Detection of Transformer Winding Fault with Oscillation Wave Signal Based on Time-Frequency Transform

In order to further improve the reliability of the transformer winding state detection, a method for analyzing winding state is proposed by using winding oscillation wave. The oscillation wave is obtained from the transient excitation generated by switching’s action under high voltage direct current (HVDC). Based on the field test, the oscillatory wave signal is de-noised by continuous wavelet transform and then its time-frequency characteristics are analyzed. The image recognition method and the matrix similarity degree are introduced to quantify the time-frequency signal characteristic. The results show that the method has high testing efficiency, and the matrix similarity can not only reflect the repeatability and regularity of oscillation waves in different amplitude and from different output side, but also better reflect the difference of transformer phase-to-phase windings. This method provides a reference for the research of transformer winding deformation detection.