Modelling CO2 diffusion coefficient in heavy crude oils and bitumen using extreme gradient boosting and Gaussian process regression

Qichao Lv, Ali Rashidi-Khaniabadi, Rong Zheng, Tongke Zhou, Mohammad-Reza Mohammadi and Abdolhossein Hemmati-Sarapardeh

Energy, 2023, vol. 275, issue C

Abstract: In this work, five machine learning models based on Gaussian process regression (GPR) and Extreme gradient boosting (XGBoost) were developed for estimating the diffusion coefficient of CO2 in heavy crude oil/bitumen. Here, Rational Quadratic, RBF, Matern, and Exponentiated Sine Squared kernels were used in GPR. A databank comprising 260 experimental data of CO2 diffusion coefficient in various heavy crude oils/bitumen were collected in broad ranges of pressures (1.72–8 MPa) and temperatures (295–363.15 °C). Input parameters of the models were the mass fraction of CO2, pressure, temperature, crude oil/bitumen density and viscosity, and the temperature at which viscosity and density of crude oil/bitumen were measured. It was found that the XGBoost model has the highest precision compared to the rest having an R2 of 0.9998 and an average absolute percent relative error of 0.68%. Analyzing the trends showed that at a certain temperature and pressure, the diffusion coefficient of CO2 is a unimodal function of gas concentration in bitumen, and also temperature and pressure have increasing effects on the CO2 diffusion coefficient, which were satisfactorily predicted by the XGBoost model. Eventually, implementing the Leverage approach proved high credit of XGBoost model and gathered experimental CO2 diffusion coefficient databank.

Keywords: CO2 diffusion coefficient; CO2-Oil system; Machine learning; Gaussian process regression; Extreme gradient boosting; Leverage approach (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (1)

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0360544223007909
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:energy:v:275:y:2023:i:c:s0360544223007909

DOI: 10.1016/j.energy.2023.127396

Access Statistics for this article

Energy is currently edited by Henrik Lund and Mark J. Kaiser

More articles in Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().