Data driven prediction of oil reservoir fluid properties

Accuracy of the fluid property data plays an absolutely pivotal role in the reservoir computational processes. Reliable data can be obtained through various experimental methods, but these methods are very expensive and time consuming. Alternative methods are numerical models. These methods used measured experimental data to develop a representative model for predicting desired parameters. In this study, to predict saturation pressure, oil formation volume factor, and solution gas oil ratio, several Artificial Intelligent (AI) models were developed. 582 reported data sets were used as data bank that covers a wide range of fluid properties. Accuracy and reliability of the model was examined by some statistical parameters such as correlation coefficient (R²), average absolute relative deviation (AARD), and root mean square error (RMSE). The results illustrated good accordance between predicted data and target values. The model was also compared with previous works and developed empirical correlations which indicated that it is more reliable than all compared models and correlations. At the end, relevancy factor was calculated for each input parameters to illustrate the impact of different parameters on the predicted values. Relevancy factor showed that in these models, solution gas oil ratio has greatest impact on both saturation pressure and oil formation volume factor. In the other hand, saturation pressure has greatest effect on solution gas oil ratio.