Vous êtes sur la page 1sur 34

SPE 100313

Production Data Analysis of Single-Phase (Gas) CBM Wells


C.R. Clarkson, SPE, Burlington Resources Canada Ltd.; R.M. Bustin, U. of British Columbia; and J.P. Seidle, SPE,
Sproule Assocs. Inc.

Copyright 2006, Society of Petroleum Engineers


Coal reservoirs have the potential for permeability
This paper was prepared for presentation at the 2006 SPE Gas Technology Symposium held anisotropy because of their naturally-fractured nature, which
in Calgary, Alberta, Canada, 15–17 May 2006.
may complicate production data analysis. To study the effects
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, as
of permeability anisotropy upon production, a 2-D, single-
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to phase, numerical CBM reservoir simulator was constructed to
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 simulate single-well production assuming various permeability
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
anisotropy ratios. Only large permeability ratios (>16:1)
for commercial purposes without the written consent of the Society of Petroleum Engineers is appear to have a significant effect upon single-well production
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 characteristics.
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.
Multi-layer reservoir characteristics may also be observed
with coal reservoirs because of vertical heterogeneity, or in
Abstract cases where the coals are commingled with conventional
Coalbed methane (CBM) reservoirs commonly exhibit two- (sandstone) reservoirs. In these cases, the type-curve and
phase flow (gas+water) characteristics, however commercial pressure transient analysis techniques are difficult to apply
CBM production is also possible from single-phase (gas) coal with confidence. Methods and tools for analyzing multi-layer
reservoirs, as demonstrated by the recent development of the CBM (+sand) reservoirs are presented. Using simulated and
Horseshoe Canyon coals of western Canada. Commercial field examples, it is demonstrated that unique reservoir
single-phase CBM production also occurs in some areas of the properties may be assigned to individual layers from
low-productivity Fruitland Coal, south-southwest of the high- commingled (multi-layer) production in the simple 2-layer
productivity Fruitland Coal Fairway in the San Juan Basin, case.
and in other CBM-producing basins of the continental United
States. Production data of single-phase coal reservoirs may be Introduction
analyzed using traditional techniques commonly used for Coal gas (CBM) reservoirs differ substantially from
conventional reservoirs. Complicating application, however, conventional gas reservoirs due to unique gas storage and
is the complex nature of coal reservoirs; coal gas storage and transport mechanisms. Unlike conventional (sand or
transport mechanisms differ substantially from conventional carbonate) reservoirs, coals act as source rocks and reservoirs
reservoirs. In addition, single-phase coal reservoirs may to gas, and the primary gas storage mechanism is through
display multi-layer characteristics, dual porosity behavior, adsorption within the microporous organic fraction, although
permeability anisotropy etc. some gas storage may occur in the free gas state (compressed
The current work illustrates how traditional single-well in pore space), within the larger pores (macropores) and
analysis techniques, such as type-curve and pressure transient fractures. CBM reservoirs are often naturally-fractured, and
analysis, may be altered to analyze single-phase (un- may be modeled as dual (fracture+matrix porosity), or even
stimulated and hydraulically-fractured) CBM wells. Examples triple porosity (fracture, matrix free- and adsorbed gas)
of how reservoir inputs to the type-curves and subsequent reservoirs. Gas transport mechanisms vary, depending upon
calculations are modified to account for CBM reservoir the scale and location within the reservoir. For example, gas
behavior are given. This paper demonstrates, using simulated transport at the scale of the matrix between natural fractures
and field examples, that reasonable reservoir and stimulation may be due to the mechanism of diffusion whereas Darcy flow
estimates can be obtained from production data analysis of occurs in the fracture system. Single or two-phase (gas and
coal reservoirs only if appropriate reservoir inputs (i.e. water) flow can exist in these reservoirs, and hence relative
desorption compressibility, fracture porosity) are used in the permeability characteristics can be important.
analysis. As the field examples demonstrate, type-curve and The unique reservoir characteristics of CBM reservoirs in
pressure-transient analysis methods for production data turn yield distinctive fluid production profiles at the single-
analysis are not used in isolation for reservoir property well and field level. Application of conventional production
estimation, but rather as a starting point for single- and multi- data analysis (decline- and type-curve analysis) is thus
well reservoir simulation, which is then used to history-match difficult to apply to CBM wells, particularly for those
and forecast coal well production (ex. reserves assignment). exhibiting 2-phase production (gas and water). Recent
studies1-2, however, have demonstrated that conventional
2 SPE 100313

decline-curve analysis can be applied to late stage CBM fluid derivative analysis to identify flow regimes, which is critical
production. A very recent study attempted to develop for ensuring that the correct equations are applied to the
production type-curves for 2-phase CBM reservoirs, including appropriate flow regimes. Additionally those authors10
the dewatering period3-4. Commercial single-phase (gas) demonstrated that the radius-of-investigation equation can be
CBM production has been recently demonstrated in the used, at the time corresponding to when the boundary
Horseshoe Canyon coals of western Canada5 and previously in dominated flow regime is reached, to calculate the drainage
various basins in the U.S. Most conventional production data area of a well.
analysis tools (i.e. type-curve analysis) are directly applicable A complication in the application of conventional
to single-phase production only, and hence single-phase CBM production data analysis techniques to coal reservoirs is the
wells are amenable to conventional type-curve analysis, given fact that some coal reservoirs may exhibit permeability
certain conditions, and with some modification. anisotropy, which is a result of the natural fracture set
Production type-curve techniques allow reservoir data and associated with commercial coal reservoirs. A recent study by
stimulation efficiency/damage information to be extracted Cui and Bustin11 discussed the effects of coal fabric on
from single-well production data, flowing pressure and initial coalbed gas production and compositional changes; among the
PVT information (initial reservoir pressure and temperature, effects studied were cleat porosity, cleat spacing and cleat
fluid composition etc.), and some basic geologic data. These permeability. The effect of permeability anisotropy was not
techniques have grown in popularity because of the relatively considered. Coal seams of higher rank (sub-bituminous and
small amount of data required to perform the analyses, the greater) commonly contain an orthogonal fracture set,
quality of reservoir information that can be yielded, and the consisting of a more continuous face cleat, and less continuous
desirability of not shutting in production data (which is butt cleat that terminates into the face cleat (Fig. 1a). There
required by other reservoir analysis techniques such as are many controls upon cleat development including bed
material balance, flow/buildup analysis etc.). thickness, coal organic composition, rank, depth of burial and
Historically, the most popular type-curves used for uplift. Further, diagenesis may impact the degree of mineral
production data analysis (PDA) are those due to Fetkovitch6-8, infilling of cleats. It is therefore possible that a range of
although recent advances in the development of type-curves fracture geometries, continuities and intensities may exist for
have greatly expanded their applicability (variable rates and commercial coalbed reservoirs, which in turn could lead to
pressures, gas reservoirs, hydraulically-fractured wells etc.). varying degrees of permeability anisotropy. In cases where
The accuracy of the extracted information from the Fetkovich the permeability contrast between cleats sets is extreme, or
type-curves is a function of how closely the producing well where one cleat set is missing entirely, the CBM wells may
conforms to the somewhat limiting assumptions used in their long-term linear flow, as has been shown for some tight gas
derivation (constant flowing-pressure, radial flow, single-layer sand wells with natural fracturing13-14. Although several
homogeneous reservoir etc.). Further, to extract reservoir and commercial numerical simulators exist that are capable of
completion information specifically for single-phase coal modeling permeability anisotropy in coal reservoirs, a 2-D,
reservoirs, the input reservoir data must, as will be single-phase, numerical coal reservoir simulator was
demonstrated, be corrected for unique CBM storage constructed for this purpose and described in the present work.
characteristics. A series of single-well coal gas production profiles,
Modern production data analysis techniques (including corresponding to varying permeability anisotropy ratios, were
modern type-curves and flowing material balance techniques) generated with the new simulator to determine the impact of
utilize the concepts of material balance pseudo-time and permeability anisotropy.
pressure- or pseudo-pressure normalized rate to account for The PDA techniques (type-curve and PTA) discussed
variable operating conditions (i.e. changing rates and flowing above are only strictly applicable to single-layer reservoirs,
pressures) and gas property variation as a function of and hence production data analysis of multi-layer reservoirs
pressure9. Unlike the Fektovich type-curves, the modern (ex. Fig 1b), or commingled multi-reservoir systems, must
PDA techniques also utilize analytical solutions for boundary- utilize other techniques. Coal reservoirs of the San Juan Basin
dominated flow. These techniques are more advanced than have been shown to exhibit multi-layer behaviour15; the multi-
the Fetkovich type-curves and should yield more accurate seam nature of Horseshoe Canyon coals may also impart
estimates of reservoir properties, stimulation/damage multi-layer reservoir behaviour. In previous work15-17, single-
information and reserves. Although the more modern well reservoir simulators, constructed with spreadsheets for
techniques can also be used for analyzing single-phase coal specific application to coal and conventional reservoirs, were
reservoirs, with appropriate modification to the time function described. It was demonstrated17 that these tank-type
to account for desorption effects, this is not discussed in detail (analytical) simulators could be coupled with Monte Carlo
here. simulation to generate a distribution of well gas (and water,
An alternative approach to production data analysis is the where applicable) forecasts from distributions of input
use of conventional pressure transient analysis (PTA) reservoir properties. This concept is extended further in this
techniques. Hagar and Jones10 demonstrated that production paper by showing how the tank models can be used to
data may be treated as an extended drawdown test (variable approximate multi-layer (with and without crossflow)
rate) and utilized the Odeh-Jones equation (which uses a behavior, and that Monte Carlo simulation can be used to
superposition time function) to extract permeability and “history-match” production data from single-phase, multi-
stimulation information from data during the pseudo-radial layer coal wells.
flow period. The technique is combined with standard
SPE 100313 3

