You are on page 1of 11

.

International Journal of Coal Geology 35 1998 369379


Computer modeling and simulation of coalbed
methane resources
G.B.C. Young
)
Adanced Resources International, Inc., Dener, Colorado, USA
Received 9 September 1996; accepted 24 July 1997
Abstract
Coal seam gas reservoirs are complex both geologically and in the mechanism of gas
.
production. Understanding these naturally fractured reservoirs for two-phase gaswater flow
conditions is often limited by a lack of data. This paper illustrates that reservoir simulation is a
powerful tool which can be used to determine key data requirements, and how variability in
reservoir properties and operating practices affect performance at the field level. The paper
presents examples of how reservoir simulation can be used to assess the efficiency of well
.
completions fracturing or cavitation , identify candidate wells for remedial treatment, examine
methane drainage in advance of mining, and assess the impact of errors in measured data on
long-term gas production forecasts. q1998 Elsevier Science B.V.
Keywords: coalbed methane; simulation; naturally fractured reservoirs; coal fields; degasification; reservoir
properties
1. Introduction
Reservoir simulation is a particularly useful tool to study gas-bearing coal beds. This
is because gas storage and production mechanisms in coal beds are more complex and
data intensive than in conventional settings. In a conventional reservoir rock, gas is
stored in the pore spaces. In a coal reservoir, methane is physically bound or adsorbed to
the solid coal surface of fracture systems. The methane does not become a free gas and
migrate to the wellbore until pressure is reduced. In addition to the material presented in
this volume, an excellent discussion of the theory of gas storage and transport in
coalbeds is provided by Law and Rice, 1993.
)
Corresponding author. Fax: q1-303-2952833; e-mail: gyoung@advres-intl.com.
0166-5162r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved.
.
PII S0166- 5162 97 00042- 6
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 370
Pressure reduction normally occurs by pumping water from the coal beds, which
initiates gas desorption. The desorbed gas must move by diffusion through the very low
.
permeability coal matrix secondary porosity in order to reach the natural fracture or
.
cleat primary porosity system. Once in the cleats, which have a high permeability
relative to the matrix and are normally saturated with water, gas and water flow
according to Darcys law and gaswater relative permeability. Even the cleat permeabil-
ity is usually too low to deliver gas in commercial quantities unless the producing wells
are stimulated by induced hydraulic fracturing or other means.
In this paper, numerical modeling techniques, software applications, and data require-
ments for coal reservoirs are described, and some specific examples of computer
modeling various types of gas wells are presented. A more general discussion on the
reservoir engineering aspects of coalbed methane can be found in Saulsberry et al.
.
1996 .
2. Modeling flow through coal beds
Although a number of approaches have been taken to model coalbed methane
.
behavior King and Ertekin, 1989 , a nonequilibrium formulation in which sorption is
pressure and time dependent is required to account for the time lag incurred during gas
diffusion through matrix pores.
The nonequilibrium models may be classified by how the diffusion process is
handled. In pseudosteady state formulations, the diffusion coefficient depends on the
geometry of the matrix elements and time, but not gas concentration, while in steady
state models gas concentration is included as a variable. As a third category, the
unsteady state approach is the most rigorous, since it includes the effects of the gas
concentration gradient, but it is also the most expensive computationally to solve.
The major difference between the results from the pseudosteady state and unsteady
state formulations is in the early-time behavior of coal-seam gas wells, which is
important in analyzing pressure transient tests. Comparison of the two approaches shows
that while the pseudosteady state method underestimates gas rates in the first several
days or so, simulated gas rates at longer times from the pseudosteady state and unsteady
.
state models are virtually the same Kolesar et al., 1990 .
.
The COMETe coalbed methane simulation software used in this study is a
.
nonequilibrium, pseudosteady state formulation based on the Warren and Root 1963
.
model of dual-porosity reservoirs Paul et al., 1990; Sawyer et al., 1990 . As applied to
.
coal beds, the orthogonal natural fractures cleats are treated as a system of connected

