RBF-SVM and its application on reliability evaluation of electric power system communication network

Support vector machine (SVM) is a novel machine learning method after the artificial neural networks (ANN). The SVM with RBF is the research hot spot in assessment area at present. Because of its good learning performance, the SVM with RBF is widely used in practical application. In this paper, the RBF-SVM and its application on reliability evaluation of electric power system communication network is researched. Through experiments, the impact of learning ability and generalization ability for the error penalty parameter C and kernel function width sigma is analyzed and compared, how the parameters affect the performance of RBF-SVM is expatiated, the pictures of the changing curve that the parameters Cand sigma affect the number of support vector (SV) and wrong recognition rate are presented. AT last, through reliability evaluation with SVM under different kernel function, compare with their assessment performance, and the performance superiority of RBF-SVM is validated.