Oil-CO2 minimum miscible pressure (MMP) determination using a stimulated smart approach

Multi-contact miscible displacement of reservoir petroleum by injection of the relatively inexpensive carbon dioxide (CO2) gas represents one of the most efficient currently available techniques of enhanced oil recovery. Experimental determination of optimal injection pressure, known as minimum miscible pressure (MMP), is very costly, laborious, and time-intensive. Therefore, the quest for a quick, cost-effective, and accurate way of determining MMP is inevitable. In this study an improved approach was followed to predict oil-CO2 MMP. First, input/output data are transformed to a higher correlated data space using an alternative condition expectation (ACE) algorithm to reduce complexity of the problem for a generalized regression neural network (GRNN). Then, transformed data are introduced to the GRNN to construct a model for estimation of MMP. Eventually, a hybrid genetic algorithm-pattern search technique is employed for optimizing the GRNN and producing more accurate estimations. This model was constructed using worldwide experimental data from open literature. The performance of the constructed model was assessed by introducing experimental data from Iran Oils as test data. A comparison between the current study and previous models, including outstanding correlations and intelligent systems, demonstrated the superiority of the proposed model.