pipes which divide the matrix into small elements the distance between cleats is about
.
0.6 cm or 0.25 in. . The matrix elements, which can be slabs, cylinders or spheres, act as
.
the source of gas that diffuses, by Ficks first law Adamson, 1973 , into the cleats.
Desorption of methane is described by a Langmuir isotherm, which relates matrix gas
. . .
content, V P , to the coal-bed cleat pressure, P, according to: V P sV Pr PqP
L L
where V is the maximum amount of gas that can be adsorbed, and P , a characteristic
L L
pressure, is a measure of residence time for a gas molecule on the surface. Both V and
L
P can be determined from gas desorption measurements on coal core samples. The
L
above equation provides the necessary boundary condition at the matrix-cleat interface.
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 371
The COMETe computer model is fully three-dimensional to account for vertical
wells intercepting multiple coal seams, structural features such as reservoir dip and
.
no-flow barriers faults and shale breaks , and gaswater gravity segregation. Other
unique features of coal reservoirs that are modeled are stress-induced changes in cleat
porosity and permeability, and matrix shrinkage due to release of adsorbed gas.
What makes a simulator useful is the availability of boundary conditions that can be
.
modeled Paul, 1996 . These conditions mainly relate to the types of wells vertical or
.
horizontal and well completions, which can be unstimulated, or stimulated by hydraulic
.
fractures vertical or horizontal, finite or infinite conductivity or open-hole cavity
methods.
3. Simulator applications and data requirements
.
Simulator applications for coalseam gas include the following: 1 test how well the
.
geologic model describes the real reservoir by matching performance history, 2
provide a basis for forecasting future production as a function of various operating
.
strategies such as variations in well spacing, 3 determine the ultimate economic
.
recovery for a field, that is the gas rate versus time, 4 confirm the physics of a
.
recovery process such as simulating a laboratory desorption test, 5 discover and
.
diagnose production problems such as wells which are performing below potential, 6
.
determine areas of the reservoir least depleted in order to properly locate infill wells, 7
.
design the best well completion scheme, such as single versus multiple seam, 8 predict
gas recovery from degasification of underground coal mines using both gob wells and
.
horizontal drainholes, and 9 determine sensitivity of simulated production to changes
in various data and identify weaknesses in critical data.
Data requirements for coalbed methane simulation may be grouped into three types:
.
1 reservoir description data, such as geometry, structure, depth, net thickness, stratifi-
.
cation and initial water saturation and pressure, 2 fluid PVT pressurevolumetem-
. .
perature data, such as gas viscosity and composition and 3 time-dependent well data,
such as fluid rates and bottomhole pressures. Some data are so important that they might
be considered a minimal data set for simulation. These minimal data are all in the
.
reservoir description category and include the following: 1 absolute cleat permeability
.
which determines rate of gas recovery, 2 initial gas content for determining gas-in-place
.
and the recovery target, 3 adsorption isotherm for determining timing of initial gas
.
show and ultimate gas recovery and 4 cleat porosity which is the site of water storage
in coal beds and determines volume of water produced.

