Pore structure heterogeneity affects the porosity and permeability variation of tight sandstone, thereby restricting sandstone gas production. In total, 11 sandstone samples were taken as a target in the northwest margin of the Junggar Basin. Then, scanning electron microscope and high-pressure mercury injection tests are used to study the distribution of a pore and fracture system in the target sandstone. On this basis, single and multifractal models are used to quantitatively characterize the heterogeneity of pore structure, and the applicability of the classification model in characterizing the heterogeneity of the pore-fracture structure is explored. The results are as follows. (1) The target samples are divided into two types, with the mercury removal efficiency of type A samples ranging from 44.6 to 51.8%, pore size mainly distributed between 100 and 1000 nm, and pore volume percentage ranging from 43 to 69%. The mercury removal efficiency of type B samples ranges from 14 to 28%, and pore diameter distribution is relatively uniform. (2) Different fractal models represent different physical meanings. The calculation results of sponge and thermodynamic fractal models indicate that the heterogeneity of pore structure distribution in the type B sample is significantly stronger than that in type A, which is inconsistent with the conclusions of the Sierpinski model. This is because the aforementioned two models characterize the complexity of pore surface area, while the Sierpinski model characterizes the roughness of pore volume. The comparison shows that there is a significant correlation between the thermal dimensionality value DT and the volume percentage of macropores and mesopores. Therefore, the thermodynamic model can better quantitatively characterize the heterogeneity of macropore and mesoporous pore distribution. (3) The results indicate that higher pore volume range is mainly influenced by mesopores and macropores. From the relationship curve between mercury removal efficiency and single fractal dimension, it can be seen that mercury removal efficiency is greatly affected by distribution heterogeneity of the lower value area of pore volume, and it has no obvious relationship with distribution heterogeneity in the lower value area of the pore volume.