PROCEEDINGS, TOUGH Symposium 2009

Lawrence Berkeley National Laboratory, Berkeley, California, September 14-16, 2009

A NUMERICAL STUDY OF MICROSCALE FLOW BEHAVIOR IN TIGHT GAS AND SHALE

GAS RESERVOIR SYSTEMS

C.M. Freeman1, G.J. Moridis2, T.A. Blasingame1

1
Dept. of Petroleum Engineering, Texas A&M University
College Station, TX 77845-3116, U.S.A.
2
Earth Sciences Division, Lawrence Berkeley National Laboratory
1 Cyclotron Rd., Berkeley, CA 94720
e-mail: [email protected]

ABSTRACT INTRODUCTION

Various attempts have been made to model flow in

tight gas and shale gas systems. However, there is Background
currently little consensus regarding the impact of
molecular diffusion on flow behavior over time in The composition of gas produced from wells in ultra-
such systems. tight gas reservoirs varies with time. In shale gas
reservoirs possessing high organic content, the
The composition of gas produced from tight gas and gradual change in produced gas composition is
shale gas reservoirs varies with time. The main attributed to preferential desorption of different gas
driver behind this phenomenon is the selective species at different rates. For example, carbon
desorption of gases from the surface of the kerogen in dioxide will desorb from kerogen more readily than
the case of shale. methane, so at earlier times the concentration of
However, a secondary but important driver is the carbon dioxide in the produced gas will be higher,
fractionating effect of the reservoir rock itself. Pore and will gradually taper off (Gao et al., 1994).
throat diameters on the order of molecular mean free
path lengths will create an exaggerated slip-flow-like However, this is not the only explanation for the
condition. When this condition prevails, Darcy’s law change in gas composition. Shale gas and tight gas
can no longer be used without adjustment. Florence reservoirs can possess a high degree of heterogeneity
et al. (2007) proposed a ‘microflow’ model for and widely varying porosity distributions. The
effective permeability correction which correctly kerogen content in shale can be considered to provide
represents flow in the Knudsen and transitional flow an entirely separate porosity system (Clarkson et al.,
regimes. 1999).

In this study we implement the microflow model into The pore throat dimensions of these porosity systems
a fluid flow modeling tool based on the TOUGH+ can vary widely, potentially ranging from 10nm to
family of codes. We examine the behavior of the 50µm. The distribution of pore throat sizes will vary
microflow correction on gas composition in ultra- reservoir by reservoir.
tight rock. We show that for very small average pore
throat diameters, lighter gases are preferentially According to the theory of gas kinetics, the mean free
produced at concentrations as high as double what path is defined as the mean distance a molecule in a
would otherwise be expected. gas will travel before interacting with another gas
molecule. When the mean free path is comparable to
Our contributions include a new, fit-for-purpose the pore throat size, a continuum flow assumption is
numerical model based on the TOUGH+ code no longer fully valid and Darcy’s law cannot be used
capable of characterizing flow behavior in without modification.
hydraulically fractured horizontal wells. This model
includes a demonstration of a “microflow” model for
The ratio of the mean free path over the feature
fluid flow in micro- and nano-scale porous media.
length, in this case the pore throat diameter, is
expressed as the Knudsen number,

……………………………………...(1)

-1-
 -2-

For Knudsen numbers less than 0.001 Darcy’s law single-component gas is conceptually the ratio of the
remains valid. For Knudsen numbers higher than this distance traveled divided by the volume of
value, an effective permeability must be computed interaction,
which corrects from the “microflow” effect.

