The low porosity and permeability characteristics of tight oil reservoirs have brought challenges to monitoring oil saturation recently. Although carbon/oxygen logging is effective for oil saturation evaluation, the statistical fluctuations of the measured energy spectrum in tight reservoirs make it impossible to distinguish the different signals between oil and water. Thus, noise adjusted singular-value decomposition (NASVD) is applied to denoise the raw energy spectrum and evaluate the oil saturation quantitatively. The energy spectrum matrix, which is composed of the energy spectrum of the measurement point and its adjacent depth points, is decomposed and reconstructed to remove noninformative signals and improve the signal-to-noise ratio of the raw energy spectrum. The parameter K evaluates the smoothness of the logging curves, reflecting the influence of the number of energy spectra and singular values on NASVD. Meanwhile, the NASVD, Savitzlcy-Golay filtering, and depth averaging methods are compared for calculating the accuracy of C/O, Si/Ca, and oil saturation with the Monte Carlo method, indicating that the NASVD is better than the other two methods for eliminating the statistical fluctuations of the raw energy spectrum. A simulation example indicates that the NASVD can control the calculation errors of tight reservoir oil saturation to within 15%, which significantly improves the accuracy of the estimated oil saturation. An oil field example indicates that the oil saturation interpretation result for tight reservoirs is in good agreement with the oil saturation from open-hole log analysis, signifying that the NASVD energy spectrum denoising method can provide a quantitative estimate of oil saturation in tight oil reservoirs.
展开▼
