Abstract: Due to the depth of deep and ultra-deep reservoirs and the ground stress reaching 200 MPa, the pore microstructure of reservoir rock will be significantly changed. However, the existing digital rock reconstruction method is based on the scanning image reconstruction at normal temperature and pressure, and cannot reflect the pore structure under high pressure conditions. Therefore, a digital rock reconstruction method based on discrete element method (DEM) is proposed to consider the effect of high stress. The first step is to convert the digital rock image at normal temperature and pressure into a discrete element model, establish the contour database by using the spherical harmonic analysis method, and establish the Clump template library in PFC^3D. According to the porosity and particle size distribution, Clump in the template library was used to build a discrete element model, and the accuracy of the model was evaluated by calculating the two-point correlation and linear path correlation function curves. The second step is to calibrate the micromechanical parameters between particles according to the results of conventional triaxial compression test, and set different true triaxial stresses to obtain the digital rock under different stresses. The third step is to analyze the pore geometry topology of digital rock under different stresses and calculate the porosity permeability. Taking Bentheim sandstone as an example, a digital core under different stresses is constructed. High stress leads to the reduction of pore and throat radius, the growth of throat, the poor connectivity, and the decrease of porosity and permeability. The results provide theoretical guidance for the pore scale simulation of deep ultra-deep oil and gas reservoirs.