Abstract: In order to accurately describe the coal seam seepage pore characteristics from Hedong mining area in Urumqi,combined with the mercury injection test and coal quality analysis test,a section fractal method was used to quantitatively discussthe pore structure characteristics and influencing factors of seepage pores in coal reservoirs, and then coupledwith the porosity and pore structure to evaluate the permeability of coal reservoir.The results show that the pore structure of No.43 and No.45 main coal seams in Hedong mining area are more complex but the connectivity is better. The mercury intrusion curves of samples from No.43 coal seam show similar mercury saturation and high mercury removal efficiency, indicating that micropores and macropores are relatively developed. The maximum mercury intrusion saturation in samples from No.45 coal seam is between 45% and 85% and the mercury removal efficiencyis low, revealing that the micropores are dominantly developed. It is concluded that the permeability of No.43 coal seam is better than that of No.45 coal seam.The fractal dimension of coal seepage pores ranges from 2 to 3, showing that there are obvious fractal features in medium and large pores, and the average fractal dimensions of the mesoporous (D2) are commonly higher than those of the macroporous (D1). It is concluded that the mesopores are more complex than the macropores in the studiedcoal reservoirs.The macropore fractal dimension decreases with the increase of the the vitrinitecontent and ash yield, but it increases with the increase of the inertinite content, and it presents an inverted "U" correlation with the increase of moisture content in the coal. The permeability of coal is determined by the coupling of seepageof both porosity and pore structure. In the 95% confidence interval, the macropore porosity (P1), mesopore porosity (P2) and fractal dimensions are strongly, moderately and weakly correlated with permeability respectively, and fractal dimensions D1 and D2 have a weak correlation.Based on support vector machine method, the porosity and fractal dimension of the seepage pores are trained as independent variables to calculate coal permeability. The calculated permeability fits well with the experimental results with a goodness-of-fit. It is concluded that the coupling results of seepage pore porosity and pore structure can be effectively used to evaluate the coal reservoir permeability in the study area.