Aplicação de método de reconstrução de sinais sintéticos de descargas parciais baseado em Decomposição Variacional de Modos com a presença de ruído de quantização

Abstract

In this work, a study was conducted on four noise removal techniques for synthetic partial discharge (PD) signals: Variational Mode Decomposition (VMD), Least Mean Squares Adaptive Filter (LMS), and Wavelet Transform with hard and soft coefficient thresholding (Hard and Soft Thresholding Wavelet - TWT). The filtered signals showed good results in terms of SNR, RMSE, and CC when combining the applied techniques. The performance of these filtering methods was evaluated on signals with different quantization levels that simulate high-frequency PD signals acquired with a low vertical resolution device. Some waveform distortions were observed, but the similarity with the original reference synthetic signal showed a correlation of ≈ 0.70. From this study, we conclude that the application of these denoising techniques demonstrates good performance. Thus, as many real signals obtained may suffer from quantization errors, the use of the presented techniques can be a reliable alternative for studies.