The effective conversion of CO2 is important for achieving the goals of peak carbon dioxide emissions and carbon neutrality. Cylindrical particles have better thermal conductivity but smaller surface area when compared with heteromorphic particles, therefore, for better bed performance, the discrete element method (DEM) is used to perform particle-scale CFD simulations of a CO2 hydrogenation system in a mixed bed of heterogeneous particles, to investigate the effect of different stacking methods on the multi-physical field distribution, CO2 conversion and CH4 yield in the bed. The results show as follows: there is a diffusion resistance within the catalyst, as the reaction progress, the material within the particles changes from stratified distribution to uniform distribution. Compared to conventional beds, random mixed beds are less prone to hot spot fluctuations before steady state and has higher yields at steady state, while columnar bed have a more uniform radial heat field and higher hot spot. Of the four regular mixed beds, the two catalyst beds with 4 holes at bottom have more uniform overall flow fields, fewer high velocity zones, larger pressure drops, a larger proportion of high temperature zones and less fluctuation of hot spot before steady state than the two catalyst beds with columnar bottom. 2-layer mixed stacking-bottom 4 holes catalyst bed has maximum steady-state exit CO2 conversion and CH4 yield. The 4-layer mixed stacking-bottom 4 holes catalyst bed has high percentage of the high temperature region, which tends to increase as the reaction progress, resulting in significant decrease in CO2 conversion and CH4 yield.