Clustering algorithm and abnormal point detection algorithm are important methods of data mining. Existing clustering and anomaly detection algorithms are mainly aimed at regular data mining, and there is no method to integrate the two algorithms for data analysis to realize stealing identification. In view of this, a new stealing identification method combines clustering algorithm and anomaly detection algorithm is proposed based on the analysis that principle of the relevant algorithms and the characteristics of the electricity data, it realizes accurate identification of stealing users through the deep excavation of the abnormal data. Theoretical analysis and experimental results show that this method can effectively improve the accuracy of stealing identification, and has certain practicability.