The above minimal properties permeability, gas content, adsorption isotherm, and
.
porosity are critical because they determine project economics. The permeability can be
evaluated with pressure transient testing, while the gas content and adsorption isotherm
can be determined by direct laboratory measurements on coal samples. However, cleat
porosity is not readily measured and should be estimated by accurate history matching
of the gas and water production and pressure data. Without at least some measurements
of them, the modeler is left with considerable guesswork and a corresponding uncer-
tainty in the computed results. In this case, a sensitivity analysis should be performed
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 372
with the simulator to assess the range of probable outcomes as a function of the
uncertain data.
4. Examples of simulation
4.1. Multi-seam fractured well
The multi-seam fractured well problem is an example of how history matching well
performance using a simulator can be used, in conjunction with field measurements, to
determine coal reservoir properties.
.
The measured data involved the zone isolation packer ZIPe tool, shich were
developed from the need to examine flows from a single completion zone in a well
.
completed in multiple coal seams Saulsberry et al., 1991 . The ZIPe tool provided an
opportunity to confirm heterogeneity of coalbed cleat permeability between the Mary
Lee and Black Creek coal groups during a history match of production and pressure
.
performance of eight wells at the Gas Research Institutes GRI Rock Creek field,
.
Black Warrior Basin, Alabama in the United States Young et al., 1993 .
For the total field, cumulative gas and water production over seven years history
were matched with the simulator to within 3 and 1%, respectively. The match on gas
.
rate for one of the multi-seam wells P5 is shown in Fig. 1. The simulated and
measured data both show that the distribution of gas rate between the two coal groups is
Fig. 1. Gas rate versus time for the Rock Creek P5 well showing the relative contribution of the Mary Lee and
Black Creek coal groups to total production.
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 373
what would be expected based on cleat permeabilities determined from pressure
transient testing. Coalbed production from the Mary Lee Group is roughly an order of
magnitude higher than that from the coal beds of the Black Creek Group. Had the Mary
Lee:Black Creek coal bed production ratio been lower, a damaged or poor Mary Lee
completion would be indicated. This turned out to be the case for several of the wells,
which when coal beds were refractured in the Mary Lee Group yielded increased gas
flows up to sevenfold.
4.2. Open-hole caity well
In portions of the Fruitland coal play of the northern San Juan Basin of New Mexico
and Colorado, open-hole cavity wells outperform hydraulically fractured wells by more
than a factor of two. A study of a cavity well at GRIs Completion Optimization and
.
Assessment Laboratory COAL project was undertaken to determine how the downhole
conditions differ between cavity and fractured wells, and how these conditions impact
long-term well performance.
.
A radial grid was constructed with the cavity well in the center Fig. 2 . Twelve rings
were used in the grid; to preserve detail, not all the rings are shown in Fig. 2. The
.
innermost, 2.4 m 8 ft diameter, ring corresponds to the estimated cavity diameter, and
was used to represent the wellbore. A history match was performed using this grid by
specifying the observed gas rates and allowing the simulator to calculate the water rate
and bottomhole pressure in the well. It became apparent during the history match that
the coalbed cleat permeability around the well had been increased due to cavitation to a
value somewhat greater than the 25 md value determined from earlier well tests. Good
Fig. 2. Schematic diagram showing the radial grid used in the simulation analysis of the COAL site cavity
well.
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 374
Table 1
Simulated cumulative gas production
9 3
.
Well type Gas, 10 m Bcf
.
Open-hole cavity 0.077 2.70
Hydraulically fractured
. .
Undamaged skin factor s0 0.077 2.70
. .
Damaged skinsq3.3 0.030 1.06
matches between simulated and observed performance were obtained with an enhanced
. .
permeability zone of 250 md to a radius of 40 m 130 ft from the well center Fig. 2 .
While the reservoir characteristics derived from the history match are not unique, the
results do represent a closely bounded set of downhole conditions from which to assess
cavity well performance.
Using the grid in Fig. 2, history matching was again performed to find the

undamaged, infinite conductivity induced fracture half-length which turned out to be 64


