Numerical Investigation of Hybrid Smart Water and Foam Injections in Carbonate Reservoirs

This contribution is a progressive effort to investigate the effect of the novel hybrid EOR method of Smart Water Assisted Foam (SWAF) technique on oil recovery from carbonates through numerical modeling. In this work, a core-scale model was utilized to provide an insight and a better understanding of the controlling mechanisms behind incremental oil recovery using a new hybrid EOR method consisting of a combination of smart water flooding and foam injection, termed as Smart Water Assisted Foam (SWAF) technology, particularly for carbonate reservoirs. A core-scale model encapsulating the physics of SWAF flooding was used to history-match experimental data and the model was further optimized utilizing the CMG simulator. For extracting the most value from this numerical investigation, a sensitivity analysis was performed to monitor the effect of influential parameters affecting oil recovery depending on the spectrum of the experimental data available. The objective functions used in the sensitivity analysis include minimizing the history-matching global error and maximizing the oil recovery profiles. Three sensitivity analysis approaches were used: Tornado-plot, SOBOL analysis, and MORRIS analysis. For generating the related proxy models, polynomial regression, and radial basis function (RBF) neural networks were investigated. Subsequently, the DECE-based and PSO-based optimization methods were employed to examine the effect of chemical design parameters such as smart water (Mg2+), surfactant aqueous solution (SAS), and foam concentrations along with the liquid production rate on the oil recovery factor during SWAF-flooding. Based on the numerical results, the experimental coreflooding data were accurately history-matched using the proposed model with a minimal error of 4.74% applying the PSO-based optimization method. Furthermore, in terms of the objective function prediction during the sensitivity analysis study, the comparative assessment of both proxy models on the verification plot reveals that the RBF neural network outperforms the polynomial regression. Consolidated findings from the three sensitivity analyses, i.e., the Tornado-plot, SOBOL, and MORRIS, outline three common parameters that significantly affect the oil recovery profiles that are liquid production rate (LigProdCon), foam (DTRAPW SAS2), and Mg2+ concentration (DTRAP Mg3) parameters. On the other hand, in terms of maximizing the oil recovery while minimizing the usage of injected chemicals during SWAF flooding, the optimal solution via the PSO-based approach is superior (97.89%) to the DECE-based optimal solutions (92.47%). This work presents one of the few studies investigating the numerical modeling of the SWAF process and capturing its effects on oil recovery. The optimized core scale model can be further used as a base for building a field-scale model and designing a successful pilot project.