| Popis: |
This study evaluates wavelet-based denoising techniques for partial discharge signals, specifically Damped Exponential Pulse (DEP) and Damped Oscillatory Pulse (DOP) types. We analyzed 63 wavelets across multiple families under four noise profiles: AM radio noise, white Gaussian noise, impulsive noise, and random noise. Performance metrics included Signal-to-Noise Ratio (SNR) improvement, energy preservation, and correlation coefficient. Our findings show significant variations in wavelet performance based on noise conditions. For AM radio noise, wavelets from the rbior and bior families achieved SNR improvements of up to 25.75 dB with excellent energy preservation. Under AWGN, several wavelets, especially for DOP signals, also reached SNR improvements of 25.75 dB. For impulsive noise, wavelets from the fk14 and dmey families performed well, particularly for DEP signals. In random noise conditions, simpler wavelets from the db family were effective for certain signal types. These results suggest significant potential for developing more effective real-time PD monitoring systems, enhancing detection sensitivity and classification accuracy, and improving maintenance strategies for high-voltage electrical equipment. |