.
m or 210 ft , that gave the same production profile and cumulative gas production as the
.
cavity well. Five-year gas production forecasts were then performed for: 1 the cavity
. .
well using the history match parameters, 2 the equivalent undamaged hydraulically
.
fractured well and 3 a damaged hydraulically fractured well. As shown in Table 1,
induced hydraulic fractures can yield performance equivalent to open-hole, cavity
completed wells if damage is avoided and highly conductive fractures are created.
4.3. Mine degasification with horizontal wells
The study of a mine degasification project for one of the Pocahantas coal seams
.
undergoing active mining in the Central Appalachian Basin eastern USA was under-
taken to determine the optimum spacing for horizontal degasification wells. Reservoir
simulation was used to evaluate different well spacings and mining advance rates to
achieve the required level of methane reduction prior to and during longwall mining
operations.
Simulated degasification was performed of a section of a coal mine longwall panel of
. .
dimensions 213=229 m 700=750 ft . Coal thickness is 1.67 m 5.5 ft . The longwall
section was represented by a 25=14=1 grid as illustrated in a plan view in Fig. 3. Gas
.
production was modeled from three unstimulated, 8.9 cm 3.5 in. diameter horizontal
.
wells located at 61 m 200 ft intervals along one coal face of the mine. Outgassing of
the coal mine face, which was exposed to atmosphere on three sides, was simulated by
.
setting a 101.3 kPa 14.7 psia boundary condition in thin rows of blocks representing
.
the three coal faces. The initial coal reservoir pressure was set at 2 MPa 300 psia .
The total time simulated was 90 days. For the first 10 days, only the coal mine face
was allowed to produce gas; no wells were producing. Thereafter, wells were brought on
.
production at 10 day intervals left to right in Fig. 3 . All wells were produced
.
unconstrained under a flowing bottomhole pressure of 207 kPa 30 psia . Peak gas rates
3 3
.
3
reached about 5.1=10 m rday 180 Mscfd for each of the wells and about 3.4=10
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 375
Fig. 3. Schematic diagram showing the plan view of the cartesian grid used in the simulation analysis of the
degasification of a longwall panel.
3
.
m rday 120 Mscfd for the coal mine face. After 90 days, the coal mine gas content
was reduced by 55% throughout the longwall panel. Additional degasification with
vertical andror gob wells would be required to further reduce coalbed gas content.
4.4. Sensitiity analysis
Sensitivity analysis was part of a GRI-industry study of the variability of Fruitland
.
coalbed properties and the effect on gas production Young et al., 1992 . The objectives
.
were to determine the sensitivity of long-term gas and water production to: 1 the
.
desorption pressure for fully saturated coals and 2 the initial gas content relative to the
.
capacity of the coal bed to hold gas degree of coal undersaturation .
If the initial coal reservoir conditions of gas content and pressure lie on the isotherm
.
Fig. 4 , the coal bed is said to be saturated, and both gas and water are produced upon
.
pumping. In this case, the desorption pressure P and initial coal reservoir pressure
d
.
P are identical. In some coals, the initial gas content lies below the sorption isotherm
i
.
Fig. 5 , which yields undersaturated coal reservoir conditions. For undersaturated
conditions, significant pressure drawdown must occur before methane can be released
from the coal bed. During this drawdown period, only water is produced while the coal
.
reservoir pressure remains above the release or desorption pressure P)P .
d
.
The isotherm used Figs. 4 and 5 represents an average curve determined from
several measured curves for the San Juan Basin wells. It is characterized by V s13.33
L
3
. .
m rtonne 427 scfrton and P s2.2 MPa 315 psia . Using the Langmuir equation
L
.
previously presented and an initial coal reservoir pressure of 9.1 MPa 1320 psia , these
3
.
constants yield an initial gas content of 10.76 m rtonne 345 scfrton .
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 376
Fig. 4. Adsorption isotherm used for the sensitivity analysis of a saturated coal.
Fig. 5. Adsorption isotherm for the sensitivity analysis of a undersaturated coal.
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 377
Fig. 6. Simulated gas rate versus time for variable desorption pressure.
Simulated results as a function of initial gas content for the three saturated cases in
Fig. 4 are shown in Fig. 6. For these simulations, initial coal reservoir pressure was
adjusted to equal the selected desorption pressure for each case. Because of the shape of
.
the isotherm, a difference in more than 7 MPa 1000 psia results in only a 15%
variation in initial gas content for fully saturated conditions. Gas recovery increases with
. .
increasing initial coal reservoir desorption pressure Fig. 6 , with most of the increased
recovery in the first 5 years.
To evaluate the performance of undersaturated coal reservoirs, initial gas content was
3
.
decreased from the saturated value of 10.76 m rtonne 345 scfrton while maintain-
.
ing the sorption isotherm and initial reservoir pressure unchanged Fig. 5 . As the
desorption pressure decreases with decreasing initial gas content, both gas and water
production decrease with the more notable differences occurring in the gas recovery. A
3
.
reduction in gas content from 10.76 to 4.53 m rtonne 345 to 145 scfrton results in a
.
93% loss in gas recovery Fig. 7 , with the greatest reduction occurring once the
desorption pressure falls below P . Fig. 7 illustrates the magnitude and timing of the
L
reductions in gas rate associated with decreasing degrees of coal gas saturation; only a
3

