Transport of small hydrocarbon molecules inside organic nanopores is of great interest in the oil and gas production from source rocks such as shale. Because of the small pore sizes, it is difficult to measure transport directly in the laboratory but require alternative approaches such as molecular simulation. Existing molecular models use nanoscale capillaries and do not describe the exact geometric structure of the organic nanopores in real source rocks, especially for mesopores (2-50nm in diameter, defined by IUPAC), which widely exist in organic solids and are crucial for fluid transport in the source rocks. Here we present a new method to simplify the existing molecular kerogen models. Pore structures can be built to reflect the exact porous structure of hydrocarbon-bearing kerogen pores for fluid transport without losing accuracy. The kerogen pore-network skeleton can be populated based on 3D digital segments obtained from high-resolution TEM tomographs, combining with other material properties, including the elemental composition and density of the organic solid. The basic idea of the simplified molecular model is to ignore the detailed chemical information of the kerogen molecules and to use a uniform representative atom to replace all various atom types, while maintaining the geometric structure intact. To test the transport simulation accuracy using molecular dynamics for this simplified model, sample pore-network structures are populated with both the existing exact molecular model and the simplified model, and the permeability of reservoir fluid flowing through the media is simulated and compared. Specifically, simple reservoir fluids are charged into the space of the 3D porous networks with grand canonical Monte Carlo simulation at given subsurface condition; then, the fluids are driven by an external force field to flow through the network using non-equilibrium molecular dynamics simulation. The simplified kerogen model can be easily applied to larger organic porous material samples to reduce spatial uncertainty, provided enough computational resources.