ACOUSTIC TAP-CHANGER MONITORING USING WAVELET ANALYSES

Summary

The acoustic signature as a diagnostic tool is a proven tool in many industrial applications. In the field of electrical equipment the On-Load Tap-Changer (OLTC) is the only mechanical part of a high voltage transformer which is responsible for voltage control that means the switching of regulating windings. The fundamental principle of an OLTC operation is the mechanical switching of contacts in vacuum or oil. In the process of OLTC operation the acoustic noise and the vibration reflect mechanical or electrical events during the switching process. The information concerning the OLTC operation can be extracted from the vibro-acoustic measurements. This contribution introduces a novel method for analysis, visualization and interpretation of the vibro-acoustic measurements on OLTCs. The measurements are provided by means of broadband acceleration sensors mounted on the transformer tank or on the OLTC head cover. The basis of the method is the time frequency analysis of recorded acceleration signals by means of continuous wavelet transformation (CWT). A two-dimensional representation of the signal in time and frequency displays all registered acoustic events with the dynamic range of up to 70 dB. Transformer vibrations, motor noise, the noise of tap selector and diverter switch as well as their timing can be identified and separated clearly in time-frequency diagrams. So an anomaly in the system can be recognized by the appearance of abnormal or by the absence of expected acoustic events as well as deviations in timing. Another major part of the method is the detailed verification of the noise signature of the diverter switch. Because of special design features of each OLTC type its signature comprises time windows with different level of deviations. The allowed variation level is predefined by the known tolerances in run-time of the relevant switching sequences of the OLTC as well as the available statistical evaluation of the data. In this way an actual signature of an OLTC under test can be compared automatically with the corresponding reference signature by means of a pattern recognition algorithm.