A Data Driven P-V-T Model to Predict the Oil Formation Volume Factor, Solution GOR and Bubble Point Pressure for Characterizing an Oil Reservoir

Abstract

Abstract Characterizing an oil reservoir requires one to understand the Pressure- Volume-Temperature (PVT) properties of reservoir fluids, especially bubble point pressure, solution gas oil ratio and oil formation volume factor because of its more often utilization in reservoir engineering studies. The current correlations are restricted by the use of sample from a particular field. As the physical properties and the composition of the crude oil varies the results becomes erroneous after a specific range. This correlation will give results only over a specific range of properties like specific gravity, viscosity, composition etc. The challenge is to develop a new approach which overcomes the current shortcomings. In this paper a new machine learning based model has been developed using Interactive Multivariate Linear Regression (I-MLR) method by integrating a large number of datasets to predict above mentioned properties. It overcomes the restriction of the previous correlations as it does not use data from any particular field. As such it is applicable over wide range of physical properties and composition. This model does not require any laboratory studies which makes it more economical. The validation of the model is done after detailed comparative study done with various commercially used empirical correlations.