Modeling the Properties of Crude Oil and Gas Systems Using RBF Network

Abstract

Abstract This paper presents a new technique to model the behavior of crude oil and natural gas systems. The proposed technique is using a radial basis function neural network model (RBFNM). The model predicts oil formation volume factor, solution gas- oil ratio, oil viscosity, saturated oil density, undersaturated oil compressibility, and evolved gas gravity. Input data to the RBFNM are reservoir pressure, temperature, stock tank oil gravity, and separator gas gravity. The model is trained using differential PVT analysis of numerous crude oil samples collected from various oil fields. The proposed RBFNM is tested using PVT properties of other samples that has not been used during the training process. Accuracy of the proposed network model to predict PVT properties crude oils and gas systems is compared to the accuracy of numerous published PVT correlations. The physical behavior of the model is also checked against experimentally measured data for the test samples. The results showed that the RBFNM is reliable and has better accuracy than the conventional PVT correlations. The model can be incorporated in reservoir simulators and can also be used to check the accuracy of future dilferential PVT reports. This study also shows that once this model is properly trained it can be used to cut expenses of frequent sampling and laborious differential PVT tests. The RBFN model can also be used to forecast PVT properties needed for reservoir and production engineering calculations such as material balance, reservoir simulation separator design, and vertical performance design. P. 35