00 positive Bewertungen00 negative Bewertungen

2 Ansichten17 SeitenCBM

Jul 16, 2019

© © All Rights Reserved

PDF, TXT oder online auf Scribd lesen

CBM

© All Rights Reserved

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

2 Ansichten

00 positive Bewertungen00 negative Bewertungen

CBM

© All Rights Reserved

Als PDF, TXT **herunterladen** oder online auf Scribd lesen

Sie sind auf Seite 1von 17

Depletion for CBM Reservoirs

C.R. Clarkson, ConocoPhillips, Z. Pan, CSIRO Petroleum Resources, I. Palmer, Higgs-Palmer Technologies,

S. Harpalani, Southern Illinois University Carbondale

This paper was prepared for presentation at the 2008 SPE Annual Technical Conference and Exhibition held in Denver, Colorado, USA, 21–24 September 2008.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been

reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its

officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to

reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract

It is well known that absolute permeability changes occur in coalbed methane (CBM) reservoirs during primary depletion or

enhanced recovery/CO2 sequestration operations. Sorption-induced strain in CBM reservoirs, also known as matrix-

shrinkage or -swelling, may dominate permeability changes at low pressures, as is the case for CBM wells undergoing

primary depletion in the Fruitland Coal fairway of the San Juan Basin.

Several analytical models have been developed to predict changes in coal permeability as a function of stress and

sorption. Most models, however, utilize an empirical method for estimating sorption-induced strain. Recently, a theoretical

model for sorption-induced strain was developed and applied to single-component adsorption/strain experimental data. The

new model was developed from basic thermodynamic principles and is more predictive than the empirically-based

approaches. In this paper, the theoretical model is expanded to incorporate multi-component adsorption models that are more

rigorous, and sometimes more accurate, than the commonly-applied Extended Langmuir equation. This improves predictions

of multi-component gas sorption-induced strain, as demonstrated by comparison to experimental data. The new sorption-

induced strain model is then used to calculate the sorption-strain component of the popular Palmer and Mansoori equation,

which in turn can be used to model permeability changes during both primary (single or multi-component gas) and enhanced

recovery operations. Finally, the coupled sorption-strain/permeability model, incorporated into an analytical simulator, is

used to predict and match permeability growth in a producing CBM well in the Fruitland Coal fairway, which has a binary

(CH4+CO2) sorbed/produced gas composition.

Matches to field-derived permeability growth using the new model are accurate but non-unique, due to lack of available

data, particularly rock mechanical properties. Given the availability of rock mechanics and adsorption isotherm data, the

rigorous thermodynamic basis of the new model should allow for more accurate predictions of coalbed permeability changes,

but further testing is required.

Introduction

Volumetric change in the coal matrix during the depletion of coalbed gases (matrix-shrinkage) and injection of non-

hydrocarbon gases is a well-known phenomenon, and has received a great deal of attention in the recent literature [select

references include Chikatamarla et al. (2004); Mitra and Harpalani (2007); Pan and Connell (2007); Karacan (2007);

Mazumder and Wolf (2008)] due to the impact on primary and enhanced recovery (or greenhouse gas sequestration)

characteristics of CBM reservoirs. Specifically, sorption-induced matrix-shrinkage/swelling, which has been quantified in

the laboratory, is known to impact the absolute permeability of coals, due to the change in the aperture dimensions of natural

fractures that segregate the coal matrix blocks (Fig. 1).

Until recently, empirical approaches have been used to describe volumetric strain changes of the coal matrix associated

with adsorption. For example, Levine (1996), used a Langmuir-like curve to fit experimentally-derived linear strain data,

that in turn was measured during single-component gas (CH4 and CO2) adsorption on coal. Recently, Mavor et al. (2004)

used an expression analogous to the Extended Langmuir equation, to calculate volumetric strain associated with mixed gas

adsorption. Other authors [Chikatamarla et al. (2004)] have noted that swelling is proportional to gas content, and have

extended this concept to the mixture of gaseous components assuming a linear combination of volumetric strain caused by

each gas component. As noted in Cui and Bustin (2005), the relationship between volumetric strain and sorbed gas content

may not always be linear.

Recently, Pan and Connell (2007), introduced a theoretical model to describe the volumetric strain associated with gas

2 SPE 114778

adsorption on coal, which was derived using an energy balance approach, assuming that the surface energy change associated

with adsorption can be equated with the elastic energy change of the solid. Those authors validated the new model against

single-component strain data, and introduced a simple multi-component gas extension of their model using the Extended

Langmuir equation. There was no attempt to incorporate the matrix strain model into existing analytical models for CBM

permeability changes as a function of pressure/sorption, such as the Palmer and Mansoori (P&M) equation (Palmer and

Mansoori 1998, Palmer et al. 2007). The strengths of the new model lie in its rigorous thermodynamic derivation, and the

non-reliance on single-component strain data to make predictions of strain associated with multi-component adsorption; the

model does require, however, estimation of rock mechanical properties for the coal solid matrix which are not commonly

measured.

The primary innovation of the current work is the extension of the Pan and Connell (P&C) model to include rigorous

multi-component adsorption calculations and the coupling of the P&C model with analytical models for calculating

permeability changes, such as the P&M equation. The current work (1) illustrates how the Pan and Connell (2007) model for

matrix strain associated with multi-component adsorption may be altered using multi-component adsorption models other

than Extended Langmuir; (2) compares the matches of the multi-component extension of the Pan and Connell model, using

various multi-component adsorption models, to experimental data; (3) demonstrates how the strain model may be

incorporated into analytical models for predicting permeability changes as a function of pressure and adsorption/desorption;

and (4) compares the modified permeability model predictions against field data.

