Abstract: Aiming at the low efficiency and accuracy of traditional power theft identification methods for low-voltage power users,a power theft identification method based on lightweight lifting decision tree and BP neural network is proposed. Firstly,the feature segmentation is carried out according to the abnormal power consumption expert feature database,and then the abnormal power consumption recognition results are extracted from the segmentation results combined with the feature index matching degree and application requirements,and a double-layer recognition model is established. Aiming at the problem that the traditional decision tree algorithm used for feature segmentation model occupies more computing resources when the eigenvalues are discretized,the leaf growth strategy is introduced and its growth depth is controlled to avoid over fitting and improve the calculation efficiency. In addition,in order to improve the efficiency of data preprocessing,Newton interpolation and 3σ The law preprocesses the missing and abnormal data in the collected user’s original power consumption data. Using the actual user data set of a grid area of Guangxi power grid for case analysis,the results show that the recognition accuracy and recognition efficiency of this algorithm are better,and its effectiveness is verified.