Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

SPE 96899

Assessment and Development of the Dry Horseshoe Canyon CBM Play in Canada
P.A. Bastian, SPE, Quicksilver Resources Inc., and O.F.R. Wirth, L. Wang, SPE, and G.W. Voneiff, SPE, MGV Energy Inc.

Copyright 2005, Society of Petroleum Engineers


mining. However, these mining operations, together with
This paper was prepared for presentation at the 2005 SPE Annual Technical Conference and water well drilling and conventional oil and gas development
Exhibition held in Dallas, Texas, U.S.A., 9 – 12 October 2005.
activity, highlighted a potentially large natural gas resource.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in a proposal submitted by the author(s). Contents of the paper, as
Numerous geological evaluations estimate the potential CBM
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to resource in Alberta to be between 500 and 550 Tscf.2-4 Prior
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at to 2000, about 200 wells had targeted these coals for
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
evaluation, but no commercial successes resulted from this
for commercial purposes without the written consent of the Society of Petroleum Engineers is effort, due to a combination of factors, including low
prohibited. Permission to reproduce in print is restricted to a proposal of not more than 300
words; illustrations may not be copied. The proposal must contain conspicuous permeability, low gas prices and completion difficulties.5 In
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
addition, the Horseshoe Canyon coals were not part of these
early exploration efforts.
Abstract Beginning in late 2000, MGV (together with EnCana)
This paper discusses how the Horseshoe Canyon coalbed embarked on an intense effort to characterize the shallower
methane (CBM) play in Western Canada was converted from Horseshoe Canyon coals in Alberta, and develop the drilling,
an under-explored, non-commercial resource to a major completion and production practices necessary to turn this
commercial play through the application and modification of untapped resource into a commercial play. During this initial
technology and analysis techniques from other basins, and multi-year exploration phase, we drilled over 100 wells,
how this play is being developed today. As of December collected a significant amount of core data, and identified
2004, production from the Horseshoe Canyon CBM play is several unique characteristics of the Horseshoe Canyon coals
estimated to be over 100 MMscfd of gas, with future that guided our decisions on how to develop this new resource.
production expected to grow significantly. The first By early 2003, MGV began producing the first commercial
commercial developments were completed in 2003 and 2004, CBM in Alberta’s history. By the end of 2004, the Horseshoe
and drilling is increasing and expected to exceed 3,000 wells Canyon coals had become a major CBM play in Alberta, with
per year in 2005.1 numerous companies developing Horseshoe Canyon projects
The Horseshoe Canyon CBM play covers a large in the corridor between Calgary and Edmonton.
geographic area of 200 miles by 50 miles, and exists in a
large, complex vertical section with numerous coal, sand, silt, Geological Setting of the Horseshoe Canyon CBM
shale and mudstone layers. In addition, the play is naturally Play
under-pressured, and the coal beds are mostly dry. Because of The Horseshoe Canyon formation is one of several coal-
these complex and unique characteristics, assessment of bearing formations that encompass a large geographic area in
commercial viability and development optimization can be Alberta, Canada (Fig. 1). The Horseshoe Canyon subcrops
confusing and only applicable over small parts of the play. beneath glacial deposits on the eastern side of the formation,
MGV Energy, Inc. (a wholly-owned subsidiary of Ft. Worth- and outcrops along river valleys where there are several
based Quicksilver Resources, Inc.) and its initial joint venture surface coal mines (Fig. 2). To the west, north and south,
(JV) partner, PanCanadian Petroleum Limited (now EnCana Horseshoe Canyon coals thin and become increasingly shaley.
Corp.), discovered the techniques to achieve commerciality Horseshoe Canyon coals were formed in a coastal plain
and pioneered many of the practices used by industry today environment on the margins of the Bearpaw Sea. Paleo-
for Horseshoe Canyon CBM development. In this paper, we shorelines run in a northwest-southeast direction sub-parallel
discuss identification of the reservoir opportunity, including to the present day Rocky Mountains. Fluctuations in relative
data acquisition and analysis. We describe geologic and sea level resulted in multiple thin seams interbedded with
reservoir models, and production forecasting methods. We continental and shallow-marine shales, and shoreface and
also cover completion and production practices, spacing fluvial sandstones. These coal seams vary in lateral
optimization and reserves estimation procedures. continuity. In some cases, individual coal seams can be
mapped over 20 to 30 miles in a depositional strike direction,
Introduction and up to 10 miles in a depositional dip direction. In other
The exploitation of Alberta, Canada’s vast coal resource has cases, coal seams pinch out within a mile.
been ongoing for many decades, but focused primarily on coal The Horseshoe Canyon CBM play includes the numerous
thin coal seams found in the large stratigraphic interval from
2 SPE 96899