Theory

Multi-Component Sorption Strain Calculations Using the Pan and Connell Model. The basic premise of the Pan and

Connell (2007) model for sorption-induced strain is that volumetric strain of the coal matrix is analogous to the dilation of

any porous body when its surface energy changes as a result of immersion in a fluid or by gas adsorption. The surface energy

of an adsorbent is reduced through the adsorption of gas; when gas desorbs, the surface energy is increased, causing the

adsorbent to shrink to reduce its exposed surface area. Pan and Connell (2007) cited the work of Scherer (1986) and Bentz et

al. (1998), who modeled the dilation of porous glass due to the adsorption of water vapour. The energy balance approach

was developed by first deriving an analytical expression for surface energy from fundamental thermodynamic expressions

that can be applied at the high pressures associated with CBM. A pore structure model relating the surface energy to elastic

energy of the solid was also introduced, allowing calculation of matrix dilation associated with adsorption. The Pan and

Connell (2007) model derivation is included in the Appendix for completeness. A fundamental equation derived by those

authors, relating strain due to adsorption and pressure compression to surface potential in the coal microporosity, coal solid

rock mechanical properties, and pore structure information, is given below:

Φρ s P

ε =− f ( x,ν s ) − (1 − 2ν s ) ………………..…..…...…………………………………………..………...………..…..(1)

Es Es

In their original work, Pan and Connell (2007) used the Langmuir model to calculate the surface potential for single-

component gases as follows (using pressure instead of fugacity for simplicity):

P

0

P

For low pressure applications, where ∫0

V a dP may be neglected, substituting Eq. 2 into Eq. 1 yields:

ρs P

ε = RTL ln(1 + BP ) f ( x,ν s ) − (1 − 2ν s ) …………..………......................................................................………….....(3)

Es Es

For multi-component gaseous mixtures, Pan and Connell noted that surface potential may be calculated using the Extended

Langmuir Model as follows:

C

Φ = ∫ V a dP − RT ∑ Li ln(1 + Bi y i P ) ……………….…...……………………...………………………..………...…….....(4)

P

0

i =1

Some authors [e.g., Hall et al. (1994)] have noted that the Extended Langmuir (EL) model may not always yield an

accurate correlation of multi-component adsorption data. The authors’ experience is that the EL equation works well for low

pressure systems, particularly for binary mixtures of CH4 and N2, but that EL performs worse for mixtures containing CO2. In

the current work, we use additional multi-component adsorption models for surface potential calculation including the Ideal

Adsorbed Solution (IAS) model (Myers and Prausnitz 1965) and the 2-D EOS model (Hall et al. 1994). In order to apply

SPE 114778 3

either the IAS or 2-D EOS models, it is assumed that surface potential may be calculated as follows:

Φ = −πA ………………….…………...………………………..………………………..…...………..….……...…….……..(5)

Note that the expression above was derived from the surface thermodynamics approach. As Myers (2002) points out, the

concept of spreading pressure ( π ) has no physical meaning in microporous solids, and surface area cannot be measured

accurately, hence the physics of adsorption is best described in the framework of solution thermodynamics (i.e. using surface

potential). Nonetheless, the mathematical equivalence between surface potential and − πA will prove useful, as will be

shown. Use of IAS Theory and 2-D EOS for calculating Eq. 5 for binary mixture of gases is discussed in the Appendix.

Coal Absolute Permeability Change Calculations Using the Palmer and Mansoori Equation. The Palmer and Mansoori

equation (Palmer and Mansoori 1998) was developed to predict absolute permeability changes in coalbed reservoirs

associated with (1) pressure changes and (2) matrix shrinkage effects due to desorption. The original version of the equation

is given below:

3 3

ka ⎛ φ ⎞ ⎡ cf ε ⎛K ⎞⎛ P Pi ⎞⎤

= ⎜⎜ ⎟⎟ = ⎢1 + (P − Pi ) + ∞ ⎜ − 1⎟⎜⎜ − ⎟⎟⎥ ……….....................................................…..………..(6)

k a ,i ⎝ φ i ⎠ ⎢⎣ φ i φi ⎝ M ⎠⎝ P + Pε Pi + Pε ⎠⎦⎥

where:

1 ⎡K ⎤

cf = − + f − 1⎥ β …………..………...............................................................................................................………..(7)

M ⎢⎣ M ⎦

Recently, Palmer et al. (2007) modified the P&M equation to attempt to account for the apparent exponential increases in

absolute permeability as a function of reservoir pressure depletion, derived from field data, as demonstrated in previous

studies (Zahner, 1997, Clarkson and McGovern, 2003, and Gierhart et al., 2007). Further, the new equation can be used to

model undersaturated CBM reservoirs by setting Pi = Pd, where Pd = desorption pressure (Pd < Pi). A simplified version of

the new equation can be obtained if it assumed that the M and K moduli are static with pressure depletion, but with a new

definition of cf:

g ⎡K ⎤

cf = − ⎢ + f − 1⎥ β …………..………...............................................................................................................……….(8)

M ⎣M ⎦

where g is a geometric term (in the range of 0-1) related to the orientation of the natural fracture system. Assuming two

orthogonal cleat sets, Palmer et al. (2007) noted that g may be low (< 0.3) for vertical cleats, and the exact explanation for the

use of this geometric factor is given in Palmer et al. (2007) and Palmer (2008). The result is that g reduces the impact of

pressure-depletion on cleat closure for coals with a matchstick-type fracture geometry and vertical or sub-vertical cleats.

Palmer et al. (2007) demonstrated that the new model (with pressure-dependent moduli) was better able to predict the actual

(field-derived) absolute permeability growth curves observed in the San Juan Basin (Zahner 1997, Clarkson and McGovern

2003, Gierhart et al. 2007).

