A New Method for Predicting Capillary Pressure Curves Based on NMR Logging in Tight Sandstone Reservoirs

Abstract

The accurate prediction of capillary pressure curves is of great significance for the evaluation of pore structure in tight sandstone reservoirs. In this paper, a hybrid model is proposed to predict the mercury injection capillary pressure (MICP) curve using multiple characteristic parameters of the nuclear magnetic resonance (NMR) transverse relaxation (T2) distribution; models based on the capillary pressure point and on the non-wetting phase saturation point are combined. Because of the high level of multicollinearity among the characteristic parameters of the NMR T2 distribution, the partial least squares method is used to solve the model coefficients. The leave-one-out cross-validation (LOOCV) method is used to determine the optimal combination of parameters of the hybrid model, i.e., the number of latent variables of the model based on the capillary pressure point, the number of latent variables of the model based on the non-wetting phase saturation point, and the location of the splicing point. The result of model self-testing, the result of the LOOCV, and the analysis of the generalization ability all indicate that the hybrid model has high accuracy and strong stability for MICP curves prediction. The prediction results in tight sandstone reservoirs further verify the effectiveness of this method.