Abstract: The fluoride salt-cooled high-temperature reactor (FHR) uses fluoride salt to cool fuel pebbles, which leads to the double heterogeneity in neutronic calculation including the random distribution of pebbles in the reactor and TRISO particles in the pebble. That is one of the special challenges in FHR design. Based on MCNP code and the solid-fueled thorium molten salt reactor (TMSR-SF1) model, the effects of different pebbles and pebble bed models used in criticality calculations on key parameters were analyzed, such as k_eff, neutron spectrum, control rod worth and temperature coefficients. The structure of pebble bed was explored by the random sequential addition (RSA) method and an equivalent body centered cubic (BCC) model with the same packing density. The distribution of TRISO particles was realized by an equivalent simple cubic (SC) model and the random perturbation implemented by the URAN card in MCNP code. The results show that the effect of the random distribution of TRISO particles is negligible compared to the random distribution of pebbles and the regular model causes more leakage than the random model, which leads to undervalue the cold state reactivity about 0.5% and overestimate the control rod worth about 5%.