the top to the base of the Horseshoe Canyon Formation, plus seam being less than 0.5 m thick. In the context of a normal
the uppermost coals of the Belly River Group, the Lethbridge CBM reservoir development, the effective dewatering of
coals. This large stratigraphic interval varies in total thickness dozens of thin coal seams over a broad depth interval would
from 400 m to 600 m, and can contain between 2 m and 30 m be considered extremely difficult, if not impractical. In
of total net coal thickness. We typically divide these coal addition, the generally accepted knowledge that the Horseshoe
seams into six informal coal groupings (Fig. 3). The bulk of Canyon coals are low rank meant that even if dewatering were
the coal is found in the lower strata from the Drumheller feasible, but difficult, perhaps the resource was insufficient to
down. justify a full-scale development.
In some areas, the interbedded sandstones (shown in Fig.
3) are conventional targets and can be commercially exploited Dry and under-pressured coals. In late 2000, we began
on their own. However, these sand intervals are more testing our first Horseshoe Canyon coal seams with a standard
commonly poor quality reservoirs that are exploited and co- water injection falloff test, a common practice in the industry.
mingled as part of the CBM completion in a well, and Results from these initial tests were very discouraging because
represent less than 20% of the total production from a co- it was very difficult to inject water into the coals. Normally,
mingled coal and sand well. this indicates an extremely low permeability. To restore some
MGV cored and canistered numerous Horseshoe Canyon communication with the reservoir, we then tried a technique
coal seams in several wells as part of its initial exploration that had been used previously in the Appalachian Basin: high-
program with Encana. Typical measurements and laboratory rate, dry nitrogen injection.6 These nitrogen injection
analyses included canister desorption tests, density treatments resulted in methane production from the coals, and
determination, proximate analysis for ash and moisture led to the discovery that the Horseshoe Canyon coals are dry,
content, and a limited amount of coal rank and maceral and very sensitive to liquid exposure. Fortunately, the
analysis. We find that the coals are largely vitrinite rich discovery that the Horseshoe Canyon coals are dry over a
(typically more than 90%). Vitrinite reflectance large geographic area meant that dewatering would be
measurements place the coal rank of Horseshoe Canyon coals unnecessary, resulting in significant savings of capital and
at Sub-bituminous A-C (Fig. 4). The face cleats are generally operating costs.
well developed, while the butt cleats are generally poorly However, we also discovered that the Horseshoe Canyon
developed, and there is no significant mineralization of the coals are severely under-pressured, and that pore pressure
cleats. Given the large number of seams, ash content varies gradients varied vertically and laterally. It was well known
widely. Individual seams can have ash contents ranging from that the shallow, conventional reservoirs in the central Alberta
