The clogging phenomenon often occurs during the reinjection of produced water due to the suspended particles, which will deteriorate the development efficiency. Many experimental and analytical methods have been introduced to solve this problem; however, few numerical approaches have been proposed to investigate the particle migration in the produced water reinjection process. Moreover, it is hard to obtain a clear understanding directly from the particle scale when the injected particles have different sizes. This paper employs a coupled lattice Boltzmann method and discrete element method (LBM-DEM) to study the aforementioned process. The method was validated by reproducing the Drafting–Kissing–Tumbling (DKT) process. Simulations of migration of injected particles with different sizes through porous media were conducted and three clogging scenarios had been identified. We investigated the impact of injected particle size distribution and porous media on particle migration and concluded the results in the polydisperse aspect. From the simulation, we can conclude that mix clogging is the scenario we should try to avoid. Besides, both critical ratio of particle diameter of porous media to median particle diameter of injected particles (D/d50) and critical standard deviation value exist. The particle size range should be as small as possible in economical limits and the D/d50 value should be larger than the critical value. Our results can provide a good guide for the produced water pretreatment, which can improve oil recovery.