AI based real-time decision making

Erik Bek-Pedersen, Morten Lind, Bjarne André Asheim

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

A real-time decision support solution for control room operators that targets increasing production efficiency by reducing plant upsets and disturbances has been developed. The solution relies on process engineering and plant knowledge combined with AI tools like ontology and rule-based reasoning. The aim is to enhance the basis for rapid and intelligent decision-making and increase human performance when abnormal situations occur. The solution is based on "Multilevel Flow Modeling" (MFM), which is a modeling language that builds on a systematic representation of relations between objectives and functions of plant equipment in a means-ends structure. The aim of MFM in this context is to develop models that allows for reasoning about causes and consequences of events, or upsets in process plants, such as oil and gas production facilities. A key feature of MFM is its compatibility with human cognition and results of MFM reasoning can therefore be communicated to operators in an intuitive way. In their daily life, control room operators are presented with process variables through a SCADA system and alarms to notify them that something is not right. Their task is to maintain the production in in the plant and avoid unnecessary disturbances and shut-downs, for which they often rely on the information they receive through the systems as well as their experience. When abnormal events occur, operators often need to respond fast, in order to avoid plant upsets, however, there are no solutions to support them with situation analysis and root cause identification. The MFM based solution samples plant data and develops "failure trees" in real-time that are presented to panel operators in a simplistic and logical way. The solution has been developed in an industry-academia collaboration, in close collaboration with two operators. Initial online testing has been done on pilot plant basis, whereas the piloting on offshore fields is currently ongoing. The modeling work is done in close collaboration with engineers and operators familiar with the targeted plant facilities, in order to secure that all relevant operational aspects are adequately covered. In parallel, a UX development path is ongoing and is part of the integrated test program. The novelty of the decision support solution as well as knowledge about how it impacts control room operators in their daily work is discussed. The learnings and experience from offshore pilot testing are also presented.

Original languageEnglish
Title of host publicationProceedings of Abu Dhabi International Petroleum Exhibition and Conference 2019
PublisherSociety of Petroleum Engineers
Publication date2019
ISBN (Electronic)978-161399672-0
DOIs
Publication statusPublished - 2019
EventAbu Dhabi International Petroleum Exhibition and Conference 2019 - Abu Dhabi, United Arab Emirates
Duration: 11 Nov 201914 Nov 2019

Conference

ConferenceAbu Dhabi International Petroleum Exhibition and Conference 2019
CountryUnited Arab Emirates
CityAbu Dhabi
Period11/11/201914/11/2019

Fingerprint Dive into the research topics of 'AI based real-time decision making'. Together they form a unique fingerprint.

Cite this