Support Vector Machine based approach for accurate contingency ranking in power system

Abstract

This paper presents an effective supervised learning approach for static security assessment. The approach proposed in this paper employs Least Square Support Vector Machine (LS-SVM) to rank the contingencies and predict the severity level for a standard IEEE −39 Bus power system. SVM works in two stage, in stage 1st estimation of a standard index line MVA Performance Index PIMVA is carried out under different operating scenarios and in stage II (based on the values of PIMVA) contingency ranking is carried out. The test results are compared with some recent approaches reported in literature. The overall comparison of test results is based on the, regression performance and accuracy levels through confusion matrix. Results obtained from the simulation studies advocate the suitability of the approach for online applications. The approach can be a beneficial tool to fast and accurate security assessment and contingency analysis at energy management centre.