Development of an artificial neural network model for predicting minimum miscibility pressure in CO2 flooding

This paper presents the development of an artificial neural network (ANN) model for the prediction of pure and impure CO2 minimum miscibility pressures (MMP) of oils. The pure CO2 MMP of a reservoir fluid (live oil) is correlated with the molecular weight of C5+ fraction, reservoir temperature, and concentrations of volatile (methane) and intermediate (C2–C4) fractions in the oil. The impure CO2 MMP factor, Fimp, is predicted by correlating the concentration of contaminants (N2, C1, H2S and SO2) in CO2 stream and their critical temperatures. The Fimp is a correction factor to the MMP of pure CO2. The advantage of using the ANN model is evaluated by comparing the measured MMP values with the predicted results from the ANN models as well as those from other statistical methods. The developed ANN models are able to reflect the impacts on CO2 MMP of molecular weight of C5+ fraction, reservoir temperature, and solution gas in the oil. The ANN model of impure CO2 MMP factor can distinguish the effects on MMP of different contaminants in the CO2 stream. It can also be used to predict the CO2 MMP of a reservoir oil and the level of contaminants in the CO2 stream which can be tolerated for a miscible injection.