Coupling of the P&C and P&M Models. In the original derivation of the P&M equation (Palmer and Mansoori 1998) the

volumetric sorption-strain associated with matrix shrinkage was assumed to follow a Langmuir-type curve. Later work

(Mavor et al. 2004) expanded the application of the P&M equation to multi-component gases, which is useful for some

primary depletion operations where the produced gas is composed of multiple components or for enhanced recovery (ECBM)

operations, by using the Extended Langmuir equation to model sorption-strain associated with multi-component

adsorption/desorption. Sorption-strain curves for the individual gas components are thus required using this approach. A

more general form of the P&M equation is obtained as follows:

3 3

ka ⎛ φ ⎞ ⎡ cf ⎤

= ⎜⎜ ⎟⎟ = ⎢1 + (P − Pi ) + 1 ⎛⎜ K − 1⎞⎟(Δε )⎥ ………................................................................................…..………..(9)

k a ,i ⎝ φ i ⎠ ⎣ φi φi ⎝ M ⎠ ⎦

Where Δε is the total volumetric strain due to single- or multi-component gas adsorption. This more general form of the

P&M equation permits the use of other methods for calculating sorption-induced strain, such as the approach suggested by

Pan and Connell (2007); for this purpose, linear strain predicted by the P&C model would have to be converted to volumetric

strain, which is assumed to be 3 x linear strain. Admittedly, strain can be anisotropic, and this may be one aspect of the P&C

model that could be improved in the future. It should be noted that the P&C model does not require that the single-

4 SPE 114778

component gas sorption strain-curves be known to calculate strain associated with multi-component gases, but does require

additional rock mechanical properties (for solid coal) and adsorption isotherm data for the individual gaseous components.

Although we have used the P&M equation in this work, it is also possible to use other analytical permeability equations that

have been developed for CBM. A summary and comparison of these equations is given in Palmer (2008). An example

would be the Shi and Durucan (S&D) equation (Shi and Durucan, 2005); this equation can be modified in a similar fashion to

the P&M model by incorporating the sorption strain calculated with the P&C model.

Example Applications

Binary Gas Sorption-Induced Strain Predictions. In this section, we first compare the predictions of the P&C model,

using different multi-component adsorption isotherm models (Extended Langmuir, IAS and 2-D EOS) as input, then compare

the predictions of the model against publicly-available experimental data.

Using Levine’s (1996) sorption strain measurements for methane and carbon dioxide as input (see Table 1), sorption-

induced strain predictions using the P&C model for the binary gas mixture (at 100 and 500 psia total pressure) are given in

Fig. 2. The P&C model predictions at low pressure (100 psi) are similar, regardless of the adsorption model used, however,

predictions using the Extended Langmuir (P&C-EL) equation diverge at higher pressure (500 psia). Clearly the P&C model

predictions for binary gas sorption-induced swelling will differ, depending on which multi-component adsorption model is

used as input. It is hypothesized that use of the more rigorous binary gas adsorption model will yield more accurate strain

calculations.

The P&C model was then applied to experimental strain data provided by Mitra and Harpalani (2007), collected on

Illinois Basin coal. In that work, coal swelling associated with CO2 and CH4 adsorption was measured, along with coal

matrix compression using helium. Further, swelling associated with CO2 injection following saturation with CH4 (while

keeping total pressure constant) was measured to evaluate differential swelling of the coal caused by CO2 adsorption. Mitra

and Harpalani (2007) then modeled the mixed gas sorption-induced strain using an equation of the same form as Extended

Langmuir for mixed gas adsorption. Langmuir-type fits to the measured single component strain were used as input. The

authors noted that the Extended Langmuir analog model for calculating sorption-induced strain did not provide an acceptable

match to the binary gas sorption-induced strain data.

In the current study, adsorption isotherms and matrix compressibility collected by Mitra and Harpalani (2007) were used

as a starting point for the analysis. The matrix compressibility was used to estimate Ks and Es, which is a key input into the

P&C model. Ideally, the single-component gas version of the P&C model would first be used to match the single-component

(CO2 and CH4) strain data at multiple pressures, and the resulting match parameters used to predict the binary-gas strain data;

however, because the pure component strain data were collected on separate samples, which were not identical in

composition and thus rock mechanical properties, an alternate approach was taken. In the mixed-gas strain experiment

performed by Mitra and Harpalani (2007), methane, in an initially methane-saturated sample, was gradually replaced by CO2

while keeping the total pressure constant. The linear strain (approximated by dividing volumetric strain by 3) corresponding

to the end-point gas compositions in the experiment (0 and 99% CO2) were first matched by the P&C model by adjusting ν s ,

and then the P&C model was used to predict the linear strain for all binary gas compositions between those two end-

members. The P&C predictions, using Extended Langmuir (P&C-EL), IAS (P&C-IAS) and 2-D EOS (P&C-2D EOS) for

multi-component sorption calculations, are compared against the experimental data in Fig. 3. The inputs to the P&C model

are provided in Table 2. The match to the binary-gas strain data using the P&C-IAS and P&C-2D EOS models is very good,

while the P&C-EL model does a poorer job predicting the strain. This limited test of the P&C model for prediction of binary

gas sorption-induced strain appears to indicate that more rigorous adsorption models are required for accurate predictions.

This will continue to be tested in the future.

Absolute Permeability-Growth Predictions Using Combined P&C/P&M Model. In a previous study (Clarkson et al.

2008), a Fruitland Coal fairway CBM well (San Juan Basin) was history-matched, using flowing material balance as a

diagnostic to assist the matching effort. Shut-in pressures (as estimated by a nearby pressure observation well), binary gas

composition (which changes during depletion due to relative adsorption effects), and flowing pressures were matched as part

of that effort. Critical to the matching success was the representation of permeability growth in the analytical simulator used

for history-matching. The Fruitland Coal CBM reservoir in this portion of the Fruitland Coal Fairway exhibits near-

exponential growth in absolute permeability as a function of depletion, as noted by previous authors (Clarkson and

McGovern 2003, Gierhart et al. 2007). In the history-matching effort, Clarkson et al. (2008) simply fit an exponential curve

to the effective gas permeability established from field data using the radial flow equation; there was no attempt to apply an

analytical permeability-growth equation, such as the P&M equation, to the data. It was also assumed that the trend in kg-

growth was identical to absolute permeability (ka) growth for the lower pressure range observed, which is supported by recent

work (Gierhart et al. 2007).