10% (very clean) to over 35%, and in some cases, higher ash region of the Western Canadian Sedimentary Basin are under-
contents are possible if the coal sample also contained some pressured, but the extent and magnitude of the under-
carbonaceous shale material. Moisture content is consistently pressuring in the Horseshoe Canyon coals had not been well
between 10% and 20%. quantified.
Fig. 5 shows the magnitude of pore pressure gradients we
The Horseshoe Canyon CBM Play - A Unique, Dry have encountered in the Horseshoe Canyon coals. Pore
Coalbed Methane Reservoir pressure gradients can range from less than 0.10 psi/ft to about
Until now, most economically attractive CBM reservoirs have 0.23 psi/ft, which is significantly below the normal freshwater
been characterized by a single (or possibly a few) very thick hydrostatic gradient (0.433 psi/ft). In addition, Fig. 5 shows
coal deposits that are relatively shallow, have high that pore pressure gradients vary from seam to seam, and do
permeability and must be dewatered. A prime example of this not necessarily increase monotonically with depth.
type of CBM reservoir is the Powder River Basin, located in
the state of Wyoming, USA. Although these coals are low Data collection and analysis. Some of the unique
rank, the thickness of these coal deposits (up to 100 ft in a characteristics of the Horseshoe Canyon coals translate into
single coal seam) yielded sufficient resource in place for a economic advantages that make these coals a viable,
commercial play, and the shallow depth and high permeability commercial resource, but they also present several challenges
reduced operating costs and allowed effective dewatering. for collecting data and evaluating the Horseshoe Canyon CBM
In Alberta, Canada, the Horseshoe Canyon coals are very resource potential.
different from the coals of a typical CBM basin like the We used special data collection methods in three technical
Powder River, and have several unique and important areas to gather information about the Horseshoe Canyon coals:
characteristics, including (1) coring, (2) pressure transient and pore pressure testing, and
• 10 to 30 or more thin seams, rather than one, or a (3) production profiles using spinner logs.
few, very thick seams, Coring. Coring and core analyses are key components to
• Broad depth (and completion) interval, evaluating an unconventional resource like CBM. Core
• Dry coals (no water in the cleats), and analysis usually establishes the rank and gas content of the
• Severe under-pressuring. coals, which leads to a proper assessment of resource in place.
Since the Horseshoe Canyon formation contained multiple,
Many thin seams, broad depth interval. The Horseshoe thin coal seams, it was necessary to collect a significant
Canyon coal section contains dozens of very thin seams, amount of core for this project. We needed good vertical and
spread out over a 300 to 500 m depth interval (Fig. 3). Seam areal coverage to properly understand the distribution of the
thickness ranges from 0.1 m to about 2.5 m, with the typical resource in place. For example, during the initial exploration
SPE 96899 3

