PVT correlations for Pakistani crude oils using artificial neural network

Abstract

Reservoir fluid properties such as bubble point pressure, oil formation volume factor and viscosity are very important in reservoir and petroleum production engineering computations such as outflow–inflow well performance, material balance calculations, well test analysis, reserve estimates, and numerical reservoir simulations. Ideally, these properties should be obtained from actual measurements. Quite often, however, these measurements are either not available or very costly to obtain. In such cases, empirically derived correlations are used to predict the needed properties using the known properties such as temperature, specific gravity of oil and gas, and gas–oil ratio. Therefore, all computations depend on the accuracy of the correlations used for predicting the fluid properties. Almost all of these previous correlations were developed with linear or nonlinear multiple regression or graphical techniques. Artificial neural networks, once successfully trained, offer an alternative way to obtain reliable and more accurate results for the determination of crude oil PVT properties, because it can capture highly nonlinear behavior and relationship between the input and output data as compared to linear and nonlinear regression techniques. In this study, we present neural network-based models for the prediction of PVT properties of crude oils from Pakistan. The data on which the networks were trained and tested contain 166 data sets from 22 different crude oil samples and used in developing PVT models for Pakistan crude oils. The developed neural network models are able to predict the bubble point pressure, oil formation volume factor and viscosity as a function of the solution gas–oil ratio, gas specific gravity, oil specific gravity, and temperature. A detailed comparison between the results predicted by the neural network models and those predicted by other previously published correlations shows that the developed neural network models outperform most other existing correlations by giving significantly lower values of average absolute relative error for the bubble point, oil formation volume factor at bubble point, and gas-saturated oil viscosity.