Phase Identification Method of Low-voltage Distribution Based on Integrated Similarity and Secondary Clustering

According to the shortcomings of the phase identification method based on single voltage or power data, this paper proposes a phase identification method based on integrated similarity and secondary clustering. This method comprehensively utilizes two measurement data of voltage and power for integrated similarity measurement, which solves the problems of single voltage data method in three-phase voltage balance and single power data method in low power consumption of users. Firstly, the redundancy of the original data is reduced by voltage data filtering and power feature extraction. Then, the users are clustered for the first time by using the filtered voltage sequence to generate the in-phase user cluster. Finally, the integrated similarity combining voltage and power similarity is used to cluster the users for the second time to realize the phase attribution discrimination of the users in the low voltage distribution networks(LVDN). Compared with the single data algorithm, the proposed method can effectively improve the accuracy of phase recognition. Compared with the existing method combined with power optimization model, this method has stronger applicability in the case of partial user data missing.