1. NARI Group Corporation(State Grid Electric Power Research Institute), Nanjing 211106, China;2. National Electric Power Dispatching and Control Center, State Grid Corporation of China, Beijing 100031, China

To meet the timeliness requirements of online transient stability assessment, a fault screening method based on support vector machine(SVM)and historical big data is proposed. Combined with the transient stability quantitative evaluation method of extended equal-area criterion(EEAC)and based on the system power angle stability mode and the generator participation factor, the feature variables are identified and the historical data are clustered. The instable samples are predicted by classification and regression. The predicted stability margin, classification stability prediction results and reliability are obtained by the applicability discrimination and model matching. Interactive parallel computing is used to screen online transient stability faults, which could avoid the inherent misjudgment of transient stability assessment by using SVM method. The effectiveness of the method is verified by an actual case.