Quality Mother Wavelet and Decomposition Level Selection Based on Noise Source Effect for Real-Time Partial Discharge in High Voltage Solid Insulation

Abstract

Partial discharge (PD) signals are non-stationary and intrinsically dynamic which results in them being very vulnerable to background noise and extraneous interference when measured. This degradation has serious impacts on the accuracy of PD detection and insulation condition determination in high-voltage systems. The paper has conducted an in-depth investigation of the denoising of PD signals obtained in real-time experimental conditions using the Discrete Wavelet Transform (DWT). This study utilizes experimentally measured PD signals obtained from laboratory tests on XLPE insulation specimens with artificial void defects. To evaluate the denoising performance under controlled conditions, additive white Gaussian noise (AWGN) and discrete sinusoidal interference (DSI) were artificially added to the measured signals. The denoising framework is based on DWT and is applied using several sets of mother wavelet and levels of decomposition. Signal-to-noise ratio (SNR), signal-to-noise ratio error (SNRE) and mean squared error (MSE) are used to measure the performance of the denoising. The findings depict the optimal use of wavelet-based denoising to reduce noise without altering the important characteristics of the PD signals. The symlet wavelet (sym5) and a decomposition level of 10 provide a good trade-off between signal fidelity and noise reduction for the signal at different noise levels. The results emphasize the significance of proper wavelet setup to attain sound PD signal augmentation and provide practical insights into the optimization of condition enhancements in high voltage insulation designs.