In the present work, the flowing pressures from the same Fruitland Coal well were re-matched by using the combined

P&C/P&M model to represent permeability-growth. The exact same inputs as were used in the original history-matching

effort were used here, except for the permeability growth-curve. Unfortunately, not all the data required for input into the

SPE 114778 5

combined P&C/P&M were available for this trial. For the P&M model portion, many of the inputs are consistent with those

used by Palmer et al. (2007) in previous efforts to match permeability-growth in the Fruitland Coal Fairway. For the P&C

model portion, we applied the original CO2 and CH4 isotherm and coal density data used in the previous history-match effort,

and used the grain compressibility (1.3x10-6 psi-1), commonly assumed by Palmer, to estimate the solid phase parameters Ks

and Es. Poisson’s ratio for the solid (ν s ) remained unknown and was varied in the matching effort; we have also set ν in

the P&M equation equal toν s , which we realize is an approximation, but is assumed to be reasonable. Young’s Modulus is

fixed at E = 300,000 psia (2.07 GPa) based on guidelines provided by Palmer et al. (2007). The other variable in the

matching effort was fracture porosity, although guidelines exist for this, based upon previous reservoir analysis. All inputs

used in the combined P&C/P&M model are given in Table 3.

The resulting permeability-growth profile, predicted from Pri = 932 psia (initial estimated reservoir pressure for this well)

using the P&C/P&M model, is compared with the field-derived kg-growth curve (see Clarkson et al. 2008) in Fig. 4. The EL

adsorption model was used to calculate binary adsorption based upon the pure component isotherm input. The match to the

field-derived curve appears reasonable, but is non-unique because of the uncertainty in key input parameters. A match using

the IAS adsorption model was also good (not shown), but required a slightly different fracture porosity as input (~ .05 % vs.

.06% using EL). These porosities are very low, but not inconsistent with values obtained from history-matching water

production from wells in the Fruitland Coal fairway.

The resulting analytical simulator history-match of flowing pressure data is shown in Fig. 5. The match appears similar

to that given previously (Clarkson et al. 2008). Matches of gas composition, and reservoir pressures were shown previously

(Clarkson et al. 2008).

Strictly-speaking, the ka,i in the P&M equation should correspond to virgin (undepleted) reservoir conditions whereas we

have set ka,i = absolute permeability at 932 psia (as estimated from field data). A second attempt at matching the

permeability-growth was tried by setting initial reservoir = 1450 psia, which is the estimated initial virgin pressure for this

portion of the Fairway; for this effort, the original gas composition had to be estimated as well. The resulting match is shown

in Fig. 6; note that ka /ka,i at 932 psia > 1 (~ 8). The permeability growth curve, derived using the exact same inputs as

before, except with lower fracture porosity of 0.03% instead of 0.06%, is very similar to the previous match. The resulting

flowing pressure match is also similar (not shown).

In summary, the combined P&C/P&M approach for predicting permeability growth in the Fruitland Coal Fairway is

promising, although the match results are non-unique and more work (and data) will be required to further constrain the

matching.

Discussion

In this work, we have modified the P&C model by incorporating rigorous multi-component adsorption models that can be

used for predicting sorption-induced strain in either primary or enhanced recovery (ECBM) scenarios. Further, we have

incorporated the P&C sorption-induced strain calculations into the P&M model for predicting permeability changes as a

function of pressure and desorption. Our current analysis indicates that: 1) incorporation of rigorous adsorption models (ex.

IAS or 2-D EOS) appears to improve P&C model predictions of binary gas (CO2 and CH4) sorption-induced strain relative to

the use of Extended Langmuir model; and 2) the coupling of the P&C model with the P&M model appears to be an effective

method for predicting permeability-growth in Fruitland Coal fairway coal wells operated under primary-recovery scenarios.

The sorption strain values calculated with the P&C model, when fit with a Langmuir-type model for strain, yield ε ∞ Pε

significantly different than those used by Palmer (4 vs. 16.5) to obtain permeability matches for fairway wells (see Palmer et

al. 2007). The high sorption-induced strain numbers used by Palmer were noted to lie beyond the range of lab data (< 8)

based upon unconstrained data. The value of this is that the P&C model has suggested a constraint on the sorption-strain

magnitude. At this time, the P&C model has only been compared with sorption-strain measured under unconfined stress

conditions. We are endeavoring to gather and analyze experimental data collected under conditions that more accurately

represent the in-situ stress/strain environment. It is possible that the amount of sorption-induced strain measured under these

more realistic conditions will be significantly less than that measured under unconfined states.

Of additional concern are the appropriate values of g and fracture porosity to be used for the Fruitland Coal. Although

Palmer et al. (2007) and Palmer (2008) proposed an explanation for setting g to a value < 1, there is still some uncertainty

around the appropriate value to use. In classically-cleated coals, with vertical cleats, the rock mechanics parameters E and ν

are anisotropic (actually transverse anisotropic), and this can explain g values that are ~ 0.3. Note that g < 1 (we used 0.3)

significantly reduces the pressure-dependent component of the P&M equation, which appears to be justified for the Fruitland

Coal examples analyzed to date, but the appropriateness for other basins remains uncertain. The fracture porosity obtained in

the matching process above (.05 to .06 %) is less than that used by Palmer (0.1%), but appears to be reasonably consistent

with mobile water porosity values mapped for Fruitland Coal wells in the area of the matched wells (typically from 0.02 - .14

%). Note that the mobile water porosity values are considered to be a lower limit to true fracture porosity.

6 SPE 114778

Conclusions

The Pan and Connell (2007) model for predicting sorption-induced strain from fundamental thermodynamic principles and

rock properties for CBM reservoirs has been expanded to include the use of rigorous multi-component adsorption models,

such as IAS and 2-D EOS. The new model appears to provide more accurate predictions of sorption-induced strain in binary

gas scenarios than the P&C model coupled with Extended Langmuir. The model will continue to be tested against

experimental data in the near future.

