A new method for capillary pressure curve prediction based on NMR echo data using integral transform, the quantum genetic algorithm, and the artificial neural network in tight sandstone

Abstract

The capillary pressure curve (CPC) is an important tool with which to investigate the petrophysical properties and pore structure of oil and gas reservoir rocks. However, the method of obtaining the CPC directly via laboratory measurements is usually time-consuming and costly, prone to the contamination or destruction of core samples, and can only be applied to limited cores. A new method adopting integral transform, the quantum genetic algorithm, and the artificial neural network (IT-QGA-ANN) based on nuclear magnetic resonance (NMR) echo data is proposed to improve CPC prediction accuracy in tight sandstone. First, 7 parameters that characterize rock properties are directly extracted from echo data using integral transform. This process does not require to obtain the transverse relaxation time (T2) distribution by inversion of echo data, thereby avoiding the influence of inversion uncertainty. These characteristic parameters are then used as the input of the ANN, and the QGA method is utilized to optimize the weights and thresholds of ANN. Numerical simulation results prove the validity of integral transform to extract characteristic parameters. A total of 18 tight sandstones are used for mercury injection and NMR tests. The results indicate that, compared with the traditional nonlinear regression and ANN methods, the proposed IT-QGA-ANN method can more accurately predict the CPC. The proposed method achieves the direct and accurate transformation from echo data to the CPC, and provides important guidance for solving complex nonlinear problems in oil and gas reservoir evaluation based on NMR.