Determination of minimum miscibility pressure in CO2-IOR projects with the aid of hybrid neuro-fuzzy system

Abstract

Gas injection process is one of the most efficient improved oil recovery methods for conventional oil reservoirs. The efficiency of this process is strongly dependent on minimum miscibility pressure (MMP) which is usually determined through very expensive and time-consuming laboratory tests. So, this paper is concerned with the use of hybrid approach, combining particle swarm optimization and adaptive neuro-fuzzy inference system (ANFIS) with fuzzy c-means clustering technique, aimed at an estimation of injected gas-reservoir oil MMP. The hybrid proposed model makes ANFIS is practical in dealing with complex and high dimensional MMP problem. The process of model building is done by considering the reservoir temperature, crude oil composition, composition of injected CO2 as input parameters and corresponding CO2-oil MMP information as target parameter. The validity of proposed model has been successfully approved by comparing with results obtained by well-known empirical correlations in published literature. The results show that the proposed model significantly outperformed those four correlations with the lowest mean absolute error of 0.0099, the lowest root mean square error of 0.0174, and the highest coefficient of determination of 0.9823. Therefore, the proposed hybrid model can be applied as an alternative method to yield more accurate results in CO2-oil MMP estimation.