Statistical correlation technique for transient signal analysis during impulse testing of transformers

Abstract

The reliable technique for identification of winding insulation faults in a transformer is necessary for a testing engineer during impulse test. In this study, a statistical correlation technique is proposed to estimate the ‘best correlation’ between transient signals for passed (withstood the impulse voltage) or failed (not withstood) conditions of the insulations. In this study, normalised reduced impulse voltage is considered initially as a reference signal. The next successive impulse test sequences due to rated test voltage are correlated as a test signal with a reference signal using a proposed technique. It makes a ‘good correlation value along with its directions’ between the reference signal and test signal based on measured the applied impulse voltage waveshape and its winding response. The fundamentals of the correlation coefficient, curve fitting techniques, conditional variance, normalisation and time delay index are integrated effectively to predict the characteristics (degree/magnitude and direction) of correlation between the signals. 2.5-MVA 11/0.433 kV (distribution transformer), 0.3-MVA 34.5/0.415 kV (earthing transformer) and 250-MVA 500/275/33 kV (power transformer) are effectively utilised to validate proposed technique. The advantage of the proposed technique is validated with partial discharge detection in windings due to an impulse voltage application using an 11 kV single-layer winding.