In the methanol-to-olefins (MTO) process, coke deposition is closely related to the selectivity of light olefins. Previous simulations of different-sized MTO reactors using two-fluid model (TFM) combined with the EMMS (energy minimization multi-scale)-based drag well predict the hydrodynamic behaviors but poorly predict the product distribution of large reactors due to the unreasonable prediction in coke distribution. In this study, the TFM integrated with the EMMS-based drag is still employed, but the solid phase is treated as a mixture of a series of species with different coke contents. Because the coke content depends on the age of catalyst particles inside the reactor, a continuous stirred tank reactor (CSTR) model mimicking turbulent fluidization while considering the age distribution of catalysts is established to predict the initial coke distribution for speeding up simulation. Compared with the previous simulation without consideration of coke distribution, this approach increases the computational time by a factor of 36% and shows no influence on hydrodynamic predictions, but the reaction quantities such as methanol conversion, mass fractions of gaseous products and selectivity of light olefins, are better predicted. (C) 2018 Elsevier Ltd. All rights reserved.