After nearly thirty years of research and development, it is now widely agreed that Low Salinity Waterflooding (LSW) provides better oil recovery than High Salinity Waterflooding (HSW). Past studies also showed that there are significant advantages in combining LSW with other conventional EOR methods such as chemical flooding (polymer flooding and surfactant flooding) or miscible gas flooding to benefit from their synergies and to achieve higher oil recovery factor and project profit. This paper presents a study of Hybrid Low Salinity Chemical Flooding as a novel EOR approach with: (1) development of hybrid EOR concept from past decades; (2) implementation of an efficient modeling approach utilizing artificial intelligent technology for mechanistic modeling of these complex EOR processes; (3) systematic validation with laboratory data; and (4) uncertainty evaluation of LSW process at field scale. The phase behavior of an oil-water-microemulsion system was modeled without the need of modeling type III microemulsion explicitly. The approach has been successfully applied to model both conventional Alkaline-Surfactant-Polymer (ASP) flooding and emerging EOR processes (LSW, Alkaline-CoSolvent-Polymer, and Low-Tension-Gas Flooding). The new development allows the mechanistic modeling of the benefits of combining LSW and chemical EOR. One of the main challenges for mechanistic modeling of these hybrid recovery processes is that several factors, e.g. polymer, surfactant, and salinity, can change the relative permeability simultaneously. To overcome this problem, Multilayer Neural Network (ML-NN) technique was applied to perform N-dimensional interpolation of relative permeability. The model was validated with coreflooding data and the effectiveness of hybrid processes were compared with conventional recovery methods. The proposed model showed good agreements with different coreflooding experiments including HSW, LSW, and Low Salinity Surfactant flooding (LSS). This model efficiently captures the complex geochemistry, wettability alteration, microemulsion phase behavior, and the synergies occurring in these hybrid processes. Results indicated that LSS is an economically attractive hybrid EOR process since it increases the ultimate recovery factor compared to the conventional approaches and reduces surfactant retention. Bayesian workflow using ML-NN algorithm is efficient to capture the uncertainties in history matching and production forecasting of LSW.

Many recent investigations showed that the prevalent thermodynamic models are incapable of predicting asphaltene precipitation without extensive data fitting. This is primarily due to lack of knowledge of the asphaltene properties, its complex nature and the large number of parameters affecting precipitation. Therefore, many authors tried to generate a simple and universe mathematical model in order to predict the amount of asphaltene precipitation. In spite of these efforts, the authors only considered temperature and type of solvents as the effective parameters in generating their scaling equations. The major disadvantage of these models is that they cannot predict the amount of asphaltene precipitation for different crude oils. Therefore, this deficiency contradicts to the universality of these models. In this work by performing experimental activities on different crude oils and analyzing their properties such as live oil GOR, Resin to Asphaltene ratio, mole percent of plus fractions and residual oil density, a new scaling equation developed in order to predict asphaltene precipitation for different oil samples. As far as this scaling equation has been generated using different samples, it can be used to estimate the amount of precipitated asphaltene at different dilution ratios and the onset dilution ratio of precipitation. It should be noted 2 that various literature precipitation data validated the predictive capability of this new scaling equation.

No abstract

Population-based optimization algorithms are shown to be excellent candidates for improving the speed and solution diversity of history matching and optimization workflows, based on their successful track records for solving real-world problems.The incorporation of reservoir engineering knowledge within these workflows, however, has been somewhat neglected. In particular, there is a lack of capability for guiding the optimization algorithms to specific regions of the search space. In a previous study, we introduced a framework for helping reservoir engineers incorporate their knowledge into history matching and optimization frameworks, by coupling a rule-based fuzzy system with a population-based sampling method.The question is how the use of this type of information in history matching affects the performance of the reservoir study during the prediction stage. This paper investigates the effect that the incorporation of reservoir engineering knowledge during the history matching of the Teal South model production data has on reservoir performance in the prediction stage.Two scenarios are considered. In Case I, we augment the history matching with reservoir engineering knowledge and then produce a forecast. In Case II, production data is history matched using differential evolution (DE), without fuzzy-logic-based engineering knowledge, then a forecast is producedThe results show that incorporating engineering knowledge of the reservoir under study during the history matching process can significantly reduce the uncertainty in the forecast, compared with the case where unrealistic parameter value ranges are used.