Background:

For an undeveloped oil and gas discovery in the North Sea, different alternatives were considered to maximise productivity from the low permeability reservoir dominating the field. The interval of interest comprises a thick stack of amalgamated sands, interbedded with thin-bedded sands, mud rich sands and shales. Major uncertainty in the reservoir is associated with the level of connectivity, along with the position of the fluid contact linked to potential compartmentalisation. In this context, the production performance for a horizontal well with multiple hydraulic fractures needed to be evaluated using a numerical reservoir simulator (Figure 1). 

Additionally, multiple uncertainties associated with the fracture design also had to be considered in the study to analyse the risk associated with the proposed development strategy and to quantify the additional potential for the field within a short deadline.

Several alternatives are available in simulation software to model hydraulic fractures, ranging from the most simplistic approach where a change in skin factor is used to more advanced solutions where local grid refinement is applied to describe both geometry and fracture properties explicitly. However, this is well known to be non-trivial task, with the more complex scenarios requiring precise definition of the geometry in the model that honours fracture dimensions along with a proper scaling of fracture properties to preserve the design conductivity. This task becomes more challenging when the base grid orthogonality and alignment deviates from basic, theoretical conditions. 

In tNavigator, the available options include modelling HFs as virtual connections or nested local grid refinement, with uniform, logarithmic or unstructured refinement types. Different objects can be used to define the fractures and distribute the properties, including polygons, arithmetic statements, and input tables. Furthermore, geometries and properties can be imported from third party software point sets or directly generated in tNavigator using the fully coupled embedded fracture simulator, where geomechanical properties and fracturing operational conditions are taken into account to model fractures propagation directly in the grid model. 

For this project, an unstructured nested LGR approach was used to incorporate the hydraulic fractures into the model, honouring the design specifications (Figure 2). The use of automated workflows in tNavigator’s pre-processor allowed to record the sequence of steps required to update different parameters affecting fractures’ geometry and attributes, providing a robust and efficient methodology for optimising the model and testing alternative designs. Finally, a sensitivity study to evaluate main uncertainties was carried out, relying on the seamless integration between pre-processor automated workflows and the AHM/Uncertainty Quantification engine.

Implemented Workflow:

A sector model of approximately 13400 x 11600 ft was created from the reference full field dynamic model and transferred for pre-processing. Fracture definition using the unstructured LGR option was implemented, with user-defined polygons as the reference for the base fracture path, whilst fracture template objects were utilised to define fracture properties including permeability, width, height, etc. All key fracture design parameters were obtained after performing an independent study for modelling hydraulic fracturing propagation using geomechanical data from the field. A near rectangular shape fracture (Figure 3) with no propagation of Stimulated Rock Volume (SRV) away from the planar fracture was assumed. The setup of multistage fractures along the wellbore was implemented using Fracture Stage objects, allowing the quick definition of a sequence of fractures based on a common template.

One of the most critical steps of the project involved the definition of the optimal LGR resolution used to model the fractures. A sensitivity test was performed on LGR resolution for 5, 6, 7 and 8 nested LGRs, evaluating model accuracy and runtime. For scenarios with 5, 6 and 7 nested LGRs, a progressive decrease in fluid production was observed, associated with overestimating effects caused by non-optimal model resolution, particularly noticeable for the 5 LGRs case. Despite simulation time increases accordingly, the use fully parallel simulation routines both in CPU and GPU allowed to maintain the runtime under reasonable limits. Negligible change was observed in the simulation results when increasing from 7 to 8 nested LGRs, while calculation time increased by a factor of 2.5. The suggested refinement level for the model was 7 nested LGRs.

Automated workflows were defined to enable the generation of discrete cases for evaluation of deterministic scenarios, along with the automated probabilistic study. Using the Latin Hypercube sampling algorithm, different uncertainties were implemented, namely fracture height, fracture half-length, fracture permeability, number of fracture stages along the wellbore and fracture angle relative to the wellbore in the horizontal plane. The ability to leverage the maximum computing power available and run the simulations in a fully parallel way was critical to complete the study involving high resolution well models within the defined framework.

Conclusion:

The approach allowed the fractures to be easily incorporated into the well model and evaluate their impact on the recovery from the field. The study suggested that all parameters except for fracture angle show positive correlation with well total productivity– e.g. the higher the amount of fracture stages the higher the incremental cumulative liquid production. However, the same parameters also present positive correlation with cumulative water production and negative correlation with cumulative oil production (figure 4).

The simulation study highlighted the importance of controlling fracture growth and ensuring that a minimum stand off from the oil-water contact is maintained to prevent excessive water production. The geomechanical work done on this field indicates that the stress contrast between the different formations is almost non-existent, therefore nothing would stop the fractures growing vertically into the aquifer and/or the overburden section. Independent hydraulic fracture modelling work carried out for this field to guide the simulation work demonstrated that even the smallest fractures could have a fracture height comparable to the total hydrocarbon column of this field. This presents a major limitation for the use of this technology (Figure 3).

The study was key to identify that despite a significant increase in well productivity index and equivalent oil production compared to a base horizontal well, the excessive increment in water production represents a major limitation for the application of this type of completion for the well. A second evaluation phase incorporating alternative completion technologies such as fishbones and multilateral wells is planned.