Well Intelligence – Optimised Production

Project Summary

Industry context/Challenge:

Currently, the optimisation of field production uses individual well models, engineering analysis and individual well tests combined to optimise each well individually.  This approach does not optimise the whole field or system to its full potential.  The Well Intelligence technology uses existing production data and machine learning to optimise the whole field.  It has the potential to improve production by 1% across the whole field.

Project overview:

The Well Intelligence technology was trialled on West Brae, the biggest field in Marathon Brae Field, typically producing around 8000bbl/day.  It is a black oil field with minimal gas production consisting of 8 production wells of varying performance with no water injection.

Testing was conducted over a 4 week period with only 5 usable test points generated.  Testing on a live operating asset proved challenging as steady state periods of production can be infrequent.  Following a review, Intelligent Plant agreed to retrospectively look at the production changes according to historical data over the last two years.  Even over that period, there were only 12 usable test points generated, demonstrating the challenge.

Analysing these verification results it can be concluded that the Well Intelligence application predominantly predicts the production values accordingly although due to the small quantity of test data, it is difficult to create a clear trend of prediction from Well Intelligence.  The test figures suggest that West Brae is fairly well optimised and the application may be capable of delivering minor production improvements.  However, these minor improvements may struggle to be realised due to the stability of the operating asset and other unrelated activities which cause a much larger impact on West Brae’s overall production.

In summary, the Well Intelligence application has succeeded in analysing historical data and recommending valve movements to optimise production. However, for West Brae due to the stability of its production and the fact it has a MPFM the application does not appear to add significant value to Marathon.

Industry value:

£26.2m with 10% uptake

Lessons learned:

• Measurement/Verification – Measuring this small, <1% uplift in amongst daily production activities on a life asset proved challenging as steady state periods of production can be infrequent.
• Automation: automation of measurement production improvement, model accuracy and improvement is critical in delivering clear, quantifiable gains
• Cultural shift: this is a significant challenge with AI as humans must trust the AI output to follow the advice given.  This is particularly evident in well control where live changes were challenging to perform
• Time spent with users: this is key in communicating change, sharing benefits and getting feedback on the technology and overall project
• Future deployment: of the technology to more suitable assets/fields is required to accurately measure the model accuracy and potential gains through the use of the technology.  A useful future output would be a clear statement of which assets/fields are suitable for this technology.