Logging Evaluation Methods for Irreducible Water Saturation in Tight High-Water-Cut Gas Reservoirs: A Case Study of the Dongsheng Gas Field in the Ordos Basin

Abstract

As a crucial parameter for the evaluation and development decision-making of high-water-cut tight sandstone gas reservoirs, irreducible water saturation (S_wi) plays a decisive role in reservoir productivity assessment and development plan design. However, traditional determination methods such as the existing core mercury injection experiment and nuclear magnetic resonance (NMR) logging inversion suffer from high costs, long cycles, and low coverage, which severely restrict gas reservoir development decisions. Taking the high-water-cut tight sandstone gas reservoir in the Dongsheng Gas Field of the Ordos Basin as an example, this paper proposes a new low-cost and continuous prediction method based on conventional logging data and the Random Forest (RF) machine learning algorithm. Through logging sensitivity analysis, the pore structure index (PSI) and density-neutron difference were introduced, and combined with the Random Forest algorithm, 5 sensitive parameters (GR, PERM, PSI, AC, DEN) were selected. Using the S_wi calculated from the NMR logging T₂ spectrum of 1,135 sets in 18 wells as the true value, a prediction model was constructed. For the test set of the model, the coefficient of determination (R²) is greater than 0.9, and the mean absolute error (MAE) is less than 0.05, indicating reliable generalization ability. The prediction results of this method in Well A of the Dongsheng Gas Field are highly consistent with the NMR resultsThis method can realize continuous evaluation of the entire well section, significantly reduce the reliance on special logging and associated costs, and provide technical support for low-cost and high-precision reservoir evaluation of tight gas reservoirs in the gas field.