Investigations using the Pan and Connell model to calculate the sorption induced-strain component of the Palmer-

Mansoori equation have demonstrated that permeability growth associated with primary depletion of Fruitland Coal fairway

wells can be matched reasonably accurately. Due to a current lack of key model inputs, such as coal solid rock mechanical

properties, the matches of field data using the new model are non-unique. Further work is required to test the model against

experimental data that accurately represents in-situ conditions. Future work will also expand the use of the combined

P&C/P&M model to field-trials of ECBM.

Perhaps the most important implication is that the combined P&C/P&M model has the potential to predict permeability

growth with depletion in other CBM plays, and to assess the additional gas rates associated with this. A second finding from

the new modeling is that the sorption strain appears to be better constrained by the P&C model. This leads to lower cleat

porosities which are consistent with other derivations, and to g values ~ 0.3, which appear to be justifiable in transversely

isotropic coals. The lower sorption strain also makes it easier to match permeabilities that increase exponentially with

depletion.

Acknowledgements

The authors wish to thank their respective companies and institutions for permission to present this paper.

Nomenclature

A = specific surface area, m2/kg

B = Langmuir constant, Pa-1

c = pore structure model constant (1.200), dimensionless

C = number of components

cf /φi = cp = natural fracture pore volume compressibility, Pa-1 or psi-1

g = pressure-strain suppression factor, dimensionless

E = Young’s modulus, Pa or psi

f = fugacity (Eq. A-12), Pa or fraction (Eq. 7)

f(x, ν s ) = structural model parameter, Eq. 1, dimensionless

G = specific Gibbs free energy, J/kg

ka = absolute permeability, md

kg = effective permeability to gas, md

kai = initial absolute permeability, md

K = bulk modulus, psi

l = length in the pore structure model, m

L = Langmuir constant, mol/kg

m = mass of adsorbent, kg

M = constrained axial modulus, psi

n = specific amount adsorbed, mol/kg

Pi = initial pressure, Pa or psi

Pi 0 = gas pressure of the pure component adsorbed at the same temperature and spreading pressure as the solution (eq. A-22)

P = pressure, Pa or psi

Pε = pressure at strain of 0.5 ε∞, Pa or psi

R = gas constant (8.314 J mol-1 K-1)

S = specific entropy, J kg-1 K-1)

T = temperature, K

U = specific internal energy, J/ kg

V = specific volume, m3 /kg

x = ratio a/l, dimensionless (Eq. 1), or sorbed phase composition (Eq. A-22)

y = gas phase composition

SPE 114778 7

Greek Symbols

β = grain compressibility, psi-1

ε = volumetric strain, dimensionless

ε∞ = volumetric strain at infinite pressure, dimensionless

π = spreading pressure, N/m

π* = reduced spreading pressure, mol/kg

φ = fracture porosity, dimensionless

φi = fracture porosity at initial pressure, dimensionless

Φ = surface potential, J/kg

γ = specific surface energy, (J kg-1 m-2)

µ = chemical potential, J/kg

ν = Poisson’s ratio, dimensionless

ρ = density, kg/m3

Subscripts

i = ith component

s = solid phase

V = volumetric

Superscripts

a = adsorbed phase

o = pure component

s = solid phase

* = state in vacuo

References

Do, D.D.: Adsorption Analysis: Equilibria and Kinetics, Imperial College Press, London (1998).

Bentz, D.P., Garboczi, E.J., Quenard, D.A., 1998. Modelling Drying Shrinkage in Reconstructed Porous Materials;

Application to Porous Vycor Glass. Modelling and Simulation in Material Science and Engineering 6: 211-236.

Chikatamarla, L.. Cui, Xiaojun, Bustin, R.M. 2004. Implications of Volumetric/Shrinkage of Coal in Sequestration of Acid

Gases. Paper 0435 presented at the International Coalbed Methane Symposium, Tuscaloosa, Alabama, 3–7 May.

Clarkson, C.R., and Bustin, R.M. 1999a. The Effect of Pore Structure and Gas Pressure Upon the Transport Properties of

Coal: I Isotherms and Pore Volume Distributions. Fuel 78: 1333–1334.

Clarkson, C.R. and McGovern, J.M. 2003. A New Tool for Unconventional Reservoir Exploration and Development

Applications. Paper 0336 presented at the International Coalbed Methane Symposium, Tuscaloosa, Alabama, 5–9 May.

Clarkson, C.R., Jordan, C.L., Gierhart, R.R., and Seidle, J.P. 2008. Production Data Analysis of CBM Wells. SPEREE 11 (2):

311–325. SPE-107705-PA.

Cui, X., and Bustin, R.M. 2005. Volumetric Strain Associated with Methane Desorption and Its Impact on Coalbed Gas

Production from Deep Coal Seams. AAPG Bulletin 89 (9): 1181–1202.

Gierhart, R.R., Clarkson, C.R., and Seidle, J.P. 2007. Spatial Variation of San Juan Basin Fruitland Coalbed Methane

Pressure Dependent Permeability: Magnitude and Functional Form. Paper IPTC 11333 presented at the International

Petroleum Technology Conference, Dubai, U.A.E., 4–6 December.

Hall, F.E., Zhou, Z., Gasem, K.A.M, Robinson, R.L. Jr., and Yee, D. 1994. Adsorption of Pure Methane, Nitrogen, and

Carbon Dioxide and Their Binary Mixtures on Wet Fruitland Coal. Paper SPE 29194 presented at the SPE Eastern

Regional Conference, Charleston, West Virginia, 8–10 November. DOI: 10.2118/29194-MS.

Karacan C. O. 2007. Swelling-Induced Volumetric Strains Internal to a Stressed Coal Associated with CO2 Sorption.

International J. of Coal Geology 72 (4): 209-220.

Levine J. R. 1996. Model Study of the Influence of Matrix Shrinkage on Absolute Permeability of Coalbed Reservoirs. In

Coalbed Methane and Coal Geology, Special Publication No. 109, 197-212. Geological Society.

