Bubble Pressure Prediction of Reservoir Fluids using Artificial Neural Network and Support Vector Machine

Abstract

Bubble point pressure is an important parameter in equilibrium calculations of reservoir fluids and having other applications in reservoir engineering. In this work, an artificial neural network (ANN) and a least square support vector machine (LS-SVM) have been used to predict the bubble point pressure of reservoir fluids. Also, the accuracy of the models have been compared to two-equation state-based models, i.e. SRK-EOS and PR-EOS and four empirical equations, i.e. Whitson, Standing, Wilson and Ghafoori et al. Compared to the experimental data, the average relative deviations (ARD) of bubble pressure prediction for these equations were obtained to be 14%, 29%, 66%, 30%, 38%, and 11%, respectively. The best semi-empirical equation has an ARD of about 11% while, the ANN and LS-SVM models have an ARD of 8% and 4.68%, respectively. Thus, it can be concluded that generally, these soft computing models appear to be more accurate than the empirical and EOS based methods for prediction of bubble point pressure of reservoir fluids.

Keywords

• artificial neural network
• bubble pressure
• empirical correlations
• genetic algorithm
• reservoir fluids
• support vector machine