Accurate determination of water saturation in tight sandstone gas reservoirs based on optimized Gaussian process regression

Abstract

Tight sandstone gas is an important field for the future development of the oil and gas industry. In the tight gas reservoir of Penglaizhen Formation in Shifang gas field, the water saturation (Sw) of sealed coring cores and the NMR irreducible Sw of conventional cores are measured by using the reservoir physical property tester and NMR analyzer. Combined with the scanning electron microscope results, the reservoir has the characteristics of complex pore structure and high-water saturation. The calculation of Sw by Archie formula, Density log, and φD-φN has a large error. Therefore, how to accurately calculate Sw is an outstanding problem that needs to be resolved. In this paper, we selected AC, EDN, XI, and φe as the parameters for modeling. We have proposed the optimized Gaussian process regression (GPR) method to calculate the Sw of tight sandstone reservoirs. The water saturation curve calculated by the optimized GPR model is in good agreement with the core saturation value. In the error results of well data in the study area, it was found that the relative error and root-mean-square error of the optimized GPR model were the lowest, and it had better calculation accuracy. In addition, in the error results of the nuclear magnetic core samples, it was also found that the optimized GPR model's MMRE was 6.65% and the RMSE was 4.95. In tight sandstone reservoirs with complex pore structure and high-water saturation, the optimized GPR model is helpful to reduce the influence of reservoir shale characteristics, low resistivity characteristics and hidden information of parameter data during the calculation of Sw. It can calculate Sw more efficiently and reflect the characteristics of irreducible Sw more accurately. The optimized GPR model developed for water saturation calculation has broad application prospects.