Long-time behaviors widely exist in gas-solid reactors, e.g., the coke deposition and catalyst deactivation in the methanol to olefins (MTO) process. Due to the limitation of the computational cost, previous simulations of MTO reactors were mainly conducted with two-fluid model at short-time scales and assumed that the catalyst are perfectly mixed. Efficient simulation methods are, therefore, highly desirable, even for evaluating the assumptions through very long-time investigation. In this study, the MTO reaction kinetics is incorporated into a coarse-grained discrete particle method, namely EMMS-DPM, to simulate a pilot-scale reactor for 8 h of physical time. The simulation is validated by comparing the obtained gaseous products and coke contents with the experimental data. The perfect mixing assumption is validated by observing the residence time distribution of the catalyst for the whole reactor. The detailed coke content could provide more accurate data for accelerating steady state simulations. Furthermore, the relation between the coke content and age of the catalyst is directly obtained for the first time, which is valuable to effective use of the catalyst before discharging. The coke content is positively correlated to the age, but is also influenced by local conditions. For example, near the distributor, the catalyst particles are not well mixed and the maximum reaction rate fluctuates significantly even at large time scales. Thus, long-time particle-scale simulation would provide more dynamic details to characterize the reaction in the critical regions of the reactor, thus being very helpful to optimize operating conditions and the reactor designs.