Vous êtes sur la page 1sur 12

SPE 84306

Realistic Assessment of Proppant Pack Conductivity for Material Selection


R. D. Barree, Barree & Associates, S. A. Cox, Marathon Oil Company, V. L. Barree, Barree & Associates, M. W. Conway,
Stim-Lab Inc.
Copyright 2003, Society of Petroleum Engineers Inc.
This paper was prepared for presentation at the SPE Annual Technical Conference and
Exhibition held in Denver, Colorado, U.S.A., 5 8 October 2003.
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
presented, have not been reviewed by the Society of Petroleum Engineers and are subject 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
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
for commercial purposes 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 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.

Abstract
Proppant selection in hydraulic fracturing is a critical
economic and technical decision that affects stimulation and
field development economics. In many cases the selection is
based on laboratory data from standardized API conductivity
tests on clean packs at specified stress and temperature. These
tests predict conductivities that are optimistic compared to
observed field performance. Often a laboratory measured
conductivity difference of only 5-10% is considered a
significant variance when applied to the producing life of a
well. The significance of these small differences, however, is
often overwhelmed by other factors affecting fracture
performance in the field.
The selection of a particular proppant should be based on
an identifiable difference in performance under field
conditions. This requires an accurate assessment of all the
damage mechanisms that can and do occur during fracturing
and their impact on final conductivity. This paper outlines the
primary damage mechanisms and their effect on conductivity,
fracture cleanup and ultimate stimulation response. The
expected variance in laboratory measurements of conductivity
is also quantified.
Introduction
In hydraulic fracture treatment design one of the most
important decisions is which proppant to use. The choice of
proppant also directly impacts overall job economics,
treatment size, and the ultimate productivity of the well. The
decision is commonly driven by balancing effective fracture
length and conductivity against reservoir flow capacity, as in
the McGuire-Sikora folds-of-increase curves, or through
estimates of dimensionless fracture flow capacity (FCD).1
Any of these methods require an accurate assessment of
proppant pack conductivity under reservoir flow conditions
along with knowledge of reservoir deliverability. Frequently

the proppant pack conductivity is estimated from standardized


API conductivity tests of proppants in a linear flow cell at a
specified pack concentration, closure stress, and temperature.
The impact of time at stress, rate of stress application, and
other factors that affect these baseline conductivity tests has
been documented.2
One aspect of these tests that has not been well
documented is the expected statistical variation in laboratory
results when all tests are conducted as similarly as possible in
terms of cell loading, stress application, aging at stress, fluid
compatibility, and flowing conditions. The choice of a
particular proppant, based on superior performance
characteristics, requires that the performance of two different
materials be statistically significant. This paper addresses the
statistical variance in laboratory conductivity measurements
on similar proppant samples at similar stress, temperature, and
flow conditions. In this case similar means as close to the
same as can be determined in the lab.
Beyond the reproducibility of laboratory measurements,
some of the principal mechanisms affecting final proppantpack conductivity are addressed. Specifically these include:

Non-Darcy flow: What is the velocity distribution in


the fracture and how does it affect conductivity loss?
Multiphase flow: What does the two- and three-phase
relative permeability curve for a proppant pack look
like, can it be measured, and how important is it?
Multiphase non-Darcy flow: How does multiphase
flow in the proppant pack change , velocity,
apparent flowing density, and viscosity? What is the
effect on final conductivity?
Gravity and viscous segregation: Do we really know
what the flow path is in a proppant pack, how much
cleans up, and what the local velocity is? How does
this affect conductivity?
Reservoir flow capacity: Does the reservoir
determine the amount of conductivity required? Is
there ever too much?

Models for fracture cleanup, filter-cake deposition and


removal, multiphase flow, regained bulk-pack permeability
and non-Darcy flow must be integrated with reservoir
transient deliverability to determine final fracture
conductivity. An integrated fracture-reservoir model is
presented that offers a method to consistently select proppants
based on realistically attainable, and predictable, ultimate

