SPE-185021-MS Waterflooding A Multi-Layered Tight Oil Reservoir Developed With Hydraulically Fractured Horizontal Wells
SPE-185021-MS Waterflooding A Multi-Layered Tight Oil Reservoir Developed With Hydraulically Fractured Horizontal Wells
SPE-185021-MS Waterflooding A Multi-Layered Tight Oil Reservoir Developed With Hydraulically Fractured Horizontal Wells
Crystal Hustak and Rous Dieva, Baker Hughes; Richard Baker, BRE Group; Ben MacIsaac, Ken Frankiw, and
Bryan Clark, Red River Oil Inc.
This paper was prepared for presentation at the SPE Unconventional Resources Conference held in Calgary, Alberta, Canada, 15-16 February 2017.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect
any position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.
Abstract
A successful waterflood can be implemented in a multi-layered tight oil reservoir developed with horizontal
multi-fractured wells. This paper forecasts the recovery factor that can be achieved in such a reservoir
as well as discusses the challenges of analyzing and modelling tight oil reservoirs developed with multi-
fractured horizontal wells.
With some unconventional reservoirs that are hydraulically fractured, a phenomenon exists whereby
material balance and simulation indicate pressure support from a water source that is not always obvious.
This phenomenon is believed to be related to the multi-layered silts/shales in the reservoir and is not
typically seen in simulation of conventional higher permeability reservoirs (Kair >10 mD). Although, the
exact petrophysical nature of the silts/shale reservoir layers in this project are not well defined at this time,
a successful production history match can be achived by incorporating their input into a simulation model.
Introduction
In the Western Canadian Sedimentary Basin (WCSB), numerous waterflood fields have been developed
successfully in the last 60 years. Typically, waterfloods are implemented in high-permeability, light
oil, conventional reservoirs. As we unlock the unconventional reserves in the WCSB, more and more
waterfloods and enhanced oil recovery (EOR) methods have been implemented in tighter oil reservoirs. As a
result, a better understanding is being gained waterflooding these tight reservoirs. Unique challenges result
when implementing a waterflood in a tight oil reservoir; thus, monitoring and analyzing data is critical in
order to implement a successful waterflood. At the same time, tight oil reservoirs are now being developed
with horizontal multi-fractured wells which can also provide benefits such as increased volumetric sweep
efficiency.
In the past several years, many questions have arisen when developing and modelling a waterflood in
tight oil reservoirs such as the Cardium, Bakken and Three Forks. This paper discusses the challenges
of modelling a tight oil waterflood developed with horizontal multi-fractured wells in a Bakken/Three
Forks reservoir package and analyzes the results of the waterflood. A Red River Oil Inc. (RROI) operated
2 SPE-185021-MS
waterflood project in southeast Saskatchewan under a current waterflood scheme will be analyzed and
modeled, and ultimate recovery factors are predicted using reservoir engineering software.
Background
The location of Fertile Section 36, which was analyzed and modeled in this paper, is illustrated in Fig. 1.
Current production/injection data for the area of interest was gathered from the public domain and from
RROI. The geology of the Bakken/Three Forks formation in the area was also evaluated. Fig. 2 shows the
type log used to describe the formation. The Bakken shale sits at approximately 1075 mTVD and provides
the cap to the reservoir package. Below the Bakken shale is the Bakken silt, which is a tighter, and higher
water saturation (Sw) rock, while the Three Forks below is a higher permeability siltstone with a lower Sw.
Through the analysis of sample cores and logs in the area, average porosity was determined to be 14%
to 16% with an average air permeability of 1 mD to 5 mD. For Section 36, an average water saturation of
approximately 50% in the Bakken Silt and 40% in the Three Forks was assumed. This section was analyzed
and modeled to determine the potential ultimate recovery from a tight oil reservoir developed with horizontal
multi-stage completions.
