Household-transformer relationship and phase identification based on the AFCM-KNN

The use of voltage curve similarity for household-transformer relationship and phase identification is easily affected by abnormal voltage fluctuations. It is also affected by the household meter voltage values being closer to the voltage of the non-affiliated transformer meter. This article proposes a method based on the AFCM clustering and KNN algorithm for identifying a household-transformer relationship and phase. By introducing the soft clustering algorithm FCM and improving it, the membership values of the household meter belonging to different transformer meters and phases are obtained, avoiding the disadvantage of the hard clustering algorithm being either 0 or 1. Based on the difference in membership values, the direct judgment method or the KNN classification method is used to determine the attribution of the transformer meter and phase. Finally, the method proposed in this paper was validated using real low-voltage transformer data. The results show that the proposed method can effectively solve the problems with high recognition accuracy and robustness.