In this work we implement the “microflow” model

proposed by Florence, et al. into a numerical model ……(3)
based on the TOUGH+ family of reservoir simulators
to examine the effect of microflow phenomena on However, this method of computing the mean free
multicomponent gas flow. path leaves out the important consequences of gas
mixtures. The average velocity of a molecule in a
Mathematical Development gas is

The flow regime of Knudsen flow, similar to the idea

of gas slippage, occurs when the mean free path of
the gas molecules is on the order of the average pore ……………………………………....(4)
throat radii. The mean free path of a molecule in a
single-component gas can be computed by The use of a single average velocity is inappropriate
for a gas mixture, where each species in the gas
possesses a different average velocity as a function of
molar mass. Therefore, we must compute the
average velocity of a gas molecule in a gas mixture.
…………………....(2)
The relative velocity between two gas particles is
expressed as
Figure 1 demonstrates the effect of pressure on mean
free path for various gases.
………………………………..(5)

Where

…………………………………....(6)

So, taking the magnitude of the relative velocity

vector,

…………………...(7)

Algebraically rearranged,

………….....(8)

Since we are not truly examining individual

molecules but rather the average properties of the
statistical ensemble of the gas, we take the averages
of the terms

…………….(9)

Figure 1. Effect of pressure on mean free path of Since and (the average velocity vectors for
various gases.
two different gas particles) are random and
uncorrelated, their dot product equals zero. Thus the
This is how the mean velocity of the molecule is formula for average velocity reduces to simply
computed. The mean free path of a molecule in a
 -3-

transitional flow, which is less strongly characterized

by the Klinkenberg correction, and Kn greater than
………………………….....(10)
10 exhibit free molecular flow which is dominated by
diffusion effects, i.e., any parameterization by
where the average velocities of the two species are in
‘viscosity’ and ‘permeability’ no longer has a
fact different, computed from Equation 4.
meaningful physical interpretation. Figure 2 shows
the impact of pressure on Knudsen number for
For one gas species in a multicomponent gas mixture,
various pore throat sizes.
the “volume of interaction” term of the expression is
modified to

…………….(11)

The term can be modified by a perfect gas law

approximation,

………….(12)

Rather than relying on a perfect gas assumption, we

can use the more accurate density computation for
density of a gas from an equation of state already
intrinsic to our numerical implementation,
Figure 2. Effect of pressure on Knudsen number for
various pore throat sizes and gases.

In fact, it is likely that flow through shale will exhibit

all of the above regimes and Knudsen numbers.

Klinkenberg demonstrated the approximately linear

………..(13) relationship between measured permeability and
inverse pressure

We now have a method by which to characterize the

mean free path of a single gas species in the
multicomponent gas within our system. Thus we can ……………………………..(15)
now examine the ramifications of different mean free
paths on gas flow in porous media with microscale The gas slippage factor is regarded as constant in the
features. The dimensionless Knudsen number will flow regime where the Klinkenberg approximation is
identify the severity of the non-Darcy ‘slippage’ valid, and is related to the mean free path
effect.

……………………………………..(16)
…………………………………….(14)
This gas slippage factor is related to the beta term via
For Knudsen numbers smaller than 0.01, the
assumption of ‘continuum flow’ is valid and Darcy’s
law can be used unmodified. Knudsen numbers in
the realm of 0.1 qualify as slip flow and can be
characterized by the Klinkenberg correction to …………………………..…..(17)
acceptable accuracy. Kn of around 1.0 qualify as
 -4-

While this beta-form of the Klinkenberg

approximation is simple and convenient for the MODEL DESCRIPTION
purposes of numerical simulation due to its
computational simplicity, a more robust, rigorously
developed “microflow” model has been proposed by The vast majority of shale gas and tight gas wells are
Florence, et al., which is valid for all flow regimes in stimulated via hydraulic fracturing. Typical
porous media, from free-molecular flow through completion strategies for these reservoirs involve
continuum flow, though not verified and probably either a vertical well with a vertical bi-wing hydraulic
invalid in high-rate inertial (Forchheimer) flow fracture or a horizontal well with multiple transverse
regimes. The basis of this model is to adjust the hydraulic fractures. At early times, the flow
theoretical permeability to an ‘effective’ permeability character is dominated by linear inflow into the
through use of the Knudsen number. fractures. Gradually the flow regime passes through
transition stages into an elliptical flow regime. To
capture the character of the early and late end
member behaviors, linear grids and radial grids are
………..(18) employed.