Group composite plots of production and injection for the area of interest are shown in Fig. 4. The
watercut of the primary vertical well was between 30% and 40%, while the horizontal wells observed a
watercut of 60% to 70% prior to water injection. Horizontal well, 191/04-36, produced prior and post-
hydraulic fracture stimulation, as shown in Fig. 5. Before the well was hydraulically fractured, a watercut of
30% to 40% was observed. After the treatment, the horizontal well produced in the range of a 60% to 70%
watercut. It is believed that the increase in watercut was due to fracing into the overlying higher SW Bakken
silts. When modeling this section, the Bakken silt layer needed to be addressed, as modeling without any
inflow from that layer resulted in the static reservoir pressure being lower than the field measured pressure.
Individual well production was analyzed in a multi-well plot as shown in Fig. 6. The first few weeks of
higher liquid and water production was partially attributed to fracture fluid being flowed back. The primary
drive for this reservoir is solution gas; therefore, over time, both oil and water production declined. It is
critical to note that water rates are decreasing with time. Also, as stated earlier, a degree of water influx was
needed to explain higher reservoir pressure and increasing watercuts. The silt zones are less permeable than
the Three Forks and therefore provide relatively weak water support compared to a classic strong water
drive system.
One higher-than-normal water producer was observed, well 191/01-36. Red River believes that this
was possibly due to a mechanical wellbore issue whereby the well was in communication with the upper
Lodgepole formation. This well has since been converted to an injector.
Injection analysis provides further clues on what is happening in the reservoir. Hall plots for the three
injectors in the section were created by Red River and quality checked by Baker Hughes. As seen in Fig.
7, no obvious outliers were observed. In the last two years, all injectors have seen a gradual decrease and
flattening in their water injection rates as initial fill up has occurred and reservoir pressure has increased to
an established maximum allowable injection pressure. As injection rates have plateaued, the VRR has been
maintained at 1.0 or higher as seen in Fig. 8.
Analogue Comparison
There are a number of units in the Bakken formation with implemented waterflood schemes to the north
of Section 36 that can be used as analogue considerations. Although variations in reservoir characteristics
exist between these analogue projects, they are suitable to compare performance trends. Table 1 and Fig. 9
provide details on reservoir properties of the analogue units and the section of interest.
A series of plots were generated to evaluate the performance of the analogue pools. Three specific
methods were used: recovery factor (RF) versus hydrocarbon pore volume injected (HCPVI), water oil ratio
(WOR) versus cumulative oil, and conformance plot.
RF Versus HCPVI
RF versus HCPVI graph is a very common tool used to analyze waterflood potential. The y-axis
demonstrates the oil being produced (represented as recovery factor of the OOIP) and the x-axis represents
the amount of water injected into the reservoir. The slope of the line representes the waterflood's efficiency.
This plot can thus be used to compare the relative efficiency of waterfloods in analogous pools. Fig. 10
shows the resulting RF versus HCPVI graph of Section 36 in comparison to a few analogue projects. Section
36 is an immature flood as compared to the Tundra waterfloods, but it does however display a similar
trending slope or waterflood efficiency as the Tundra units. The Tundra units have a lower water saturation
compared to Fertile Section 36 due to the thinning or absence of the overlying high water saturation Bakken
silts. Despite this difference, Section 36 appears to be following Tundra Unit 1 in its recovery trend (similar
slopes or waterflood efficiency for section 36 compared to unit 3 and 5).
SPE-185021-MS 9
It is also important to note that Tundra's Units No. 1, No 2, and No. 3 have been developed with vertical
producers at 40-acre spacing and horizontal injectors at 400 m spacing. Section 36 and Tundra's Unit
No. 5 were both developed exclusively with horizontal wells at approximately 200 m inter-well spacing,
alternating between injecting and producing wells. It can be observed that Section 36 and Unit No. 5 are
both flatter initially than the vertical-horizontal waterfloods. This trend can be attributed to the fact that
cumulative voidage of the horizontal-horizonal floods is higher than that of the vertical-horizontal floods
at project initiation, and thus these projects take longer to ‘fill up’ and to show quantifiable response.
