An application about kernel principal component analysis(KPCA) is proposed for feature abstract in electric systems transient stability assessment(TSA). The kernel function’s parameter is optimized by using the algorithm of particle swarm optimization(PSO). Emluator of TSA model based on support vector machine(SVM) with power system EPRI36 is also given. The result shows that the method not only has good prediction accuracy, but also reduces the input dimension greatly.