Understanding the relationship between a fluid flow and the structural properties of a bed is important for various industrial applications and a proper description of particle-fluid flow. This work presents a pore-scale study of fluid flow through randomly packed beds by a network model. The model incorporates the inertial effect and thus can be applied to flows of high Reynolds numbers. The results show the structure heterogeneity associated with the connection of pores is important in determining the distribution of fluid flow. The statistical distribution of normalized drag forces on individual particles is Gaussian, related to the statistical distribution of pore properties. The mean drag force of a bed by the pore-scale model agrees well with that from a previous study by the lattice-Boltzmann method. These findings support the application of the pore network model for simulating particle-fluid flows under a wider range of Reynolds numbers.