Conformance Plots
A conformance plot depicts RF versus net water injected/movable pore volume (NWI/MPV). This plot
displays the difference between injected and produced water as a function of movable pore volume on the
x-axis versus recovery factor on the y-axis. As the trend curves to the left or into negative values, more
water is being produced than injected, indicating either: primary connate water production, water influx due
to nearby patterns, bottom water, aquifer or trapped formation water. This particular graph does not take
time into account. The RF is normally extrapolated to a NWI/MPV of 1, which is a theoretical best-possible
recovery from the reservoir, assuming 100% volumetric sweep.
In Fig. 12 all of the waterfloods see a trend that curves to the left into negative net water injected/
MPV. This trend represents the time period before injection begins where the project is under primary
production and therefore the net water injected term becomes negative. More water has been produced from
the formation under primary recovery relative to the movable pore space if the plot trends farther left. As
seen from Fig. 12, Fertile Section 36 had more primary water production than the analogue units due to
production contributions from the higher water saturations in the Bakken silts that are not present in the
analogue units. Section 36 displays a very similar trend to the Tundra units as the injection begins and the
curves trend towards positive NWI/MPV. This suggests that eventually Section 36 could be extrapolated to
a similar ultimate recovery depending upon water injection volumes.
SPE-185021-MS 11
Red River provided geological mapping for the entire Section 36 area within Township 6 and Range 30
W1. The maps include a top-of-structure map for the cap above the Bakken silt, net pay, Phi H and KH
maps for the Bakken silt and Three Forks formations. Gross pay was estimated to be 4 meters in the Bakken
12 SPE-185021-MS
silt and 8 meters in the Three Forks. For the purpose of this simulation, it was assumed that the formations
were in communication and fluids were able to flow between the two formations. Table 2 lists the reservoir
properties. The initial OOIP was calculated to be 1,444 e3m3 (9 MMBbl) in the simulation for the section.
There are no special core or relative permeability tests performed for Section 36 or in the surrounding
areas. The relative permeabilities were obtained using correlations and adjusted for water saturation and
water production. A connate water saturation of 20% and a residual oil saturation of 30% were used for the
Bakken silt and the Three Forks. No gas-oil relative permeability data is available, thus, correlations were
also used. Table 3 shows the relative permeability values.
To initialize the model with the appropriate amount of mobile water, capillary pressure curves were used.
By introducing a capillary pressure curve, the initial water saturation of each formation was initialized.
However, there is no capillary pressure lab test data available for either formation. This meant that
assumptions had to be made on the curvature of the capillary pressure trend as well as the absolute values
of the capillary pressure forces as seen in Fig. 14. The curves are history matched to an average water
saturation of 50% in the Bakken silt and 40% in the Three Forks. A free water level of 595 m or 30 m below
the grid was assumed for this simulation.
SPE-185021-MS 13
No PVT laboratory tests are available for Section 36; therefore, PVT data used in the model was taken
from a waterflood simulation study conducted in 2005 by Tundra Oil and Gas Ltd., found on the Manitoba
government database. Table 4 lists the properties used as the basis of the PVT data.
History Match
The simulation model was set to run with liquid rate constraints and monthly timesteps for all wells
throughout the historical time period (June 2001 through August 2015). There was no recorded gas
production data because the produced gas rate is too low to effectively measure. Injector wells were
constrained on historical water injection volumes.
A water layer was introduced into the Bakken silt to account for the increased water production after
hydraulic fracturing. As mentioned before, well 191/04-36 had production from an openhole completion
(pre-fracture) between December 2005 and June 2008. Fig. 15 shows that oil was more mobile than water
during this period. After post-hydraulic fracture, water volumes were more movable and more easily
produced, indicating the hydraulic fractures must have acessed water that was not accessible before the
fracture process. Observations from core suggest that tighter facies within the complex stratified Bakken
silts could account for this water. The water layer was initialized with a very strong capillary pressure curve
14 SPE-185021-MS
that held the water saturation within that layer set at 100%. The water was held and only produced through
the hydraulic fracture channels in the simulation.
