Abstract: There are abundant complex pores in the coal matrix, which provide a large amount of storage space and migration channels for methane. Diffusion is involved in the gas migration from the pore to the fracture. The scale of the matrix determines to some extent the resistance to diffusion into the fracture, influencing the difficulty of methane diffusion. This study took the intrinsic connection between gas diffusion and coal matrix scale as the starting point, the quantitative relationship between methane concentrations and mass exchange rate within the dual pore structure of coal at different desorption moments was obtained by processing the data of the desorption transient process of particle coal. The matrix shape factor was calculated combined with the time-varying diffusion coefficient. A transient diffusion-based matrix scale calculation method for granular coal was proposed and experimentally validated. The results shown that the initial diffusion coefficients of larger particle coals with intact matrix morphology were essentially unchanged compared to smaller particle coals, so the value of the initial diffusion coefficient can characterize the degree of matrix destruction to some extent. The matrix shape factor decreased with the extension of desorption time and can be divided into sharp-decreasing phase, slow-decreasing phase and stable phase, in which the stable phase matrix shape factor can accurately reflect the matrix shape in the proposed steady state at the late stage of diffusion and is the most suitable for solving matrix scale. The method can reflect the change pattern of matrix scale variation during the damage process of granular coal pulverization and provide a basis for explaining the existence of diffusion-limited particle size. The three experimental particle size matrix scales of the Nuodong coal samples increased with the increasing particle size, 0.059 mm, 0.287 mm and 0.457 mm, respectively, and were non-differential in the large particle size range, proving the accuracy of the method. The particle coal matrix scale can be used to correct the calculation parameters of K1 values, rendering the gas loss calculation model equally applicable in coal samples with a high degree of pulverization.