Digiwell aims at developing new methods, algorithms, and tools for oil production with minimum energy consumption, maximized profit and under uncertain information. This will be achieved by investigating:
- The impact of uncertainty in models
- Data driven models (Machine learning)
- Open-source multiphase flow network models
- Reservoir, network model integration
- Models for optimization of efficiency and emissions while operations
The project aims at pushing the knowledge front describing profit and energy consumption and how it varies with operational parameters and uncertainty. By combining these cost functions with developed tools of physics models of oil-fields both on short term (minutes-hours) and long term (years), this leads to quantitative short term and long term models which are suitable for developing and testing control algorithms.
To maximize short term profit, large scale algorithms for coordinating control for multiple wells under uncertainty will be developed and tested in the modelling tool, while adaptivity of the control algorithms will be tested on the long-term models. Feedback control reduces the detrimental effect of uncertainty, and data reconciliation algorithms will be developed to maximize the information content.
Because the control architecture determines attainable feedback performance, methods will be developed to select actuators and sensors in new fields. New, hybrid methods combining data driven methods with physics models will help reduce the limitations of physics models for poorly understood, new fields.
Digiwell is a Knowledge Building Project financed by the Norwegian Research Council (PETROMAKS2/308817), Equinor and Kongsberg Digital. The project will be executed by research partners USN, SINTEF, UiO, Imperial Colelge of London.