A short note on permeability


The permeability of a porous medium, specifically a proppant
pack, is of critical importance in optimizing fracture
performance. To understand proppant pack conductivity it is
first necessary to understand the nature of permeability.
Darcys original work was conducted on unconsolidated
sand packs. His experiments showed that, at least over the
range of conditions he studied, there was a linear relationship
between imposed potential gradient and volumetric flow rate.3
Details of local velocity distribution within the pack, actual
flow-path length, and path geometry were not considered.
Instead superficial velocity and overall sample area and
length were used to describe the flow.
Darcy designated the linear constant of proportionality (k)
as the system permeability. Since that time many people have
stated that permeability is an intrinsic material property, when
it is actually an experimental proportionality constant that
describes the overall (statistical) performance of a particular
experiment where the porous medium, fluid, and flow
conditions are fixed. The description of the porous medium
includes its grain size distribution, packing arrangement, and
other factors.
An examination of Darcys contrived units of permeability
(cp-cm2/atm-sec) shows that the constant, k, is used to
conveniently describe the observed linear flow relationship
suggested by his data, and not a material property. In
determining the permeability of a proppant pack under varying
stress conditions this subtle distinction is important. The
permeability of any proppant pack is dependent on its packing
arrangement which is not always the same, for the same
material. As closure stress increases, the packing geometry
changes, grains begin to fail and the entire pore and grain
system changes along with pack permeability. It should be
expected that permeability of a proppant pack behave more as
a statistical measure of flow capacity, dependent on the
experimental conditions and initial packing state, and not as an
intrinsic material property.
Proppant pack conductivity
Proppant pack conductivity is defined as the product of pack
permeability and width. Given the expected variance in
permeability already discussed, the conductivity should also
be dependent on initial grain packing and grain size
distribution. It should also change with stress as the packing
density and arrangement change and grains break. The same
packing re-arrangement will also affect the pack width.
Conductivity, like permeability, is a representation of a
specific packing condition and grain size distribution, not an
intrinsic material property. To describe conductivity we must
rely on a statistically significant number of observations under
similar packing conditions and along similar loading paths.
The observations can be separated into apparent pack width
and permeability. The statistical range of error in each
measurement can be quantified.

Statistical variation in proppant data


The choice of one proppant over another should be based on
an expectation of measurable difference in their performance
or cost. To quantify the difference in performance among
materials it is necessary to understand the expected variations
in performance of one material under as close to identical
packing and stress history conditions as possible. Data used in
this study were obtained in cells packed according to
procedures set forth in API-RP61.4 The recommended
procedure states that the pack should be leveled with a blade
and not vibrated or tamped. Stim-Lab has amassed a large
amount of data on many proppant types and sizes, under these
standardized test conditions, that can be used to estimate the
statistical variation in both width and permeability.
Variation in pack width
Numerous measurements of pack width versus stress have
been conducted on many proppant types and sizes with the
same proppant loading (pounds of proppant per square foot)
and stress ramp. Data from a representative series of tests is
shown in Figure 1 for 2#/ft2 16/30 white sand.
White Sand (16/30) 2#/ft2
0.250
0.240
0.230
Pack Width, inches

fracture conductivity. The inclusion of the effects of all


identifiable damage mechanisms is necessary to economically
optimize any stimulation design.

SPE 84306

0.220
0.210
0.200
0.190
0.180
0.170
0.160
0.150
0

2000

4000

6000

8000

Closure Stress, psi

Figure 1: Pack width versus stress for 16/30 white sand

The largest variation in width is caused by the original,


unstressed packing arrangement. Influence of the initial
packing state is carried through the remaining stress ramp and
affects pack compression and re-arrangement under increased
loading. In the test set shown, as in all standard tests, great
care has been taken to ensure that the initial grain packing is
as consistent as possible.
1000 psi
Mean
Standard Error
Standard Deviation
St Dev (% of Mean)
Minimum
Maximum
Count

0.228
0.001
0.003
1.139
0.225
0.235
13

2000 psi
0.224
0.001
0.003
1.139
0.222
0.23
13

4000 psi
0.217
0.001
0.003
1.393
0.211
0.223
13

6000 psi
0.205
0.001
0.003
1.561
0.202
0.213
11

8000 psi
0.195
0.001
0.002
1.051
0.194
0.201
11

Table 1: Statistical analysis of pack width results for 16/30 white


sand

Even with standardized cell loading procedures there is a


consistent variation in measured pack width using the same
proppant. A summary of a statistical analysis of the data from

