Upscaling is an effective solution to overcome time and resource limitations in dynamic reservoir simulation. In this
study, eight absolute permeability upscaling methods were evaluated, namely: arithmetic, geometric, harmonic-arithmetic,
renormalization, pressure-solver, unbiased, sequential unbiased, and sequential pressure-solver methods. The
last three methods, introduced here for the first time, are numerical approaches in which the diffusivity equation is
solved over local fine grid-blocks under the corresponding boundary condition of the coarse grid-blocks, and the
equivalent permeability is then calculated using Darcy’s equation. By implementing six water-flooding scenarios
over three synthetically generated reservoirs, the effects of a broad range of mobility ratios (0.2 to 20), capillary
pressure, and reservoir heterogeneity on the performance of upscaling methods were investigated. Moreover, it was
observed that average boundary pressure errors were high and increased rapidly with up-scaling level from 2% at level
2 to 35% at level 16. Furthermore, a similar trend was observed for the average reservoir pressure, but it has much
smaller error which goes up to 7% at highest upscaling level. Therefore, it is necessary to apply near-well/boundary
upscaling methods to improve pressure distribution. The overall error never exceeded 0.5% with respect to heterogeneity
and capillary pressure effects, whereas at MR = 0.2, it increased to about 5–6% for oil and 1.5% for water production
rate. In most scenarios, numerical methods produced smaller overall error compared with analytical methods.
Ultimately, in general, unbiased and sequential unbiased estimation were found to be the superior numerical methods
and arithmetic and harmonic-arithmetic methods were identified as the superior analytical methods.