Enhanced oil recovery (EOR) using CO2 injection is promising with economic and environmental benefits as an active climate-change mitigation approach. Nevertheless, the low sweep efficiency of CO2 injection remains a challenge. CO2-foam injection has been proposed as a remedy, but its laboratory screening for specific reservoirs is costly and time-consuming. In this study, machine-learning models are employed to predict oil recovery factor (ORF) during CO2-foam flooding cost-effectively and accurately. Four models, including general regression neural network (GRNN), cascade forward neural network with Levenberg–Marquardt optimization (CFNN-LM), cascade forward neural network with Bayesian regularization (CFNN-BR), and extreme gradient boosting (XGBoost), are evaluated based on experimental data from previous studies. Results demonstrate that the GRNN model outperforms the others, with an overall mean absolute error of 0.059 and an R2 of 0.9999. The GRNN model's applicability domain is verified using a Williams plot, and an uncertainty analysis for CO2-foam flooding projects is conducted. The novelty of this study lies in developing a machine-learning-based approach that provides an accurate and cost-effective prediction of ORF in CO2-foam experiments. This approach has the potential to significantly reduce screening costs and time required for CO2-foam injection, making it a more viable carbon utilization and EOR strategy.
Bibliographical noteFunding Information:
This research was supported by the National Research Foundation of Korea (NRF) grants (No. 2022H1D3A2A01096314 ) and the Korea Institute of Energy Technology Evaluation and Planning (KETEP) funded by the Ministry of Trade, Industry & Energy (MOTIE) of the Republic of Korea (No. 20214710100060 and No. 20212010200010 ).
In the petroleum industry and underground gas storage, ML-based models have been utilized for a variety of purposes, such as reserve appraisal in both traditional and non-traditional reservoirs [ 28–35], assessment of natural gas compressibility , prediction of reservoir quality , history matching of simulation models for oil production forecasts in fluvial channels , lithofacies and petrophysical predictions in carbonate reservoirs [39,40], prediction of cumulative oil production in shale formations , microbial enhanced oil recovery , pore pressure estimation using petrophysical well log data , and distribution 3D geostatistical models . ML models have been used to predict oil recovery factors in several studies. Van Si et al.  built an ANN model for predicting the oil recovery factor (ORF) for CO2-EOR. Esene et al.  performed the ORF prediction using ANN, least-squares support vector machine (LSSVM), and gene expression programming (GEP) for a carbonate water-injection process. In their work, the ANN yielded the most accurate prediction performance with an R2 of 0.99. Recently, Larestani et al.  developed a series of ANN models and decision trees to predict the ORF and the net present value of chemical flooding projects, with the CFNN-LM model generating the highest predictive performance. These studies demonstrate the potential of ML models to predict recovery factors and optimize oil recovery processes. Table 1 highlights the employed machine learning approaches for prediction oil recovery performance in EOR projects.This research was supported by the National Research Foundation of Korea (NRF) grants (No. 2022H1D3A2A01096314) and the Korea Institute of Energy Technology Evaluation and Planning (KETEP) funded by the Ministry of Trade, Industry & Energy (MOTIE) of the Republic of Korea (No. 20214710100060 and No. 20212010200010).
© 2023 Elsevier Ltd
- CO-Foam experiments