Optimization of improved oil recovery in carbonates through LS-WAG-CO2 flooding using proxy models

Abstract

Due to the high concentration of carbon dioxide (CO2) as contaminant, combined to the high GOR (gas-oil ratio) of the fluids present in the Brazilian pre-salt reservoirs and because environmental reasons, researches on the advanced recovery technique like the water alternating with CO2 injection (CO2-WAG) have become highly relevant on the national scenario. In fact, this method has been applied by Petrobras in the Lula's field since 2011, and whose expected increase in the recovery factor is still unknown by the company. This will also happen with the other companies that will operate in the pre-salt cluster. Recently, a new recovery method was proposed by joining the low-salinity waterflooding with the water-alternating-CO2 injection (WAG), resulting in the LS-WAG-CO2, combining the advantages of both methods for the recovery in the carbonate reservoirs (Teklu et al., 2016). Preliminary results indicate that the high solubility of CO2 in the water with low salinity is the main reason for the mobilization of the residual oil when compared to the conventional WAG. However, the results presented are still very scarce and require further research in this area. Due to complexity of the problem, since the operation of many wells and critical parameters involves many control variables in the optimization process and, added to the high processing time of compositional simulation, this methodology will search efficiency and robustness in the modeling and optimization process. For this, the proposed methodology involves developing an efficient workflow of optimization parameters, considering in the process: 1) variables related with LS-WAG-CO2 flooding like salinity concentration, cycle and ratio WAG, among others; 2) integration between all of them and efficiency in the optimization workflow; 3) generate a proxy model using machine learning and 4) integration between optimization method with proxy created. The optimization method that will be tested is ensemble-based method (Chen & Reynolds, 2015), which have been highlighted in the recent literature. Once achieved greater efficiency and high speed in the optimization process, a benchmark model like UNISIM-II (Correia et al., 2015) will be used to apply the methodology. At the end of this work is expected to obtain a methodology that provides the best parameters for the advanced recovery operation through the water injection with low salinity alternating with CO2 injection, maximizing the NPV, seeking to understand the main mechanisms that govern this new EOR method, with reasonable simulation time. This can fill an important gap in the literature and real applications, which means that many companies fail to evaluate clearly the real benefits and advantages of the operation in optimal conditions of LS-WAG-CO2 flooding. (AU)