Reconstruction of 3D multi-mineral shale digital rock from a 2D image based on multi-point statistics

Abstract

Introduction: Shale oil and gas reservoirs contain a variety of inorganic and organic pores that differ significantly from conventional reservoirs, making traditional experiments ineffective. Instead, the pore-scale imaging and modeling method, regarded as a novel and practical approach, is proposed to characterize shale microstructure and petrophysical properties. Therefore, it is of great significance to accurately reconstruct the three-dimensional (3D) microstructure of the porous medium, that is, the digital rock. However, microstructural images of shale at high-resolution, obtained through scanning electron microscopy (SEM) are constrained in the two-dimensional (2D) scale.Method: In this work, a novel iterative algorithm to reconstruct 3D multi-phase shale digital rock from a 2D image using multi-point statistics has been proposed. A multi-grid data template was used to capture the conditional probabilities and data events. The novelty of this work stems from an accurate representation of different types of pores and the mineral characteristics of shale rock from 2D images.Result: A series of simulations were conducted to reconstruct 2D shale digital rock from a 2D segmented training image, 3D shale digital rock from a 2D segmented training image, a 2D gray training image to reconstruct 2D shale digital rock, and a 2D gray training image to reconstruct 3D shale digital rock.Discussion: To corroborate the accuracy of the reconstructed digital rock and evaluate the reliability of the proposed algorithm, we compared the construction image with the training image with the two-point correlation function, geometry, morphological topology structure, and flow characteristics. The reconstruction accuracy indicates that the proposed algorithm can replicate the higher-order statistical information of the training image.