Transformer Differential Protection with Neural Network Based Inrush Stabilization

Abstract

Application of artificial neural networks (ANN) for transformer differential protection stabilization against inrush conditions is presented. Three versions of the stabilization scheme are described. The best of them employs three ANNs fed with transformer terminal currents that has proven to be superior over the two other ANN schemes. The final solution combines the classification strengths of neural networks with commonly used second harmonic restraint, thus being a hybrid classification unit. To determine the most suitable ANN topology for the inrush classifier a genetic algorithm was used. The developed optimized neural inrush detection units have been tested with EMTP-ATP generated signals, proving better performance than traditionally used stabilization algorithms.