Production Data Analysis Techniques – Theory and Fetkovich used 1/2 to reduce the discontinuity between the
Methodology transient and depletion stems; the inflow equations in the
Several techniques are illustrated in this study for application present work, however, reference the average reservoir
to single-well, single-phase CBM gas production. These pressure during drawdown, so 3/4 was used for consistency.
include type-curve analysis of production data, pressure The calculation of q Dd and t Dd for late-time (boundary-
transient analysis of production and flowing pressure data and
dominated) flow was derived from the Arps equations21.
reservoir simulation. Reservoir simulation is applicable to all
Calculations of the early-time (transient) production of a
cases discussed for the type-curve and PTA techniques, and is
single-phase (slightly-compressible) fluid used in the
also useful for modeling complex reservoir behaviour, such as
permeability anisotropy and multi-layers. This discussion is derivation of the Fetkovich type-curves are obtained from the
limited to single-well applications. constant-flowing wellbore pressure solution for a well
centered in a closed, circular (homogenous) reservoir with
uniform thickness. The original solution for this problem was
Type-Curves. There has been considerable advancement in
derived by van Everdingen and Hurst22 in Laplace space. In
the development of production type-curves over the past
couple of decades. For example, modern techniques allow the this work, the approximate numerical curve fits to the transient
analysis of producing wells that are subject to variable flowing portion of the van Everdingen and Hurst solution, as provided
by Edwardson et al.23 were used to model the transient stems
pressures, that have been hydraulically-fractured18, that
for the Fetkovich type-curves. The type-curve set, resulting
produce from reservoirs of variable geometries14 etc. Well-
from the combination of early- and late-time solutions, is
tests analysis techniques have historically developed constant
given in Fig. 2.
rate analytical solutions to various well-test scenarios, and
modern production type-curve developers have wisely drawn The type-curve matching procedure involves matching the
from the vast inventory of well-test solutions available. The well production data, on log-log paper, with a type-curve;
from this match one can obtain information about reserves,
Fetkovich type-curves6 represent an early attempt to utilize
completion efficiency and the reservoir. Details of the
analytical solutions (transient flow regime) for the purpose of
Fetkovich type-curve matching procedures have been
extracting reservoir and stimulation/damage information from
field production data, and are somewhat limited in their discussed elsewhere24 and will not be reviewed here. Once the
application. The authors realize these limitations, but chose to match of production data to a type-curve has been made, the
start with the simple Fetkovich-style curves to illustrate how permeability, skin and drainage radius and pore volume may
be obtained from a match point using the equations given in
coal properties need to be accounted for in order to yield
Appendix A.
quantitative information. In future work, the application of
For a coal reservoir, it is the bulk volume of the reservoir
more advanced production type-curve approaches, such as
those of Blasingame et al.19 and Agarwal et al.20, will be Vb that is of interest (not just the fracture pore volume), which
attempted. is obtained by dividing the obtained V p (Eq. A-5) by the
The Fetkovich type-curves are generated through the
coupling of early-time analytical solutions to the radial flow fracture porosity φ . Eqs (A-5 and A-6) then become:
equation with the empirical Arps21 decline-curves for
boundary dominated flow only; dimensionless variables are ⎛ t ⎞ ⎛ q ⎞ 2000 Psc T
Vb = ⎜⎜ ⎟ ⎜ ⎟
used in the graphical representation of the curves. The
Fetkovich type-curves were programmed into a spreadsheet ⎝ t Dd
⎟ ⎜
⎠ match ⎝ q Dd
⎟ ( ) [ ( )] ....(5)
⎠ match φμ g ct i Tsc m(Pi ) − m Pwf
and used the following definitions of dimensionless variable
groupings:
re = Vb / πh ………..……...................................................(6)
T
qD =
[
0.000703kh m(Pi ) − m( Pwf ) ]
q g ……….………..…...(1) Once the drainage area has been established, the gas-in-
place may be determined. For a conventional gas reservoir
this calculation is as follows:
q Dd = q D [ln(re / rwa ) − 3 / 4] ……….………........................(2)
43560 Ahφ (1 − S wi )
Gi = …...…….......................................(7)
0.00634k B gi
tD = t ………………..........................................(3)
φμ gi c ti rwa
2

where A (acres) is derived assuming a circular drainage area,


2 with re as the drainage radius. For a coal reservoir, the gas-in-
t Dd =
[(r / r
e wa ) 2
]
− 1 [ln (re / rwa ) − 3 / 4]
t D …........................(4)
place calculation must include adsorbed gas:

Note that use of pseudopressure has been used to Gi = 1.3597 Ahρ cVi ………..…….........................................(8)
accommodate gas wells at all PT conditions. In the current
work, the 3/4 is used on the right hand side of Eq. 2 and right where Vi is the volume of gas adsorbed at initial reservoir
hand term of the denominator in Eq. 4 instead of 1/2
pressure, and is assumed to be corrected to in-situ conditions
4 SPE 100313

(corrected for in-situ ash and moisture content). If a compressibility effects (by incorporated cd into total
significant amount of gas is assumed to be stored in the
compressibility term), which like gas compressibility and
secondary porosity (fracture) system, the gas-in-place
calculation must be corrected to include a free-gas storage viscosity, is a function of pressure:
term:
( )i ∫0 μdtc
t
t a = μ g ct …………….......................................(12)
⎛ 43560φ (1 − S wi ) ⎞ g t
Gi = Ah⎜ + 1.3597 ρ cVi ⎟ …………............(9)
⎜ B gi ⎟
⎝ ⎠ where μ g , ct are calculated at average reservoir pressure.
For application to single-phase (gas) coal reservoirs, two The dimensionless time function (Eq. 3) would then
modifications to the dimensionless variables must be incorporate t a instead of t . Spivey and Semmelbeck27 utilize
considered: t a , corrected for desorption effects, in the development of an
Desorption. Coals are dual-porosity type reservoirs where analytical model to describe radial flow (at constant flowing
gas is stored in the matrix via adsorption (+ some possible pressure) for dewatered coal seams and fractured shales.
free-gas storage) and transported through the secondary In order to account for changing operating conditions, material
porosity (fractures or cleats) system to the wellbore. The gas balance pseudo-time (given below) should be used:
in the matrix may or may not be in equilibrium with that in the
fractures as gas pressure is lowered in the fractures. If (μ g ct )i t qdt
equilibrium exists (i.e. instantaneous desorption), the
desorption effect (gas entering the fracture set from the
t ca =
q ∫ 0 μ g ct
……………....................................(13)

matrix) must be accounted for in the total compressibility term


in the dimensionless time equation (Eq. 3). The expression for The calculation of t ca is iterative9 and, for coal, the procedure
total compressibility for a single-phase (gas) coal reservoir
would be need to include material balance corrected for
(note we have assumed 100% gas saturation for simplicity) is
desorption effects, as discussed in Spivey and Semmelbeck27.
as follows25:
Fracture Porosity. The porosity term in the calculation of
dimensionless time must correspond to the fracture porosity.
ct = c g + c f + c d ……………............................................(10) As the transport of gas in coal is commonly considered to
occur only through the fracture system (in the equilibrium
The expression for desorption compressibility is26: scenario), the effective gas pore volume of the fracture system
should be considered only. It would be tempting to use an
B g ρ cV L PL effective porosity that yields the same OGIP as the coal
cd =
( )2 φ
………………….......................(11) reservoir with adsorbed gas storage, so that the OGIP equation
32.0368 PL + P for a conventional gas reservoir could be used (as is done in
the application of the conventional Fetkovich type-curves) but
where it is assumed that VL is corrected to in-situ conditions, as will be shown, this would yield erroneous results from the
and that P is equal to the average pressure of the natural type-curve analysis.
fracture system. Also note that φ is the fracture porosity,
Pressure Transient Analysis. A recent paper10 discussed the
corrected for Swi.
application of standard pressure transient methods to the
Previous authors have noted that the desorption
analysis of conventional and tight gas sand production and
compressibility can dominate the total compressibility term for
coal reservoirs25, and thus must be accounted for if flowing pressure data. Those authors noted that the
equilibrium exists between matrix and fractures. In cases production of gas reservoirs is analogous to an extended
drawdown test, most commonly with variable rate production.
where no significant influx occurs from the matrix to the
Using superposition, the multi-rate data can be transformed
fracture, desorption compressibility can be ignored. An
into an equivalent single-rate drawdown test, and analyzed
example of this would be in the case of well test analysis of
using conventional techniques. Derivative curves can be
limited duration for coal reservoirs with a wide fracture
spacing or a very small diffusion coefficient in the coal matrix. used to analyze flow regimes, as with conventional pressure
In this work, the dimensionless time function (Eq. 3) is simply transient analysis. Hagar and Jones10 utilized the Odeh-Jones
equation for a multi-rate, drawdown test (in the infinite-acting
adjusted by including the desorption compressibility
flow regime) assuming compressible fluids:
calculated at initial reservoir pressure conditions, into the total

