In view of the current situation that anti-theft technology is often analyzed by a single algorithm, which results in unsatisfactory anti-theft effect, a recognition method for low-voltage anti-theft users is proposed in this paper. Firstly, the technical line loss part of the line loss in the station area is separated. Then, K-means clustering algorithm is adopted to analyze the processed line loss data to identify the station area where the line loss rate fluctuates abnormally or is continuously high, and defines the time dispersion according to the clustering result to measure the suspected degree of electricity theft. Then, it analyzes the users under the abnormal station area, and studies the possible relationship between the change of electricity quantity of single users and the change of line loss rate in the station area through the correlation analysis. The outlier algorithm and K-means clustering algorithm are used to analyze the daily electricity consumption data of users, judge the suspected electricity theft of a single user, and determine the specific electricity theft behavior. The research results show that this method can identify the electricity theft of low-voltage users more effectively, which provides a new way for electricity theft identification and remediation.