Power large users are the key users of power sup.ply enterprises and their potential value and development trend inpower market environment are closely related to the profit of powersupply enterprises.In order to identify valuable user behavior andvalue characteristics, a large user segmentation method based on APand K.means algorithm is proposed.First of all, from the existing in.dicators to extract the key sub.indicators, and consider the recentlarge and long.term power consumption rate of electricity, put for.ward to assess the development potential of large users of the break.down of indicators.Secondly, the AP and K.means are combined tosegment the large power users to solve the problem of finding the ini.tial clustering center and the number of clusters.At the same time, itis easy to avoid the local optimization when K.means clustering prob.lem; finally, based on the user data of a region in Zhejiang province,the author analyzes the feasibility of the proposed method.