SPE 84306

the tests shown above is presented in Table 1. These data


indicate that a variation in expected pack width of less than
2% is expected under ideal packing conditions.
Data for 20/40 mesh white sand is presented in Figure 2
and Table 2. Similar data sets have been analyzed for many
other proppant types and sizes. When the tests are conducted
under ideal conditions by the same laboratory using the same
packing techniques the typical standard deviation is less than
2% of the mean value for a set of measurements. Including
data from other laboratories increases the maximum expected
standard deviation to 5% of mean. In general proppant pack
width should be predictable to within 5% for any specified
concentration as long as the pack is relatively consolidated.
White Sand (20/40) 2#/ft2
0.250
0.240

proppant pack consisting of 5 #/ft2 intermediate-strength


ceramic.
Loss of width to embedment is difficult to quantify and
cannot be observed without disassembly of the proppant pack.
It affects both available flow width, hence internal fluid
velocity in the pack, and conductivity. Its effects are
sometimes accounted for as a decrease in permeability and
sometimes as a change in pack width. The way internal width
losses are handled can significantly affect velocity-dependent
(non-Darcy) conductivity estimates.
The figure also illustrates a second source of effective
pack-width loss that is less readily apparent from outside
measurements: the width loss from spalling of the formation
into the proppant pack. Spalling is essentially the extrusion of
formation material into the proppant pack. The formation
material is usually crushed formation grains or fines generated
by the embedment process as the harder proppant grains are
forced into the fracture wall.

Pack Width, inches

0.230
0.220
0.210
0.200
0.190
0.180
0.170
0.160
0.150
0

1000

2000

3000

4000

5000

6000

7000

8000

9000

Closure Stress, psi

Figure 2: Pack width versus stress for 20/40 white sand

Table 2: Statistical analysis of pack width results for 20/40


white sand

Secondary changes in pack width


The changes in width shown above represent the
measurable decrease in external or overall pack width. The
primary cause of these changes in width is grain rearrangement as packing structure changes under load, and
grain failure at contact points.
Another mechanism that affects external pack width is
grain embedment into the walls of the fracture or facings of
the laboratory test cell. The Stim-Lab data is generated using
hard Ohio sandstone plates as the pack facings. This sandstone
has a Youngs Modulus of about 5 MMpsi. Embedment at
10,000 psi closure stress reduces the pack width by about 1/8
of a grain diameter on each face, or roughly 0.006-0.008
overall.
Embedment losses increase dramatically for softer rocks.
For nearly unconsolidated sands the pack width loss to
embedment can reach one full grain diameter or more on each
face. Figure 3 shows embedment in a high concentration

Figure 3: Embedment and spalling in soft formations

Variation in permeability
As with the measurement of pack width, the variation in
initial packing conditions also influences the measurement of
pack permeability, but to a larger degree. Again, taking as
much care as possible to form a consistent initial pack, the
data in Figure 4 show a typical variation in permeability as a
function of closure stress for 16/30 mesh white sand.
Permeability of 16/30 White Sand
1000.0

Permeability, darcy

1000 psi 2000 psi 4000 psi 6000 psi 8000 psi
Mean
0.226
0.222
0.215
0.205
0.195
Standard Error
0.000
0.000
0.000
0.001
0.001
Standard Deviation
0.001
0.001
0.002
0.003
0.003
St Dev (% of Mean)
0.561
0.491
0.783
1.503
1.313
Minimum
0.224
0.221
0.211
0.199
0.191
Maximum
0.228
0.225
0.217
0.208
0.199
Count
13
13
13
11
11

100.0

10.0
0

2000

4000

6000

8000

10000

12000

14000

Closure Stress, psi

Figure 4: Permeability vs. stress for 16/30 white sand

The statistical analysis summary for the data set shown is


presented in Table 3. Numerous data sets gathered under ideal

SPE 84306

packing conditions on various proppants indicate that a


standard deviation of +/- 10% of the mean expected
permeability at low stress is common. The trend of
permeability reduction with stress is also slightly affected by
initial packing. The changes in onset of pack compaction and
grain failure are included in the observed variance in
permeability at higher stresses. With all materials, the error
as a percentage of the mean increases and easily reaches +/25% at high stress. This is a natural consequence of the
properties of granular materials under stress. Therefore, a
representation of the properties of a granular pack is, by
definition, statistical in nature and must be the product of
replicate measurements.
1000 psi
Mean
Standard Error
Standard Deviation
St Dev (% of Mean)
Sample Variance
Minimum
Maximum
Count