phase of this project, we collected and analyzed over 38 m of interpretable results. In addition, we adopted this test design
core in the area northeast of Calgary known as the Palliser for pore pressure isolation testing to quantify the complex
block. pressure regime that exists in these reservoirs (see Fig. 5).
For coring the Horseshoe Canyon coals, we used a However, testing CBM wells this way was significantly more
wireline retrievable coring system to quickly return the cores expensive than a typical injection falloff test, and in some
to surface, which reduces the amount of lost gas as the core cases, required weeks or months to complete.
makes its way to the surface. Once samples were selected and Production profiling. Considering the complexity of this
canistered, we allowed some samples to remain canistered for reservoir, it would be difficult to predict how much production
weeks or months to better understand the desorption rate of is coming from any particular interval in a well. However, we
these low rank coals. needed to couple interval production rates and profiles to other
A typical core analysis of a Horseshoe Canyon coal seam well information, such as depth, thickness, gas content and
included measurements of pressure, to better understand the potential of this resource.
• Gas content (in situ and dry, ash-free basis), Production profiling using spinner logs was a cost
• Ash and moisture content, effective solution to this problem, and has become part of our
• Coal density, and routine operating practice in many of our Horseshoe Canyon
• Gas composition. CBM wells. Results from spinner logs are integrated with
On select core samples, additional analyses include coal other regional and well data to estimate the permeability of the
maceral composition, and vitrinite reflectance for determining coal intervals, and to properly distribute productivity to layers
coal rank. in our predictive models.
Results from our core analysis of the Horseshoe Canyon
coals show a wide range of values for key measurements (i.e., Completion and production practices
ash content, gas content, density, etc.), both vertically from The most common practices used in the completion,
seam to seam in a wellbore, and laterally between wells. This stimulation and production of water-saturated coals are: (1)
variability is consistent with other aspects and properties of hydraulic fracturing using proppant and gelled water or foam,
the Horseshoe Canyon coals, making this a very complex and (2) dewatering using artificial lift. However, the layered,
CBM play. dry and under-pressured characteristics of the Horseshoe
Pressure transient and pore pressure testing. Since each Canyon coals required us to replace this conventional wisdom
Horseshoe Canyon CBM well may have dozens of perforated with a completely different set of operating practices.
coal seam intervals, it was impractical to test every seam in
every well. Furthermore, since we could not inject water into Stimulation using nitrogen and coiled tubing units. A
the Horseshoe Canyon coals, we needed to develop other typical Horseshoe Canyon CBM well is drilled with water or
options for estimating initial pressure, permeability and skin air, cased and cemented, and then perforated to open the coals
factor. to the wellbore. But, the unique nature of the Horseshoe
We attempted injection falloff tests using pure nitrogen Canyon coals precluded traditional stimulation practices. We
gas, but determined that the stress sensitivity of the cleat had already established that injecting water into the coals was
system in the coals, and non-Darcy flow effects during the very damaging, which eliminated water-based fracturing
injection period, resulted in difficult and non-unique fluids, and even high-quality foam systems. And, the low
interpretations. We also considered using very low injection reservoir pressures in these coals meant that there may not be
rates for these tests to mitigate the problems mentioned above, enough natural energy to flow back the fracturing fluid after
but found that we could not accurately measure such low gas stimulation. Without water or some other viscous fluid, we
injection rates at the surface. Considering the problems we could not transport proppant into the coal seams to stimulate
were having with injection-falloff tests, we began the wells. Finally, it was difficult to effectively stimulate
experimenting with conventional drawdown-buildup tests. (conventionally) the large number of open coal intervals in a
A conventional drawdown-buildup test had several typical Horseshoe Canyon well.
advantages that were important for testing Horseshoe Canyon Recognizing the damaging effects of liquids on these coals
coals. Since the coals are dry, we did not have concerns with and the practical limitations of conventional stimulations, we
multi-phase flow, and expected the interpretations to be turned to a technique that had been used previously in the
relatively straightforward. Also, a drawdown-buildup test Appalachian Basin: high-rate, dry nitrogen injection
would produce native reservoir fluid (i.e., methane gas) at a stimulations.6 By using coiled tubing units equipped with a
rate consistent with the natural flow capacity of the reservoir, down hole fracturing isolation tool, we could individually treat
reducing or virtually eliminating the stress-sensitivity and non- each of the open coal seam intervals in a well, ensuring that
Darcy flow effects associated with the high-rate nitrogen every completion interval received some amount of
injection falloff tests. stimulation.
Since a typical Horseshoe Canyon completion involved During the exploration phase of this project, we used a
many (sometimes dozens) of open coal seams, we designed two-step completion program to evaluate stimulation
drawdown buildup tests that isolated individual coal seams in effectiveness: 1) a small-volume, lower-rate dry nitrogen
a well, testing each isolated interval in sequence (Fig. 6). This injection “breakdown” treatment, followed by 2) a high-
simultaneous, multi-test design resulted in successful pressure volume, high-rate dry nitrogen stimulation treatment. We
transient tests in a single wellbore, yielding good data and typically flared wells after each of the two stimulations to
evaluate the effectiveness of each style of treatment. Fig. 7
4 SPE 96899

