Based on the meso scale model of concrete, a simulation method combining COMSOL with MATLAB was used to inverse heat conduction problems of concrete in steady state. With the verified simulation method, effects of randomly distributed aggregates on thermal conductivity of concrete were analyzed, and influence of aggregate content, aggregate gradation, aggregate shape and thermal conductivity of the dispersed phase were also discussed. It has been shown that coarse aggregate distribution is an important factor causing randomness of thermal conductivity of concrete. The average thermal conductivity of concrete increases with the increased volume content and thermal conductivity of aggregate, while aggregate size and particle shape have little impact on the average thermal conductivity of concrete. Thermal conductivity increases with the increasing aggregate volume content, but remains unchanged if aggregate volume content exceeds 35%. It is also found that the thermal conductivity of concrete changes more significantly with the increase of proportion of large particle aggregates, the axial length of aggregates and the difference of thermal conductivity between dispersed phase and cement mortar.