Connecting Multilayer Semantic Networks to Data Lakes: The Representation of Data Uncertainty and Quality

Abstract

Summary Drilling oil and gas wells is a complex process involving many disciplines and stakeholders. This process occurs in a context where some pieces of information are unknown, or are often incomplete, erroneous, or at least uncertain. Yet, during drilling engineering and construction of a well, drilling data quality and uncertainty are barely addressed in an auditable and scientific way. Currently, there are few or no placeholders in engineering and operational databases to document uncertainty and its propagation. The Society of Petroleum Engineers (SPE) has formed a cross-disciplinary technical subcommittee to investigate how to describe and propagate drilling data quality and uncertainty. The subcommittee is a cooperation between the drilling system automation, wellbore positioning, and drilling uncertainty prediction technical sections. As the topic is vast and complex, a systematic method was adopted, where multiple user stories or pain points were generated and ranked with the most compelling user story analyzed in detail. From this approach, a series of multidisciplinary workflows (drilling data generators) can now be captured and described in terms of data quality and propagation of uncertainty. The paper presents details of one user story focused on capturing the description of the quality and uncertainty of depth measurements. Multiple use cases have been extracted from this single user story exemplifying how multiple stakeholders and disciplines manage, communicate, and understand the notion of wellbore depth and its relative uncertainty. Current data stores have the main objective of recording the results of processes but have very limited capabilities to store how the interdisciplinary processes generated and cross-related these results. The study explores the use of semantic networks to capture those multidisciplinary data relationships. A minimum vocabulary has been created using just a few tens of concepts that has sufficient expressiveness to describe all the extracted use cases, showing that the semantic network method has the potential to describe a broad range of complex drilling-related processes. The study also demonstrates that use of a multilayered graph, employing other notions that do not expressly refer to the processes that generated the data, can capture the description of how uncertainty propagates between each of those concepts.