compares the post-treatment production performance of each complex reservoir over time. What is needed is a probabilistic
the two stimulation treatments in one of our exploration wells. method that is coupled to numerical reservoir simulation.
This figure shows that the initial “breakdown” treatment
resulted in some gas flow from the coals in the well, but that Monte Carlo reservoir simulation. Our solution to the
the subsequent high-volume, high-rate treatment resulted in a problem of forecasting Horseshoe Canyon CBM well
250% productivity increase. We applied this two-step performance is Monte Carlo reservoir simulation (MCRS).
stimulation technique to most of the wells in our initial Fig. 8 shows the basic schematic workflow of the MCRS
exploration program, and saw productivity gains of 200% to process. Some main aspects of MCRS are:
400% as a result of the second stimulation treatment. • MCRS combines numerical reservoir modeling (to
Eventually, we dropped the “breakdown” treatments from our address reservoir complexity) with probabilistic
completion program, and currently perform only the high- methods (to address uncertainty and risk).
volume, high-rate nitrogen stimulations to complete our wells. • MCRS preserves general trends seen in reservoir
data, while also accounting for the scatter that is
No dewatering or artificial lift needed. A typical CBM evident in those data.
development might have one or two primary coal seams that • Reservoir properties are randomly sampled from
must first be dewatered before gas production can begin. user-defined statistical distribution functions, and
Since the Horseshoe Canyon coals are dry, dewatering and combined into a realization that represents a single
artificial lift are unnecessary, representing a significant cost well at some pre-defined spacing.
savings for any Horseshoe Canyon CBM development. • With MCRS, each Monte Carlo iteration represents a
However, the many dry, thin coal seams must be co-mingled distinct “realization” of a single-well reservoir
in a well to achieve commercial production rates. simulation, with a complete rate vs. time forecast.
All of these factors make operating a Horseshoe Canyon • Hundreds or thousands of iterations can be performed
CBM well much more economic. Following nitrogen to adequately sample all input distributions.
stimulation, Horseshoe Canyon CBM wells are placed on • Results can be consolidated into a range of rate vs.
production almost immediately, with nothing more than time projections, driven by known reservoir and
production casing in the wellbore and a wellhead at the geologic data.
surface. For large scale projects, wells are tied-in to a The idea of MCRS is not new. Purvis7 demonstrated the
common, low-pressure gathering system leading to a viability of MCRS with case histories including a CBM
compression facility. reservoir, and Roadifer8 applied this technique to perform a
large parametric study related to CBM well performance.
Modeling reservoir performance Recently, Jalali and Mohaghegh9 discussed a simple, single-
As the commercial success of the Horseshoe Canyon CBM well CBM reservoir simulator that integrates probabilistic
play started to become apparent, some key questions began to methods. Our MCRS process uses a standard, commercial
emerge: Monte Carlo simulation software package wrapped around a
• What is the future performance and ultimate recovery commercial reservoir simulator.
of CBM in the Horseshoe Canyon formation?
• What is the best way to develop this new CBM play Model construction for MCRS. Unconventional reservoirs
(i.e., optimum well spacing)? are sometimes described as statistical plays, where it is
• How do we answer these questions confidently using recognized that
exploration-phase geologic data and limited, early- • Many wells will be drilled in the project,
time well performance? • Production performance from wells will be highly
As with other large, unconventional resource basins, the variable, and not necessarily predictable from well to
ultimate development of the Horseshoe Canyon CBM play well, and
will involve thousands of wells, and so far, only a small • The aggregate performance of all the wells in the
fraction of those wells have been drilled. The highly complex project should lead to an economically viable project.
nature of the Horseshoe Canyon coals has resulted in a wide The basic objective of MCRS is to characterize the
variety of well performance profiles that are not necessarily reservoir and production performance statistically to mimic
predictable from location to location, or correlatable to the development of a statistical play. To do this, we describe
common sub-surface measures such as depth or thickness. input reservoir properties (tops, net pay, permeability, etc.) as
Any estimates of resource potential must account for this continuous ranges of possible values, determined from
local variability, and include the range and confidence of available well data. Most of the time, we can use simple,
expected outcomes. Traditional analysis methods, such as statistical distribution functions having a mean or median, a
volumetric or decline curve analysis, are not suitable for standard deviation, and minimum and maximum values.
analyzing such a complex reservoir, and would likely result in Figs. 9 and 10 are examples of actual well data that are
conservative estimates of resource potential. Probabilistic calibrated to distribution functions, and illustrate how we
methods, such as Monte Carlo simulation, can help account convert discrete reservoir information into continuous
for the statistical variability of the Horseshoe Canyon coal statistical functions. Fig. 9 is a probability plot for depths to
reservoirs, but lack the robust reservoir model that can the top of the informal coal groups we use to describe the
adequately forecast the production performance of such a
SPE 96899 5

