Background:

The Freya reservoir is located in the Clair Field, roughly 47 miles to the West of Shetland. The Clair Field was discovered in 1977, and Freya discovered a few years later in 1980. The structure is comprised of old Devonian sandstone with a relatively low porosity and matrix permeability. The reservoir has very heavily faulted and fractured sandstones with multiple sets of intersecting complex and rugose faults and fracture zones from the Devonian to Carboniferous age.

Production performance is expected to be affected by controlling factors such as faults, network permeability and wells placement. Exploring the range of uncertainties and their impact on production profiles and produced volumes was key for the project.

Static Workflow:

Data was imported into tNavigator, including Horizons and faults. These were picked on a time converted depth seismic 3D volume. Logs and markers were imported too. The faults initial framework was not suitable for gridding, so the faults were recalculated from pointsets using convex hull to capture their structure while keeping them robust enough for modelling. Some of the faults were planar while many were curving, even horizontal at some places. Therefore, some manual editing was needed to remove some more erroneous fault sticks and create a solid grid build. The faults are thought to be strongly associated with fracture permeability so were required to be included in the model. Editing was carried out to retain structure while ensuring they were not causing artefacts in the grid build. The main fault laid to the western flank of the model, data was cleared from one side so the fault became a structural boundary of the reservoir. 

The initial model defined cell dimensions of 50m x 50m in X and Y, and a variable cell vertical thickness depending on the layer thickness which produced an array of 251,86,124 in x,y and z. When new surfaces from seismic were introduced, the grid was updated, changing layering, applying fixed cell thicknesses within zones and cutting the grid down to the main area of interest, all to reduce errors in simulation caused by the complex faulting. The revised model prevented instabilities in property interpolation and dynamic simulation and cut the computation time down drastically. A distance to fault property was created after the model was complete and was used to create a scalar to apply to the permeability property.

Dynamic Workflow:

Dynamic simulations were run for the Devonian Freya reservoir using two static models, with complementary and consistent outcomes. The dynamic model building was carried out by third party consultants whose approach was to build the dynamic model using the legacy methodology of adding data to a data file, proven completely compatible with the software. Some of the dynamic data was gathered from discovery well 206/10a-1 and some was interpreted as analogue data from observations reported on Clair field (WoS). A total of 3 horizontal wells were added along the western flank of the model; 2 producers of 1.5 km each and an injector of 2.5 km. 

There are few uncertainties in the Freya model (Model 1); volume in place and the aquifer support being the most important ones. Uncertainty was added in the both the static and dynamic models including different combinations of wells and properties. Uncertainty was also added to rock properties such as porosity and permeability to determine their effect on production rates. 

The different scenarios (base case, low and high case) were simulated to understand the expected uncertainty based mainly in fluid properties ranges (PVT). It was found that the well placement played a crucial role in the forecasted rates as the permeability contrast between the naturally fracture network and the matrix is large. 

Total simulation times for Model 2 improved after faults, surfaces and gridding method were revised on the static model. Well placement on Model 1 was modified as results showed the field being more heavily fractured in its southern region where one of the producers was initially placed. Several well placement sensitivities to capture impact from fracture permeability, changes in porosity, different oil viscosity and oil densities and a range of oil water contacts were run before identifying the most optimal well trajectory, critical for oil and water forecasted rates. 

The fast simulation speed meant the heterogeneity could be captured in the reservoir while still running at an efficient time in order to complete the project. 

Integrated Workflow

Outcome:

The integration of the software with powerful hardware enabled the running of different realisations to find the optimal set of parameters needed for a successful field development. Running realisations using a range on the static properties meant wells which would not be commercially viable were identified. The synergy between static and dynamic modelling was leveraged though automated workflows, allowing adjustments of static model parameters that could be improved during simulations to stabilise results and optimise run times.

The resultant model was able to capture the complex faulting of the reservoir at a fine geological scale (1-1.5m vertical layers) to demonstrate the effect the structural geology had on production rates in the area. As a result, it was concluded that the location of the current exploration well was not optimal and was unlikely to flow oil to surface from that position. 

This Greenfield project was feasible by proposing a workflow that included all aspects of the reservoir geology powered by a solution that allows process updates from static to dynamic modelling. Compared to the traditional workflow of juggling separate software’s together to achieve a result, this approach of using a fully parallel software allows the running of multiple calculations, meaning no time is lost waiting for workflows to be computed. The full utilisation of hardware was key to the creation of multiple realisations of a complex faulted model without disruptive long run times.