Abstract: Because the background harmonic voltage fluctuation leads to the decline of system impedance parameter identification accuracy, and finally leads to the unreasonable division of harmonic responsibility, a harmonic responsibility division method based on weighted least squares based on modal clustering is proposed in this paper. K-medoids clustering algorithm is used to cluster the sample points to obtain the typical operation modes; The weighted least square method is used to identify the modal parameters under different modes and share their harmonic responsibility. The IEEE13 node simulation platform is built by MATLAB to simulate the harmonic voltage fluctuation under different background. The accuracy of identifying system impedance parameters by using this algorithm and traditional algorithm is compared to verify the rationality of this algorithm.