On-line dynamic security contingency screening using artificial neural networks

Abstract

On-line dynamic security analysis has now become realistic due to advances in computer technology and algorithms for security assessment. Details of pattern recognition based electro-mechanical stability screens which have been implemented within a dynamic security assessor are presented. Use of statistical functions of features is shown to overcome the dimensionality problem of applying pattern recognition techniques to large power systems. The low computational cost of this approach coupled with efficient operation has resulted in a significant step towards achieving full on-line dynamic security assessment.