High-voltage equipment in electrical systems is subject to electrical insulation degradation, which promotes the evolution of partial discharge activity (PDs), a key factor in operational failures of these assets, resulting in significant losses. Therefore, the analysis of PD signals for accurate representation of operational conditions is essential, as it aids in making assertive decisions in predictive maintenance of equipment, as well as understanding their impacts. The procedure for acquiring PDs involves measurements using invasive or
non-invasive devices, which show the occurrence of this phenomenon under field or laboratory operating conditions. However, this acquisition is subject to interference, leading to PD signals embedded in noise, which may arise from the electromagnetic nature of the equipment used or from external sources. This highlights the need to implement noise reduction and signal reconstruction techniques that ensure good representation of PDs, allowing for precise analysis by minimizing the loss of signal characteristics for further studies. Thus, in this work, a study was conducted on four noise reduction techniques for PD signals: Variational Mode Decomposition (VMD), Adaptive Filtering with Least Mean Squares algorithm (LMS), Wavelet Transform with Hard Thresholding (HTWT), and Wavelet Transform with Soft Thresholding (STWT), for reconstructing real PD signals obtained from measurements with High-Frequency Current Transformer (HFCT). The results from the
signal filtering process were evaluated using metrics such as Root Mean Square Error (RMSE), Signal-to-Noise Ratio (SNR), and Correlation Coefficient (CC), demonstrating the importance of filtering methodology with the incorporation of VMD and the variation of parameters and filtering sequences established.
|
