A method based on compound artificial neural network（ANN）to predict critical clearing time（CCT）margin and to do transient stability assessment is derived in this paper.It consists of probabilistic neural network（PNN）and radial basic function（RBF）network to integrate their advantages.PNN network is first used to classify sampling data and it can consider the overlapping influence in adjacent samples in the classification.Then RBF network is used to forecast the CCT margin based on the classification results.A self-checking procedure and an amending procedure are added to improve the prediction accuracy.New England 39 bus system is used to do the validation.Numerical studies reveal that the proposed method is better than the traditional BP and RBF methods not only on predictive accuracy but also on calculation speed.So the given method is helpful to power system transient stability assessment.