Imagine the time when we are no longer concerned about the digital-transformation issues we face today, such as data availability, security, and many others. This would be the time when companies make the best out of digital infrastructure. In this era, employees would not struggle to collect, gather, and transform the data. Instead, automated algorithms would enable them to quickly leverage information hidden in data sources. We may, however, face other kinds of issues such as ensuring that computers and humans collaborate in the most-efficient way, providing business managers the best decision-making framework. This era might be here sooner than expected.
The upstream industry is not short of innovative ideas where machine-learning algorithms have enabled new concepts since a wide variety of digital tools for fast prototyping and implementation have become available. However, companies need to establish digital operating models that support not only agile innovation but also robust deployment of scaled-up solutions. This will require a multiparty framework for planning resources to support and sustain such applications from idea to reality. Such a framework shall assure that the portfolio of ideas is translated into a coherent applications road map, with clear delivery milestones. Change management is required to embed these new tools in daily business processes, amalgamating with other solutions along with providing user support and managing the evolution of information-technology infrastructure.
As evidence of such progression, the October JPT’s Data Analytics Technology Focus section presents synopses of papers that address common-use cases in petroleum engineering and geosciences.
Paper SPE 196657 shows automating 3D image processing that can significantly improve the quality and speed of classifying the morphology of heterogeneous carbonate rock. The implications for fluid-flow modeling are tremendous.
The objective of paper SPE 196110 is to enable the prediction of the spatial variation in new target wells’ decline type curves for gas production on the basis of existing data of production, completion, and geological parameters through an automated machine-learning approach.
Paper OTC 29815 proposes a machine-learning/artificial-intelligence approach using sequence-mining algorithms for predicting the next drilling operation and classifying it on the basis of textual descriptions, allowing automatic pattern discovery within drilling reports.
As suggestions for additional reading, paper SPE 196011 investigates various deep-learning approaches to predict the long-term well performance on the basis of a moderate duration of early-life well monitoring data.
The variable grid method presented in paper SPE 196094 provides a robust method for quantifying and representing uncertainty in spatial data analyses, simultaneously offering information about the analysis and the associated risks, knowledge of which is critical for decision-making in upstream.
Online library OnePetro already offers a massive collection of new-use cases for machine-learning and artificial-intelligence algorithms applied to typical petroleum engineering and geoscience problems. Our next challenge is to make them part of the day-to-day business processes, transforming the way we work.
