Structures of randomly packed beds of cylindrical shape porous pellets composed of tiny ceria (CeO2) particleswith wide rangesof pellet porosities and number density probabilities are generated byan improvedcontrollable structure generation scheme (CSGS) including deviations of both pellet porosities and sizes. Three dimensional (3D) mesoscopic scale thermal conduction through the structures immersed in carbon dioxide (CO2)gasissim-ulated by a parallel non-dimensional lattice Boltzmann method (P-NDLBM). Effective thermal conductivity (ETC) in each direction and 3D temperature distributions with effects of pellet number density probabilities, porosities, and deviations are illustrated. Althoughthepelletnumber density probability and pellet porosity are equivalently important for changing of macroscopic global porosity, they have different effects on temperature distribution patterns and the overall ETC. A new form of ETC correlation related to the pellet number density probability and pellet porosity is proposed. The present ETC correlation based on the two pellet parameters shows a slight improvement comparing to conventional forms based on the macroscopic global porosity.