( ) = m ' n (q j − q j −1 ) log(t
compressibility term (Eq.10). This was done because most
m(Pi ) − m Pwf
commercial type-curve matching software packages available
at present do not allow the modification of the time function qn ∑ qn
n )
− t j −1 + m ' s ….(14)
j =1
calculations directly; the initial total compressibility term can,
however, be overridden. In order to utilize more advanced
PDA techniques for single-phase (gas) coal reservoirs, the where:
pseudo-time function (defined below for a conventional gas
reservoir) would need to be adjusted to account for desorption
SPE 100313 5

⎛ k ⎞ anisotropy on production profiles of single-phase coal


s = log⎜ ⎟ − 3.23 + 0.869s ……............................(15) reservoirs and to extract single-zone reservoir and completion
⎜ φμ c r 2 ⎟ data from multi-layer reservoirs.
⎝ gi ti w ⎠
Permeability Anisotropy. Recent papers by Arevalo-
Villagran et al.13-14 demonstrated that tight gas with parallel
and equivalent time (te) may be defined as: natural fractures could exhibit long-term linear flow

(q j − q j −1 ) log(t
characteristics and demonstrated that anisotropic reservoirs
n
te = ∑ qn
n )
− t j −1 …….…............................(16)
can be transformed into equivalent isotropic reservoirs using
the transformations shown in Fig. 3. Referring to Fig. 3, the
j =1
effect of parallel fractures would have the effect of changing
the reservoir geometry into a narrow, linear channel-type
A Cartesian plot with the left-hand-side of Eq. 14 (rate reservoir. The effects of permeability anisotropy in coal
normalized pseudopressure drop) versus te will yield a straight reservoirs are investigated using a numerical reservoir
line from which effective permeability and skin can be simulator. A 2-D numerical simulator (“COALGAS2D”),
estimated (when in infinite-acting flow regime). The capable of modeling single-phase (gas) coal production (single
permeability and skin can be estimated from the slope and or multiple wells) from a heterogeneous reservoir, was
intercept, respectively, as discussed in10. constructed for this purpose. For completeness, the
In order to strictly apply Eq. 14, the pseudoradial flow mathematical model and brief description of its numerical
regime must be identified. The use of log-log plots in solution is discussed in Appendix B. The transformations
combination with derivative curves can be used for the described shown in Fig. 3 were used to model anisotropic
purpose of identifying flow regimes, as is commonly done in permeability effects and coupled with reservoir simulation to
well-test analysis. In this case, the derivative is as follows: illustrate the effects of permeability anisotropy upon single-
well gas production profiles (see examples).
⎛ m(Pi ) − m Pwf
Δ⎜⎜
( ) ⎞⎟ Multi-layer Reservoirs. An example of how numerical
simulation and analytical models can be used to extract

ΔP (q ) qn
= ⎝ ⎠ ……..………......................(17) individual-layer reservoir data from a multi-layered reservoir
log(Δt ) Δt e was discussed in a previous paper15, using a multi-well field
example. The current work focuses on the use of single-well,
multi-layer analytical simulators15-17 to extract individual-layer
As with conventional derivative analysis, smoothing
data from single-well production and pressure data. The
functions may be applied to improve interpretation. We have
simulators described in previous work15-17 assume no-
found, as with Hager and Jones10, that derivative smoothing is
crossflow, but have recently been extended to include
a critical step in the use of PTA techniques applied to
crossflow between layers using the communicating reservoir
production data, due to the lower accuracy of field production
(CR) model described by Payne28.
data relative to well-test pressure transient data. In some cases
The multi-layer analytical simulators have been coupled
that we have examined for conventional well production data,
with Monte Carlo simulation to extract reservoir and
the inability to effectively identify flow regimes from poor
stimulation data, using only total well (commingled)
quality derivative data has lead to inaccurate analysis. Ideally,
production data and surface flowing pressure data. A similar
however, the derivative curves should enable the identification
procedure to that employed by Roadifer et al.29 can therefore
of various flow regimes, such as zero slope for infinite-acting
be used to converge on a unique solution by adjusting the
flow, ½ slope for linear flow (ex. fractured wells), and unit
input distributions of the reservoir properties that are being
slope for boundary-dominated flow.
matched (the reader is referred to Roadifer et al.29 for details).
One other useful calculation that can be made using
A least squares objective function is used to quantify the
standard pressure transient analysis is radius-of-investigation.
difference between the actual (commingled) production profile
The calculation (again for infinite-acting flow) is as follows:
and the simulated profile at each Monte Carlo iteration. We
refer to this technique as Monte Carlo/Analytical
kt Simulation/Least Squares (MCASLS) approach. Ideally, a
ri = ………………........................................(18)
948φμ gi c ti unique set of reservoir properties (and stimulation values) for
each layer exists within the final input distributions, resulting
This calculation can be used to estimate how far a transient in a simulated (commingled) production profile that matches
has traveled at a particular time during production and is the actual profile.
particularly useful for estimating drainage radius, Although additional data such as production logging data
corresponding to the onset of boundary-dominated flow. for estimation of individual-layer rates or individual-layer
In this work, the equivalent time calculations were shut-in pressure data would be useful to constrain the
programmed in MATLAB® while all other analyses were modeling inputs to achieve unique matches of the individual
performed within a spreadsheet. layer data (ex. see Bastian et al.5 ), these data have been
historically lacking for some fields studied by the authors.
Reservoir Simulation. Reservoir simulation (numerical and
analytical) is used to quantify the effects of permeability
6 SPE 100313

Example Applications The Fetkovich type-curves, with proper selection of coal


input, and modified algorithms for GIP, can therefore be used
Fetkovich Type-Curve Analysis of Production Data. to yield reasonable estimates of reservoir and stimulation
Simulated and field examples are used to illustrate the use of parameters, in the case of constant operating conditions
Fetkovich type-curves for yielding reservoir and stimulation (flowing pressure). Golan and Whitson24 discuss how to
information from single-phase coal wells. Type-curve manipulate the Fetkovich type-curve to account for abrupt
matching has historically been used by the authors to establish changes in backpressure (using the principle of superposition),
initial estimates of permeability and skin, and in some cases continuously changing flowing bottomhole pressure (rate
drainage area (single-well analysis), to be input into single- normalization) and re-stimulated wells (reinitialization).
well or field reservoir simulation. For example, for some Modern production type-curves9 have been designed to
previous studies performed by the authors, permeability account for the effects of changing backpressure through the
estimates derived from type-curve analysis were used to use of material balance time and rate-normalization.
develop kh maps that served as a starting point for reservoir Example 2 (Field, Horseshoe Canyon Coal ). This field
simulation studies. example is from a well producing from the Horseshoe Canyon
If the permeability and skin values (or drainage area in the coals of the Western Canadian Sedimentary Basin. Individual
single-well case) are adjusted during the history-match coal seams (multiple seams exist) were stimulated through
procedure, these new values are checked to see if they fall coiled tubing using a proppantless, N2 fracture-stimulation
within the range of values that could be estimated from the procedure. In the field in which the well is located, the
type-curves, or well-test data, if available. This is illustrated Horseshoe Canyon coals are of low rank (sub-bituminous) and
with the two (single-well) field examples given below. are unusual (for a low rank coal) in that they produce very
In the examples that follow, the spreadsheet-based type- little water and exhibit single-phase (gas) reservoir behaviour.
curves were used to match simulated and field single-phase A Fetkovitch type-curve match of the production data for
coal gas production data. The type-curve matches were this well is given in Fig. 5. Both the spreadsheet-based type-
checked against other available software, such as that provided curve matching software developed in this work and the New
by Munoz et al.30 (referred to in this work as New Mexico Mexico Tech. type-curve software30 were used, with very
Tech.’s type-curve software). This software package has the similar results. Input reservoir data is given in Table 3. The
ability to input user-defined porosity and total compressibility, well has been producing at a relatively stable backpressure
which is critical for matching coal production data. (estimated at sandface) of 20-25 psia, hence the constant-
Example 1 (Simulated). A commercial numerical flowing wellbore pressure assumption used in the Fetkovich
simulator31, and a single-well analytical simulator were used curve derivation is reasonable here.
to generate simulated gas production profiles (radial flow), The type-curve match results (using re/rwa = 1000 and 5000
which were then matched to the Fetkovich type-curves (Fig. stems) are given in Table 4. A drainage area of 75 – 80 acres
4). Input data used in the simulators are given in Table 1. was obtained, depending on which type-curve matching
The simulated well exhibits both transient and boundary- software was used. Due to the high permeability of the
dominated flow behavior. The early-time data from the reservoir, the well spent very little time in the transient flow
numerical model (Fig. 4a) appears to best match the re/rwa = period thus it is difficult to get a unique estimate of
1000 stem, whereas the early-time data from the analytical permeability and skin from the type-curve match. The re/rwa =
model appears to best match the re/rwa = 5000 stem (Fig. 4b). 5000 match is more consistent with the well-test (flow and
Results, obtained from matching either stem, and using Eqs. buildup test performed upon initial completion) derived
A-1 to A-3, are given in Table 2. The match results using the estimates (Table 5). The slightly positive skin is not
the re/rwa = 5000 are closest to the input values into the unexpected for this stimulation style, which is designed
simulator. We have included a higher resolution of transient mainly to remove near-wellbore damage associated with
stems (by including the re/rwa = 5000 stem) than is often drilling the well than to create large effective hydraulic
afforded with commercial software stems (often only re/rwa = fractures.
1000 and 10000 stems are shown). If this is the case, the re/rwa These estimates of permeability and skin (from type-
= 10000 stem would have been chosen for the analytical curve match and well test) and drainage area (from type-curve
model match with permeability and skin estimates of 1.1 mD match) were then input into an analytical simulator, and along
and 0.8, respectively. with the estimated flowing pressures (converted from surface)
During the boundary-dominated flow period, the depletion were used to history-match the production data for the well
stem corresponding to b=0.8 appears to be the best match to (Fig. 5b). In the history-matching procedure, the drainage
the simulated production data. The pore volume and bulk area was adjusted to match historical production decline. The
volume of the reservoir at the onset of boundary-dominated final result (80-acres) agrees very well with the type-curve
flow can be estimated using Eqs. A-5 and 5, respectively derived estimates of 75-80 acres, hence the history-matching
(Table 2). The estimated OGIP, obtained using Eq. 8, is confirmed the type-curve match results.
essentially equal to the actual value in the simulator. The It is worth mentioning that the pore volume estimates from
OGIP estimated using the pore volume alone (3.21 x 104 the type-curve match is < 0.2 MMcf, resulting in an OGIP
Mscf) is much smaller than adsorbed gas OGIP; this is estimate of ~ 1 MMscf, which is only a small fraction of the
because a fracture porosity of only 0.3% was used, translating more correct estimated OGIP (~ 152 MMscf) using the coal
into a very small pore volume. bulk volume and adsorption isotherm. This example is
extreme as we have assumed such a small value for fracture
SPE 100313 7

