Exploring data-driven methods to drive production optimization

Solution Seeker and Lundin are long time partners, working closely together to explore and exploit better data utilization for production optimization purposes.

Solution Seeker acts as a digitalization partner to Lundin in a dedicated workflow designed to make sure opportunities for increased production are seized. We have worked on numerous use cases over the course of the partnership, all centering around leveraging data to improve production.

In Lundin’s view, Solution Seeker’s laser focus on Artificial Intelligence (AI) and Machine Learning (ML) strictly applied to production data is a competitive advantage and one of the main reasons for the collaboration. Lundin has one field under operation; the Edvard Grieg field on the Norwegian Continental Shelf.

“The fact that they are working on and learning from numerous fields globally is another reason,” says Kjartan Berg, a Petroleum Engineering Manager at Lundin.

One of the main topics of the collaboration is the exploration of the physics based versus data-driven approach to rate estimation, and the combination of these methods into what is called Hybrid AI. The goal is to achieve highly accurate, predictive, and low maintenance models with quantified uncertainty. High quality real-time flow rates are of crucial importance to the operation of a petroleum asset, and especially to the optimization of production and recovery.

Solution Seeker provides expertise on data-driven methods, and is a pioneer in offering the market a purely data-driven VFM. ”Succeeding with data-driven VFMs requires addressing some very specific challenges, the main one being little and sparse data to learn from,” says Solution Seeker’s CCO Sheri Shamlou. Solution Seeker has proprietary data-mining technologies tailored to automatically extract exactly the right information from the ever-increasing flow of raw data from the sensors on a field. This often rare, but high quality input data is fed to NeuralCompass, our proprietary ML framework for constructing and training deep neural networks. This framework also allows for incorporating physical truths for Hybrid models, and Transfer Learning for learning across wells.


But well flow rates do not exist in a vacuum. There are numerous steps and workflows both upstream and downstream the well rate estimation task that have to be taken into account to achieve high quality rates that can make a difference in real-time operations. Well testing is one such important workflow. Optimizing well tests is another benefit of the collaboration for Lundin, through the use of Solution Seeker’s Well Test application. “This application has greatly simplified our well testing workflow. We now have a handle on the uncertainty of the test parameters, a systematic overview of all previous tests with easy access to the information and we receive recommendations on which well should be tested next,” says production engineer Christina Berge at Edvard Grieg.

Both companies are excited about the continued partnership, preparing for the connection of tie-ins Solveig and Rolvsnes to the Edvard Grieg facilities later in 2021.