Optimization of drilling parameters and efficiency improvement using gradient boosting decision trees

Abstract This study aims to optimize drilling parameters through Gradient Boosting Decision Trees (GBDT) to improve drilling efficiency. The workpiece materials used in the study primarily consist of medium to hard rock formations, and the measured output is the Rate of Penetration (ROP) in meters per minute (m/min). First, historical data from actual drilling operations were collected, including key parameters such as weight on bit, rotary speed, standpipe pressure, hook load, and inlet flow rate. Next, a GBDT model was built, and hyperparameter tuning was performed using cross-validation to optimize parameters such as learning rate, tree depth, and the number of weak learners, improving the model’s prediction performance. The model’s prediction accuracy was evaluated using indicators such as Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) and compared with traditional regression models to verify the advantages of the GBDT model in ROP prediction. The results show that: (1) At sampling points 0 meters and 10 m, the GBDT model predicted ROP values of 0.24174 m min −1 and 0.32231 m min −1 , respectively, both of which were highly accurate and clearly superior to the predictions made by other models. (2) In the analysis of performance metrics for ROP prediction, the R 2 value reflects the model’s ability to explain data variability. The GBDT model showed the best performance with an R 2 value of 0.90, indicating its effectiveness in capturing trends in the data. These findings suggest that GBDT not only helps optimize relevant parameters in drilling operations but also provides data support for achieving sustainable energy development. This study offers a practical solution for improving drilling efficiency and reducing costs, highlighting the significant application potential of modern data science technologies in the oil and gas industry.