High-voltage direct current system embedded in large-scale AC power systems will further increase the impact of commutation failure,DC blocking and other faults on system security and stability. The interaction between AC system and DC system will make the system dynamics more complex. In this paper, an intelligent analysis method based on electromechanic-electromagnetic hybrid simulation and machine learning is proposed to analyze the complex dynamic response characteristics of power system after large disturbances under different operation modes. A two-stage clustering model based on principal components analysis (PCA) dimensionality reduction,density-based spatial clustering of applications with noise (DBSCAN),K-means and other algorithms is built in this method,which can automatically cluster a large number of hybrid simulation dynamic curves in high-dimensional space, give corresponding identification and severity,and extract typical dynamic patterns of AC-DC systems under different faults. The leading security and stability problems in each mode are labeled and identified. The effectiveness of the proposed method has been verified in the planning model of East China Power Grid in 2025. The results show that the proposed method can effectively extract the system dynamic mode under different faults and support the dynamic mechanism analysis of AC/DC system under subsequent complex faults.