Support vector machines for on-line security analysis of power systems

The pattern recognition approach for security analysis (SA) of power systems has been presented as a promising tool for on-line applications. This paper applies a learning-based nonlinear classifier, which is a support vector machine (SVM) for SA. Three single SVM are trained to classify the state of the system: secure, alert and emergency. The final classification is obtained combining the output of each classifier with a Bayesian rule. The effectiveness of the proposed approach has been demonstrated on two IEEE test systems.