The function α is a rarefaction coefficient parameter, A Peng-Robinson equation of state is used to model
a dimensionless adjustment parameter of the form the gas density as a function of pressure. Gas
viscosity is modeled by the Chung et al. method.
Isothermal conditions are assumed because with a
very low flow rate of gas, the effect of Joule-
………………..(19) Thompson cooling will be minimal.

This formula, Equation 18, will vary the ‘effective’ For all cases, flow from a fracture face was simulated
permeability experienced by each individual gas for a period of 2 days through 1µD rock. The molar
species, due to the fact that each gas species in the initial composition of each gas component is equal,
mixture will have a different mean free path and thus i.e., all molar gas compositions are initially one third.
a different Knudsen number. Implementation of this The initial reservoir pressure was 145 psia and the
theory will enable modeling of the ‘fractionating’ fracture-face pressure was 14 psia. Porosity is 10%.
effect of reservoirs with extremely small pores, A sensitivity analysis was performed on the pore
wherein smaller molecules flow preferentially faster throat radius.
than larger ones. Figure 3 shows the impact of
various pore throat dimensions on adjusted Table 1. Sensitivity parameters.
permeability for various gases.

RESULTS
The pressure transient progresses less than 0.8 meters
in any given case, as shown via the purple line in
Figure 4. According to these sensitivities, the gas
deviation from the expected (initial) gas composition
is plotted in Figure 4.
Figure 3. Effect of various pore throat dimensions
on adjusted permeability for various
gases at varying pressure.
 -5-

Figure 4. Results of sensitivity analysis; deviation of Figure 5. Variation of produced gas mole fraction
gas composition from initial gas with pore throat radius.
composition, with pressure transient.

These results satisfy our expectation that methane CONCLUSIONS

will flow more quickly than propane and ethane with
decreasing pore throat diameter. In the 10 nm pore Decreasing pore throat size is shown to accelerate
throat diameter case (corresponding to a pore radius production of lighter, smaller gases at the expense of
of 5.0e-9m) the methane composition at the fracture heavier ones. Pore throats of 10 nm diameter cause
face is 1% less than expected, since this region has the produced gas to contain more than twice the
already been evacuated of methane while the relative concentration of methane than would be expected
concentration of ethane and propane remains high. without microflow adjustment.

The shape of the pressure transient is strongly The lighter gases are shown to deplete from the near-
impacted and somewhat distorted by the small pore fracture region more quickly than heavier gas
throat effects. This nonlinearity presents obstacles to species. Correspondingly, the rate and degree of
numerical convergence. Furthermore, this sharper propagation of the pressure transient into the
pressure transient and may explain the sharp decline reservoir is increased by the microflow adjustment
in production rate seen at early times in ultratight gas caused by smaller pore throats.
reservoirs.
Implementation of the microflow permeability
Through the short duration of production simulated, adjustment model into a robust fluid flow modeling
the gas composition in the produced gas stream tool demonstrates (a) promise for future work in
remains relatively constant in a given case. The predicting compositional change with time in
simulation time in these cases has not progressed ultratight reservoirs and (b) potential utility in
long enough to demonstrate a strong change in gas sampling changes in gas composition with time in
concentrations over time in a given case. However, these reservoirs in order to better characterize the
between cases, the composition of the produced gas rock properties.
varies dramatically. Figure 5 shows the variation of
gas composition in the gas produced from each of the This model provides a strong basis for future work,
sensitivity cases. such as to incorporate a desorption model and
simultaneously modeling multicomponent desorption
With the very small pore throat sizes the composition and multicomponent microflow-adjusted gas flow.
of methane in the produced gas stream is as high as
68% by mole despite the fact that only 33% by mole
of the gas in the pores is methane. If this concept is ACKNOWLEDGEMENTS
scaled up to examine reservoir-scale behavior, it This work was supported by RPSEA (Contract No.
becomes clear why only methane is produced from 07122-23) through the Ultra-Deepwater and
ultratight reservoirs – the other gas species are left Unconventional Natural Gas and Other Petroleum
behind.
 -6-