Mavor, M.J., and Gunter, W.D. 2004. Secondary Porosity and Permeability of Coal vs. Gas Composition and Pore Pressure.

Paper SPE 90255 presented at the SPE Annual Technical Conference and Exhibition, Houston, Texas, 26–29 September.

Mazumder S. and Wolf, K.H. 2008. Differential Swelling And Permeability Change in Coal in Response to CO2 Injection for

ECBM. International J. of Coal Geology 74 (3): 122-138.

Mitra, A., and Harpalani, S. 2007. Modeling Incremental Swelling of Coal Matrix with CO2 Injection in Coalbed Methane

Reservoirs. Paper SPE 111184 presented at the SPE Eastern Regional Meeting, Lexington, Kentucky, 17–19 October.

8 SPE 114778

Moffat D.H. and Weale, K.E. 1955. Sorption by Coal of Methane at High Pressures. Fuel 34: 449-462.

Myers A.L. and Prausnitz, J.M. 1965. Thermodynamics of Mixed-Gas Adsorption. AIChE J. 11 (I): 121.

Myers A.L. 2002. Thermodynamics of Adsorption in Porous Materials. AIChE Journal 48 (I): 145-160.

Palmer, I. and Mansoori, J. 1998. How Permeability Depends on Stress and Pore Pressure in Coalbeds: A New Model.

SPEREE 1 (6): 539–544. SPE-52607-PA.

Palmer, I., Mavor, M. and Gunter, B. 2007. Permeability Changes in Coal Seams During Production and Injection. Paper

0713 presented at the International Coalbed Methane Symposium, Tuscaloosa, Alabama, 5–9 May.

Palmer, I.D. 2008. Permeability Changes in Coal: Analytical Modeling. IJCG accepted for publication.

Pan Z. and Connell, L. D. 2007. A Theoretical Model for Gas Adsorption-Induced Coal Swelling. International J. of Coal

Geology 69 (4): 243-252.

Scherer, G.W. 1986. Dilation of Porous Glass. Journal of the American Ceramic Society 69 (6): 473-480.

Shi, J.Q., and Durucan, S. 2005. A Model for Changes in Coalbed Permeability During Primary and Enhanced Methane

Recovery. SPEREE 8 (4): 291–299. SPE-87230-PA.

Zahner R. 1997. Application of Material Balance to Determine Ultimate Recovery of a San Juan Fruitland Coal Well. . Paper

SPE 38858 presented at the SPE Annual Technical Conference and Exhibition, san Antonio, Texas, October.

SPE 114778 9

-1 3

Gas Type L (mmol/g) B (MPa ) Matrix Φ (%) ρs (g/cm ) Es (GPa) vs

CH4 1.257 .294 6.0 1.300 1.54 .372

CO2 1.488 .953 6.0 1.300 1.54 .372

TABLE 2 — INPUT PARAMETERS FOR P&C MODEL MATCH TO MITRA AND HARPALANI (2007) DATA

-1 3

Gas Type L (mmol/g) B (MPa ) Matrix Φ (%) ρs (g/cm ) Es (GPa) vs

CH4 0.629 .271 6.0 1.600 .824 .484

CO2 1.326 .725 6.0 1.600 .824 .484

TABLE 3a — INPUT PARAMETERS FOR P&C MODEL MATCH TO CBM WELL DATA

-1 3

Gas Type L (mmol/g) B (MPa ) Matrix Φ (%) ρs (g/cm ) Es (GPa) vs

CH4 0.627 .725 6.0 1.570 2.23 .43

CO2 0.928 .215 6.0 1.570 2.23 .43

TABLE 3b — INPUT PARAMETERS FOR P&M MODEL MATCH TO CBM WELL DATA

β (psi )

-1

10 SPE 114778

Matrix

Block

Figure 1. Diagram illustrating matrix-shinkage concept associated with gas desorption from coal.

0.50%

0.40%

P&C-IAS: 500psi

P&C-EL: 500psi

Linear Strain (%)

P&C-IAS:100psi

P&C-EL: 100psi

0.20%

P&C-2D EOS: 100psi

0.10%

0.00%

0% 20% 40% 60% 80% 100%

CO2 Concentration (%)

Figure 2. Sorption-induced strain predictions using the Pan and Connell (2007) model for a binary mixture of methane and carbon

dioxide. Three binary gas adsorption models are used in the predictions: Extended Langmuir (EL), Ideal Adsorbed Solution (IAS),

and 2-Dimensional Equation-of-State (2-D EOS). Note the similarity of the predictions at low pressure (100 psi), but divergence of

the EL model at high pressure (500 psi).

SPE 114778 11

1.00%

0.90%

0.80%

0.70%

Linear Strain (%)

0.60% Experimental

P&C-IAS

0.50%

P&C-EL

0.40% P&C-2D EOS

0.30%

0.20%

0.10%

0.00%

0% 20% 40% 60% 80% 100%

CO2 Concentration (%)

Figure 3. Sorption-induced strain predictions using the Pan and Connell (2007) model for a binary mixture of methane and carbon

dioxide. Three binary gas adsorption models are used in the predictions: Extended Langmuir (EL), Ideal Adsorbed Solution (IAS),

and 2-Dimensional Equation-of-State (2-D EOS). Note the good agreement between model predictions using 2-D EOS and IAS.

100

kg/kg@932psi

10

1

50 150 250 350 450 550 650 750 850 950

Estimated Reservoir Pressure (psia)

Figure 4. Comparison of permeability ratio predicted by coupled Pan and Connell (P&C)/Palmer and Mansoori (P&M) equation and

gas permeability obtained from field data (see Clarkson et al 2008). Base permeability is anchored to the value estimated from field

data at the initial pressure of the well (932 psia).

12 SPE 114778

800

700

600

500

Pwf, psia

400

300

200

100

0

0 1 2 3 4 5 6 7 8 9 10 11

Time, Years

Data Model

Figure 5. Analytical simulator match to bottomhole flowing pressures using the coupled P&C/P&M permeability prediction (Fig. 4).

