Microscopic flow simulation of shale multi-scale digital core based on image classification

Abstract

The shale reservoir is characterized by complex pore structures spanning nano- to micron-scale, with fluid flow behavior varying significantly across these scales. To address the challenges of simulating multi-scale flow in digital cores, this study develops a novel micro-flow simulation method based on automatic microstructure classification using the K-means clustering algorithm. By coupling pore-scale data from SEM-Maps and CT imaging, multi-scale digital cores were constructed and fluid flow simulation performed using the Darcy-Brinkman-Stokes approach. Results demonstrate that neglecting the multi-scale flow effect will underestimate the apparent permeability, particularly for cores with poor connectivity of micro-scale pores. Multi-scale simulations reveal pronounced permeability anisotropy, with horizontal and vertical permeability differing by two orders of magnitude due to the distribution of sub-resolution pores and micrometer-scale pores. This study highlights the critical role of sub-resolution pores and fracture geometry in accurately predicting flow behavior in shale reservoirs, offering insights for optimizing reservoir simulation and management strategies.