Audio Detection and Fault Diagnosis of Ultra-High Voltage Transformers

This research work addresses the critical issue of audio detection and fault diagnosis in Ultra High Voltage (UHV) transformers, essential components of power transmission systems. Various faults such as short-circuits, insulation problems and oil temperature anomalies can have a significant impact on power converters and the entire power grid. This study proposes a novel approach using audio signals for real-time fault detection and diagnosis. The proposed methodology involves the extraction of audio signals from ultra-high voltage transformers, taking into account factors such as peak frequency, frequency offset, and total harmonic distortion (THD). A deep neural network is used for systematic analysis and interpretation of these features to enable accurate fault detection. Moreover, this study presents a thorough verification process using an experimental dataset, demonstrating superior detection accuracy compared to existing models. The dataset includes diverse samples of normal and fault scenarios, and the proposed model consistently achieves accuracy levels between 98% and 100%. The results suggest the effectiveness of the proposed audio-based methodology for enabling reliable fault detection in ultrahigh-voltage transformers.