NEURAL NETWORKS AS A TOOL FOR RECOGNITION OF PARTIAL DISCHARGES

In this paper the application of three different neural networks (NN) to recognize partial discharge (PD) sources is studied. Results of PD measurements on simple two-electrode models as well as on models of ar tificial defects in industrial objects are presented. The PD were meas ured using conventional discharge detection and PD patterns were proce ssed by previously developed statistical tools. Satisfactory results i n the past have shown that using mathematical descriptors, the propert ies of the phase-position distributions can be analyzed. Therefore the se descriptors were used as input patterns for back-propagation networ k, Kohonen self-organizing map and learning vector quantization networ k. All three NN, as used in this work, recognize fairly well the PD pa tterns of those insulation defects for which they were trained. On the other hand, the NN might misclassify those PD patterns for which they were not trained. The classifications of PD patterns by NN can be inf luenced also by the structure of the particular NN, the value of conve rgence criterion, and the number of learning cycles.