Data-Driven Prediction of D-Exponent in Drilling Operations Using XGBoost and LightGBM Algorithms
Keywords:
Machine learning, Artificial Intelligence, Extreme gradient boosting, LightGBMAbstract
Accurate prediction of key parameters in drilling operationsparticularly the D-Exponent plays a crucial role in optimizing processes and reducing operational costs. In this study, real-time drilling data from two directional wells. including parameters such as Weight on Bit (WOB), Equivalent Circulating Density (ECD), Rotary Speed (RPM), and other relevant variables were used to evaluate the performance of two machine learning algorithms: XGBoost and LightGBM. To train and evaluate the models, 80% of the dataset was used for training and the remaining 20% for testing. The results demonstrated that both models are capable of accurately predicting the D-Exponent value. Specifically, the XGBoost model achieved an accuracy of 98% with a Mean Absolute Error (MAE) of 0.073, while the LightGBM model reached an accuracy of 97% with an MAE of 0.074. These methods not only enable real-time prediction but also support drilling engineers in making timely and effective decisions to enhance operational efficiency and reduce costs. The findings of this study suggest that integrating artificial intelligence and machine learning into drilling processes can significantly improve productivity, enhance safety, and minimize operational risks in the oil and gas industry.
Downloads
References
[1] Real-Time Artificial Intelligence-Enhanced Machine Learning for Drilling Parameter Prediction.
(2024). SPE Annual Technical Conference and Exhibition (ATCE), Society of Petroleum Engineers.
https://onepetro.org/SPEATCE/proceedings abstract/24ATCE/24ATCE/563683.
[2] Jorden, J. R., & Shirley, O. J. (1966). Application of drilling performance data to overpressure
detection. Journal of Petroleum Technology, 18(11), 1387-1394. https://doi.org/10.2118/1387-G.
[3] Bourgoyne, A. T., Millheim, K. K., Chenevert, M. E., & Young, F. S. (1986). Applied Drilling
Engineering (2nd ed.). Society of Petroleum Engineers.
[4] Wang, Y., Zhang, L., & Li, J. (2024). Pore pressure prediction based on conventional well logs and
machine learning methods. Journal of Petroleum Science and Engineering, 210, 110345.
https://doi.org/10.1016/j.petrol.2024.110345.
[5] Zhou, F., Fan, H., Liu, Y., Ye, Y., Diao, H., Wang, Z., Rached, R., Tu, Y., & Davio, E. (2022).
Application of XGBoost algorithm in rate of penetration prediction with accuracy. Presented at the
-1 -0.5 0 0.5 1 1.5 2 2.5 3 measured depth (ft)
d - exponent
xgboost prediction
xgb_pred measured
-1 -0.5 0 0.5 1 1.5 2 2.5 3
measured depth (ft)
d - exponent
lightgbm prediction
lgbm_pred measured
https://oil.bcnf.ir Page 10
International Petroleum Technology Conference, Riyadh, Saudi Arabia, 21-23 February.
https://doi.org/10.2523/IPTC-22100-MS.
[6] Yu Ling, Wu Tiejun. Ensemble learning: Boosting algorithm overview [J]. pattern recognitionand
artificial intelligence, 2004, 17 (01): 52–59.
[7] Friedman J H 2001 Greedy function approximation: a gradient boosting machine Annals of statistics
1189–1232.
[8] Liu, J.-J., & Liu, J.-C. (2022). Permeability predictions for tight sandstone reservoir using
explainable machine learning and particle swarm optimization. Geofluids, 2022, Article ID 2263329.
https://doi.org/10.1155/2022/2263329.
[9] Yu Ling, Wu Tiejun. Ensemble learning: Boosting algorithm overview [J]. pattern recognitionand
artificial intelligence, 2004, 17 (01): 52–59.
