v-SVM for transient stability assessment in power systems

In this paper, support vector machines (SVMs) are studied in the application of transient stability assessment in power systems. SVMs have the following advantages: automatic determination of the number of hidden neurons, fast convergence rate, good generalization capability, etc. SVMs use the principle of structural risk minimization, and thus reduce the dependency of experience unlike neural networks and have better generalization and classification precision. Furthermore, SVMs are solved by the 2nd order convex programming and the final solution of SVMs is sole and optimal. The performance of SVMs depends on the type of kernel functions and the parameters of kernel functions, which are determined by experience or experiments. So the effects of kernel functions and the parameters of kernel functions are analyzed by experiments in the paper. In addition, Experiments corroborate the superiority of v-SVM applied in TSA in power systems by comparing with BP and RBE.