Unsupervised neural net classification of power system static security states

Abstract

This paper studies the application of Kohonen's self-organizing feature map to power system static security assessment. The Kohonen classifier maps vectors of an N-dimensional space to a two-dimensional neural net in a nonlinear way, preserving the topological order of the vectors which, in general, is not known a priori. The classification of line-loading patterns by the Kohonen network is demonstrated for two different test systems. The generalization capability of the Kohonen network permits the correct classification of system states which have not been encountered during the training phase. This feature is extremely important for power system operation where it is unrealistic to expect that all possible cases will be encountered during off-line simulation.