Reservoir fluid PVT properties modeling using Adaptive Neuro-Fuzzy Inference Systems

Abstract

Knowledge of prediction of PVT properties of reservoir oil is of primary importance in many petroleum engineering studies such as inflow performance calculations, production engineering studies, numerical reservoir simulations and design of proper improved oil recovery techniques. Ideally these parameters should be determined experimentally in laboratory under the reservoir conditions such as pressure and temperature. But owing to the fact that experimental methods are very expensive and time consuming, numerical models are developed for prediction of PVT properties. In this study several predictive models, based on a large data bank from different geographical regions were developed to predict the reservoir oil bubble point pressure as well as oil formation volume factor (OFVF) at bubble pressure. Developed models were successfully applied to the data set and the predicted values were in a good agreement with experimental values. Also a comparative study has been carried out to compare the result of this study to previously proposed correlations in terms of accuracy. Results show that the proposed models are more accurate than the available approaches.