Prediction of bubble point pressure for Sudan crude oil using Artificial Neural Network (ANN) technique

Abstract

In this study, a new Artificial Neural Network predictive model was developed to determine the bubble point pressure P_b for Sudanese oil field using ANN tools in MATLAB software. Because of limitations of the experimental procedures and the time taken to obtain bubble point pressure value from the reservoir fluid samples analysis, an alternative is required where many researchers have been conducting research on the use of Artificial Neural Network (ANN) techniques. In the present study ANN model was developed and evaluated using 151 experimental data sets for Sudanese oil field, and more 61 data sets are used to compare the developed model with universal and regional published models. Comparing with universal and global empirical models, the developed model of ANN for P_b pressure has better precision index of correlation 94.63% with MSE and RMSE of 180 and 156, respectively. However, the results show that of ANN has a lower performance than regional PNN model, as the PNN shows index of correlation 97.57% with MSE and RMSE of 88 and 101, respectively. This difference may be due to the limitation in number of variables and number of data points used in each model developed. Thus, the ANN developed model in this study might be improved to predict the P_b especially for Sudan oil fields and similar oil field properties in regional by increasing the data point used in ANN model training.