Partial Discharge Denoising Technology for GIS Equipment Based on CEEMDAN

Addressing the problems with noise interference in partial discharge (PD) signals of gas - insulated switchgear (GIS) and the shortcomings of traditional denoising methods, such as mode mixing and residual noise, the article introduces a dual - threshold collaborative denoising approach that utilizes Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN). To enhance the signal decomposition process, Gaussian white noise is incorporated, which significantly alleviates the modal aliasing issue. Subsequently, a two - stage filtering framework is established. In the initial stage, components dominated by noise are eliminated using a correlation coefficient threshold. Then, in the subsequent stage, multiple intrinsic mode functions (IMFs) are chosen based on a variance contribution threshold. In the single and double exponential attenuation and oscillatory PD model experiments, the signal quality is significantly improved after noise reduction. The technique offers high - quality signals for GIS partial discharge detection, aiding in the enhancement of diagnostic accuracy and dependability for faults.