Denoising of ultrasonic echo signals of transformer windings based on Empirical Mode Decomposition-Soft Threshold Denoising-Normalized Minimum Mean Square
Yuebo Jia, Qingze Cao, Zhigang Wang, Hongbo Guo · 2025
As a critical component in power systems, the operating condition of transformer windings directly affects their performance and operational reliability. Ultrasonic echo signal detection, due to its high sensitivity and non-destructive advantages, is widely used for winding fault diagnosis. However, in practical applications, ultrasonic echo signals are often affected by environmental noise, electromagnetic interference, and signal reflections, significantly reducing the accuracy and stability of detection. To address this issue, this paper proposes a combined denoising method that integrates Empirical Mode Decomposition (EEMD), Soft Threshold Denoising (ST), and Normalized Minimum Mean Square (NLMS) filtering. This method first uses EEMD to decompose the original signal into modal components, extracting high-frequency noise components and applying soft threshold functions for denoising. Subsequently, the processed components are reconstructed with low-frequency components to form an initial denoised signal, and NLMS adaptive filters are introduced to further suppress residual noise. Simulation results show that the proposed EEMD-ST-NLMS method outperforms traditional methods in terms of signal-to-noise ratio (SNR) and mean square error (MSE), significantly improving the quality of ultrasonic signals and providing a reliable data foundation for high-precision detection of transformer winding faults.