The porosity in the water layer was kept the same as the porosity within the silt. A pore volume multiplier
had to be added to the layer to match the historical pressure points and the historical water production
volumes. Downhole pressures provided by Red River on new infill drills showed an initial pressure in the
range of 4,000 – 6,000 kPa. This was not achievable in the model without the pressure support provided
through water volume in the silts. The reservoir pressure as compared to the pressures of the infills after
the fracture can be seen in Fig. 17. The permeability in the layer was determined by the amount of water
influx that was required to history match the water production. Wells 191/08-36, 191/09-36 and 192/09-36
required more water and the permeability was adjusted accordingly; see Fig. 18.
SPE-185021-MS 15
Fig. 19 displays the history match on watercut. After the hydraulic fractures were applied and the water
from the silt was accessed, the watercut rose from 30% to approximately 70%.
16 SPE-185021-MS
Forecasts
To determine the ultimate recovery potential from this waterflood, a number of forecasts were conducted.
A summary of the forecast scenarios are listed below. All forecasts were run until January 2045, providing
a 30-year forecast. The forecast scenarios evaluated are constrained as follows:
1. Basecase: All producers and injectors are forecasted at their last simulated bottom-hole pressure.
2. Case 1 – Infill: An infill well was added in the west edge of the section where additional drainage
can be accessed. It produces for one year before being converted into an injector. All injectors and
producers are forecasted at their last simulated bottom-hole pressure.
3. Case 2 – Infill + Injection Rate: Same as Case 1, but injectors (including the infill) are now injecting
at 40 m3/d, assuming an optimization of the injectors in the flood occurs.
4. Case 3 – Residual Oil Sensitivity: Change in residual oil saturation from 30% to 20%, based off
the same forecast set up as Case 2. This case was conducted to analyze the possibility of lower
residual oil saturation due to the higher initial water saturation. Work has been done on the Viking
formation in Saskatchewan showing the concept of lower residual oil saturation with higher initial
water saturation1.
The recovery is separated by formation as the majority of the production is forecasted from the Three
Forks. The Bakken silt is only accessed through the hydraulically induced fractures.
The summary table shows a recovery factor of 15% if the section continues to produce as-is, with the
majority of the production coming from the Three Forks (19% RF). With an infill in the west edge of the
section, the recovery increases to 18% for the section and 21% in the Three Forks. Case 2 shows a case
with an infill and double the injection rate, which would yield a recovery of 20% overall and 23.9% in the
Three Forks. The sensitivity forecast on residual oil, Case 3, changes the history match, therefore caution
must be taken when comparing this case to the others. However, if the residual oil saturation is lower than
the assumed 30%, a higher recovery would be achieved.
A comparison of all the forecast cases can be found in Figs. 20 through Figs. 22. The figures display the
oil rate and cumulative oil production as well as a comparison of RF versus HCPVI.
18 SPE-185021-MS
Conclusions
The work described in this paper suggests that a recovery factor of 15% to 25% can be achieved in a
waterflood implemented in a multi-layered tight oil reservoir developed with multi-stage horizontal wells.
Additional mobile water accessed by hydraulic fracturing is believed to be sourced from multi-layered
silts and shales within the Bakken-Three Forks reservoir section. The increased watercut of hydraulically
fractured wells versus lower watercut openhole completions represents an unanticipated water volume
(pressure support mechanism) that was required to balance this reservoir model.
In scenarios such as these, the more data collected, the better a model can be quality checked. Additional
pressure tests, relative permeability as well as PVT data are important pieces of information that would
further narrow down the range of possible recoveries from this waterflood project.