Pattern recognition of PD in large turbine generators with a neural network system

In this paper, a neural network system used for pattern recognition of partial discharge (PD) is described. The neural network is a three-layer artificial neural system with feed forward connections, and its learning method is back propagation algorithm incorporating with an external teacher signal. Digital PD pulse signal can be obtained by a PD pulse digitized record system. Combination of the discharge magnitude, the phase angle of applied voltage at which PD occurs, and the numbers of pulse counts are taken as the input of the neural network system. After learning typical input patterns, the neural network may discriminate unknown patterns successfully. Some new results are given, and practical application of neural network for pattern recognition of PD in large turbine generators is also discussed