Abstract:To cope with the difficulty and high cost of low-voltage users phase identification, this paper proposes a method based on the voltage curves cluster analysis using K-medoids algorithm. First, measure the similarity between different users smart meter voltage curves by calculating the Pearson correlation coefficient; Secondly, select the initial clustering centers based on the density information of the voltage curve similarity distribution. Then, cluster all the users into three group users by the K-medoids clustering algorithm, and each group users voltage curves are more similar; Finally, identify the user phase. The method is applied in a power grid company, in the three-phase seriously unbalanced transformer area carryout the user phase identification, and then adjust some users phase according to different phase load, thus reducing the unbalance degree of transformer three-phase load, which has obtained the good effect.