Engineered/low-salinity water injection technology is one of the latest promising techniques for improving recovery in both sandstone and carbonate reservoirs. In this paper, the optimization of engineered water injection technology is investigated using three 5-spot reservoir models including heterogeneous with channeling, heterogeneous with gravity underride, and homogeneous. The net present value was chosen as an objective function for the study. 18 design parameters were selected for the study including reservoir, operational, and economic parameters. The machine learning and artificial intelligence tools were utilized. Response Surface Methodology and Designed Exploration and Controlled Evolution algorithms were implemented for sensitivity analysis and optimization studies, respectively.
The sensitivity study showed that oil price, tax rate, and initial oil saturation are among the most significant design parameters on the net present value for the three models investigated. Moreover, the findings showed that developing the gravity underride model requires more attention as being the most sensitive model with 13 influential design parameters. The optimization study highlighted the need for engineered water injection in the secondary mode to achieve the best profitability out of the three models. Also, it is recommended for an operator to invest in the homogeneous, followed by the channeling, and least the gravity underride models due to the corresponding net present values of $1.44 million, $1.39 million, and $0.96 million, respectively.
The study highlighted the importance of selecting the most suitable objective function for achieving the project profitability by comparing net present values vs. oil recovery as objective functions. In addition, this study can be used as a guide for using artificial intelligence tools to understand the most influential engineered water injection design parameters that affect profitability, and hence field scale developments can be conducted with more certainty and lower risk.