510.54
12.79
46.13
9.04
2128.13
422.92
585.28
13

2000 psi
450.08
8.34
30.09
6.68
905.11
395.24
505.91
13

4000 psi
279.69
9.30
33.52
11.98
1123.50
242.26
338.57
13

6000 psi
123.73
7.14
23.67
19.13
560.42
88.79
150.00
11

8000 psi
46.62
3.60
11.93
25.59
142.27
26.04
63.94
11

Table 3: Statistical analysis of permeability results for 16/30


white sand

Similar data for the 20/40 mesh white sand tests are shown
in Figure 5 and Table 4. The variations in measured
permeability are very similar for these, and other, proppant
types. Changes in pack concentration have little apparent
affect on the error in permeability measurement.
Permeability of 20/40 Whie Sand

Permeability, darcy

1000.0

100.0

10.0
0

2000

4000

6000

8000

10000

Closure Stress, psi

Figure 5: Permeability vs. stress for 20/40 white sand


1000 psi
Mean
Standard Error
Standard Deviation
St Dev (% of Mean)
Sample Variance
Minimum
Maximum
Count

223.77
6.33
20.99
9.38
440.71
186.83
264.58
11

2000 psi
191.65
4.77
15.81
8.25
249.94
168.60
226.01
11

4000 psi
135.86
4.36
14.48
10.66
209.58
114.13
157.23
11

6000 psi
68.73
4.09
12.26
17.83
150.22
50.73
84.81
9

8000 psi
29.98
2.48
7.44
24.82
55.38
20.47
39.59
9

Table 4: Statistical analysis of permeability results for 20/40


white sand

Overall conductivity variations


Taken together the variations in pack width and
permeability show that conductivity variations of +/- 20%
about a mean are within laboratory accuracy for a given
proppant type and size
Packing of granular materials
The effects of stress on observed pack width and
permeability are caused by the development of a non-uniform
stress distribution in the proppant pack. This observation has
been confirmed by several studies of the properties of granular
materials in various areas of industrial research. In these areas,
as in hydraulic fracturing, the packing arrangement, flow
capacity, mechanical structure and ability to withstand stress
are key properties of granular materials. Pioneering work with
photoelastic (birefrigent under stress) disks, clearly shows that
granular systems are dominated by stress chains corresponding
to the paths along which the majority of the force is carried.5
The stress chains lead to high intergranular stresses at grain
contacts, and to areas of very low stress in the same pack.
Observations confirm that forces in granular systems are
inhomogeneous and intermittent, if the system is deformed by
the application of an external load such as the closing of the
fracture. This leads to local grain failure even under conditions
where the average stress appears to be below the strength of
the granular material. The result is an increasing inconsistency
in behavior of granular packs at elevated stress.
Mechanisms affecting field conductivity
Statistical variations in proppant pack conductivity are a
function of packing arrangement and the development of a
non-uniform stress field in the pack, when subjected to outside
loads. When a proppant pack is placed in a fracture several
other damage mechanisms also affect the final conductivity.
These include the effects of filter-cake, gel residue, non-Darcy
flow, and multiphase flow including gravity and capillary
gradients. The impact of these potential damage mechanisms
is cumulative.
Filter-cake and gel residue damage
The data shown above represents the conductivity of a
clean proppant pack containing only compatible brine such as
2% KCl. Additional changes to conductivity result from the
addition of a polymeric frac fluid. During fracturing an
aqueous suspension of polymer deposits a filter-cake of
concentrated polymer on the fracture wall.6 The thickness of
the filter-cake can reach almost 0.055 (1.4 mm), or more than
one proppant grain diameter for 20/40 mesh proppant.7 The
ultimate thickness of the filter-cake depends on formation
permeability, fluid efficiency, local shear rate causing erosion
of the deposited cake, and many other factors.8 The filter cake
deposition decreases the effective fracture width during
pumping and causes the proppant to be contained between the
layers of gel cake on the fracture walls. The polymer in the
filter cake can reach concentrations equivalent to 300-400
lbs/Mgal.
During fracture closure the filter cake is extruded into the
proppant pack and invades the pore space of the pack. Using
an average porosity of 40% for the proppant pack, the 0.055
wall filter cake can extrude to a depth of up to 0.137 (2.33

