Bubble_Point_Correlation_With_ML : This Technical Paper presentation placed 2nd (1st runner up) in the 2022 Africa Regional Student Paper Contest
This repository seeks to show a practical application of machine learning in the oil and gas sector, particularly in reservoir systems
Bubble point pressure (BPP) is a key parameter for reservoir simulation and management. An accurate correlation of this property, in the absence of experimental PVT analysis, serves as guidance in the development of reservoir strategy to be put in place for a particular field. In this study, a bubble point pressure (BPP) correlation was derived by intrinsically linearizing a nonlinear multiple regression, with the best coefficients (global minimum) extracted by fitting a Linear Regression Machine Learning model (white box) to the data. The new correlation was developed, validated, and tested using 314 measured PVT data points from the Niger Delta Region. 60% of the data was used for training, 10% for validation, and 30% for testing. With the wide range of API gravity obtained from the data points, the data was further subdivided into four classes: extra-light crude for API > 45, light crude for 31.1 < API ≤ 45, medium crude for 22.3 < API ≤ 31.1, and heavy crude for API ≤ 22.3. Statistical measures such as root mean squared error, percent average absolute relative error, and percent average relative error were employed to compare the performance of the new correlation with the existing ones. The new correlation outperformed the existing ones from the result of the statistical evaluation metrics employed. It is believed that this new correlation developed, can accurately estimate bubble point pressure (BPP) over a wide range of oil volatility.