Particles in industrial fluidized beds are usually continuously distributed rather than monodisperse in size, and the separation of particles is a key factor in determining the performance of the fluidized reactor. In this paper, a bubble-based Energy Minimum Multi-Scale (EMMS) drag model was proposed to simulate polydisperse segregation in a gas–solid fluidized bed. Particle size distribution (PSD) was split into a sufficient class of characteristic particle sizes. Each size was treated as a solid phase, and the total drag force could be estimated by considering the influence of all solid phases and mesoscale structures. The EMMS drag model was then coupled with the Eulerian multi-fluid model (MFM) to simulate the segregation behavior of a polydisperse particle system. The effects of discrete particle size classes, different drag models, and gas velocities on the simulation results were also discussed under different operating conditions. The results showed that the present drag model was in better agreement with the experimental data in terms of axial solids concentration and mean particle size distribution than other models. In bubbling beds, the segregation patterns of polydisperse particles can be well predicted by this method.