In this paper, the Shanghai industrial users in some areas is studied using data mining techniques to analyze its behavior of electricity. According to user profile data acquisition and integration of electricity data, the data is repaired and normalized. Considering two factors that the number of clusters and selection of the initial cluster centers to improve the K-means algorithm, the improved K-means algorithm is used in data classification to extract all types of users clustering characteristic curve,then analyze the typical characteristics of behavior of electricity, and compared with the traditional K-means algorithm and relevant indicators is introduced to test clustering effect. The results show that improved K-means clustering algorithm can realize the different types of user classification function and can be more accurately and effectively analyze the behavior of users of electricity .