Abstract: With the upgrade of the feed lines for low-voltage distribution systems, the consumer-transformer relationships and the phase information of customers change frequently. In order to correct the errors of the consumer-transformer relationship and the phase document caused by inefficient checking or untimely updating, a consumer-transformer relationship and phase identification method based on the voltage feature extraction and the clustering algorithm is proposed in this work. Firstly, the adaptive piecewise aggregate approximation (APAA) algorithm is used to extract the features of voltage curves. Secondly, the improved density-based spatial clustering of application with noise (DBSCAN) algorithm is used to detect the unmatched consumer-transformer relationship and identify the phase information of the normal users. The accuracy of the clustering results is boosted by determining the parameters adaptively and testing the correlation of the noise-making users in the improved DBSCAN algorithm. Some actual cases in the low-voltage distribution systems are studied to verify the accuracy of the proposed method.