Utilizing Machine Learning to Unlock Subsurface Production Potential

Abstract One of the ways to unlock subsurface production potential in a brown field is via Behind Casing Opportunities (BCO). The conventional method of assessing remaining hydrocarbon potential by deploying logging tools downhole may be challenging in some fields, especially those with highly deviated or horizontal wells. A novel BCO identification process discussed in this paper leverages Machine Learning (ML) model within an Artificial Intelligence (AI) framework that integrates petrophysical data, dynamic production attributes, and production potential using automated calibrated models to identify unperforated zones in existing wells and spot those with significant production potential. The methodology involves defining target variables, selecting input features, choosing a modeling approach, and characterizing uncertainties. The datasets encompass well construction details, reservoir petrophysics properties, dynamic production attributes, and the production history of neighboring wells. Employing sophisticated algorithms like Artificial Neural Networks (ANNs) and random forests, the model undergoes dynamic training to refine its accuracy and reliability. These optimal settings are then preserved for final deployment, with a detailed analysis of uncertainties to support informed decision-making. This paper introduces a novel application of an Artificial Intelligence and Machine Learning (AI-ML) framework for BCO identification, providing unparalleled insight for the O&G industry. It features an in-depth validation process, the step-by-step deployment of innovative methodology, and revolutionary strategies to enhance production in challenging wellbore conditions and complex reservoir environments.