Neural networks for contingency evaluation and monitoring in power systems

Abstract

In this paper an analysis of the applicability of different neu- ral paradigms to contingency analysis in power systems is presented. On one hand, unsupervised Self-Organizing Maps by Kohonen have been implemented for visualization and graphic monitoring of contingency severity. On the other hand, supervised feed-forward neural paradigms such as Multilayer Perceptron and Radial Basis Function, are implemented for severity numerical evaluation and contingency ranking. Experiments have been performed with successfully result in the case of Kohonen and Multilayer Perceptron paradigms. © Springer-Verlag Berlin Heidelberg 2001.