A dynamic cluster structure-dependent (CSD) drag coefficient model is proposed to be consistent with the temporal-spatial characteristics of dynamic clusters through the convective accelerations and temporal accelerations of the dense phase and the dilute phase. The CSD drag coefficient is determined by a nonlinear niicro-meso-grid scales equation set which consists of three momentum conservation equations, two mass balance equations, one equation for volume fraction balance, and an extreme value of a function in combination with the bivariate extreme value (BEV) theory. Flow behavior of gas and particles is predicted by means of gas-solid two-fluid model coupled with CSD drag model and kinetic theory of granular flow. The distributions of independent variables of dense phase and dilute phase are predicted in the riser. Numerical analysis suggests that the inertial difference between the dense phase and the dilute phase affects flow behavior of dynamically temporal-spatial clusters in risers. The simulated solids volume fraction, cluster existence time fraction and frequency of cluster occurrence are compared to experimental measurements in the literature. (C) 2017 Elsevier B.V. All rights reserved.