1000

100

kg/kg@1450psi

10

1

50 150 250 350 450 550 650 750 850 950

Estimated Reservoir Pressure (psia)

Figure 6. Comparison of permeability ratio predicted by coupled Pan and Connell (P&C)/Palmer and Mansoori (P&M) equation and

gas permeability obtained from field data (see Clarkson et al 2008). Base permeability is anchored to the estimated virgin pressure

of the field (1450 psia). Note that the permeability ratio at 932 psia is > 1, unlike Fig. 4.

SPE 114778 13

Appendix

The derivation of the Pan and Connell model (2007) for predicting coal swelling starts with a derivation of surface potential

energy, which is then related to the solid elastic energy using a pore structure model.

Surface Potential Energy. Myers (2002) derived fundamental thermodynamic expressions to describe the adsorption

process in microporous solids, which are analogous to solution thermodynamics. Because the coal matrix porosity is

commonly dominated by micropores (Clarkson and Bustin 1999), it is natural to extend Myers approach to coal. Myers

(2002), however, neglected the change in solid volume in his formulation, whereas Pan and Connell (2007) allowed for the

change in the solid volume due to either compression or adsorption.

Following Myers’ (2002) work, the derivation of an analytical expression for surface potential energy starts with the

differential equation describing the internal energy of the (microporous) solid + adsorbed phases (or entire condensed phase)

containing multiple gas components:

C

dU = TdS − PdV + ∑ μ i dni + μdm ……………………………………….………………………………..……………..(A-1)

i =1

C

U = TS − PV + ∑ μ i ni + μm ……………………..………………………………...………………………..……...……..(A-2)

i =1

Assuming the mass of the adsorbent is constant and writing the extensive variables in terms of per unit mass of adsorbent

(mass extensive variables):

C

U = TS − PV + ∑ μ i ni + μ …………………..………………………………………….……………...………………....(A-3)

i =1

Similarly we may express the internal energy of the solid adsorbent in its pure standard state (without adsorption):

U s = TS s − PV s + μ s …..……………………………………………………….…………………..…..………………....(A-4)

Mass extensive variables for the adsorbate may be obtained by subtracting the variable for the solid phase from the

condensed (solid + adsorbate) as follows:

U a = U −U s

Sa = S −Ss

V a = V −V s

nia = ni

The internal energy of the adsorbate may be obtained by subtracting Eq. A-4 from A-3:

C

U a = TS a − PV a + ∑ μ i nia + Φ ……………..………………………………………………..…………………………..(A-5)

i =1

where

Φ = μ − μ s ………………………….……………………………………………………...…………………………...…..(A-6)

In the derivation of Eq. A-5, Pan and Connell (2007) did not neglect the adsorbate volume changes, because it is assumed

a

that the solid adsorbent is compressible, and that V is the difference in the apparent volume change of the adsorbent due to

adsorption and mechanical compression. In differential form, Eq. A-5 becomes:

14 SPE 114778

C

dU a = TdS a − PdV a + ∑ μ i dnia ………………………………………………..………………………………………..(A-7)

i =1

C

G a = U a + PV a − TS a = ∑ μ i nia + Φ ……………………………………………………………..……………………..(A-8)

i =1

or in differential form:

C

dG a = − S a dT + V a dP + ∑ μ i dnia …………………………..……………………..……………………………………..(A-9)

i =1

C

Φ = G a − ∑ μ i nia ……………………………..………………………………….…………………………..…………..(A-10)

i =1

C C C

dΦ = dG a − ∑ μ i dnia − ∑ nia dμ i = − S a dT + V a dP − ∑ nia dμ i ……………………………………………………....(A-11)

i =1 i =1 i =1

C C

dΦ = V a dP + ∑ nia dμ i = V a dP − RT ∑ nia d ln f i ………………………………………………..………...…………..(A-12)

i =1 i =1

Integrating:

P⎛ C ⎞

Φ = ∫ V a dP − RT ∫ ⎜ ∑ nia d ln f i ⎟ ……………………...………………………………………..………...…………..(A-13)

P

0 0

⎝ i =1 ⎠

P P na

Φ = ∫ V a dP − RT ∫ df ………………….…………...………………………………….……..………...…………..(A-14)

0 0 f

a

Neglecting V for now, and assuming a perfect gas, Eq. A-14 simplifies to:

P na

Φ = − RT ∫ dP ………………….…………...……………………………….………...………..………...…………..(A-15)

0 P

Note that the expression above is mathematically equivalent to the negative product of spreading pressure and surface area,

which is used in the surface thermodynamics approach:

P na

Φ = −πA = − RT ∫ dP ………………….…………...………………………………...………..………...…….……..(A-16)

0 P

As Myers (2002) points out, the concept of spreading pressure has no physical meaning in microporous solids, and surface

area cannot be measured accurately, hence the physics of adsorption is best described in the framework of solution

thermodynamics (i.e. using surface potential). Nonetheless, the mathematical equivalence between surface potential and

− πA will prove useful, as will be shown.

SPE 114778 15

The integral in Eq. A-16 can be evaluated analytically only for a few isotherms (for most isotherm forms, numerical

integration is required); for example, using the Langmuir equation, Eq. A-16 becomes (Do, 1998):

P

which was derived in Pan and Connell’s original work (2007). Note that because of the mathematical equivalence between

surface potential and − πA (Eq. A-15), more sophisticated isotherm models, such as the 2-D EOS models, may be used to

calculate surface potential, should the Langmuir equation prove inaccurate for certain adsorption systems.

