Direct application of computational fluid dynamics and discrete element method (CFD-DEM) in industrial applications is unfeasible due to the enormous number of particles. As a result, the coarse-grained CFD-DEM (CFD-CGDEM) is proposed, which could save time and money by lumping a cloud of real particles into a coarse particle. However, due to its under-prediction of collisional energy dissipation, CFD-CGDEM typically over-predicts granular temperature and solid stresses when compared to CFD-DEM. As a result, in CFD-CGDEM simulations, a coarse-grained model capable of increasing energy dissipation is necessary. This work developed a coarse-grained model called discrete relaxation model based on the granular kinetic theory in the homogeneous cooling system (HCSs). By putting dissipation forces on the particle-pair, the discrete relaxation model could increase the energy dissipation between them. The proposed model eliminates errors in the estimation of local average solid phase velocity and granular temperature, as compared to Yu et al.'s relaxation model (Ind. Eng. Chem. Res., 2021, 60(15): 5651-5664). A posteriori simulations on homogeneous cooling systems and bubbling fluidized beds were used to assess the proposed model. It was discovered that CGDEM with the proposed model produced a more accurate forecast of the instantaneous granular temperature in HCSs than CGDEM with the usual coarsening model which was unable to improve energy dissipation. Furthermore, when compared to CFD-CGDEM with Yu et al.'s model and that with usual coarsening model, CFD-CGDEM with the proposed model better reproduced the time-averaged fields generated by CFD-DEM simulation for the considered bubbling fluidized bed. This emphasized the significance of increasing energy dissipation in CFD-CGDEM simulations, as well as the potential of the proposed model to considerably increase simulation accuracy.