A new correlation for accurate prediction of oil formation volume factor at the bubble point pressure using Group Method of Data Handling approach

Abstract

Listen

Translate article

Pressure-Volume-Temperature (PVT) crude oil properties play a significant role in reservoir evaluation and field planning. PVT properties are usually determined through laboratory experiments on representative fluid samples. However, during the preliminary stages of exploration and appraisal, such data might not be available; hence, it is frequent to use empirical correlations to predict PVT properties. Oil formation volume factor (FVF) is one of the PVT properties used to convert the measured oil flow rate from surface conditions to reservoir conditions. In this paper, the Group Method of Data Handling (GMDH) has been used to predict the oil FVF at the bubble point pressure as a function of gas solubility, reservoir temperature, oil API gravity, and gas specific gravity. A total of 625 data sets were collected from published literature. Then, the data were divided into four sets: training, validation, testing, and deployment, with the ratio of 2:1:1:1. The results of the proposed correlation are compared against seven other correlations used in the petroleum industry. Also, trend analysis has been performed to confirm that the proposed correlation is physically sound. From the results, the proposed correlation is proven to accurately predict the oil FVF at the bubble point pressure with an average absolute percentage error AAPE of 1.333% and correlation coefficient of 0.995 for the deployment set. • This paper developed a new correlation to estimate the oil formation volume factor at the bubble point pressure. • The Group Method of Data Handling (GMDH) approach has been used successfully in this study. • A total of 625 data samples were used to train, validate and test the model’s performance from different literature sources. • The proposed correlation’s performance was compared with six other correlations used in the petroleum industry. • Different statistical indices were used to confirm the model superiority, including trend analysis and group error analysis. • Statistical analyses show that the model has outperformed all correlations and models that were tested against. • The GMDH approach provides an easy-to-use correlation compared to some hard-to-retrieve models. • FVF is estimated as a function of only gas solubility, the reservoir temperature, and the gas-specific gravity.