Wavelet ANN Based Stator Ground FaultProtection Scheme for Turbo Generators

Abstract

This article proposes a protection scheme for turbo generators to detect stator ground faults in particular closer to neutral. The scheme considers total generated voltage of the machine and makes use of wavelet multi-resolution analysis. A fault index is defined in terms of peak and valley values of the highest-level wavelet decomposition, which is compared with a threshold value to discriminate from other transients, like sudden unbalance, in load. To identify the faulty coil, an ANN employing back propagation algorithm is used with the lowest level wavelet coefficients as inputs. The results of digital simulation are presented for different fault impedances and fault locations.