porosity, but illustrates how incorrect estimates of OGIP can application of PTA techniques to analyze production data.
be arrived at using type-curve matching software designed for Examples of field applications (to conventional gas wells) are
conventional reservoirs (and conventional OGIP calculations). given in Hager and Jones10. The lack of and quality of the
Horseshoe Canyon coal wells typically (as in this example) gathered flowing pressure data for the coal examples
are completed in multiple thin coal seams, which are often discussed in the previous section precluded the use of the PTA
stimulated individually; these wells have the potential to technique to extract quantitative information, thus only
exhibit multi-layer reservoir characteristics because of simulated examples of coal production data are given.
differences in coal seam productivities and gas content. In Example 1 (Simulated). In this example, the synthetic
this example, we have assumed a single-layer system and as production profile was generated using a numerical simulator
such have derived average permeability and skin values for (and assuming radial flow) and the input of Table 1. The
contacted reservoir in the wellbore. Although these data may derivative curve is shown in Fig. 8, from which the infinite-
be useful for providing an initial single-layer equivalent acting flow regime (zero slope, t < 2000 hrs) and boundary-
production profile for the well (appears to be reasonable for dominated flow regime (near unit slope, t > 4000 hrs) are
the first year of the life of the well in this example), the long evident. The late-time straight line does not exhibit an exact
term performance prediction may be inaccurate if multi-layer unit slope (on log-log plot) as expected during boundary-
reservoir behaviour is exhibited. Some operators5 have used dominated flow. It is postulated that the cause of this is due to
spinner log information along with other reservoir data to the fact that the data were plotted vs. time, and not pseudo-
distribute productivity to layers, which is a more rigorous and time (Eq. 12) – if pseudo-time were used instead of time, a
potentially accurate approach if multi-layer reservoir unit slope is expected.
behaviour exists. A cartesian plot of rate normalized pseudopressure vs.
Example 3 (Field, Fruitland Coal ). This example is from equivalent time is then generated, and a straight line fit
a well producing from the Fruitland coal of the San Juan through the data (Fig. 8) corresponding to the time-period for
Basin. The well was part of field study performed to assess infinite acting flow (obtained from derivative analysis). The
the need for infill drilling (320 to 160-acre spacing) in areas same plots were generated using data from the analytical
outside the prolific Fruitland coal Fairway15. In the simulator for comparison (Fig. 9). The permeability and skin
aforementioned study, type-curve analysis, combined with values obtained from the slope and intercept of the diagnostic
single-well history-matching, was used to provide initial plot are given in Table 7 and are in reasonable agreement with
estimates of kh and skin for input into the field reservoir the actual values of 1 md and 0 skin.
simulation performed for the study area. Little well-test data Using Eq. 18 to calculate the radius-of-investigation, and
were available in the area within or adjacent to the study area, noting that the time at the start of the unit slope (boundary-
therefore type-curve/single-well analysis was the only viable dominated flow), the well drainage area (radial) was estimated
source for reservoir/stimulation information. to be 163-165 acres, consistent with the actual drainage area of
Flowing wellhead or sandface pressure was unavailable for 160-acres.
the first 21 months of the life of this well, so calculated Example 2 (Simulated). This example uses a numerical
sandface pressures were used from the reservoir simulation simulator to generate a production profile (radial flow) that
history-match, discussed in a previous paper15. The flowing incorporates a sorption time of ~ 110 days to establish the
sandface pressure data is plotted in Fig. 6 to demonstrate that impact of non-equilibrium behaviour upon the analysis (Fig.
the pressures are declining in this example (unlike the 10). A large sorption time was chosen to exaggerate the
previous example where they were fairly constant) – for the effect; most sorption times that the authors have encountered
type-curve match estimates of reservoir properties, the first for commercial reservoirs have been considerably smaller than
year average flowing pressure was used. this. All other input are the same as in Table 1.
The type-curve matching process was challenging because The derivative curve demonstrates classic dual porosity-
little data resided on the transient portion of the curve, making type behaviour with a characteristic “dip”, followed by a
it difficult to establish a unique permeability/skin estimate. It transition into boundary-dominated flow. In this example,
is known, however, that no fracture stimulation treatment was there is no well-established radial flow period (zero-slope),
performed on this well, and wells that were similarly although a fit through the first two data points yields
completed yield positive to slightly negative skins. The re/rwa permeability (0.9 mD) and skin (0.2) values reasonably close
= 10000 stem (Fig. 7) was therefore chosen for the match to actual values (1 mD and 0, respectively).
resulting; permeability/skin estimates are given for re/rwa =
5000 and 10000 in Table 6, for both the spreadsheet type- Reservoir Simulation Analysis of Permeability Anisotropy
curves developed in this work, and the New Mexico Tech. and Multi-layer Reservoirs.
type-curve set. The single-well history-match, using the Example 1 (Permeability anisotropy, Simulated). A
analytical simulator with the type-curve match estimates of series of model runs using the COALGAS2D numerical
permeability and skin as input, is also given in Fig. 7. simulator were generated for varying anisotropy ratios (2:1,
Although the parameter estimates are somewhat non-unique, 4:1, 16:1 and 100:1) and compared. The model parameters for
they still result in a reasonable match of the well production the isotropic case are given in Table 8. The drainage area (40-
data. acres) and geometric average permeability (1 mD) was fixed
in all cases, and equations displayed in Fig. 3 were used to
Pressure Transient Analysis of Production and Flowing convert the reservoir geometry for each anisotropy case. The
Pressure Data. Simulated examples are used to illustrate the resulting forecasts are shown in Fig. 11. Significant deviation
8 SPE 100313

