A DATA-DRIVEN SURROGATE MODELLING FRAMEWORK FOR FIELD-WIDE PRODUCTION OPTIMIZATION IN MATURE OIL FIELDS | IJCT Volume 13 – Issue 3 | IJCT-V13I3P110
International Journal of Computer Techniques
ISSN 2394-2231
Author
Umar Alhaji Mohammed, Abdullahi Usman, Ahmad Usman Ahmad
Abstract
Production optimization in mature oil fields is challenging due to decreasing reservoir performance and the limitations of conventional physics-based simulation tools, which are computationally expensive and inflexible for decision-making. This study proposes a data-driven surrogate modelling framework for field-wide production optimization using machine learning techniques. A multi-well dataset from the Volve oil field, comprising approximately 8,000 daily production records from five wells, was used. The data were pre-processed and enhanced through physics-informed feature engineering, including lagged and cumulative production variables to capture temporal dynamics and reservoir depletion effects. Several regression models were evaluated, including Linear Regression, Random Forest, XGBoost, and Gradient Boosting methods. The Histogram Gradient Boosting model achieved the best performance, with a coefficient of determination (R²) of 0.9995 and low prediction errors. The trained surrogate model was then applied to evaluate multiple operating scenarios involving choke settings and pressure conditions. The results show that the proposed approach achieved a field-wide production improvement of 0.42% compared to baseline operations, outperforming conventional manual optimization. Although the improvement is modest, it is significant for mature fields operating near optimal conditions. The findings demonstrate that data-driven surrogate models can provide efficient and flexible decision-support tools for real-time production optimization while reducing dependence on computationally intensive simulation workflows.
Keywords
Surrogate modelling, production optimization, mature oil fields, machine learning, data-driven modelling, digital oilfield, field-wide optimization
Conclusion
In this study, a machine learning-based surrogate model was used to improve production prediction and optimization in a mature oil field. The results show that the model can learn the relationship between operational variables and production output using historical data. Among the tested models, Histogram Gradient Boosting has performed best and achieved better predictions compared to the linear models.
The study also shows that the surrogate models can be used to test different operating conditions without relying fully on complex physics-based simulators. This makes the process faster and more flexible. The optimization results indicate that the AI-based approach was able to find small additional production gains compared to the manual method.
Another important point is that combining petroleum engineering knowledge with machine learning can improve the model’s performance. Features such as lagged production and cumulative production helped the model to capture real system behaviour. This shows that domain knowledge is still important when applying machine learning in engineering problems.
This study contributes by developing a multi-well modelling approach, testing different machine learning models, and applying the best model for production optimization. It also provides a simple workflow that can be reused in similar oil field problems.
References
[1]O. T. Adewale, S. O. Olatunji, and A. A. Adebayo, “Machine learning approaches for production forecasting and choke optimization in the Niger Delta oil wells,” Journal of Petroleum Science and Engineering, vol. 215, p. 110764, 2026.
[2]T. Ahmed, Reservoir Engineering Handbook, 5th ed. Gulf Professional Publishing, 2019.
[3]S. Ahuja, N. Kumar, and R. Singh, “Integrated production system modelling for optimization of mature oil fields,” Journal of Petroleum Exploration and Production Technology, vol. 12, no. 3, pp. 873-885, 2022.
[4]S. A. Alarifi, “Machine learning-based prediction of multiphase flow through wellhead chokes,” Journal of Petroleum Science and Engineering, vol. 208, p. 109312, 2022.
[5]A. Alkindi and S. Linthorst, “Integrated production system modelling for reservoir management and production optimization,” SPE Production & Operations, vol. 27, no. 2, pp. 128-138, 2012.
[6]A. S. Arief, A. Nugroho, and B. Setiawan, “Feature engineering strategies for oil production prediction using machine learning models,” Energy Reports, vol. 7, pp. 8423-8432, 2021.
[7]O. Bello, S. Adebayo, and T. Ibraheem, “Digital oilfield technologies and data-driven production optimization in upstream petroleum operations,” Energy Informatics, vol. 6, no. 1, pp. 1-16, 2023.
[8]G. E. P. Box, G. M. Jenkins, G. C. Reinsel, and G. M. Ljung, Time Series Analysis: Forecasting and Control, 5th ed. Wiley, 2015.
[9]D. Bukharbayeva, Z. Liu, and Y. Kurniawan, “Surrogate modelling techniques for reservoir and production optimization: A review,” Journal of Petroleum Science and Engineering, vol. 189, p. 107021, 2020.
[10]T. Chen and C. Guestrin, “XGBoost: A scalable tree boosting system,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 785-794.
[11]M. J. Economides, A. D. Hill, and C. Ehlig-Economides, Petroleum Production Systems, 2nd ed. Pearson, 2013.
[12]J. O. Eli, M. A. Adewumi, and B. Ogunyomi, “Integrated production system modelling for improved field performance,” Journal of Petroleum Technology, vol. 65, no. 4, pp. 74-82, 2013.
[13]T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction, 2nd ed. Springer, 2009.
[14]International Energy Agency, World Energy Outlook 2023.
International Energy Agency, 2023.
[15]J. Kim, D. Lee, and S. Park, “Artificial intelligence applications in geoenergy systems: A review of digital oilfield technologies,” Applied Energy, vol. 355, p. 121876, 2025.
[16]M. H. Kutner, C. J. Nachtsheim, J. Neter, and W. Li, Applied Linear Statistical Models, 5th ed. McGraw-Hill/Irwin, 2005.
[17]X. Ma, Y. Zhang, and H. Li, “Machine learning-based production prediction using operational and reservoir variables,” Journal of Natural Gas Science and Engineering, vol. 115, p. 104987, 2025.
[18]S. O. Mepaiyeda, M. Adewumi, and M. O. Onyekonwu, “Production optimization in the Niger Delta mature fields using integrated production system modelling,” SPE Production & Operations, vol. 39, no. 2, pp. 223-237, 2024.
[19]M. K. Ng, R. Hauge, and O. Sævareid, “Oil production forecasting using machine learning: A case study of the Volve field dataset,” Journal of Petroleum Science and Engineering, vol. 196, p. 107733, 2021.
[20]A. Samad, M. Khan, and S. Ahmed, “Machine learning models for oil production prediction using the Volve dataset,” Energy AI, vol. 14, p. 100273, 2025.
[21]
A. Sircar, K. Yadav, and S. Mishra, “Production optimization in offshore oil fields through choke management and machine learning,” Energy Reports, vol. 7, pp. 4583-4595, 2021.
A. Adeyinka, O. Oriola, and O. S. Tomomewo, “An innovative approach for oil well bottomhole pressure forecasting using Kolmogorov-Arnold Neural Networks (KANs): A case study in an offshore oilfield,” Deep Resources Engineering, vol. xx, pp. 100233, 2025.
How to Cite This Paper
Umar Alhaji Mohammed, Abdullahi Usman, Ahmad Usman Ahmad (2026). A DATA-DRIVEN SURROGATE MODELLING FRAMEWORK FOR FIELD-WIDE PRODUCTION OPTIMIZATION IN MATURE OIL FIELDS. International Journal of Computer Techniques, 13(3). ISSN: 2394-2231.
