Insights into the Effects of Pore Structure, Time Scale, and Injection Scenarios on Pore-Filling Sequence and Oil Recovery by Low-Salinity Waterflooding Using a Mechanistic DLVO-Based Pore-Scale Model

Abstract

Summary Despite the proven advantage of the low-salinity waterflooding (LSWF) technique, mechanistic understanding of the underlying phenomena at pore-scale remains uncertain. Hence, the corresponding models have limited predictability. In this study, wettability alteration via electrical double-layer (EDL) expansion is captured in a pore-scale model using a multispecies, multiphase computational fluid dynamics simulator. A combination of a pore-doublet and snap-off model is used to evaluate the low-salinity effect (LSE) in two geometries with different pore-throat size distributions. Contact angle is calculated intrinsically within the model using the concept of disjoining pressure through the implementation of the Derjaguin-Landau-Verwey-Overbeek (DLVO) theory and augmented Young-Laplace equation. The results illustrate that even in a simple pore structure, various pore-filling sequences and recoveries are obtained based on the pore geometrical factors, time effects, backward mixing, and injection scenarios. Secondary LSWF results in higher ultimate oil recovery since both small and large pores are accessible to flow and breakthrough is delayed, giving more time for more efficient displacement. Regarding the pore-throat geometry, the case with larger pores connected via larger throats generally exhibits higher ultimate recoveries. However, the geometry with larger pores connected by small throats results in higher incremental recovery via tertiary LSWF. Moreover, an optimal time scale exists in secondary LSWF due to the snap-off phenomenon, while faster LSE results in higher recovery in tertiary mode. The proposed model is capable of mechanistically capturing and predicting LSE and its subsequent flow dynamics, which exhibits a higher recovery factor by LSWF compared to the commonly used linear wettability model. Thus, this approach improves the predictive capability of the previous models as it does not require contact angle data and arbitrary interpolation schemes.