headings

Accord & RGU Knowledge Transfer Partnership:

To research, develop, evaluate and implement an innovative Bayesian based data reconciliation and Gross Error Detection (GED) approach for hydrocarbon allocation and attribution using a cloud-based software approach.

What is a Knowledge Transfer Partnership?

In 2019, Accord entered a collaboration with Robert Gordon University and Innovate UK through a Knowledge Transfer Partnership (KTP).  This is a government initiative to drive productivity and economic growth by supporting businesses to develop and realise potential of new ideas.  At the centre of this partnership is the KTP Associate, an early career graduate recruited to deliver and steer commercialization of a product.

ktp2

The aim of this KTP is to produce commercial software that can help clients identify measurement errors and/or value leakage in the hydrocarbon allocation process. 

What makes this KTP exciting and novel?

Uniquely, the software will use historical data to inform the validity of current measurement quality and to glean useful information from dislocated yet related measurement points.  It will be informative yet simple to use and easy to integrate or interface with existing client systems. Established data reconciliation techniques use mathematical methods, such as least-squares methods, to adjust the measurements and achieve a mass or energy balance (Figure 1).  Those data points that are adjusted more than an expected amount are flagged up as potential errors for the hydrocarbon accountant to investigate further.

fig1

Conventional approaches only look at a snapshot in time but there are patterns in data that measurement engineers and hydrocarbon accountants can see that give them a feel of how the system or a particular meter is performing over time.   

How can abstract empirical knowledge be integrated into an allocation system in a scientifically robust way?  

By using historical data and Bayesian networks and sophisticated testing methods, it is possible to develop a model that learns how different measurements are related even in variable conditions.  For example, as illustrated in Figure 2, if a separator temperature T1 goes up then it expects an increase in fuel gas F2, but it will also know that if well F3 is flowing then that combination may have a different expected outcome and therefore flag potential issue with a measurement.  

fig2

What progress has been made so far?

This past year, KTP Associate Daniel Dobos, has been working with RGU Knowledge Base supervisors, Professor John McCall and Research Fellow Dr Thanh Nguyen of the School of Computing to research and develop novel methods for identifying gross errors in networks.  Currently the team is testing the software with new analytical methods to identify errors and anomalies in measurements in a proxy offshore installation and developing a demo program using test data.   Concurrently, Daniel has been gaining domain knowledge at Accord, supported by Technical Director Phil Stockton, Dr Allan Wilson and Dr Helen Corbett. 

The next steps are to take a proof-of-concept model and test using real data and bring the software to market.  The partnership is now seeking interested parties to collaborate on this next phase.   This would involve sharing data and any findings under a confidentiality agreement and participating in a small number of workshops to exchange feedback on progress and direction relating to the following:

  • Top level specification
  • Proof of Concept Studies
  • Workflow Development
  • System Refinement

Any insights from the research would be made readily available to the contributors.   This is a great opportunity for interested parties, who see the potential value in such a product, to feed into its development from an end user perspective.   

How will it benefit contributors?

Potential benefits we see for contributors are:

  • The ability to assess allocation data quality;
  • Potential for early fault detection helping prevent mismeasurements and corrective action arising from misallocations.

What does it involve?

To develop the software, we would request historical data sets for key areas of allocation systems. The data should cover quantity and quality measurements across the process, such as

  • Flow measurements or estimates;
  • Feed compositions; and,
  • Process conditions.

We would expect data to be subject to Confidentiality Agreements with each contributor.

Ultimately, we would offer support to contributors in a trial of the product, monitoring and analyzing results, and providing feedback.

As the KTP is a collaboration with RGU, it is intended to publish research results in academic papers and/or industry workshops.  Accord would seek permission before publishing any data and results related to each contributor.

If you wish to engage with Accord and RGU on this endeavour, please feel free to contact us at enquiries@accord-esl.com