3
.
1.56 m rtonne 50 ft rton or 14% reduction in initial gas content results in an
approximate 60% fall in the gas rate during the initial period of production. Thus, a
small error in measured gas content can have a relatively large effect on forecasted gas
rates. As the area of effective pressure reduction around the well expands with continued
production, larger volumes of the coal reservoir are lowered below the desorption
pressure and the gas rates for the undersaturated coal beds show some increase. The
difference between the pressure at which gas can begin to desorb and the specified
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 378
Fig. 7. Simulated gas rate versus time for variations in initial gas content.
.
bottomhole pressure of 689 kPa 100 psia decreases with decreasing gas content for
undersaturated coal beds. This reduction in pressure differential has less impact on water
production because the water storage capacity of the coal bed is not as affected by the
lower desorption pressure.
5. Conclusions
Reservoir simulation is a versatile and powerful tool which can yield insight into gas
well performance in complex coal seam gas reservoirs. Pseudosteady state, nonequilib-
rium sorption formulations are required to model the mechanisms of gas release and
transport. The primary constraint in the application of a simulator is the lack of adequate
data with which to characterize reservoir performance. It is therefore essential that
reservoir simulation be used not as a stand alone analytical tool, but as an integral part
of other data collection programs designed to assist in the evaluation of coalbed methane
reservoir producibility and operating practices.
A major benefit of reservoir simulation is an economic one: a simulator can be run
many times at little expense while the field can be produced only once. Simulation can
provide several other benefits as well. For example, simulation provides rates of fluid
recovery with time, simulation can produce a better reservoir description than geology or
production testing alone, and simulation applies on any scale laboratory cores, single
wells, or entire fields. Because of the increasing need to justify economics and support
( )
G.B.C. YoungrInternational Journal of Coal Geology 35 1998 369379 379
decisions with accurate technical data, simulation has become an indispensable tool for
managing andror predicting reservoirs.
Acknowledgements
Advanced Resources International wishes to gratefully acknowledge the guidance and
support of GRI, in particular R.A. McBane and R.A. Schraufnagel, for most of the
research results discussed in this paper. On a more personal note, the author deeply
appreciates the insights and invaluable contributions of J.E. McElhiney and G.W. Paul
to an improved understanding of the principles presented here.
References
Adamson, A.W., 1973. A Textbook of Physical Chemistry. Academic Press, New York, pp. 67, 696.
King, G.R., Ertekin, T.M., 1989. A survey of mathematical models related to methane production from coal
seams. In: Proc. Coalbed Methane Symp. Univ. of Alabama, Tuscaloosa, pp. 125155.
Kolesar, J.E., Ertekin, T.M., Obut, S.T., 1990. The unsteady-state nature of sorption and diffusion phenomena
in the micropore structure of coal. SPE Formation Evaluation, pp. 6371.
Law, B.E., Rice, D.D., 1993. Hydrocarbons from coal. AAPG Studies in Geology No. 38, Tulsa, OK, 394 pp.
Paul, G.W., 1996. Simulating Coalbed Methane Reservoirs. In: Saulsberry, J.L., Schafer, P.S., Schraufnagel,
.
R.A. Eds. , A Guide to Coalbed Methane Reservoir Engineering. Gas Research Institute, Reference No.
GRI-94r0397, pp. 6.16.48.
Paul, G.W., Sawyer, W.K., Dean, R.H., 1990. Validation of 3D Coalbed Simulators. In: Proc. 65th SPE Annu.
Tech. Conf. and Exhibition. SPE Paper 20733, New Orleans, LA, pp. 203210.
Saulsberry, J.L., Lambert, S.W., Dobscha, F.X., 1991. Determining production from individual coal groups in
multi-zone wells with a zone isolation packer. In: Proc. Coalbed Methane Symp. Univ. of Alabama,
Tuscaloosa, pp. 353357.
Saulsberry, J.L., Schafer, P.S., Schraufnagel, R.A., 1996. A Guide to Coalbed Methane Reservoir Engineering.
Gas Research Institute, Reference No. GRI-94r0397.
Sawyer, W.K., Paul, G.W., Schraufnagel, R.A., 1990. Development and Application of a 3-D Coalbed
Simulator. In: Proc. 65th CIMrSPE Int. Tech. Meeting. CIMrSPE Paper 90-119, Calgary, Alberta.
Warren, J.E., Root, P.J., 1963. The behavior of naturally fractured reservoirs. Soc. Petrol. Eng. J., 245255.
Young, G.B.C., Paul, G.W., McElhiney, J.E., McBane, R.A., 1992. A parametric analysis of Fruitland coalbed
methane reservoir producibility. In: Proc. 67th SPE Annu. Tech. Conf. and Exhibition. SPE Paper 24903,
Washington, D.C., pp. 461473.
Young, G.B.C., Paul, G.W., Saulsberry, J.L., Schraufnagel, R.A., 1993. Characterization of coalbed methane
reservoirs at the Rock Creek project site, Alabama. In: Proc. Int. Coalbed Methane Symp. Univ. of
Alabama, Tuscaloosa, pp. 705714.

You might also like