Super-Resolution Reconstruction of Shale Digital Core Images Guided by Fluid Flow Mechanisms

Summary

Digital core techniques are crucial for analyzing reservoir rock parameters in seepage simulation, but it is usually expensive to obtain high-resolution (HR) core images. Instead, image super resolution (SR) reconstruction techniques, by generating HR images from low-resolution (LR) ones, offer a promising solution for restoring details in rock images. However, they have certain limitations in preserving structural accuracy and permeability for porous media. In this study, we propose a new method that incorporates fluid flow information in the SR technique to correct pore structures in generated rock images. By adjusting pore geometry using the velocity field, our approach aligns flow parameters and pore characteristics in the SR images more closely with those in the true HR images. We validate the method using paired scanning electron microscopy (SEM) images from five distinct regions, as well as 3D digital rock images of matrix shale, fractured shale, and propped fractures. Comparative analyses of porosity, velocity fields, and equivalent permeability demonstrate that our method effectively reproduces fine-scale fluid flow in porous media. Moreover, the method exhibits superior applicability in seepage simulations due to its ability to resolve finer pore structures. In this controlled study, representative HR images are used for supervised residual learning and flow-based calibration of the correction parameter. By bridging the gap between image reconstruction and flow simulation, our study offers a more accurate and efficient approach for analyzing reservoir rock parameters, ultimately improving the prediction of oil and gas production.