A New Model for Predicting Minimum Miscibility Pressure (MMP) in Reservoir-Oil/Injection Gas Mixtures Using Adaptive Neuro Fuzzy Inference System

One of the critical concerns in determining the effectiveness of Enhanced Oil Recovery (EOR) is the understanding and evaluation of the Minimum Miscibility Pressure (MMP). Minimum miscibility pressure is the lowest possible pressure required to attain the mixing of injected fluid and the hydrocarbons in the reservoir into one phase. It is believed that optimum recovery and better sweep efficiency could only be achieved by reaching this minimum pressure. MMP determination, however usually depends on reservoir condition, reservoir fluid composition, and injected gas properties. The reservoir fluid composition could be represented by the Molecular weight C₇₊. However, Reservoir condition is represented by a reservoir temperature that affects MMP response. Selection of hydrocarbon gasses as the injection fluids is represented by the injected gas composition (Mole fraction C₂–C₆, and Mole fraction C₁). Determination of the minimum pressure could be either through experimental or empirical approaches. The objective of this study is to provide an empirical correlation to estimate the MMP by using Adaptive Neuro-Fuzzy Inference System (ANFIS). To develop the model, a code is generated under MATLAB environment. A total of 177 data points have been used in training the proposed model while 98 data sets have been used for testing the model performance. The proposed ANFIS correlation is then being compared with other previously published correlations. The best model currently used by industry has scored an average absolute percent error (AAPE) equivalent to 15% while the proposed ANFIS model managed to score 4.12%. By using the new ANFIS model, the study was able to produce a reliable and accurate correlation for estimating Minimum Miscibility Pressure as compared to other previously tested correlations.