Horseshoe Canyon section, while Fig. 10 shows total coal However, by adopting a statistical approach and using
thickness for those same coal groups. MCRS processes, we could make performance predictions that
Once all the reservoir properties have been described we consider reasonable and technically sound.
statistically, individual well models are constructed by Fig. 12 is an example of how our estimates of original gas
sampling the statistical distribution functions in one of two in place from well data compare to those computed from the
ways: statistical model. This comparison is part of our calibration
• By sampling directly from the distribution functions process to ensure that our statistical model is properly
of known well data (such as tops and thicknesses initialized prior to reservoir simulation.
from logs, as illustrated in Figs. 9 and 10), or Fig. 13 shows the 40-year, rate vs. time curves for the
• By calculating mean values from correlation average well in five different spacing scenarios for one of our
equations, and applying statistical scatter (standard Horseshoe Canyon projects. These “type curves” can be used
deviation) quantified from analysis of known data. to represent the average well in a large-scale development of a
statistical play. Alternatively, since we preserve the entire set
Calibrating the model. The main objective in calibrating a of rate vs. time forecasts from the MCRS model, we can also
MCRS model is to match the population of outputs from the select a forecast for any probability, such as P(10) and P(90).
many reservoir simulation runs (initial fluids in place, rate vs. Fig. 14 illustrates the sensitivity of recovery to well
time) to estimates derived from available field and well spacing for one of our Horseshoe Canyon development
information. This calibration includes both static and dynamic projects. For 320 ac and 160 ac well spacing, 40-year
data to ensure that inputs to the model are consistent with cumulative gas recovery ranges from 1.1 Bscf per section to
observed sub-surface well information, and that the resultant over 1.4 Bscf per section, representing recoveries of 44% and
simulated population of rate vs. time curves is consistent with 54% of original gas in place, respectively. From our analysis,
the observed production performance of wells in the field. we find that the appropriate initial spacing for Horseshoe
During the history matching of well production Canyon CBM developments is project specific, and ranges
performance, we alter the distributions of reservoir parameters from four to eight wells per section, depending on resource in
to modify the model, rather than changing discrete place and initial well productivity.
(deterministic) values for a particular well. Beyond that, the MGV has used analysis results such as those shown in
calibration of the MCRS model is very similar to that of a Figs. 13 and 14 to perform full-cycle economic analysis,
conventional deterministic model. spacing optimizations, and to help in the estimation of
Fig. 11 presents the results of a MCRS model calibration, reserves.
showing how the average well from the MCRS simulation
closely matches the average production performance of wells Conclusions
in a project area. The Horseshoe Canyon CBM play in central Alberta is fast
becoming the largest commercial “dry” CBM development in
Forecasting performance. Once the MCRS model is the world. The complex geology and layering of the reservoir,
calibrated, we can forecast the expected performance of a well combined with the dry and under-pressured nature of the
or project, knowing that the forecasts are derived from models coals, presented unique challenges for analyzing and
that are consistent with our geologic characterization and field developing this resource.
observations. However, since we are treating the modeling Special processes for completion, production and reservoir
probabilistically, we are able to answer questions such as analysis were required to properly evaluate the potential of
• What is the likely distribution of initial gas in place this CBM resource, including:
per well (the size of the resource)? • Simultaneous, stacked, multi-seam pressure transient
• How much gas will a typical well produce over time and pore pressure tests,
(recovery factors)? • Stimulations using dry, nitrogen gas and coiled
• What is the optimum well spacing for a particular tubing units,
project area? • No dewatering, artificial lift, water handling or
Additionally, we can use the results of the MCRS model to disposal, and
estimate a rate vs. time profile computed at almost any • Statistical methods (Monte Carlo Reservoir
probability (i.e., P10, P50, or P90) for economic analysis and Simulation) for reservoir analysis and forecasting
reserve bookings. Applying these processes during the exploration phase of
the project allowed us to make critical business decisions
Reserves and recovery from the Horseshoe Canyon regarding the viability of large-scale development of the
coals Horseshoe Canyon CBM resource.
Throughout this paper, we have discussed the very complex By early 2003, we had begun commercial production of
nature of the Horseshoe Canyon coalbed reservoirs. This Horseshoe Canyon CBM gas. Development of this resource is
complexity leads to a wide variability in initial well accelerating, with many operators set to drill thousands of
performance, and makes it difficult to predict the long-term wells over the next several years. Total production of
performance of large-scale development projects. Horseshoe Canyon CBM from all operators now exceeds 100
MMscfd, and is expected to grow rapidly over the next several
years.
6 SPE 96899