SPE 84306

100
90
80

% Regained Permeability

mm) in the pack. The concentrated gel can extrude to a depth


of more than 4 grain diameters into the pack from each wall.
Under the worst-case conditions the concentrated gel cake can
fill a 2 lb/ft2 proppant pack.
The polymer gel residue behaves essentially as a solid
once it has come to rest, and may develop a substantial yieldpoint that resists flow. This makes displacement of the gel by
produced fluids very difficult. The presence of the gel cake
occludes pore space and flow capacity within the proppant
pack and, primarily, decreases the available cross-sectional
flow area or effective pack width. Open channels for fluid
flow through the proppant pack, or between the gel-invaded
zones, represent a small fraction of the total pack crosssection. This results in decreased conductivity and increased
flow velocity in the available pore channels.
As with formation spalling, width losses to gel filer-cake
extrusion are internal losses and cannot be easily quantified
from external pack-width measurements. The impact on
conductivity damage and the manner in which it is
apportioned between pack width and permeability damage is
not reported in any standardized way. Specific measurements
have been made to determine the effects of gel filter-cake on
cleanup and proppant pack conductivity. These tests show that
the presence of a filter-cake has a much more significant
impact on the conductivity of a narrow (low concentration)
pack than on a thicker (high concentration) pack. Given the
same overall regained permeability in the portion of the pack
actually available for flow, this observation is consistent with
a decrease in effective flow width inside the pack.
Data from a series of cleanup tests in 1, 2, and 4# lb/ft2
proppant packs using the same fluid and proppant is shown in
Figure 6. In these tests a dynamic filer-cake was first
deposited in a fixed-width cell. The proppant slurry was then
placed in the conditioned cell and the closure stress increased
with continuous leakoff until the cell was closed on the
proppant pack. The closure stress was then ramped at
consistent rate to the test conditions. Regained pack
permeability to 2% KCl brine was then measured at a series of
increasing flow rates. In general the removal of gel damage
and filter-cake can be correlated to the pack effective, or
pseudo-Reynolds number (Rep) established during clean-up.
The correlating parameter is termed a pseudo-Reynolds
number because the term v/ is not a dimensionless group. A
true Reynolds number requires the inclusion of a characteristic
length in the numerator.
The data points on the plot show the actual measured
regained permeability compared to a clean pack with no gel
residue. The lines are generated by a model of the cleanup
process that accounts for the presence of gel filter-cake
through a reduction of pack width that is independent of the
reduction in bulk permeability. Both the filter-cake and the
dispersed bulk gel residue are assumed to clean-up as a
function of increasing Rep. Clean-up rates for the two damage
mechanisms are different. The amount of damage from the
two mechanisms is a strong function of pack width, given the
same filter-cake and gel residue in the same proppant. The
conductivity multiplier resulting from regained permeability is
given by Frp and the factor for filter-cake loss is Ffc.

70
60

4#/ft2 Pack
50
40

2#/ft2 Pack
30
20
10

1#/ft2 Pack
0
0.01

0.1

10

100

Effective Reynold's Number, rho*v/mu

Figure 6: Variation in cleanup in different concentration packs


caused by filter-cake

One problem with the interpretation of these results is that


the actual flow velocity, hence the effective pseudo-Reynolds
number is not known since the open channel width is
indeterminate. During clean-up the actual flow width changes
as filter-cake is eroded or displaced. The lab data can only be
based on an apparent (external) pack width and superficial
velocity. Note that the combined effects of filter-cake and gel
residue can account for conductivity damage ranging from
about 20% up to a factor of 10 or more. In extreme cases,
where the entire pack is filled with filter-cake material, the
conductivity can be completely eliminated.
Non-Darcy flow effects
The same difficulty in accounting for superficial versus
interstitial velocity occurs when dealing with non-Darcy flow
effects in a fracture containing gel residue, filter-cake, and
possibly multiple flowing phases. Non-Darcy flow is usually
represented by the Forchheimer equation.

dP
dL

+ v 2 (1)

Darcys Law predicts a linear relation between pressure


gradient and flow rate. At high flow rates a deviation from
linearity is commonly observed and a second-order term is
invoked to account for the non-linearity. In Forchheimers
equation, the first term represents the Darcy linear pressure
gradient in terms of superficial velocity (v), fluid viscosity (),
and the inverse of permeability (k). The second term
represents the deviation from linearity caused by inertial
losses as flowing fluid expands and contracts while moving
from pore throats to pore bodies and back again. The nonlinear pressure gradient is expressed as a function of fluid
density (), the square of the superficial velocity, and a
constant (). As with the Darcy permeability, the constant is
a contrived proportionality constant with units of atm-sec2/g.
Note that, as with the definition of permeability, the constant
contains the gravitational acceleration and accounts for
conversion of mass to force, among other things.
The primary difficulty in estimating the effect of nonDarcy flow is that only the superficial velocity, as defined by