from the base isotropic case only appears to occur for higher 14); note that the well was re-stimulated at ~ 4.5 years, with a
ratios (16:1, 100:1), but not for lower permeability ratios (2:1, resulting increase in production (lower skin).
4:1), indicating that even if some anisotropy exists, the effect As no flowing wellhead or sandface pressure data was
(on well production profile) would be very difficult to available for this well, the production data subsequent to the
distinguish from the isotropic case. When super-imposed well test was history-matched using a (2-layer) analytical
upon a Fetkovich type-curve plot (which assumes radial flow), simulator; the well-test data (permeability, skin, reservoir
the effect of higher permeability anisotropy (i.e. 100:1) can be pressure) for each layer were input into the simulator, along
easily diagnosed (Fig. 11b) – the higher permeability with other hard reservoir data (coal thickness, fracture porosity
anisotropy translates into longer transition times from transient etc., see Table 11), and a flowing sandface pressure profile
to boundary-dominated flow (characteristic of linear flow) was created to match the production data. This flowing
which cannot be analyzed effectively with the Fetkovich type- pressure profile, which appears consistent with wells in the
curves. There are, however, type-curves available for area for which flowing pressure data were available, was used
analyzing wells with permeability anisotropy/linear flow in subsequent analysis (see Fig. 14). This approach of course
characteristics32. assumes that the well-test derived properties are accurate.
The maximum reported permeability anisotropy for CBM In this example, the skin values for each layer were
reservoirs thus far is 19:133, with most anisotropies reported to hardcoded (along with coal thickness, fracture porosity, and
be < 8:1. It is therefore expected that in most commercial gas content) and the values for permeability and drainage area
CBM fields, permeability anisotropy would not have a were varied in the history-matching multi-layer (2) MCASLS.
significant impact upon isolated single-well analysis, but Only a handful of well-test-derived permeabilities were
could certainly impact development strategy (spacing and available from wells offset to the example well, precluding a
well-orientation with respect to fracture direction) and statistical fit of a distribution to those data. Triangular
ECBM/CO2 sequestration project design34. distributions were therefore used as input for absolute
Example 2 (Commingled sand and coal, Simulated). In permeability (both layers), with the most likely (peak) set at
this example, a simulated (using analytical simulator) 50 mD, which roughly corresponds to the average of the offset
production profile of a commingled sand and coal reservoir (2- well data. The minimum and maximum were set at 1 and 100
layer) is analyzed to demonstrate how individual layer mD, respectively, representing a broad range in possible
reservoir and stimulation (skin) values can be estimated by permeabilities. A discrete uniform distribution was used for
using the MCASLS approach discussed above. The example drainage area, with a minimum of 80 and maximum of 640
is kept simple (2-layers, simplistic input distributions) to acres. 5000 Monte Carlo iterations were generated, with
demonstrate the concept. No crossflow between reservoir corresponding production profiles. The least squares objective
layers and a constant flowing bottomhole pressure is assumed. function value was used to screen the best matches; the top
Table 9 contains all relevant data input into the analytical five (smallest objective function value) matches along with the
simulator to generate the forecast. manual history-match used to derive the flowing bhp pressure
In the current example, the “unknowns” are the individual profile is given in Fig 14. Note that the history-match as
layer absolute permeabilities, skins, and drainage areas; input performed for actual production data up to the point at which
distributions (assumed discrete uniform, see Table 10) for the the well was re-stimulated (4.5 years), then a forecast was
unknowns were created in the history-matching (Monte generated with an assumed skin for post-restimulation
Carlo/analytical) simulator as a starting point for the analysis. production.
The distributions were sampled at each Monte Carlo iteration All history-matches (manual and Monte Carlo) are in good
(5000 total iterations) to generate a range of individual layer agreement with each other, and the average permeability for
and commingled production profiles (Fig. 12). The iteration each layer used in the matches are in reasonable agreement
that led to the minimized objective function yielded individual with the well-test derived values (see Table 12). 3 of the top 5
layer reservoir and skin values identical to the values input iterations have drainage areas for the two layers (320-acres),
into the analytical simulator. The final “matched” layer which is the same as the manual history-match, whereas 2
properties and production profiles (shown on Fetkovich type- iterations have a 640-acre drainage area for the lower layer.
curves) are shown in Fig 13. Although a 640-acre drainage area is possible, it is less likely
This example illustrates that in the simple case of a two- than 320-acres (well-spacing of comparable producing wells at
layer system (with reasonably constrained input distributions), the time of production history), hence these two iterations
with no crossflow between reservoir layers, a unique set of would be discounted in the final selection of best matches.
layer properties may be extracted. If a broader, continuous
range for the input distributions is assumed, then a procedure Discussion
such as that described by Roadifer et al.29 may be used to The type-curve matching methodology for single-phase coal
reduce the uncertainty in the input distributions, and provide a reservoirs described above represents a simple modification of
range of outcomes that represent a history-match. existing technology. All the commercial software products
Example 3 (Multilayer coal, Field). In this second that the authors have used for performing production data
Fruitland coal example, a well was selected for which analysis have the ability to input total compressibility and
individual layer (2 coal layers) well-test data had been porosity (but lack the correct OGIP calculation for coal), and
collected, and analyzed to obtained permeability, skin and therefore can be modified in a simplistic sense to account for
reservoir pressure (Table 11). The production history coal reservoir behaviour.
subsequent to the well-test was used to history-match (Fig.
SPE 100313 9

More sophisticated techniques, that rigorously account for anisotropy ratio) with a sorption time of zero and 110 days.
gas and desorption isotherm properties as a function of The early time production is significantly suppressed relative
reservoir pressure, and changing operating conditions (i.e. to the equilibrium (zero diffusion) scenario. In the next
wellbore pressure) can easily be developed through the example (Fig. 17), we use a numerical simulator (using radial
modification of the conventional pseudo-time or material grid) to generate a production profile with a sorption time of
balance pseudo-time, as discussed above. An example 110 days , match the profile with the Fetkovich type-curves,
application of more rigorous techniques to the Horseshoe and compare the results to the equilibrium (zero sorption time)
Canyon coal well data, is given in Fig. 15. In this example, case (input parameters given in Table 1). The early-time
we have applied flowing material balance and an adjusted production data for the non-equilibrium case appears to trend
Fetkovich type-curve, both of which account for desorption towards higher re/rwa stem values, indicative of a higher skin.
compressibility in the pseudo-time function35. The flowing In other words the effect of diffusion/non-equilibrium appears
material technique yields an OGIP (which accounts for to manifest itself as an apparent positive skin.
adsorbed gas) of 147.5 MMcf, which is very similar to the In all example cases given above, we also have assumed
value obtained from the Fetkovich match given in Table 4. no permeability changes as a function of depletion/desorption.
The permeability and skin values obtained from the adjusted Should these effects be present, it is anticipated that they
Fetkovich type-curves are also consistent with the previous would have a significant impact upon the interpretation of the
match although the permeability from the adjusted Fetkovich data using type-curves, if the type-curves are not first
curve is a little lower (8.8 mD vs 9.2 mD). corrected to include such reservoir behaviour. An illustration
Using the conventional (without pseudo-time) Fetkovich of the effect is given in Fig. 18 where we have type-curve
type-curves, the impact of inputting the wrong porosity and matched a simulated production profile (generated with a
total compressibility upon derived reservoir and stimulation numerical simulator, input parameters given in Table 1) with a
values can be significant. As discussed earlier, most existing permeability-growth curve (~ 2-fold increase from 600 to 60
commercial software packages for production type-curve psia). The calculated values of permeability and skin, using
analysis use a conventional (compressed gas in pore volume) the un-corrected type-curve, and the re/rwa = 5000 stem is 1.2
calculation to estimate GIP, and hence input of coal fracture mD, and 0.1, yielding a slightly higher permeability than
porosity (usually < 1%) will yield very small estimates of GIP actual (1 mD). The boundary-dominated flow profile does not
for coal. It would be tempting to increase the porosity to conform to the b-stem value match (0.8) for the no-
match the adsorbed GIP estimate, but this will result in permeability-growth case, and an increase in gas rate is
incorrect skin estimation. For example, referring to example 1 noticeable at early times, even though a constant flowing
of the Fetkovich type-curve match examples above, use of the pressure was assumed. Further work is required to establish
conventional V p calculation (Eq. A-5) results in an OGIP the impacts of different stress and desorption-dependent
permeability curves on type-curve match results.
estimate of 3.21 x 104 Mcf, whereas the correct (adsorbed gas)
OGIP is 1.60 x106 Mcf. If a 15% porosity is used in the Conclusions
conventional OGIP calculation (instead of the fracture This paper discusses the use of several common reservoir
porosity of 0.3%), the conventional OGIP estimate matches engineering tools for analyzing single-phase coal gas
the adsorbed OGIP, however, using the type-curve match for production data, including type-curves, pressure transient
the analytical model-generated production profile ( re/rwa = analysis, analytical (combined with Monte Carlo) and
5000 stem match), a skin value of 2.0 is obtained (compared to numerical simulation. The major conclusions are:
.1 using fracture porosity). Use of the incorrect total 1) Conventional type-curve and pressure transient
compressibility in the type-curve calculations likewise has an analysis techniques can be applied successfully to
impact upon the estimated skin. For example, if only gas non-complex reservoirs (preferably reservoirs lacking
compressibility is considered in the calculation, the type-curve strong multi-layer behavior, strong permeability
calculations would yield (using re/rwa = 5000 stem of anisotropy, large sorption times or permeability
analytical model match) a skin value = -1.5 is obtained changes with pressure/desorption), provided efforts
(compared to .1 for the case including desorption are made to account for desorption (through total
compressibility). compressibility term) and the correct reservoir data
The impact of two other potentially important coal values are input (ex. fracture porosity)
reservoir properties upon type-curve matching results have not 2) The type-curve and other production diagnostic tools
been discussed: diffusion (non-equilibrium between matrix should not be used in isolation; the reservoir and
and fractures) and absolute permeability changes as a function stimulation data resulting from those analysis should
of depletion/desorption. All type-curve examples shown thus be validated using using single- or multi-well
far assume equilibrium between matrix and fractures (zero (preferably the latter to account for well to well
sorption time). The common production type-curves interference effects if it occurs) reservoir simulation
(Fetkovich, Blasingame, Agarwal/Gardner etc.) assume single methods that involve a detailed calibration (i.e.
porosity reservoir behaviour, and thus could lead to an history-matching)
incorrect analysis if dual porosity effects are significant. A 3) The effect of permeability anisotropy (associated
simulated example is presented in Fig. 16, where a single well with natural fracturing) upon single-well production
production profile was generated with the COALGAS2D profiles appears to be significant only for large
numerical simulator (cartesian grid assuming 4:1 permeability (>16:1) anisotropy ratios
10 SPE 100313

