This thesis investigates the impact of physical mixing in controlled/low salinity waterflooding (LSWF) on optimal slug injection taking into account uncertainties in the reservoir description. The main motivation is that uncertainty can be detrimental to optimal field-scale implementation with poor scale-up performance even if cm-scale laboratory experiments suggest favourable recoveries. Mixing of the injected low salinity (LS) brine with high salinity (HS) formation brine potentially compromises the low salinity effect especially if injected as slugs. Other physical effects including reservoir heterogeneity and crossflow exacerbate such interaction between the distinct brines. It is challenging to capture this physical mixing accurately in reservoir simulations of LSWF because of numerical dispersion.
In this thesis, a new methodology, based on ‘effective or pseudo’ salinity-thresholds, is proposed to allow accurate and representative field-scale simulation of the effect of physical mixing accounting for numerical errors. The effect of inter-well mixing is quantitatively evaluated on LS slug performance, its displacement stability, and the optimal amounts to inject considering associated flow and geological uncertainties. The influence of LS injection start-time and flood or well pattern is examined. Injection optimization and performance uncertainty are assessed through validated experimental designs and response surfaces, in addition to a practical and effective approach to select representative geological realizations to minimize computational resources. Crossflow effects on LS displacement are explored for different mobility ratios with the more realistic non-piston-like shock-front using dimensionless numbers.
The procedure to constrain numerical dispersion shows improved effectiveness both in 1D and higher dimensions. Incremental recovery and the optimal slug injection are observed to be relatively insensitive to LS injection start-time for the line-drive, but not the quarter five-spot pattern, until tertiary-mode displacement becomes established. A favourable influence of the spatial distribution of heterogeneity is seen for both flood patterns, due to transverse dispersion. The optimal slug sizes from existing studies are, at best, only suitable as thresholds for stable LS displacement of the injected slug. For a cumulative water injection of 2 PV, LS slug design would require at least 0.8 PV optimal size and 0.24 PV minimum stability-threshold size across flood patterns. This can vary with the cumulative water injected and flood pattern but remains generally favoured by an early injection. The optimal slug size is found to be largely insensitive to heterogeneity for the Q5S flood pattern. It is found that the slug size dependence of performance and LS displacement stability – under the effects of physical mixing in multidimensional models – is a function of well arrangement. An unfavourable mobility-ratio case is identified for viscous-dominated crossflow effects on LS recovery efficiency. Ways to improve the flood design for favourable displacements given this mobility-ratio case are presented.