For an insight into the catalytic reforming process in the exhaust reformer, three-dimensional mesoscale models of the fixed bed packed with different particle shapes were constructed by Computational Fluid Dynamics (CFD) and Blender software. Six particle shapes viz. sphere as the reference, 1-hole sphere, cylinder, Raschig ring, trilobe and 4-hole cylinder were chosen. Combined with the empirical formula of porosity distribution and experimental data, the accuracy of the random stacking model was verified. The effects of particle shape on flow heat transfer and catalytic reforming process in the exhaust reformer were investigated to provide theoretical support for selecting the optimum particle shape and optimal operating conditions of exhaust gas-methane reforming reaction. The results showed that the shape of catalyst has a noticeable effect on the flow resistance and uniformity of the flow field in the random packing area. Among the particle shapes considered in the present work, the fixed bed packed with 1-hole spheres is the optimal option for the exhaust reformer, which offers the highest methane conversion, hydrogen yield per unit pressure drop and the most uniform flow field. Meanwhile, under different methane-to-oxygen ratios (i.e., M/O), the methane conversion increases with steam-to-methane ratio (i.e., S/M). The volume ratio of hydrogen to carbon monoxide (i.e., H2/CO) decreases continuously with the increase of GHSV. Ultimately, the optimal operating conditions for exhaust reformer were determined: M/O is 2, S/M is 1 and GHSV is 8000 h−1, which can reach a peak hydrogen production of 24.4%.