Advanced Fault Detection in Power Transformer using SFRA, Statistical Indicator and ANFIS

Abstract Power transformers are vital components of electrical power systems, and their failure can lead to severe operational and economic consequences. Mechanical stresses, short circuits, and insulation aging contribute to transformer faults by inducing electromagnetic forces that result in axial and radial deformations. If left undetected, these deformations can weaken insulation integrity, reduce the transformer’s short-circuit endurance, and cause complete failure. To ensure reliability, an effective fault diagnosis method is required. sweep frequency response analysis (SFRA) is a powerful tool used to detect mechanical and electrical anomalies by comparing measured frequency responses with a healthy reference. This study utilizes SFRA in conjunction with mathematical tools such as lumped parameter models and an adaptive neuro-fuzzy inference system (ANFIS) to improve fault detection accuracy. ANFIS enables automated classification of transformer faults based on statistical indicators such as skewness, kurtosis, cross-correlation, and absolute difference sum (DABS). MATLAB-based simulations validate the effectiveness of the proposed approach by analyzing SFRA responses under various fault conditions. The results demonstrate that integrating machine learning with SFRA significantly enhances the precision of fault classification, making transformer health assessment more efficient. This research provides a step towards automated fault monitoring, reducing the reliance on manual expert interpretation and enhancing transformer lifespan.