method based on the analysis of matrix characteristic roots. The clustering number is determined by the analysis result and the method is applied to the power load curves clustering. Firstly, the unified mathematical expression of the same type of load curves is established based on the changing pattern of load curves. Then, the equivalent relation between the clustering number of load curves and the number of larger characteristic roots is demonstrated. Besides, two criterions are presented to calculate the number of larger characteristic roots; and the number is the clustering number. After that, the load matrix is standardized and clustered by K-means algorithm. At last, the clustering result is evaluated by two indexes, the distance within classes and the distance between classes; and case simulation is given to prove the effectiveness and reasonability of the method proposed.