Aiming at the problem that the traditional distribution network topology identification method can not meet the efficiency and accuracy in complex environment, a low-voltage distribution network topology identification method based on wavelet transform (WT) is proposed. The connection topology of single-phase and three-phase users is accurately identified by using the data of smart meters and the measured values of substation feeders, combined with the multi-scale feature extraction technology of wavelet transform. Through the experimental verification, under the conditions of different network observability and high penetration of renewable energy, the topology identification shows superior robustness and efficient computational performance, and can effectively adapt to the complex situations of unbalanced and incomplete observation in the power system. The results show that the method has high recognition accuracy and is suitable for the practical application of the power distribution network.