Effective thermal conductivity (ETC) is an important parameter in packed-bed systems and is affected significantly by packing structure. Heat flow through a packed bed can be divided into three parallel paths: the solid-fluid-solid path, the solid-solid path and the fluid path. The models to evaluate ETC in packed beds for the first two paths were previously developed. In the present work, a comprehensive model to consider the heat transfer through the voids of a randomly-packed bed is further developed using the Delaunay tessellation. The results show that heat conduction through the fluid-filled voids becomes significant when the solid-to-fluid conductivity ratio decreases to a certain level (<5). The effect of packing structure on the ETC of packed beds is then investigated based on the packing generated by the discrete element method (DEM). Four cases with different heat transfer mechanisms or paths are considered. The results indicate that the probability density distribution of heat flow is generally dependent on the packing structure, represented by porosity. With an increase in porosity, the heat conduction and radiation via the solid path decrease, while the heat conduction via the fluid-filled voids increases. (C) 2019 Elsevier B.V. All rights reserved.