Machine Learning for Enhanced Production Optimisation and Management

Abstract This study introduces an innovative framework for harnessing Machine Learning (ML) within production engineering. The objective is to offer engineers a comprehensive framework for utilising ML-based modelling across core production engineering tasks to elevate operational efficiency, reduce expenses, and advance sustainability efforts. Methodology: 1. Build a Model Register: Catalogue existing modelling approaches used across the production engineering function, detailing inputs, outputs, assumptions, dependencies and associated decision-making. 2. Data Review: Assessment of data accuracy, completeness, and interdependencies. 3. Task Mapping: Mapping current tasks to existing approaches with the addition of an ‘ML Opportunity’ column marked as enhance, augment, or replace. 4. ML Opportunity Matrix: A matrix calculation to prioritise ML applications. 5. Implementation Strategy: Strategy for incorporating ML into existing workflows. Results: This study presents a framework for incorporating Machine Learning (ML) into production engineering. By systematically cataloguing existing models and assessing data processing and utilisation across various decision-making processes, we identified pivotal areas where ML can deliver value, such as in enhanced production optimisation and control, turndown management, and maintenance and intervention planning. The creation of an "ML Opportunity Matrix" prioritised the selection of ML techniques for specific engineering tasks, pinpointing supervised learning as a critical tool for enhanced estimation and predictability in the management of dynamic, integrated production systems. The proposed framework presents a systematic approach for prioritising ML-based modelling for specific production engineering tasks as well as identifying the opportunity for alignment and collaboration with associated disciplines such as reservoir engineering, drilling and completions engineering, and control systems engineering. ML-based modelling is recommended as a significant enhancement to traditional simulation and physics-based modelling techniques, for assets that have collected operational data, to enable more efficient, data-driven decisions in production system optimisation and management. Benefits: This paper introduces a novel framework for production engineering teams to be able to systematically identify, evaluate, and integrate Machine Learning (ML) solutions into core production engineering tasks. This study equips engineers with a framework for understanding the modern digital tools available to develop strategies that not only optimise operations but also yield cost reductions and environmental improvements.