Abstract: It has been found in the recent studies that the CO₂ geological storage is one of the efficient methods to achieve a large amount of CO₂ reduction, while the understanding of the transport law of CO₂ in the rock pore space is fundamental to the flexibility of a safe and long-term CO₂ geological storage scheme. The pore network model based on the core micro-structure reconstruction can not only reflect the real core pore and throat distribution law, but also reflect the distribution of pore space and its development characteristics, which is of importance in the field of multi-phase flow research in porous media. The Berea sandstone, which is more common in the reservoir, was selected to obtain two-dimensional CT images based on the micro-focus X-ray computed tomography (micro-CT) technology, and the digital images of core samples were reconstructed in ImageJ software through a series of image processing processes such as size cropping, noise reduction filtering and threshold segmentation, and binarized in the Matlab software via extraction functions. Then, via the maximum sphere algorithm, the pore-throat topological equivalent network in the experimental core samples was extracted through three main steps, including searching the maximum sphere, establishing the maximum sphere connectivity, identification of pores and throats and parameter calculation, and visualized by using Amira software to equate the pores and throats to a spherical rod models. By comparing the structural parameters of core porosity, coordination number, pore and throat geometry and shape factor obtained via different image processing and pore network extraction methods, a digital core reconstruction and pore network model extraction method that can better reflect the spatial characteristics of real core pores was determined, such as model 7: the mean value is set to 1.5 and median value is set to 2 during filtering. The Otsu algorithm was selected for threshold segmentation and the minimum aperture was set to 1 when the digital core was extracted by the maximum sphere method. The pore structure parameters of Berea core were measured as follows: the coordination number was distributed within 30 and the peak value was around 5. The pore radius was distributed within 80 μm and concentrated around 20 μm. The throat radius was distributed within 60 μm and concentrated around 10 μm. The pore shape factor and throat shape factor were distributed within 0.07, the peak of pore shape factor was around 0.03, and the peak of throat shape factor was around 0.035. Finally, based on the pore network two-phase seepage simulation program developed by Imperial College of Technology, the seepage simulation of CO₂-replaced brine in porous media was carried out under the reservoir conditions of 50 ℃ and 12.4 MPa by characterizing brine with 0.103 mol/kg NaCl solution. Then, the capillary pressure curve and relative permeability curve during the drainage and imbibition cycle were analyzed in detail. The influence of the pore network extraction method on the multi-phase flow process in porous media was elucidated, and the reliability of the pore network modeling method and the accuracy of the obtained modeling parameters were demonstrated again.