4) Multi-layer effects upon coal production profile can Vp = pore volume of the reservoir, ft3
be significant; a methodology utilizing multi-layer Z = gas supercompressibility factor, dimensionless
analytical simulation combined with Monte Carlo z * = gas supercompressibility factor, adjusted to account for
and a least squares objective function is demonstrated adsorption effects (Eq. B-3), dimensionless
to be useful for yielding individual layer reservoir
and stimulation properties Greek Symbols
5) The techniques discussed in this work (other than αc = volume conversion factor, 5.615
reservoir simulation) do not account for non-
βc = transmissibility conversion factor, 0.001127
equilibrium between matrix and fractures in coal
γg = gas gravity, air = 1
reservoirs and permeability changes during depletion
μg = gas viscosity, cp
and hence may yield inaccurate results in those cases
φ = porosity, dimensionless, fraction
σ = coal matrix shape factor, ft-2
In future work, we would like to apply additional τ = sorption time, days
analytical techniques for production data analysis of single- ρc = coal bulk density, g/cm3
phase coal reservoirs including modern type-curves (ex.
Blasingame, Agarwal-Gardner). We would also like to Subscripts
continue to explore the effects of more complex coal reservoir D = dimensionless
behavior (permeability anisotropy, multi-layer effects, dual dD = type-curve dimensionless
porosity, permeability changes during depletion) upon the i = initial
results obtained from these simple analytical techniques and m = matrix
possibly develop new analytical techniques to handle the more n = number
complex reservoir characteristics. sc = standard conditions
Nomenclature
A = drainage area, acres Acknowledgments
Ax, Ay = gridblock cross-section area, normal to direction The authors would like to thank the management of
given in subscript (Eq. B-1), ft2 Burlington Resources for their time to review and permission
Bg = gas formation volume factor, rcf/Mscf or RB/scf to publish this paper. The authors also thank Louis Mattar of
Fekete for his helpful comments.
B g* = gas formation volume factor, adjusted to account for
adsorption effects (Eq. B-2), RB/scf References
cd = desorption compressibility, psi-1 1. Seidle, J. : “Coal Well Decline Behavior and Drainage Areas:
cf = formation compressibility, psi-1 Theory and Practice,” paper SPE 75519 presented at the 2002
cg = gas compressibility, psi-1 SPE Gas Technology Symposium, Calgary, Alberta, 30 April –
ct = total compressibility, psi-1 2 May.
2. Mavor, M.J., and Pratt, T.J.: “Powder River Basin Ft. Union
D = diffusion coefficient, ft2/d
Coal Reservoir Properties and Production Decline Analysis,”
G = gas-in-place, Mcf paper SPE 84427 presented at the 2003 SPE Annual Technical
h = formation thickness, feet Conference and Exhibition, Denver, Colorado, 5-8 October.
k = absolute permeability, md 3. Aminian, K. et al.: “Type-curves for Coalbed Methane
m’ = slope (Eq. 14) Production Prediction,” paper SPE 91482 presented at the 2004
m(P)= pseudopressure, psi2/cp SPE Eastern Regional Meeting, Charleston, West Virginia, 15-
P = reservoir pressure, psia 17 September.
PL = Langmuir pressure constant, psia 4. Aminian, K. et al.: “Type-curves for Production Prediction and
Pwf = flowing bottomhole pressure, psia Evaluation of Coalbed Methane Reservoirs,” paper SPE 97957
presented at the 2005 SPE Eastern Regional Meeting,
qg = gas surface flow rate, Mscf/d
Morgantown, West Virginia, 14-16 September.
re = drainage radius, ft 5. Bastian, P.A. et al.: “Assessment and Development of the Dry
ri = radius of investigation (Eq. 18), ft Horseshoe Canyon CBM Play in Canada,” paper SPE 96899
rw = wellbore radius, ft presented at the 2005 SPE Annual Technical Conference and
rwa = effective wellbore radius, ft Exhibition, Dallas, Texas, 9-12 October.
s = skin factor, dimensionless 6. Fetkovich, M.J.: “Decline-curve Analysis Using Type-curves,”
Sw = water saturation, dimensionless, fraction JPT (June 1980) 1065.
t = time, days 7. Fetkovich, M.J. et al.: “Decline-Curve Analysis Using Type-
ta = pseudo-time (Eq. 12) curves – Case Histories,” SPEFE (December 1987) 637; Trans.,
AIME, 283.
tca = material balance pseudo-time (Eq. 13)
8. Fetkovich, M.J. et al.: “Useful Concepts for Decline-Curve
te = equivalent time (Eq. 14) Forecasting, Reserve Estimation, and Analysis,” SPERE
T = temperature, °R (February 1996) 13.
V = adsorbed gas volume, corrected to in-situ conditions, 9. Mattar, L., and Anderson, D.M.: “A Systematic and
scf/ton Comprehensive Methodology for Advanced Analysis of
Vb = bulk volume of the reservoir, ft3 Production Data,” paper SPE 84472 presented at the 2003 SPE
VL = Langmuir volume constant, scf/ton
SPE 100313 11

Annual Technical Conference and Exhibition, Denver, 28. Payne, D.A.: “Material Balance Calculations in Tight Gas
Colorado, 5-8 October. Reservoirs: Pitfalls of P/z Plots and a More Accurate
10. Hager, C.J., and Jones, J.R.: “Analyzing Flowing Production Technique,” paper SPE 36702 presented at the 1996 SPE
Data with Standard Pressure Transient Methods,” paper SPE Annual Technical Conference and Exhibition, Denver,
71033 presented at the 2001 SPE Rocky Mountain Petroleum Colorado, 5-9 October.
Technology Conference, Keystone, Colorado, 21-23 May. 29. Roadifer, R.D. et al.: “History Matching (Reservoir Parameter
11. Cui, X., and Bustin, R.M.: “Controls of Coal Fabric on Coalbed Estimation) for Coalbed Methane Reservoirs via Monte Carlo
Gas Production and Compositional Shift in both Field Simulation,” paper 0364 presented at the 2003 Intl. Coalbed
Production and Canister Desorption Test,” SPEJ (March 2006). Methane Symposium, The U. of Alabama, Tuscaloosa,
12. Laubach, S.E. et al.: “Characteristics and Origin of Coal Cleat: Alabama, 5-9 May.
A Review ”; Intl. J. of Coal Geology (1998) 35, 175. 30. Munoz, J.D., Chen, H.Y., and Teufel, L.W.: “A Type-Curve-
13. Arevalo-Villagran, J.A. et al.: “Production Analysis of Long- Based Spreadsheet Program for History Matching and
Term Linear Flow in Tight Gas Reservoirs: Case Histories,” Forecasting Tight-Gas Production,” paper SPE 71067 presented
paper SPE 71516 presented at the 2001 SPE Annual Technical at the 2001 SPE Rocky Mountain Petroleum Technology
Conference and Exhibition, New Orleans, Louisiana, 30 Conference, Denver, Colorado, 21-23 May.
September - 3 October. 31. Schlumberger GeoQuest: ECLIPSE Reference Manual (2004A).
14. Arevalo-Villagran, J.A. et al.: “Transient Analysis of Tight Gas 32. Cox, D.O., Kuuskra, V.A., and Hansen, J.T.: “Advanced Type-
Well Performance - More Case Histories,” paper SPE 84476 curve Analysis for Low Permeability Gas Reservoirs,” paper
presented at the 2003 SPE Annual Technical Conference and SPE 35595 presented at the 1996 Gas Technology Conference,
Exhibition, Denver, Colorado, 5 - 8 October. Calgary, Alberta, 28 April – 1 May.
15. Clarkson, C.R., and McGovern, J.M.: “Optimization of Coalbed 33. Massarotto, P., Rudolph, V., and Golding, S.D.: “Anisotropic
Methane Reservoir Exploration and Development Strategies Permeability Characterisation of Permian Coals,” paper 0359
Through Integration of Simulation and Economics,” SPEREE presented at the 2003 Intl. Coalbed Methane Symposium, The
(December 2005) 502. U. of Alabama, Tuscaloosa, Alabama, 5-9 May.
16. Clarkson, C.R. and McGovern, J.M.: “Study of the Potential 34. Smith, D.H. et al.: “Simulating Carbon Dioxide
Impact of Matrix Free Gas Storage Upon Coalbed Gas Reserves Sequestration/ECBM Production in Coal Seams: Effect of
and Production Using a New Material Balance Equation,” paper Permeability Anisotropies and the Diffusion-Time Constant,”
0113 presented at the 2001 Intl. Coalbed Methane Symposium, SPEREE (April 2005), 156.
The U. of Alabama, Tuscaloosa, Alabama, 14-18 May. 35. Rapid OppSource Inc.: PDA Theory Manual, version 3.0
17. Clarkson, C.R. and McGovern, J.M.: “A New Tool for 36. Lee, W.J., and Wattenbarger, R.A.: Gas Reservoir Engineering,
Unconventional Reservoir Exploration and Development Textbook Series, SPE, Richardson, Texas (1996) 5.
Applications,” paper 0336 presented at the 2003 Intl. Coalbed 37. King, G.R.: “Material-Balance Techniques for Coal-Seam and
Methane Symposium, The U. of Alabama, Tuscaloosa, Devonian Shale Gas Reservoirs with Limited Water Influx,”
Alabama, 5-9 May. SPERE (February 1993) 67.
18. Poe, B.D. et al.: “Advanced Fractured Well Diagnostics for 38. Ertekin, T., Abou-Kassem, J.H., and King, G.R.: Basic Applied
Production Data Analysis,” paper SPE 56750 presented at the Reservoir Simulation, Textbook Series, SPE, Richardson, Texas
1999 SPE Annual Technical Conference and Exhibition, (2003), 7.
Houston, Texas, 3-6 October. 39. Manik, J., Ertekin, T., Kohler, T.E.: Development and
19. Blasingame, T.A., McCray, T.L., and Lee, W.J.: “Decline-curve Validation of a Compositional Coalbed Simulator,” J. Cdn. Pet.
Analysis for Variable Pressure Drop/Variable Flowrate Tech. (April 2002) 41, No. 4, 39.
Systems,” paper SPE 21513 presented at the 1991 SPE Gas 40. King, G.R., Ertekin, T., and Schwerer, F.C.: “Numerical
Technology Symposium, Houston, Texas, 23-24 January. Simulation of the Transient Behaviour of Coal Seam
20. Agarwal, R.G. et al.: Analyzing Well Production Data Using Degasification Wells,” SPEFE (April 1986) 165; Trans., AIME,
Combined Type-curve and Decline-curve Concepts,” SPEREE 284.
(October 1999) 478. 41. Kohler, T.E., and Ertekin, T.: “Modeling Undersaturated Coal
21. Arps, J.J.: “Estimation of Primary Oil Reserves,” Trans., AIME Seam Gas Reservoirs,” paper SPE 29578 presented at the 1995
(1956) 207, 182. SPE Rocky Mountain Regional/Low Permeability Reservoirs
22. van Everdingen, A.F., and Hurst, W.: “Application of the Symposium, Denver, Colorado, 20-22 March
Laplace Transformation to Flow Problems in Reservoirs,”
Trans., AIME (1949) 186, 305. Appendix A
23. Edwardson, M.J. et al.: “Calculation of Formation Temperature The following equations are used to extract permeability, skin,
Disturbances Caused By Mud Circulation,” JPT (April 1962) drainage area and pore volume (at the onset of boundary-
416; Trans AIME, 225.
dominated flow), using the matches of production data to the
24. Golan, M. and Whitson, C.H.: Well Performance, second
Fetkovich type-curves36:
edition, Tapir, Norway (1996).
25. Mavor, M.J.: “Coalbed Methane Reservoir Properties,” A Guide
to Coalbed Methane Reservoir Engineering, Gas Research Inst. ⎛ q ⎞ ⎛ T ⎞⎛ ⎛ re ⎞ 3⎞
k = ⎜⎜ ⎟ ⎜ ⎟⎜ ln⎜ ⎟ − ⎟
Report GRI-94/0397, Chicago (1996).
26. Bumb, A.C., and McKee, C.R.: “Gas-Well Testing in the ⎝ q Dd

