Detection of Short-Circuited Turns in Transformer Vibration Signals using MUSIC-Empirical Wavelet Transform and Fractal Dimension

Abstract

In electrical systems, the transformers have an important role; therefore, their correct operation is fundamental. However, they can present malfunctions due to different types of faults. The short-circuited turns (SCTs) are one of the main causes of transformer damages, which can scale into most serious faults. In this regard, there are several works in literature to diagnose this type of fault through the analysis of vibration signals. In this work, the vibration signals are analyzed to diagnose the transformer condition; however, the analysis of a vibration signal with no relevant information and a high level of noise is not a simple task. In this work, the problem is addressed using the MUSIC-empirical wavelet transform (MEWT). This technique allows extracting the relevant features from vibration signals. Additionally, three fractal dimension algorithms (FDAs): Katz, Sevcik, and Petrosian, are proposed and investigated as potential fault indicators. To test the proposed methodology, a modified transformer to emulate different SCTs fault conditions is used. These conditions are healthy, 5, 10, and 15 SCTs. The results show that MEWT removes properly noise and irrelevant information, allowing the identification of the fault condition using the FDAs as indicators.