Multivariate Graphical Analysis of Transformer FRA Signatures: An Integrated Amplitude–Phase Approach for Automated Winding Fault Diagnosis

Power transformers serve as critical components in electrical transmission networks, continuously exposed to severe electromechanical stresses that lead to winding faults including Axial Displacement (AD), Radial Deformation (RD), and inter-disk Short Circuit (SC). Early detection of these faults is crucial for preventing catastrophic failures and ensuring grid reliability. While Frequency Response Analysis (FRA) has gained global acceptance as a sensitive diagnostic tool, its primary challenge lies in the subjective interpretation of complex frequency responses. This paper presents an automated, quantitative framework that leverages two parallel visualization methodologies: Principal Component Analysis (PCA) for optimal feature compression and noise reduction within hybrid amplitude-phase feature space, and Fuzzy Clustering Analysis (FCA) for managing uncertainties in transitional and combined fault conditions. The proposed approach implements a structured two-stage identification process: initially distinguishing healthy from faulty states, followed by specific fault classification. Experimental validation on a 1 MVA transformer under various fault severity levels demonstrates outstanding performance, with PCA and FCA achieving accuracies of 98.9% and 99.0%, respectively. The dual visualization system independently presents results from both methods, significantly reducing dependency on specialized expertise and enabling clearer, more objective interpretation. This research represents a significant step toward standardizing FRA interpretation and offers a practical, reliable solution for transformer condition assessment in industrial applications.