You have access to this full article to experience the outstanding content available to SPE members and JPT subscribers.
Although improvement in hydraulic-fracture properties and infill drilling remains the focus of recovery enhancement from the Bakken, low oil recoveries and steep initial decline rates are experienced in primary-recovery operations, even after application of multifractured-horizontal-well technology. Therefore, many pilots have been executed to determine the viability of waterflooding for maintaining oil rates and improving recoveries through reservoir-pressure maintenance and sweep-efficiency enhancement. This paper presents the performance results from one of the waterflood pilots in the Viewfield Bakken.
A section of the Bakken reservoir (the geology of which is described in detail in the complete paper) deemed to be representative of the waterflood performance in Viewfield is considered for modeling. This section has been developed by use of multifractured horizontal wells completed in the Middle Bakken (main target reservoir) with a well spacing of 200 m (eight wells per section, named A through H). All eight wells started oil production within a similar time frame, and, after approximately 1 year of production, every other well was converted to a water injector.
Reservoir-Fluid Model. Conventional pressure/volume/temperature (PVT) analysis was conducted by a commercial laboratory on 12 surface-separator oil and gas samples. Recombination of fluids at reservoir temperature (156.2°F) yields a final gas/oil ratio of 810 scf/STB. Subsequently, a series of constant-composition-expansion and differential-liberation tests was conducted on the recombined fluid to determine oil-saturation pressure, oil-formation-volume factor, oil density, and oil and gas viscosity as a function of pressure. The Peng-Robinson equation of state and modified Pedersen viscosity correlation were tuned to replicate the PVT properties of oil and gas as a function of pressure.
Reservoir Grid Model. On the basis of the well tops and reservoir net-pay values, reservoir structure for the study area was generated. It is known that the minimum horizontal stress is aligned in the northwest direction and at approximately 50° with respect to the east/west horizon. Therefore, reservoir gridding is rotated at this angle to mimic the hydraulic-fracture orientation along the horizontal-well laterals. Grid size in the horizontal direction is 65×65 ft, and the total thickness of the reservoir is approximately 28 ft, which is divided into nine layers of equal thickness.
High-resolution (inch-by-inch) measured profile permeability and porosity values of Bakken core extracted from the study area were used to model the vertical variations in these properties. Because the measurement resolution cannot be reflected in the vertical grid resolution of the simulation model, these properties were averaged and scaled up to the grid resolution (approximately 3 ft). It is known that the initial pressure in the Viewfield Bakken is approximately 2,600 psi, which is well above the oil-saturation pressure of 1,480 psi, so the only remaining parameter to evaluate is the initial water saturation.
The Bakken reservoir in the study area is overlain by the Lodgepole aquifer. To account for this, two layers are added to the top of the model. Unfortunately, only sparse data are available for the rock properties of the aquifer, so an assumption has been made that those properties are similar to those of the target reservoir, with the exception of a water saturation of 100%.
Horizontal wells in the pool have been typically completed with liners and therefore are in contact with the reservoir through hydraulic-fractured stages only. In the study area, six wells out of eight have been fractured at 230‑ft spacing between 16 stages, while the spacing of the other two wells (Wells A and C) is increased to 460 ft between eight stages. It was assumed that the fracture planes cover the entire vertical height of the model, including the top aquifer layers (this assumption is based on the water production of the wells during the primary-recovery period). The fracture half-length is correlated with injection tonnage used in the fracture operations and ranges between 295 and 360 ft.
To capture the physics of the fluid flow within the fracture pathways, logarithmically spaced local grid refinement is used (Fig. 1 above). With this arrangement, the middle fine grid designated for the fracture plane is assigned a high permeability and the grid sizes increase on both sides of the fracture plane symmetrically.
Once the fluid and geological models were prepared, a few sensitivity runs were performed to identify the primary tuning parameters affecting history-match quality, at the field and individual-well levels. While it is common practice to manipulate permeability data, locally or even globally, to achieve a history match, this study incorporated original and detailed data to replicate actual reservoir conditions better. After a few trials, it was realized that the following parameters have a profound effect on the results and hence were used as tuning parameters:
The final results of the history match at the field level demonstrate that, even under the unusual and complex arrangement of fluids in this reservoir (i.e., an overlying aquifer instead of the typical configuration of bottomwater), an acceptable match within a reasonable number of iterations was accomplished. Further scrutiny reveals that the simulated water-production rate between Day 600 and Day 800 is less than the actual production rates.
It is well-known that fracture conductivity is variable between wells. Because the fracture planes (middle fine grid of the local grid refinement) all have a constant width of 1.64 ft, the conductivity (fracture width times fracture absolute permeability) value becomes a function of absolute permeability. The range of conductivity values is 40 md-ft up to 85 md-ft, with the wells with higher water cuts having higher conductivity values. In other words, higher conductivity of the fractures in these wells increases the contribution of the aquifer, and hence water production, in these wells.
By the end of primary production, and with eight active producers, oil recovery factor reaches a value close to 6%. With the well conversion and waterflood operation, the rapid decline in oil-production rates ceased and rates stabilized at approximately 200 STB/D at the field level, which is indicative of waterflood success in this region. At the end of the entire production period (including waterflood), the recovery factor improved to 11% and the average reservoir pressure increased to approximately 2,000 psia.
Once a calibrated (history-matched) model has been obtained, it can be used to answer a variety of what-if questions that aim at optimizing current operations or can provide insight into development planning in analogous sections in the same pool. For instance, the model can be used to evaluate how the major design elements—well spacing, fracture spacing, and fracturing-job tonnage—can be changed to maximize value.
The calibrated model is used to assess the long-term performance of primary and waterflood recovery in the Bakken. Furthermore, the effect of enhanced-oil-recovery schemes is illustrated. It should be noted that neither operational constraint (e.g., infrastructure for handling high water or gas volumes) nor economic constraint (e.g., initial capital investment) is considered in these scenarios.
The Bakken has not been considered by some operators as a candidate for waterflood until recently, mainly because of its perceived low injectivity as a result of low permeability. However, this study, which incorporates realistic permeability and other reservoir data, suggests a clear and distinct production advantage with the use of waterflooding operations.
The complete paper contains a detailed discussion of gasflooding performance in this part of the Bakken. That discussion establishes that carbon dioxide (CO2) flooding outperforms immiscible-gas flooding in all cases. In comparison with CO2, at Bakken reservoir temperature and pressure, the immiscible gas studied has a much smaller density and viscosity, which leads to inferior results.
In this study, details of the numerical-simulation history match of a waterflood pilot are provided. The primary conclusions of the modeling efforts are
Water-Injection Operations and Gas-Injection Sensitivities in the Bakken Formation
01 October 2017