The present invention relates to a spectral clustering-based voltage source converter based high voltage direct current transmission (VSC-HVDC) AC-DC system optimal splitting section search method. The method is characterized by comprising the steps of constructing an undirected edge weight graph G of a power system according to a power system structure; calculating an adjacent weight matrix W and a degree matrix A according to trend data; determining whether the system is unstable due to large disturbance; dividing a generating unit into k homolog groups according to a rotor power angle rocking curve after a generator is disturbed, and updating the adjacent weight matrix W and the degree matrix A; updating the adjacent weight matrix W and the degree matrix A according to a VSC-HVDC converter station placement; calculating a standardized Laplacian matrix LN; and clustering, according to the homolog group number k, feature vectors corresponding to the first k smallest feature values of the standardized Laplacian matrix LN, so as to determine the optimal splitting section, wherein the feature vectors are calculated by using a k mean value clustering algorithm. The method is scientific and reasonable, has high applicability and high reliability.