• Onzaca, Christian
• Severino, Sabatino
• Raimondi, Lorenzo
• Vignati, Emanuele
• Mariotti, Pamela
• Vimercati, Silvia

Abstract

<jats:title>Abstract</jats:title><jats:p>Reservoir Management requires the definition of a strategy that analyzes the risk assessment to identify surveillance activities and corrective measures. In Eni, a best practice has been established to regulate this strategy through the Reservoir Management Plan (RMP). This paper will present the application of Natural Language Processing (NLP) technologies to improve the definition of RMP. Results will be discussed, highlighting benefits and advantages in extracting cross-document information and running comparisons between different RMPs.</jats:p><jats:p>Natural Language Processing refers to the branch of computer science concerned with the ability to understand text and spoken words, like human beings. NLP algorithms have been applied to RMPs to pull structured information adding useful numerical data. The main strategy for managing unstructured data, re-elaborating information upon specific request and implementing cross-document queries are discussed. Hundreds of RMP documents have been collected, hoarding extensive information in compartmentalized storages. NLP algorithms can be applied to unlock this hidden potential by capitalizing lessons learnt between Business Units and partaking the experiences acquired in similar assets.</jats:p><jats:p>The information extracted from unstructured data includes insights on well and reservoir surveillance activities and corrective measures in different assets. A tool has been developed to enable a rapid screening of the key parameters in different assets, highlighting how similar risks can be mitigated. The tool analyzes documents and forms, looking for data and relationships among entities in the text. In particular, the NLP model extracts tables, table cells, and the items within and returns JSON formatted results.</jats:p><jats:p>This task is performed by using open-source NLP libraries and custom logic, in order to adapt common algorithms to RMP application. The procedure is orchestrated by a cloud based ETL (Extract Transform and Load) and data integration service to create workflows for moving and transforming data.</jats:p><jats:p>The workflow is scheduled and executed for each RMP document collected. Insights are collected and presented in a dashboard. Two main applications will be discussed to present the additional value brought by NLP algorithms in Reservoir Management. The mitigation strategies and relevant lessons learnt from analogues can be efficiently collected, analyzed and integrated in the planning and scheduling of reservoir management activities.</jats:p><jats:p>To the authors’ knowledge, this paper presents the first successful implementation of NLP techniques to Reservoir Management. Natural Language Processing algorithms allow unstructured data to be searched, organized and mined, highlighting the strength of combining data and leveraging the main insights without having to read through all the RMP documents.</jats:p>

Topics

• impedance spectroscopy
• strength