The invention discloses a support vector machine based transient power angle stability fault screening system and method. Characteristic quantity can be selected based on transient power angle stability quantitative evaluation results of samples in a historical sample set, so that the problems of the complex calculating of a feature selection algorithm and characteristic quantity mis-selection andomission; a support vector machine can be adopted to perform supervised learning and training on different stable modes after the stable modes are classified, so that the accuracy of a prediction model can be enhanced; a regression prediction model which takes a transient power angle stability margin as a variable can be adopted to sample sets that only having stable samples; the applicability ofa transient power angle stability prediction model can be judged based on the characteristic quantity of a current operation mode, so that the mis-judgement problem of a machine learning algorithm can be avoided; and whether anticipated faults and corresponding power flow modes are brought into the sample sets based on the detailed transient power angle stability quantitative evaluation results and predicted value difference, and the number of instability samples in the sample sets can be properly increased, so that the rationality of the samples in the sample sets can be guaranteed.