Well modeling using machine learning methods for integrated modeling

.Background. Bottom hole pressure prediction is crucial issue in integrated field modeling. The well is connection element between surface network and reservoir. It must describe the movement of a two-phase fluid in the wellbore. To control the well production, it is necessary to describe the dependence of the bottom hole pressure on the fluid and well parameters. The classical approach is direct calculation by empirical correlation — physical equation constructed from experimental data. It requires large computing power, expert opinion and as a result large time resources.

Aim. This article proposes a new approach to well modeling. Using machine learning model describe the well depend on fluid properties and production parameters.

Materials and methods. The well model was implemented using the “Random Forest” assembly of “Decision Trees” using the gradient boosting technique. The model was tested on synthetic and real data from various fields.

Results. The developed model was tested on synthetic and real field data. The proposed approach outperforms current solutions in terms of speed and prediction score. It also allows to reduce usage of expensive licenses.  In case of enough data the need to create models in simulator is lost.

Conclusions. Due to its high predictive ability, the proposed algorithm will be introduced into production processes as a well model for the needs of integrated asset modeling.

1. Al Shehri F. H. et al. Utilizing Machine Learning Methods to Estimate Flowing Bottom-Hole Pressure in Unconventional Gas Condensate Tight Sand Fractured Wells in Saudi Arabia // SPE Russian Petroleum Technology Conference. — Society of Petroleum Engineers, 2020.

2. Beggs D. H., Brill J. P. A study of two-phase flow in inclined pipes // Journal of Petroleum technology. — 1973. — V. 25. — No. 05. — Pp. 607–617.

3. Duns H., Ros N. C. J. Vertical flow of gas and liquid mixtures in wells // 6th World Petroleum Congress. — OnePetro, 1963. 4. D. I. Ignatov, K. Sinkov, P. Spesivtsev, I. Vrabie, V. Zyuzin, Tree-based ensembles for predicting the bottomhole pressure of oil and gas well flows, in: International Conference on Analysis of Images, Social Networks and Texts, Springer, 2018. — Pp. 221–233.

4. Hagedorn A. R., Brown K. E. Experimental study of pressure gradients occurring during continuous two-phase flow in small-diameter vertical conduits //Journal of Petroleum Technology. — 1965. — V. 17. — No. 04. — Pp. 475–484.

5. Kanin E. A. et al. A predictive model for steady-state multiphase pipe flow: Machine learning on lab data // Journal of Petroleum Science and Engineering. — 2019. — V. 180. — Pp. 727–746.