⎠ match ⎝ [ ( )]
⎜ 0.000703h m(Pi ) − m Pwf ⎟⎜ ⎜ rwa
⎠⎝ ⎝

⎠ match 4 ⎟⎠
Presence of Desorption for Coalbed Methane and Devonian (A-1)
Shale,” SPEFE (March 1988), 179
27. Spivey, J.P., and Semmelbeck, M.E.: “Forecasting Long-Term
Gas Production of Dewatered Coal Seams and Fractured
Shales,” paper SPE 29580 presented at the 1995 SPE Rocky
Mountain Regional/Low-Permeability Reservoirs Symposium,
Denver, Colorado, 20-22 March.
12 SPE 100313

⎛ t ⎞ 0.00634k allows for the modeling adsorbed gas storage. Bg* , as used in
2
rwa = ⎜⎜ ⎟

⎝ t Dd ⎠ match ⎛
1 ⎛ r ⎞
2 ⎞⎛ ⎛ r ⎞ 3⎞ Eq. B-1b, is given by:
φμ g c t ⎜⎜ ⎜⎜ e ⎟⎟ − 1⎟⎜ ln⎜⎜ e ⎟
⎟ − ⎟
2 ⎝ rwa ⎠ match ⎟⎜⎝ ⎝ rwa ⎠ match 4 ⎟⎠
⎝ ⎠ Psc T z *
(A-2) B g* = …............................................................(B-2)
α c z scTsc P
⎛ r ⎞ z * is obtained from King37, modified with the following
s = ln⎜⎜ w ⎟ …………………..………………………….(A-3)
⎟ assumptions: water and formation compressibilities are
⎝ rwa ⎠ negligible, water influx is negligible, and water saturation = 0.
The equation for z * as used in this work is:
⎛ r ⎞
re = rwa ⎜⎜ e ⎟
⎟ …...…….……………………………(A-4)
⎝ rwa ⎠ match z
z* = ……...................................(B-3)
0.031214 zTPscV L ρ c
1+
⎛ t ⎞ ⎛ q ⎞ 2000 Psc T z sc Tsc (PL + P )φ
V p = ⎜⎜ ⎟ ⎜ ⎟
⎝ t Dd
⎟ ⎜
⎠ match ⎝ q Dd
⎟ (μ c ) [
T ( )]
⎠ match g t i sc m(Pi ) − m Pwf
The above equation for z assumes a Langmuir-type
*
(A-5)
isotherm.
Procedures similar to those described in Ertekin et al.38
Assuming a circular drainage area:
were used in the numerical solution of Eqs. B-1a and B-1b.
For equation B-1a, following Manik et al.39, the sorption rate
re = V p / πhφ ………...…………………………...……(A-6) is calculated using a non-equilibrium, dual porosity
formulation, assuming a pseudosteady-state between matrix
Appendix B and fractures. The advantage of the Manik et al.39 approach
A finite difference computer model was constructed to predict over the popular King et al.40 method is that oscillations in the
reservoir pressures, and gas production rates from producing sorption rate calculations are avoided (see Kohler and
wells, in a 2-D heterogeneous, isotropic, dual-porosity Ertekin41). The non-equilibrium is expressed as:
reservoir described by the following mathematical model:
∂Vm 1
= [Ve − Vm ] …......................................................(B-4)
∂ ⎛⎜ Axkx ∂P ⎞⎟ ∂ ⎛ Ayky ∂P ⎞⎟ Vφ ∂ ⎛ 1 ⎞ ∂t τ
βc Δx + ⎜ βc Δy + qsc + qm = b ⎜ ⎟
∂x ⎜⎝ μg Bg ∂x ⎟⎠ ∂y ⎜⎝ μg Bg ∂y ⎟⎠ αc ∂t ⎜⎝ Bg ⎟⎠ where Ve is the adsorbed phase volume (within a gridblock)
(B-1a) (scf) matrix-cleat boundary and Vm is the adsorbed phase
volume (within a griblock) in the coal matrix. Integrating and
Eq. B-1a describes flow of gas through a 2-D reservoir with
re-arranging:
an incompressible pore volume. q sc (scf/D) represents gas
production (or injection) rate from a well and q m represents ⎛ − Δt ⎞ k +1 ⎡ ⎛ − Δt ⎞ ⎤
Vmk +1 = Vmk exp⎜ ⎟ + Ve ⎢1 − exp⎜ ⎟⎥ ..................(B-5)
⎝ τ ⎠ ⎝ τ ⎠⎦
gas flow from the coal matrix.

A simplification of Eq. (B-1a) can be made if we assume
that the matrix and fractures are in equilibrium:
1
where τ =

∂ ⎛⎜ Ax k x ∂P ⎞⎟ ∂ ⎛ Ay k y ∂P ⎞⎟ V φ ∂ ⎛⎜ 1 ⎞⎟
βc Δx + ⎜ βc Δy + qsc = b
∂x ⎜⎝ μ g Bg ∂x ⎟⎠ ∂y ⎜⎝ μ g Bg ∂y ⎟⎠ αc ∂t ⎜ Bg* ⎟
⎝ ⎠ The source term in Eq. B-1a then becomes:
(B-1b)
qm =
Δt
[
−1 k +1
]
Vm − Vmk …......................................................(B-6)
This allows a conventional single-phase (single-porosity)
simulator to be converted to a single-phase (equilibrium) coal
simulator through the use of Bg* in place of B g , which
SPE 100313 13

TABLE 1. INPUT PARAMETERS FOR SIMULATED 1-PHASE COAL GAS PRODUCTION.