SPE 84306

0.1
fg=0.939

0.09

0.005
0.0045
0.004
0.0035
Beta, atm-sec 2/g

the external pack width, can be estimated from the volumetric


flow rate. The impact on interstitial velocity of internal packwidth changes caused by gel filter-cake and spalling, along
with areal variations in flow, caused by gel residue and
saturation profiles, cannot be readily determined. Instead, the
relationship of observed superficial velocity to total pressure
drop must be used to estimate an effective value of that may
be a function of many variables. The apparent value of is
used to describe the change in apparent conductivity at high
velocity, but does not account for the increased interstitial
velocity caused by occlusion of pore space and elimination of
flow area.
The inclusion of a second (or third) flowing phase in the
pore system, or the presence of an immobile gel residue also
changes the geometry of the pore system and affects the actual
flow path. In some cases this results in a decrease in inertial
losses that partially offsets the expected increase in non-Darcy
pressure drop.
The data plotted in Figure 7 were obtained in a series of
flow experiments on one proppant pack under conditions of
steady-state flow with gas and water at various fractional
flows. Once steady-state conditions were established the total
flow rate was varied at constant fractional flow and the
apparent permeability was determined. The results are shown
as a Forchheimer plot with a separate line for each fractional
flow. The slope of each line is for that flow condition and
the intercept is the inverse of the apparent Darcy permeability.

0.003
0.0025
0.002
0.0015
0.001
0.0005
0
0.92

0.93

0.94

0.95

0.96

0.97

0.98

0.99

Gas Fractional Flow, fg

Figure 8: Apparent variation of with gas fractional flow

The impact of non-Darcy flow effects on overall pack


conductivity, or total pressure drop, can be expressed as a
multiplier on the Darcy pressure drop or divisor of pack
conductivity. The appropriate multiplier, the non-Darcy factor
(Fnd), is obtained by factoring the term v/k out of both terms
in the Forchheimer equation. The total non-Darcy pressure
drop can be expressed in terms of the expected Darcy pressure
gradient as shown in equation (2).


dP

= 1 + k v dP
...(2)

dL Darcy
dL total

Inverse Permeability, 1/k

0.08
fg=0.970

0.07
0.06
0.05

fg=0.988

0.04

fg=0.994

0.03
0.02
0.01

fg=1.0

0
0

10

15

20

25

30

35

Pseudo Reynold's Number, v /

Figure 7: Forchheimer plot for various gas fractional flows

The data clearly shows decreasing effective gas


permeability and increasing as gas fractional flow decreases.
The first result is the expected impact of two-phase relative
permeability. The second result implies that the presence of
the water in the pore system changes both the velocity
distribution and the interconnected pore system in which the
phases flow.
Plotting the slopes of each curve against their respective
gas fractional flow values generates an empirical correlation
of with fg. Because of the limited data available, it is not
clear whether this correlation is general, but it provides insight
into one possible impact of saturation on the internal flow
dynamics of a proppant pack.

In equation (2), the term (1+kv/) is the non-Darcy factor


(Fnd) which is the ratio of the total pressure gradient to the
Darcy pressure gradient. Detailed numerical modeling of the
flow profile along the length of the fracture shows that the
appropriate superficial velocity for use in equation (2) is about
2/3 of the velocity at the fracture-wellbore interface. This
adjustment accounts for the decrease in velocity toward the tip
of the fracture as overall volumetric rate decreases.
It is interesting to note that the non-Darcy pressure-drop
multiplier, Fnd, is 1+Re where the term k serves as the
apparent characteristic length in the dimensionless group, as
suggested by Jones.8,9 Application of typical values for the
terms in Fnd shows that the ratio of the total pressure drop to
the Darcy pressure drop can range from 1.0 to as high as 15 or
more under normal field conditions.
Multiphase flow and relative permeability
Under virtually all realistic field conditions, the proppant
pack contains more than one fluid phase, whether there is only
one mobile phase or several. The proppant is placed in a
fracturing fluid that may be aqueous or oil-based. This fluid
always leaves a residual saturation behind, determined by the
relative-permeability and capillary pressure curves for the
specific proppant pack. Additional liquids may be produced
from the reservoir as mobile phases or through condensation
in the proppant pack, or even in the wellbore. Wellbore fluids
can be introduced into the proppant pack during shut-in
periods or when the well rate falls below the loading rate for
the tubular configuration.

