Prediction of CO2‐Oil Minimum Miscibility Pressure Using Soft Computing Methods

Abstract

AbstractSearching for computational approaches for determination of the minimum miscibility pressure (MMP) is highly requested during the miscible gas injection process. New models, namely, the stochastic gradient boosting (SGB) algorithm and two distinct hybrid artificial neural network (ANN) models were used to predict CO2 MMP as a function of reservoir temperature, mole percent of volatile oil components, mole percent of intermediate oil components, molecular weight of pentane‐plus fraction in the oil phase, mole percentage of CO2 in injected gas, volatile components, and intermediate components in the injected gas based on 144 published data points. The SGB model was found to provide the better performance. The reservoir temperature turned out to be the most important factor.