Modelling rate of penetration in drilling operations using RBF, MLP, LSSVM, and DT models

Area of Science:

• Petroleum Engineering
• Data Science in Energy

Background:

• Drilling cost is a major challenge in the oil and gas industry, significantly influenced by drilling time.
• Optimizing the rate of penetration (ROP) is crucial for reducing drilling duration and associated expenses.

Purpose of the Study:

• To develop and evaluate smart models for predicting the rate of penetration (ROP) to aid in drilling operations planning.
• To compare the performance of various machine learning models and an empirical correlation for ROP prediction.

Main Methods:

• Utilized 5040 real-world data points from a South Iranian oil field.
• Modeled ROP using Radial Basis Function, Decision Tree (DT), Least Square Support Vector Machine (LSSVM), and Multilayer Perceptron (MLP).
• Trained MLP with Bayesian Regularization Algorithm (BRA), Scaled Conjugate Gradient, and Levenberg-Marquardt algorithms; employed Gradient Boosting (GB) for DT.

Main Results:

• The Decision Tree-Gradient Boosting (DT-GB) model achieved the highest accuracy with an R² of 0.977.
• Least Square Support Vector Machine (LSSVM) and Multilayer Perceptron-Bayesian Regularization Algorithm (MLP-BRA) also demonstrated strong performance with R² values of 0.971 and 0.969, respectively.
• Sensitivity analysis indicated that depth and pump pressure are the most influential factors on ROP.

Conclusions:

• The DT-GB model provides a statistically valid and highly accurate method for ROP prediction.
• The developed models and empirical correlation offer valuable tools for optimizing drilling operations and reducing costs.
• Depth and pump pressure are key parameters to consider for ROP enhancement.