Diagnostic of transformer winding deformation fault types using continuous wavelet transform of pulse response

Abstract

Impulse Frequency Response Analysis (IFRA) provides an alternative technique for detection of mechanical type faults in power transformers among researchers. However, the conventional frequency response curve obtained from the time domain transient response only contains the frequency domain information of signal, and the time domain information is neglected. Furthermore, the traditional approach of using the fast Fourier transform (FFT) in processing the non-stationary signal could result in lack of useful information. This paper proposes a novel winding deformation fault diagnostic technique which is modified from IFRA, the detected transient signals of pulse response are processed based on the continuous wavelet transform (CWT) algorithm, and the wavelet time-frequency diagram are plotted and analyzed. The simulation of a simple transformer high frequency model and experiment analysis show that the various types of winding deformation can be distinguished by the proposed technique. The data processing result by the wavelet transform conforms with the variation pattern of frequency response to different winding fault types. Besides, the verification result of experiment indicates that the wavelet time-frequency diagram obtained by the proposed technique shows better performance than the frequency response curves obtained by the conventional FFT on processing the transient signals.