Frequency Response Analysis (FRA) has become a reliable tool to detect mechanical deformation within power transformers. Many researchers have investigated the impact of various mechanical winding and core deformations on transformer FRA signature using either simulation analysis or practical testing to establish a standard code for FRA signature interpretation. None of them however, have given attention to the impact of power transformer insulating oil degradation on the transformer FRA signature. This paper investigates the effect of insulating mineral oil degradation on power transformer FRA signature. In this regard, the physical geometrical dimension of a single-phase transformer filled with insulating mineral oil is simulated using three dimensional finite element analysis to emulate the real transformer operation. Transformer FRA signature is measured and analysed with various health conditions of the insulating oil. Results show that, insulating oil degradation has a significant impact on the transformer FRA signature.
Digital image processing (DIP) technique has been growing rapidly as an essential tool to interpret various image features for many applications of science and engineering. Condition monitoring and diagnosis are considered the main areas that relay on DIP. Frequency response analysis (FRA) technique has become a popular and reliable diagnostic tool in detecting various winding deformations within power transformers. However, interpretation of FRA signatures still requires high expertise because of its reliance on graphical analysis. This paper presents a new technique for the interpretation of transformers FRA signatures. The proposed technique relies on incorporating both magnitude and angle of the FRA signature in one polar plot, which is manipulated to extract some unique features using DIP techniques. The proposed technique can assess in identifying and quantifying various winding deformation within power transformers. The proposed technique is easy to implement in any frequency response analyser.