Reliable reservoir prediction is essential for optimized production and reservoir management. The prediction is normally done by reservoir simulation. Reservoir simulators solve fluid flow equations of reservoir numerically on homogenized coarse blocks of reservoir model. The original fine grids are generated by primary geological blocks which are output of geological software. The upscaling is necessary since geological software by means of statistical methods create models with millions and even billion of grid blocks and dynamic simulation on these models is practically not possible. In this study, we introduced different types of hybrid grid by nature-inspired method, which is the basis of scale-up model and then implement upscaling procedure. The simulation results on the geological structure well compared with the results of upscaled models. The results confirm that nature-inspired method consumes less run time with nearly accuracy of fine model.