Applying Artificial Neural Networks with Bourgoyne and Young Model to Predict Rate of Penetration in Al-Garraf Oil Field

Abstract

Estimating the rate of penetration (ROP) for oil well drilling is essential for cost-effective and safe drilling operations; drilling companies have been aiming for ROP estimation since the industry's first decade. To achieve this goal, among numerous models, the Burgoyne and Young model (BYM), an equation-based approach, was developed and widely used to predict the ROP based on multiple linear regression (MLR). Artificial Neural Network (ANN) is a machine learning technique that analyzes drilling data and makes ROP predictions. Many studies have been conducted worldwide to employ BYM, and others have aimed to improve it for different circumstances. ANN has also shown effectiveness in these fields. This study proposes an approach that combines the benefits of Feedforward Neural Networks (FNN) from the ANN model and BYM equations to enhance ROP prediction. Integrating BYM with FNN leverages the equation-based model and harnesses the power and efficiency of machine learning. These results significantly improved accuracy and efficiency in predicting ROP in oil wells. ROP modeling input parameters include total measured and true vertical depth, Weight on Bit, Rotation Per Minute, standpipe pressure, pump flow, equivalent mud weight, bit size, nozzle size, formation pressure, and bit jet impact force, which are recalculated by BYM equations and fed to both MLR and FNN. When tested on real-time data from Al-Garraf oil field, the outcomes demonstrate higher R2, lower residuals, and zero P-value compared to MLR, which validate the approach accuracy and provide precise ROP prediction in future drilling plans.

Keywords

• Artificial neural networks
• Bourgoyne and young model
• Multiple linear regression
• Rate of penetration
• Al-Garraf oil field