Voiceprint Recognition-based Power Transformer Fault Detection for Enhancing Safety Maintenance of Electrical Systems

Power transformers play a crucial role in maintaining the stability and security of power systems. Traditional fault detection methods suffer from subjectivity and a lack of sensitivity to subtle and ultrasonic signals, limiting their accuracy and efficiency. To address these challenges, this paper proposes a voiceprint recognition-based method for transformer fault detection. By analyzing acoustic signals and leveraging deep learning, specifically an x-vector model combined with a Time Delay Neural Network (TDNN), the proposed method effectively distinguishes between normal and abnormal transformer operating conditions. Experimental results demonstrated that the x-vector-based approach outperformed existing methods such as i-vector, ResNet+self-attention, and d-vector models, achieving superior fault detection accuracy. This method provides a more robust and efficient solution for real-time transformer monitoring, contributing to the enhanced safety and reliability of power supply systems.