Acknowledgements
The discovery and success of the Horseshoe Canyon CBM
play would not have been possible without the skill,
commitment and hard work of the geologists, engineers and
staff members at MGV Energy, Inc. This paper is dedicated to
all those whose individual contributions helped uncover a
world-class CBM resource.
The authors also wish to thank the management of MGV
Energy, Inc., and Quicksilver Resources, Inc. for their
considerable support in allowing this work to be published.

References
1. Zelinsky, T.: “Service Companies Will Benefit from Increased
CBM Activity Next Year,” Nickles Daily Oil Bulletin, Dec. 17,
2004.
2. Beaton, A.P.: “Production Potential of Coalbed Methane
Resources in Alberta,” 2003 EUB/AGS Earth Sciences Report
2003-03.
3. Canadian Gas Potential Committee, 2001 Report, “Natural Gas
Potential in Canada”.
4. Hughes, J.D.: “In Situ Assessment of Coalbed Methane
Resources in the Alberta Plains Volume I – Unconstrained
Cases”, Geologic Survey of Canada report to MGV Energy,
Inc., 2001.
5. “Canadian Coalbed Methane Forum General Catalogue of CBM
Data in Alberta and British Columbia (as of January 1, 1993),
Volumes I and II,” Sproule Associates Ltd, report to Canadian
CBM Forum, January 1993.
6. Stidham, J.E., Tetrick, L.T. and Glenn, S.A.: “Nitrogen Coiled-
Tubing Fracturing in the Appalachian Basin,” paper SPE 72382
presented at the 2001 Eastern Regional Meeting, 17-19 October,
Canton, Ohio
7. Purvis, et. al.: “Coupling Probabilistic Methods and Finite
Difference Simulation: Three Case Histories,” paper SPE 38777
presented at the 1997 SPE Annual Technical Conference and
Exhibition, 5-8 October, San Antonio, Texas
8. Roadifer, et al.: “Coalbed Methane Parametric Study: What's
Really Important to Production and When?,” paper SPE 84425
presented at the 2003 SPE Annual Technical Conference and
Exhibition, 5-8 October, Denver, Colorado
9. Jalali and Mohaghegh: “A Coalbed Methane Reservoir
Simulator Designed and Developed for the Independent
Producers,” paper SPE 91414 presented at the 2004 SPE Eastern
Regional Meeting, 15-17 September, Charleston, West Virginia

SI Metric Conversion Factors


ft × 0.3048 E+00 = m
ft3 × 2.831 685 E–02 = m3
psi × 6.894 757 E+00 = kPa
ac × 4.046 781 E–01 = ha
sec (640 ac) × 3.861 004 E–01 = km2
SPE 96899 7

Fig. 1 – Distribution of coals in Alberta, Canada, including the Horseshoe Canyon formation (courtesy of Alberta EUB and AGS).
E d ge of D eform e d B e lt

D ru m he lle r V alle y
(R e d D e er R iver)

W e st East
y C oal
A r d le om p son
on - Th
S co llard F m C a rb ver
W ea e
b r id g owk
i
H o rses ho e
e lle r L e th Pak
C an yo n D r u m h Coa
l
r
al Tabe s
Bas S h a le n v il
l
ra d o M an
B elly R ive r C o lo e
Lea P
a rk
C o a ls
G ro u p M c K a y n v i ll e Low
er
r M an
M ilk R iver U ppe z o ic
o
P a le
M an n ville G p
Fm
s s in
K o o te n a y G p N ik a n a
F ernie G p

Fig. 2 – Schematic cross-section of coals in the Western Canadian Sedimentary Basin.


