Improved field scale simulations of Low Salinity Waterflooding using analytical solutions, numerical analysis and upscaling

Abstract

Low Salinity Water Flooding (LSWF) is an emerging enhanced oil recovery (EOR) process that has been increasingly studied due to its potential to reduce residual oil saturation in oil reservoirs. This thesis focuses on modelling LSWF at the field scale. In particular, it aims to improve the representation of flow behaviour and solute transport at the inter well scale. We present in depth analyses of numerical issues that may appear in LSWF simulations and derive two novel upscaling methods. A black oil simulator was used to model this process where the research focused on the fluid flow behaviour at the reservoir scale rather than the underlying mechanism. Two sets of relative permeability were used to simulate the process. Switching from one set to another was salinity dependent. The flow behaviour was first examined in 1D homogenous models. Then, 2D-layered models were studied. Also, we investigated models where permeability and porosity were randomly distributed. We found that an interaction between dispersion and effective salinity range led to a retardation in the fractional flow solution of LSWF with an outcome similar to that produced from adsorption in polymer and other cEOR models. We suggest an empirical correlation to predict that effect which was derived later by linking the fractional flow solution with the advection-dispersion equation. Also, we observed that salinity was transported faster than predicted by the traditional advection-dispersion equation. We modified the advection dispersion equation to capture the advection velocity of the salinity front. We extended the fractional flow solution of chemical flooding and the advection dispersion equation from 1D to 2D non-communicating layers where in such systems the frontal velocities vary as a function of time. We found that dispersion induced by geological heterogeneity affects fluid flow in a manner similar to numerical dispersion. Thus, we can use the numerical dispersion as a proxy for the effects of physical dispersion. In terms of the numerical issues, we found that LSWF models resulted in pulses, false retardation and overestimated recovery factor. Pulses were apparent for stable numerical solutions. We analysed the parameters that induce numerical error and examined the resulting errors which were reduced by upscaling. Two upscaling methods were developed in this thesis where these methods succeeded to control the numerical issues for coarse scale models so that the production data matched the fine scale models. A simplified modelling process is another outcome of this upscaling method where one can use the traditional water flooding models to simulate LSWF.