Eulerian-Eulerian models incorporated with the kinetic theory of granular flow were widely used in the simulation of gas-solid two-phase flow, while the effects of mesoscale structures such as particle clusters and gas bubbles could not be considered adequately if traditional homogeneous drag models were adopted in the coarse grid simulations. The energy minimization multiscale (EMMS) model has been proved to facilitate calculating a structure-dependent drag coefficient by considering particle clustering phenomena, which can be coupled with the two-fluid model (TFM) to improve the accuracy of coarse-grid simulation of gas-solid circulating fluidized beds. However, the original EMMS drag model cannot be further applied to the simulation of gas-solid fluidized beds with solids flow rate smaller than zero, e.g., turbulent fluidized beds and cocurrent downward flow, because the original cluster diameter correlation gives rise to a value smaller than single particle diameter or even negative value at extremely low solids fluxes or downward gas-solid flow. In this study, a new proposed cluster evolution equation is proposed by quantifying local clustering dynamics to replace the original cluster diameter correlation. The newly formulated EMMS drag model can be used to avoid a negative cluster diameter to be involved in calculating interphase drag force in the overall fluidization regime. The improved EMMS drag law is incorporated into the Eulerian-Eulerian model to simulate gas-solid turbulent fluidized beds and cocurrent downer reactors, since they both were widely used in many industrial processes. By analyzing local hydrodynamics as well as the axial and radial heterogeneous distributions in the two kinds of fluidized beds, it is clarified that the simulation using the improved EMMS drag model shows a better agreement with the experimental data than the computation using the homogeneous drag law. (C) 2019 Elsevier Ltd. All rights reserved.