Key points are not available for this paper at this time.
Abstract Efficient drilling operations necessitate accurate risk assessment to prevent wellbore failure and consequent nonproductive time (NPT). In this work, the authors present a comprehensive strategy to enhance subsurface well planning decisions using artificial intelligence (AI) and machine learning techniques, optimizing the estimation of integrated drilling contracts value, specifically Authorization for Expenditure (AFE) models. In this study, authors address the challenge of quantitative risk assessment during drilling operations, aiming to predict minor impacts that can lead to catastrophic wellbore failure. The primary objective is to develop an AI-driven interactive dashboard that assimilates historical drilling experience to provide superior decision-making for well planning. The approach involves analyzing historical offset well drilling events to identify critical risks leading to NPT. By leveraging the Gower similarity (GS) algorithm, a similarity matrix is established between observations from offset wells and a new drilling contract. Thirteen key features, including section depths, expected rate of penetration (ROP), mud characteristics, mud density, and dog leg severity (DLS), influence the clustering process. The resulting similarity matrix informs an unsupervised hierarchical clustering algorithm, optimized through Silhouette analysis. Subsequently, Monte-Carlo simulation is executed on derived risk categories to yield more precise AFE estimates. This novel approach is validated using a diverse offset well database from 20 countries, with a specific application in a Middle Eastern country. Analyzing data from more than 252 historically drilled wells across five fields, the study predicts drilling risk categories and associated NPT for a targeted well. Visualizations, including interactive charts and maps, illustrate the distribution of risk categories among offset wells. Post-outlier removal, risk categories from 139 offset wells are selected for Monte-Carlo simulations. Predictions are presented in terms of occurrence probabilities and total NPT, promoting a more informed AFE. Collaboration with drilling domain experts and blind tests further corroborate the approaches effectiveness. In this study, the authors pioneer a real-time monitoring methodology for drilling events and risks, harnessing evolving machine learning and AI advancements. By leveraging historical data and expert insights, it successfully improves the cost-effectiveness and safety of drilling practices. The approach stands as a testament to the power of AI in revolutionizing drilling operations.
Hussain et al. (Mon,) studied this question.