Resources Research and Development Program as Industrial & Engineering Chemistry Research 27,
authorized by the US Energy Policy Act (EPAct) of 671-679, 1988.
2005. The authors are indebted to Dilhan Ilk of
Texas A&M University for his many suggestions and Clarkson, C.R.. and Bustin, R.M. 1999. Binary gas
for the careful review of the manuscript. adsorption/desorption isotherms: effect of moisture
and coal composition upon carbon dioxide selectivity
SYMBOLS AND ABBREVIATIONS over methane. International Journal of Coal
Geology. 42(2000) 241-271.
mean free path
average pressure Florence, F.A., Rushing, J.A., Newsham, K.E., and
temperature Blasingame, T.A. 2007. Improved Permeability
viscosity Prediction Relations for Low-Permeability Sands.
SPE paper presented at the SPE Rocky Mountain Oil
ideal gas constant
& Gas Technology Symposium held in Denver,
molecular mass of gas molecule Colorado, 16-18 April.
mean velocity of a molecule in a gas
effective molecule diameter Gao, C., Lee, J.W., Spivey, J.P., and Semmelbeck,
time of flight M.E. 1994. Modeling Multilayer Gas Reservoirs
number of molecules per unit volume Including Sorption Effects. SPE paper 29173
presented at the SPE Eastern Regional Conference &
the average of component number 1 Exhibition, Charleston, West Virginia, 8-10
the mean free path of component number 1 November.

the average of the ith component

Hill, D.G., and Nelson, C.R. 2000. Gas Productive
the molecular diameter of the ith component Fractured Shales: An Overview and Update. Gas
TIPS 6 (3): 4-13.
Avogadro’s number, 6.02e23
mass fraction of component i in the gas mixture
Javadpour, F., Fisher, D., and Unsworth, M. 2007.
total density of the gas mixture Nanoscale Gas Flow in Shale Gas Sediments. JCPT.
partial density of the gas mixture 46 (10): 55-61.
the total number of present gas components
the gas component index Mattar, L., Gault, B., Morad, K., Clarkson, C.R.,
Freeman, C.M., Ilk, D., and Blasingame, T.A. 2008.
Knudsen number
Production Analysis and Forecasting of Shale Gas
characteristic length of flow geometry (average Reservoirs: Case History-Based Approach. SPE
pore throat diameter) paper 119897 presented at the SPE Shale Gas
Production Conference held in Fort Worth, Texas,
permeability to liquid 16-18 November.
measured permeability
gas slippage factor Medeiros, F., Ozkan, E., and Kazemi, H. 2007.
approximately 1 Productivity and Drainage Area of Fractured
effective pore radius Horizontal Wells in Tight Gas Reservoirs. SPE paper
rarefaction coefficient parameter, dimensionless 108110 presented at the Rocky Mountain Oil & Gas
Technology Symposium held in Denver, Colorado,
a constant valued at 4.0 16-18 April.
a constant valued at 0.4
a constant valued at Medeiros, F., Ozkan, E., and Kazemi, H. 2006. A
Semianalytical, Pressure-Transient Model for
Horizontal and Multilateral Wells in Composite,
REFERENCES Layered, and Compartmentalized Reservoirs. SPE
paper 102834 presented at the SPE Annual Technical
Chung, T.H., Ajlan, M., Lee, L.L., and Starling, K.E., Conference and Exhibition held in San Antonio,
“Generalized Multiparameter Correlation for Texas, 24-27 September.
Nonpolar and Polar Fluid Transport-Properties”,
 -7-

Raghavan, R.S., Chen, C., and Agarwal, B. 1997. An

Analysis of Horizontal Wells Intercepted by Multiple
Fractures. SPE paper 27652 presented at the SPE
Permian Basin Oil and Gas Recovery Conference
held in Midland, Texas, 16-18 March.

Schettler, P.D., and Parmely, C.R. 1991.

Contributions to Total Storage Capacity in Devonian
Shales. SPE paper 23422 presented at the SPE
Eastern Regional Meeting held in Lexington,
Kentucky, 22-25 October.