Artificial intelligence help creating a more productive control room

By Marianne Bom

A close collaboration between researchers and a small Norwegian start-up leads to a new product, the ‘AlarmTracker’, ready for the control rooms of oil and gas platforms in 2019. The device supports the operator in making the right decisions in abnormal situations. Its objective is to secure a steady production, leading to an expected five percent increase of oil production.

At least two times every ten minutes an alarm sounds in the control room of an oil platform. It is telling the operator that somewhere in the complex system of tubes, valves, pumps, compressors, sensors and software there is a problem that need to be fixed to keep the oil and gas flowing.

Does that sound stressful? Then take into consideration that two alarms every ten minutes represent ‘a good day at the office’, leaving the operator time to reflect and react. When something is really critically wrong you have constant alarms going, and unfortunately, present state-of-the art control systems are not smart enough to tell the operator the root cause of the inferno, nor what to do about it. All it can do is to bombard the person in charge with data on high pressure here, low temperature there, a pump stopping here, and another starting up.

“When things begin to go wrong the alarms begin to sound and you often run into a situation as an operator where you can see the alarm, but you don’t know the underlying cause. Then you try to combat the problem, and maybe you fix it intermediately. We want to identify the root cause of any situation, show the future consequences, and suggest actions of what to do to avoid the situation to escalate,” says Programme Manager Erik Bek-Pedersen, DHRTC.



A result of university-industry collaboration

Erik Bek-Pedersen is part of a university-industry collaboration team of 24 scientists doing research into the monitoring, control and supervisory control of oil and gas production. On the team are also six experts from the Norwegian start-up company Eldor Technology. They collaborate with the academic staff to develop the ‘AlarmTracker’, a software system based on fundamental research in artificial intelligence and placed on top of the control system.

Explained in a simple way, the software solution knows the plant, and when the alarms sound it offloads the operator by suggesting the underlying root-cause and consequences of the disturbance. It does also recommend actions. The software serves as a digital twin of the process and is used by the operator to get advice on what to do. Thus, it is expected to ease the stress of the control room operators and to improve the productivity of the oil platforms.

The research and development is done with funding from the Norwegian Research Council under the Demo2000 program, Eldor Technology, DHRTC, Maersk Oil and ConocoPhillips UK. The modelling and testing of the industrial case is taking place at Aalborg University’s water treatment pilot plant in Esbjerg on a water injection-system, at a Danish Maersk-operated field, and on a gas treatment-system at a ConocoPhillips field off shore UK.

Based on 25 years of research

Eldor will have the first ‘AlarmTracker’ product ready for the market in 2019, says CEO Bjarne André Asheim.

“I expect that by implementing this technology in the control rooms, the oil companies will be able to increase oil production by five percent,” says Bjarne André Asheim, based on research showing current high losses due to unplanned upsets in the process industry. The product is made possible due to more than 25 years of research done at DTU Elektro on human-machine interaction and control rooms. Emeritus Professor Morten Lind’s group has developed the ‘Multilevel Flow Modelling’ which is the core technology in the AlarmTracker. Morten Lind expects future complex industrial systems in the process industry be developed with similar integrated artificial intelligent systems, based on knowledge of process physics and control technology, including also knowledge from cognitive sciences on human problem solving.

“We provide a link from the technical designers to the operators who are going to use the system, and help them to make complex decisions which require understanding of the goals and functions of the process and its operation: What is the cause of the failure, what are the possible consequences and the possible actions which can be taken. Most likely in future there will also be fewer alarms because the built in artificial intelligence will help the operators in avoiding critical situations,” says Emeritus Professor Morten Lind, DTU Electrical Engineering

AAU Energiteknik (Esbjerg) contributes to the research and development with knowledge on advanced monitoring and control systems as well as with experimental facilities. DTU Compute supports the development of the AlarmTracker with their experience on data analytics.

A model of situation management

Management of situations in a complex system – such as alarms in a control room of an oil platform – involves two overall phases of decision-making: situation analysis and action planning and execution. Each phase can be decomposed into subtasks as illustrated. The aim of the ‘AlarmTracker’ is to help operators execute informed decision to keep an optimal oil and gas production.