8 SPE 96899

Fig. 3 – Groupings used to characterize the coals in the Horseshoe Canyon formation.

North America Vit. Ro. %


0.15
Peat
Low

0.3
Lignite
• Measured Ro values between
Sub-Bit. C 0.4 0.39 and 0.56
B
0.5 • Rank: Sub-bituminous A-C
C A 0.6
High Vol

B 0.7
Bit.

0.8
A 1.0
Medium

1.2
Med. Vol Bit.
1.4

Low Vol Bit. 1.6


1.8
Semi-Anthracite

Anthracite 3.0
High

4.0
Meta-Anth.
Modified after ASTM

Fig. 4 – Coal rank of Horseshoe Canyon coals is Sub-bituminous A-C.


SPE 96899 9

Pore pressure gradient, psi/ft 0.5

0.4
Normal freshwater
gradient

0.3

0.2

0.1

0.0
200 250 300 350 400 450 500 550 600
Depth from surface, m
Fig. 5 – Pore pressure gradients demonstrate under-pressuring in Horseshoe Canyon coals.

Packer with tandem


gauges

Horseshoe Canyon
perfs
Bridge plug to isolate
perfs

Fig. 6 – Schematic of multi-zone pressure transient test in a Horseshoe Canyon CBM well.
10 SPE 96899

400

350

300
Gas rate, Mscfd

250
Post-stimulation
rate (~ 130 Mscfd)
200

150

100

50 Post-breakdown rate
(< 50 Mscfd)

0
0 50 100 150 200 250 300 350
Time, days
Fig. 7 – Example comparison of production performance after “breakdown” and stimulation.

Monte Carlo Sampling and Simulation Engine

Sample distribution
functions for
reservoir properties

Build numerical
One simulation input file
Monte
Carlo
Iteration
Run numerical
simulation

Results post-
processing and data
capture

Fig. 8 – Workflow schematic of Monte Carlo reservoir simulation process.


SPE 96899 11

100
Symbols - well data
150
Depth to middle of coal group, m

Lines - distribution
200 function fit

250

300

350

400

450

500

550

600
1 2 5 10 15 20 30 40 50 60 70 80 85 90 95 98 99
Probability, % less than

Fig. 9 – Distribution of coal reservoir depths calibrated to statistical functions.

8
Symbols - well data
7
Lines - distribution
function fit
6
Coal thickness, m

0
1 2 5 10 15 20 30 40 50 60 70 80 85 90 95 98 99
Probability, % less than

Fig. 10 – Distribution of coal reservoir thicknesses calibrated to statistical functions.


12 SPE 96899

200
Avg well rate from field data
180
Avg well rate from MCRS model
Gas production rate, Mscfd

160

140

120

100

80

60

40

20

0
0 30 60 90 120 150 180 210 240 270 300 330 360 390 420
Time, days
Fig. 11 – History match of well performance data using a Monte Carlo reservoir simulation model.

6
Original sorbed gas in place, Bscf/sec

5 Deterministic estimate from well data

MCRS estimate
4

0
1 2 5 10 15 20 30 40 50 60 70 80 85 90 95 98 99
Probability, % less than

Fig. 12 – Calibrating original gas in place with Monte Carlo reservoir simulation.
SPE 96899 13

180

640 ac/well
160
320 ac/well
160 ac/well
140
80 ac/well
Avg daily gas (Mscfd)

40 ac/well
120

100

80

60

40

20

0
0 5 10 15 20 25 30 35 40
Year
Fig. 13 – Average well production forecast from Monte Carlo reservoir simulation model for various well spacings.

2.5

40-year EUR per well

2.0 40-year EUR per section


Recovery factor
Recovery indicator

1.5

1.0

0.5

0.0
640 320 160 80 40
Drainage area per well, ac
Fig. 14 – Example estimated ultimate recovery (EUR) by well and by 640-ac section for a Horseshoe Canyon
development project.

You might also like