Winding Condition Assessment of Power Transformers Based on Vibration Correlation

Abstract

In this paper, a winding condition assessment model using vibration signals is presented, which can be used to diagnose power transformers online. The basic principle of this model is based on the correlation analysis of winding vibrations. In the model, the fundamental frequency vibration analysis is used to separate the winding vibration from the mixed signal. Then, a health parameter is proposed via principal component analysis. Another parameter is also proposed to detect the fault locations for suspected faulty transformers. In laboratory tests, the model is validated on a specifically designed 110-kV transformer. During the tests, man-made winding deformations are simulated to compare the vibrations under different conditions. The model has also been tested on several in-service power transformers. The preliminary study shows that the proposed model is feasible to assess the power transformer winding condition.