INPUT PARAMETER PARAMETER VALUE

THICKNESS (FT) 40

BULK DENSITY (G/CM )3 1.55

CLEAT POROSITY (%) 0.3

GAS GRAVITY 0.554

ABSOLUTE PERMEABILITY (MD) 1.0

INITIAL RESERVOIR PRESSURE (PSIA) 600

RESERVOIR TEMPERATURE (ºF) 120

LANGMUIR VOLUME (SCF/TON, IN-SITU) 250

LANGMUIR PRESSURE (PSIA) 661

DRAINAGE AREA (ACRES) 160


WELLBORE DIAMETER (INCHES) 7.875
SKIN 0
FLOWING BOTTOMHOLE PRESSURE (PSIA) 100

TABLE 2. TYPE-CURVE MATCH OF SIMULATED PRODUCTION DATA.

RESERVOIR PROPERTY FETKOVICH TYPE-CURVE MATCH


RE/RWA=1000 RE/RWA=5000

PERMEABILITY (MD) 0.8 1.0


SKIN -1.5 0.1
PORE VOLUME (FT3) - 8.34 X 105
BULK VOLUME (FT3) - 2.78 X 108
OGIP* (MSCF) - 1.60 X 106

* CALCULATED USING EQ. 8, NEGLECTING FREE-GAS STORAGE

TABLE 3. RESERVOIR DATA FOR HORSESHOE CANYON COAL WELL EXAMPLE.

INPUT PARAMETER PARAMETER VALUE

THICKNESS (FT) 49.9

BULK DENSITY (G/CM )3 1.33

CLEAT POROSITY (%) 0.1

GAS GRAVITY 0.55

INITIAL RESERVOIR PRESSURE (PSIA) 86

RESERVOIR TEMPERATURE (ºF) 67

LANGMUIR VOLUME (SCF/TON, IN-SITU) 155

LANGMUIR PRESSURE (PSIA) 547


14 SPE 100313

TABLE 4. TYPE-CURVE MATCH OF HORSESHOE CANYON COAL WELL DATA.

RESERVOIR PROPERTY FETKOVICH TYPE-CURVE MATCH


RE/RWA=1000 RE/RWA=5000

PERMEABILITY (MD) 7.3 9.2


SKIN -1.4 0.2
PORE VOLUME (FT3) - 1.74 X 105
BULK VOLUME (FT3) - 1.74 X 108
DRAINAGE AREA (ACRES) - 80
OGIP* (MSCF) - 1.52 X 105

* CALCULATED USING EQ. 8, NEGLECTING FREE-GAS STORAGE

TABLE 5. RESULTS OF INITIAL WELL TEST (FLOW AND BUILDUP) FOR HORSESHOE CANYON WELL.

INPUT PARAMETER PARAMETER VALUE

PERMEABILITY (MD) 9.2

SKIN 0.2

P* (PSIA) 86.1

TABLE 6. TYPE-CURVE MATCH OF FRUITLAND COAL WELL DATA.

USING SPREADSHEET TYPE-CURVES GENERATED IN THIS WORK:

RESERVOIR PROPERTY FETKOVICH TYPE-CURVE MATCH


RE/RWA=5000 RE/RWA=10000

PERMEABILITY (MD) 9.9 10.8


SKIN -0.9 -0.3

USING NEW MEXICO TECH. TYPE-CURVES:

RESERVOIR PROPERTY FETKOVICH TYPE-CURVE MATCH


RE/RWA=5000 RE/RWA=9000

PERMEABILITY (MD) 9.9 9.9


SKIN -0.9 -0.4
SPE 100313 15

TABLE 7. PRESSURE TRANSIENT ANALYSIS OF NUMERICAL AND ANALYTICAL MODEL-GENERATED


PRODUCTION DATA.

RESERVOIR PROPERTY PTA


NUMERICAL MODEL ANALYTICAL MODEL

PERMEABILITY (MD) 1.0 1.0


SKIN -0.3 0.2
*DRAINAGE AREA (ACRES) 165 163

* CALCULATED USING EQ. 18, AND THE TIME AT THE START OF THE UNIT SLOPE (APPROXIMATE).

TABLE 8. INPUT PARAMETERS FOR SIMULATED 1-PHASE COAL GAS PRODUCTION (USING COALGAS2D
NUMERICAL SIMULATOR).

INPUT PARAMETER PARAMETER VALUE

THICKNESS (FT) 20

BULK DENSITY (G/CM ) 3 1.55

CLEAT POROSITY (%) 0.3

GAS GRAVITY 0.554

ABSOLUTE PERMEABILITY (MD) 1.0

INITIAL RESERVOIR PRESSURE (PSIA) 600

RESERVOIR TEMPERATURE (ºF) 120

LANGMUIR VOLUME (SCF/TON, IN-SITU) 250

LANGMUIR PRESSURE (PSIA) 661

DRAINAGE AREA (ACRES) 40


WELLBORE DIAMETER (INCHES) 6.25
SKIN -1
FLOWING BOTTOMHOLE PRESSURE (PSIA) 150

TABLE 9. INPUT PARAMETERS FOR SIMULATED 2-LAYER COAL/SAND GAS PRODUCTION.

INPUT PARAMETER PARAMETER VALUE


COAL SAND
THICKNESS (FT) 10 10

BULK DENSITY (G/CM3) 1.45 -

CLEAT (COAL)/TOTAL (SAND) POROSITY (%) 0.3 8.0

GAS GRAVITY 0.554 0.554

ABSOLUTE PERMEABILITY (MD) 1.0 0.5

INITIAL RESERVOIR PRESSURE (PSIA) 1000 1000

RESERVOIR TEMPERATURE (ºF) 100 100

LANGMUIR VOLUME (SCF/TON, IN-SITU) 302 -

LANGMUIR PRESSURE (PSIA) 600 -

DRAINAGE AREA (ACRES) 160 320


WELLBORE DIAMETER (INCHES) 6.25 6.25
SKIN -1 -2
FLOWING BOTTOMHOLE PRESSURE (PSIA) 100 100
16 SPE 100313

TABLE 10. INPUT PARAMETER DISTRIBUTIONS (DISCRETE UNIFORM DISTRIBUTION ASSUMED FOR SAND
AND COAL LAYERS) FOR MCASLS APPLICATION TO SIMULATED 2-LAYER COAL/SAND GAS PRODUCTION.
ALL OTHER RESERVOIR AND OPERATIONS DATA FIXED ACCORDING TO TABLE 9.

DISCRETE UNIFORM
INPUT PARAMETER DISTRIBUTION DATA

ABSOLUTE PERMEABILITY (MD) {.1, .5, 1, 1.5, 2}

SKIN {1, 0, -1, -2}


DRAINAGE AREA (ACRES) {80, 160, 320, 640}

TABLE 11. PARAMETERS USED IN (MANUAL) HISTORY-MATCH OF FRUITLAND COAL WELL (2-LAYER) GAS
PRODUCTION. PERMEABILITY, SKIN AND INITIAL RESERVOIR PRESSURE FOR EACH LAYER OBTAINED
FROM FLOW/BUILDUP TEST PERFORMED JUST PRIOR TO HISTORY-MATCH PERIOD.

INPUT PARAMETER PARAMETER VALUE


UPPER COAL LOWER COAL
THICKNESS (FT) 18 17

BULK DENSITY (G/CM3) 1.44 1.44

CLEAT POROSITY (%) 1.0 1.0

GAS GRAVITY 0.554 0.554

ABSOLUTE PERMEABILITY (MD) 55 41

INITIAL RESERVOIR PRESSURE (PSIA) 193 145

RESERVOIR TEMPERATURE (ºF) 89 89

LANGMUIR VOLUME (SCF/TON, IN-SITU) 334 334

LANGMUIR PRESSURE (PSIA) 823 823

DRAINAGE AREA (ACRES) 320 320


WELLBORE DIAMETER (INCHES) 6.25 6.25
SKIN -1.61 1.5
SPE 100313 17

TABLE 12. RESULTS OF TOP 5 ITERATIONS FOR HISTORY-MATCH, BASED UPON MINIMIZED LEAST SQUARES
OBJECTIVE FUNCTION. THE INPUT PARAMETERS USED FOR THE MANUAL HISTORY-MATCH ARE ALSO
GIVEN.

ITERATION INPUT PARAMETER PARAMETER VALUE


UPPER COAL LOWER COAL
1 PERMEABILITY (MD) 51.5 41.4
DRAINAGE AREA 320 320
(ACRES)

2 PERMEABILITY (MD) 53.6 39.0


DRAINAGE AREA 320 320
(ACRES)

3 PERMEABILITY (MD) 51.3 30.9


DRAINAGE AREA 320 640
(ACRES)

4 PERMEABILITY (MD) 49.5 41.7


DRAINAGE AREA 320 320
(ACRES)

5 PERMEABILITY (MD) 55.6 28.8


DRAINAGE AREA
320 640
(ACRES)

MANUAL HM PERMEABILITY (MD) 55.0 41.0


DRAINAGE AREA
320 320
(ACRES)
18 SPE 100313
SPE 100313 19
20 SPE 100313
SPE 100313 21
22 SPE 100313
SPE 100313 23
24 SPE 100313
SPE 100313 25
26 SPE 100313
SPE 100313 27
28 SPE 100313
SPE 100313 29
30 SPE 100313
SPE 100313 31
32 SPE 100313
SPE 100313 33
34 SPE 100313

Vous aimerez peut-être aussi