Machine Learning Models for Predicting Flow Rate for Niger Delta Oil Wells

• Olusiji A. ADEYANJU

  Author

• Joseph O. OLAIDE

  Author

Most choke correlations used to determine oil production rates through chokes are invalid for most fields. Due to the complexity of multiphase flow behaviour, varying process conditions of oil wells, and limited data used to develop these correlations, they do not accurately predict oil flow rates in many of the Nigerian oil wells. This study presents an analytical method for improved oil flow rate estimation in Nigerian oil wells employing machine learning using the typical Gilbert choke input parameters (flowing tubing head pressure, choke size, and gas-liquid ratio) with the addition of flowing well temperature and basic sediment and water content. Six non-linear machine learning models were developed to estimate oil flow rate. These are: CatBoost, TabNet, Random Forest, XGBoost, Support vector machine with radial basis function kernel, and Gaussian process regression with radial basis function kernel. All six models outperformed existing choke correlations using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute percentage error (MAPE) evaluation metrics, with CatBoost and Random Forest returning the best performance. The CatBoost model achieved an R² of 97%, an RMSE of 353 STBD, and a MAPE of 7.81%, while the Random Forest model achieved an R² of 97%, an RMSE of 369 STBD, and a MAPE of 8.55%. A parameter importance and sensitivity analyses showed that basic sediment and water content and choke size have the highest impact on the oil production rate determination. A consistent negative trend was observed in the sensitivity analysis for the basic sediment and water parameters, an indication of the need to minimize basic sediment and water levels for optimal oil production estimation. The developed model will be of significant assistance to petroleum industry operators in the Niger Delta region of Nigeria for quick effective estimates of oil flow rates.

• FJET_21_32_32

Copyright (c) 2026 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.