There are thousands of transient stability faults in large scale power system. Transient stability analysis cannot be finished in given time required by online assessment. To meet the requirement of calculation time of on-line transient security and stabiliby assessment, an transient power angle security and stability contingency screening method base on clustering analysis of power grid operation history data.Extracting key feature quantities based on power grid operation mode and transient power angle stability assessment results in historical data.Determine the number of clusters and the initial clustering center by vector quantization method that takes into account the stable mode.Optimization of clustering center points using K-Medoids algorithm. The transient power angle margin is quickly estimated for the faults of the transient power angle stability after classification. Finally, a severe contingency set consisting of the faults of the transient power angle instability and the estimated margin below the threshold is obtained. According to the clustering analysis of actual history data, the validity and practicability fo the proposed method can be verified.