Coalbed methane (CBM) is an unconventional energy source.Its productivity is mainly controlled by the fracture network of the coal seam.Therefore, the key to accurately evaluate CBM productivity is to simulate the complex fracture network in the reservoir.In comparison, many studies about the simulation of fracture networks are based on the assumptions that fracture is smooth, they do not accord with the actual fracture.Here, a stochastic rough fracture network (SRFN), which is closer to the real coal fracture network, is established based on the pixel probability decomposition method.The relationship between the fractal geometric characteristics of the reconstructed fracture networks and permeability are discussed.The main findings through the simulation results were drawn as follows: fractal dimension, porosity, and permeability show good power-law relationships with fracture d.; the greater the fracture d., the denser is the fracture distribution, so effective seepage channels form easily and improve reservoir permeability.The permeability of the fracture network decreases with the increase of fracture dip angle.With a rise in fracture d. (14.653-80.135 m/m2), the permeability of a coal reservoir estimated using the SRFN model was 19.2-32.6% lower than that of the SFN model, indicating that the latter overestimate reservoir permeability.The research results of this paper are of great significance to further accurately predict the permeability of coal reservoir.