The extension of Eq. A-18 to multiple gas components, as would be necessary in modeling primary depletion of CBM

reservoirs containing gas mixtures, or for ECBM/sequestration processes, is relatively straight forward. Pan and Connell

(2007) used the Extended Langmuir equation, and arrived at the following equation:

C

Φ = ∫ V a dP − RT ∑ Li ln(1 + Bi y i P ) ……………….………………………...……………………..……...…………..(A-19)

P

0

i =1

Other multi-component adsorption models may be used in the calculation of surface potential by using the mathematical

equivalence between surface potential and − πA . For example, the popular Ideal Adsorbed Solution (IAS) Theory may be

used to estimate − πA ; the model assumes that the reduced spreading pressures of the single components are equal to the

mixture:

π 1* = π 2* = ....π n* = π * …...…………………………………..………...………..……………………….…...…………....(A-20)

where:

πi A Pi0 na

π i* = =∫ dP ……………...…….…………...……………...………..…………………….……...…………....(A-21)

RT 0 P

and:

which is analogous to Raoult’s Law for ideal solutions. Note that fugacities may be used in the above equations to account

for gas-phase non-idealities.

πA may then be calculated using Eqs. A-20 to A– 22; assuming Φ = − πA , then IAS Theory may be used to calculate

surface potential. An example application is given using the equality of Eq. A-20, assuming a binary mixture of gases at low

a

pressure, neglecting V , and using the Langmuir isotherm for single-component gases, which allows an analytical

calculation for the integral given in Eq. A-21:

( ) ( )

RTL1 ln 1 + B1 P10 = RTL2 ln 1 + B2 P20 = πA ……….……….………...………………………..…….……...…………..(A-23)

where:

Py1

P10 (π ) = ………………………….………..……...……………………………....…...………..………...…………..(A-24)

x1

and:

Py 2 P(1 − y1 )

P20 (π ) = = ………………………….………..……...……………………………….………...…………..(A-25)

x2 (1 − x1 )

16 SPE 114778

If P, y (y1 = 1-y2), and the single-component isotherm parameters are known, then the equality of Eq. A-23 may be solved

iteratively by adjusting x1 = 1-x2. The surface potential can then be calculated by assuming Φ = − πA . The 2-D EOS

model, as discussed in Hall et al. (1994), may also be used to calculate πA . The 2-D EOS model for adsorption (single and

multi-component) is given in its general form below:

⎡ ⎤

⎢ Aπ +

1 + Uβ

αn 2

n + W (β n )2

[ ]

⎥ 1 − (βn ) = nRT …………………...............................……………………...…………..(A-26)

m

⎣⎢ ⎥⎦

Where n is the specific amount adsorbed; α, β are 2-D EOS constants; and U, W and m are model coefficients. This

equation, when coupled with fugacity and equilibrium adsorption relationships, may be used to model single-component

adsorption. Mixing rules are required for multicomponent adsorption (see Hall et al. 1994).

There are therefore several multi-component adsorption models, which have been used to correlate multi-component gas

adsorption data for coal, which also may be used in the calculation of surface potential.

Pore Structure Model. The next step in the development of the Pan and Connell (2007) sorption-induced strain model, was

the selection of a coal matrix pore structure model. For this purpose, the pore structure model of Scherer (1986), used

originally to model strain induced in glass due to water and water vapour adsorption, was adopted, due to the assumption of a

glassy, strained, cross-linked macromolecular structure for the coal matrix. Scherer calculated linear strain for the glass-

adsorption system, assuming isotropic behaviour, and an energy balance between elastic and surface energy, as follows:

γAρ s

ε =− f (x,ν s ) …………………………………..…...………………………………………..………...…………..(A-27)

Es

where:

f ( x, v s ) =

(3 − 5v s )(2 − 3cx)

c = 1.2

x = a/l

and

Noting that at high pressure, strain is induced in the coal solid due to compression, and combining it with sorption-induced

strain given with Eq. A-27, results in:

Φρ s P

ε =− f ( x,ν s ) − (1 − 2ν s ) ………………..…..…...………………………………………..………...…………..(A-30)

Es Es

where:

Φ = γA

Eq. A-30 thus provides a method for calculating linear strain associated with sorption/compression, assuming the matrix

behaves according to the porous-glass model used by Scherer. The rock properties used in Eq. A-30 (Young’s Modulus,

Poisson’s Ratio) are for the solid phase only, which are not commonly measured. Pan and Connell (2007) provided several

auxiliary equations relating the rock properties derived for the matrix including the elasticity of the pores (which are more

commonly measured) to the rock properties of the solid only. Eq. A-30 is general in that numerous methods are available for

calculating sorption potential, for single-component and multi-component adsorption. For example, for single component

gases, assuming Langmuir-type adsorption, the surface potential is given by Eq. A-18, and the linear strain is given by:

SPE 114778 17

ρs ( ) P

ε = RTL ln(1 + BP ) f ( x,ν s ) −

P

(1 − 2ν s ) − 3 f x,ν s ∫0 εdP ……………..…………………….……...…….....…..(A-31)

Es Es (1 − φ )E s

In the derivation of Eq. A-31, it was assumed that V a can be related to strain as follows:

3ε s

Va = ……………………………….....…..…...……………………..…………………….……...…..………..(A-32)

ρ s (1 − φ )

At low pressures (neglecting compression of the matrix), Eq. A-31 is simplified to:

ρs P

ε = RTL ln(1 + BP ) f ( x,ν s ) − (1 − 2ν s ) ……………..………............................................................………….....(A-33)

Es Es

Pan and Connell (2007) validated Eq. A-31 and A-33 against coal strain (induced by single-component gas adsorption)

data of Levine (1996) and Moffat and Weale (1955). They also mentioned that Eq. A-31 and A-33 can be extended for

multi-component adsorption using the Extended Langmuir model in the sorption potential calculation (Eq. A-19).

As noted previously, in addition to the Extended Langmuir model, which may not be accurate in all mixed-gas/coal

adsorption scenarios), other more thermodynamically-rigorous and accurate models may be used for the calculation of

surface potential, which appears in Eq. A-30.

## Viel mehr als nur Dokumente.

Entdecken, was Scribd alles zu bieten hat, inklusive Bücher und Hörbücher von großen Verlagen.

Jederzeit kündbar.