Spectral denoising approach using enhanced wavelet packet-optimized EEMD algorithm

Abstract

The quantification of hydrocarbon gases within formation fluids is paramount for the detection and evaluation of oil and gas reservoirs through well-logging geophysical techniques. Nonetheless, the spectral data are frequently fraught with complexity and noise contamination, attributable to the heterogeneous nature and extensive concentration spectrum of alkane gases, coupled with environmental interferences. This pervasive noise can markedly distort the absorption spectra, consequently compromising the accuracy of alkane gas quantification. The imperative challenge lies in the precise denoising of acquired infrared spectra while meticulously preserving the signal-to-noise ratio and spectral resolution. This present an innovative denoising methodology for infrared spectral analysis, leveraging an advanced wavelet packet coupled with an optimized Ensemble Empirical Mode Decomposition algorithm. This approach initially employs a bivariate correlation analysis to discern and isolate the noisy components within the Intrinsic Mode Functions. Subsequently, it harnesses the sample entropy of the noise signals in tandem with the Grey Wolf Optimization algorithm to ascertain the most efficacious threshold set for each IMF component. The methodology culminates in the formulation of threshold parameters through the synthesis of the sample entropy of wavelet packet coefficients with the correlation coefficients of the noise, thereby tailoring the threshold function to the distinct noise attributes of each wavelet coefficient. Empirical findings demonstrate that our approach outperforms conventional denoising techniques, achieving superior signal-to-noise ratios and resolution, with an average perturbation to characteristic peaks of less than 0.3 %.