Data-Driven Optimization and Machine Learning-Powered Automation for Intervention Opportunity Management in Oil Wells

Abstract

The document presents a data science workflow that leverages machine learning (ML) and artificial intelligence (AI) to promote innovative methodologies for automating and optimizing the selection of wells as workover candidates. It integrates multiple disciplines, including production, engineering, reservoir, and geology. The digital tool identifies, classifies, and ranks opportunities based on incremental production and profitability, thereby maximizing asset field development plans, improving decision-making efficiency, and ensuring strategic alignment. Focusing on oil-producing wells and waterflooding patterns, the workflow aims to oversee production forecasting, minimize downtime, and provide robust recommendations. The process integrates engineering expertise, economic analysis, and optimization strategies by leveraging diverse datasets, including production and injection rates, pressures, petrophysical data, well coordinates, well history and intervention events, well schematics, and economic parameters to calculate incremental net present value (NPV) for each well opportunity. Data was migrated to a collaborative digital platform to facilitate the development of back-end solutions and seamless integration with various platforms or web applications. This incorporates ML models to predict fluid production and water-cut (WC) estimation. The workflow prioritizes intervention candidates based on production potential and NPV, enabling efficient, consistent, and data driven decision-making by engineers. The solution successfully identified and ranked over 300 opportunities across the field. It effectively highlighted the top wells for intervention, providing detailed remedial proposals, decline curves analyses, and production gain estimates. This capability supports asset development plans that can be tailored to various scheduling horizons—weekly, quarterly, annually, or on-demand. The digital tool also introduced innovative features, including filtering, sorting, and categorization of opportunities based on the field hierarchy and the type of financial commitment required: capital expenditure (CAPEX) or operational expenditure (OPEX). These functionalities closely align with the project's defined work scopes and other specific operational attributes. During the second half of 2024, the tool underwent rigorous testing to validate the recommended candidates. Notably, wells suggested by the tool for stimulation were also chosen by the engineering team using traditional methodologies. Similarly, candidates identified for zone changes matched those selected by engineers, further validating the tool's effectiveness. The ML models used for production forecasting achieved a Coefficient of Determination or R-squared (R2) score of 0.9192, demonstrating the robustness, predictability, and reliability of the AI-driven model. Overall, the results confirm that the system enables timely, data-driven decision-making, optimizing well selection, and maximizing operational benefits.