Drilling Data Based Approach for Real-Time Rate of Penetration Prediction for Motorized Bottom Hole Assembly Using Artificial Intelligence

Abstract Predicting and optimizing the drilling rate of penetration (ROP) poses a significant challenge due to its dependence on various factors, prompting increased attention towards achieving precise ROP estimations given its direct influence on overall drilling expenses. Among the factors influencing ROP, the driving mechanism of the bottom hole assembly (BHA) plays a pivotal role. Motorized BHAs offer versatile applications beyond directional drilling, including optimization of ROP and mitigation of downhole vibration. While several models have been proposed to forecast ROP for rotary and rotary steerable system BHAs, limited attention has been directed towards motorized BHAs. In this study, a novel artificial intelligence (AI)-based model employing gradient boosting regression (GBR) was developed to predict ROP for motorized BHAs, leveraging surface drilling parameters, mud characteristics, and motor output features. The dataset used for model training, validation, and testing was sourced from six wells spanning two adjacent fields in the Egyptian Western Desert, comprising over 5,800 data points. Mean absolute percentage error (MAPE) served as an evaluation metric for prediction accuracy, while the correlation coefficient (R) quantified the extent of agreement between real and predicted ROP values. Results demonstrated that the GBR model accurately estimated ROP for motorized BHAs, exhibiting a high correlation (R of 0.95) between predicted and real values. The GBR-based model consistently performed well without exhibiting underfitting or overfitting issues. Furthermore, the developed model enables exploration of the impact of different drilling parameters on motorized BHA ROP, thereby facilitating ROP optimization, reduction of open hole exposure duration, and overall drilling cost minimization.