To better understand the characteristics of coal pores and their influence on coal reservoirs, coal pores in eight main coalfields of North China were analyzed by mercury porosimetry and scanning electron microscopy (SEM). Fractal characteristics of coal pores (size distribution and structure) were researched using two fractal models: classic geometry and thermodynamics. These two models establish the relationship between fractal dimensions and coal pores characteristics. New results include: (1) SEM imaging and fractal analysis show that coal reservoirs generally have very high heterogeneity; (2) coal pore structures have fractal characteristics and fractal dimensions characteristic of pore structures are controlled by the composition (e.g., ash, moisture, volatile component) and pore parameters (e.g., pore diameter, micro pores content) of coals; (3) the fractal dimensions (D1 and D2) of coal pores have good correlations with the heterogeneity of coal pore structures. Larger fractal dimensions correlate to higher heterogeneity of pore structures. The fractal dimensions (D1 and D2) have strong negative linear correlations with the sorted coefficient of coals (R2=0.719 and 0.639, respectively) that shows the heterogeneity of coal pores; (4) fractal dimension D1 and petrologic permeability of coals have a strong negative exponential correlation (R2=0.82). However, fractal dimension D2 and petrologic permeability of coals have no obvious correlation; and (5) the model of classic geometry is more accurate for fractal characterization of coal pores in coal reservoirs than that of thermodynamics by optimization.