Abstract: In recent years, due to the increasingly fierce competition in the distribution transformer market, many small and medium-sized enterprises "replace copper with aluminum" to produce transformers, leaving a hidden danger to the stability of the power system. At present, there is a lack of methods and techniques for non-destructive identification of transformer materials. The harmonic resistance method requires simple equipment, low cost, and has high engineering practical value, but it lacks scientific and rigorous algorithm support. In order to identify the material of transformer windings accurately and efficiently, we proposed an identification algorithm based on the Kmeans clustering and optimized on the basis of existing research. This method performs cluster analysis on the harmonic resistance of transformer winding obtained by statistics, then the algorithm is optimized through variance selection method and PCA dimension reduction. The optimization algorithm is verified to be feasible. The application analysis results show that the algorithm can effectively identify the material of transformer windings and has broad engineering application prospects.