INCORPORATING ROCK BEHAVIOR IN REAL-TIME DRILLING ANALYSIS

Abstract

The primary objective of drilling is to access subsurface hydrocarbon reservoirs and establish a pathway for transporting hydrocarbon fluids to the surface. One of the key goals during drilling is to improve the rate of penetration (ROP) while avoiding problems that cause non-productive time (NPT). To manage overall well costs effectively and ensure successful drilling to the target zone, it is essential to comprehend and foresee these drilling challenges, identify their root causes, and plan for suitable solutions in advance.In previous studies, drilling efficiency has been analyzed to improve ROP and reduce costs. Mechanical Specific Energy (MSE) has frequently served as a tool for evaluating drilling efficiency in real time, leveraging surface drilling parameters due to their availability. However, MSE does not account for variations in rock strength across different lithologies, which directly affect drilling efficiency. No existing studies have incorporated rock strength at the bit into real-time MSE calculations, nor have they effectively addressed the occurrence of drilling dysfunctions like vibrations at the bit-rock interface, which make rock type determination essential. Consequently, a significant gap exists in optimizing real-time drilling efficiency by integrating rock strength at the bit. This study addresses these gaps by developing a real-time drilling optimization system that incorporates rock strength, improving MSE reliability, and minimizing drilling costs. To address these gaps, the following tasks have been performed: predict compressional wave velocity and lithology at the bit using drilling and petrophysical parameters as input variables, estimate rock strength at the bit using established correlations, normalize the MSE with estimated rock strength values, integrate the rock strength in hybrid ROP model and evaluate drilling efficiency using both MSE and hybrid ROP models. This study used a dataset of five wells from the Norwegian continental shelf to develop and validate the model. To perform model development and validation, various traditional machine learning (ML) models were developed. For prediction of compressional wave velocity, four regression models were evaluated, with results showing that the Random Forest (RF) regression model achieved the highest accuracy, with an R2 of 98%. For prediction of lithology at the bit, five supervised learning models were studied, with the RF classifier showing the best accuracy at 86%. Using these predictions, uniaxial compressive strength (UCS) and confined compressive strength (CCS) were calculated, allowing for MSE normalization with CCS values at each depth interval. The hybrid ROP model also utilized the estimated CCS value at the bit to generate an estimated ROP graph as a tool for measuring drilling efficiency. This hybrid ROP model was developed, and sensitivity analysis was performed using operational and bit design parameters. The sensitivity analysis indicated that parameters such as CCS, WOB, RPM, flow rate, bit diameter, and bit nozzle size had the most significant impact on ROP. Additionally, the sensitivity analysis for bit-rock interaction models, using hard rock geothermal well data, demonstrated that stick-slip severity was influenced by rock properties and operational parameters. The analyses revealed that parameters such as UCS, number of blades, WOB, RPM, and bit diameter have a significant impact on stick-slip vibration, emphasizing the importance of fine-tuning operational parameters to improve ROP. This real-time drilling optimization system shows that MSE and hybrid ROP models, when integrated with at the bit rock strength, can precisely quantify drilling efficiency. This integrated approach aims to deepen the understanding of dynamic interactions between drill bits, formations, and drilling conditions, ultimately identifying more effective and efficient drilling practices.