Double-Layer Improved K-Means Based Electricity Industry Classification Method

Abstract

In the traditional industry classification, load characteristics of users in the same industry are quite different, seriously affecting their application. Thus, this paper focuses on the commonality of user behavior and proposes a novel industry classification method based on two-layer improved k-means algorithm, which optimizes the selection of cluster number. In the upper-level clustering, daily load curve of each enterprise is firstly plotted so that the improved k-means algorithm is applied to obtain the typical daily load curve of each enterprise. In the lower-level clustering, the improved k-means algorithm is used to cluster the typical daily load curves of each enterprise to obtain new industry classification. Simulations are carried out with a dataset of a province in China and the results show that the proposed method can effectively achieve industry classification depend on similar electricity consumption behavior.