SPE 84306

Historically the effects of relative permeability in the


proppant pack have been largely ignored. In recent years some
attempts have been made to characterize the relative
permeability of a proppant pack, at least for two-phase gaswater flow. Figure 9 shows a Cartesian plot of a typical
relative permeability curve for 20/40 mesh light-weight
ceramic proppant. These data were obtained by Stim-Lab
through a series of steady-state and unsteady-state
displacements under 4000 psi closure stress using a 2#/ft2
proppant pack.
The rapid decline in brine permeability at very low gas
saturation is fundamentally important to proppant pack cleanup. Once the gas saturation in the pack becomes mobile, the
rate of water desaturation decreases dramatically. At the same
time, the relative permeability to gas remains at a low value
until the gas saturation can increase above about 40%. At this
gas saturation the water relative permeability is less than 0.05
and its mobility is very small.
1
0.9

0.7
0.6
0.5

g
.(4)
=
fg

Given an observed gas fractional flow it is sometimes


difficult to estimate the saturations in the proppant pack or the
effective permeability for the multi-phase system. It is the
overall system effective permeability that determines the
effective conductivity of the proppant pack. As an example,
note that at 40% gas saturation, for the gas-water relative perm
data shown in Figure 9, fg is 0.997, krg is approximately 0.224
and krw is 0.032. Under the conditions given, the water flow
rate is almost insignificant, yet the overall system permeability
is only 22% of the expected proppant pack permeability,
ignoring all other damage factors. To put this effect in field
terms, a gas well producing water at 10 bbl/MMSCF with a
flowing bottom-hole pressure of 2000 psi has a gas fractional
flow of 0.995 and an expected gas relative permeability of
0.189. This will reduce the overall proppant pack conductivity
by a factor of more than five. Figure 10 is a plot of the
conductivity multiplier generated by the relative permeability
data set from Figure 9 over the range of gas fractional flow
from 0.9 to 1.0. The data in the plot show that a 10%
volumetric liquid flow can cut proppant pack conductivity by
a factor of 20.

0.4
krw

0.3

krg

0.2
0.1
0
0

0.2

0.4

0.6

0.8

Gas Saturation

Figure 9: Gas-water rel-perm data for a typical proppant pack

It is important to stress the difference between relative


permeability and fractional flow. The fractional flow of a
phase is defined as the phase flow rate divided by total flow
rate. It is determined by the ratio of phase mobilities which
include both the relative permeability and viscosity of each
flowing phase. The fractional flow of gas, fg, is given by
equation (3).

fg =

kw g
1+

k
g w

.(3)

The total multi-phase system pressure gradient is


determined by the total mobility of the system, which varies
with the fractional flow. The multi-phase conductivity
multiplier, or inverse of the multiphase pressure-drop
multiplier, is given by equation (4). This relationship is
derived from the calculation of the pressure drop under singlephase flow divided by the pressure drop for multi-phase flow
using Darcys Law.

Effective Conductivity Multiplier

Relative Permeability

0.8

k +k

Fmp = g w w g

0.1

0.01
0.9

0.92

0.94

0.96

0.98

Gas Fractional Flow, fg

Figure 10: Conductivity reduction caused by multiphase flow

Gravity segregation and capillary effects


The equilibrium water saturation established in a proppant
pack in a vertical hydraulic fracture is determined by the
balance of capillary, viscous and gravitational forces
established during production. In general, the gravitational and
capillary gradients are large enough to dominate the system in
linear flow, and the viscous gradients induced by flow, are
relatively small by comparison. The vertical gravity head for a
gas-water system is approximately 0.4 psi/ft while the lateral
viscous pressure gradient for a typical gas well producing 1.0
MMSCF/D may be in the range of 0.2 psi/ft in the fracture.
Capillary forces are important at all discontinuities in the
reservoir-fracture system. At each boundary between the
reservoir and proppant pack, and the proppant pack and
wellbore, there is a distinct change in the capillary pressure

SPE 84306

caused by a change in the average pore size of the medium.


This causes a discontinuity in the pressure gradient of the
wetting phase that must be dissipated by a buildup of
saturation at the boundary. The increased wetting phase
saturation, and associated gradient away from the boundary, is
called the capillary end-effect. The impact of the high wetting
phase saturation and its generation is described by several
authors.11-13 While these end-effects are usually only
considered important in laboratory systems, they are important
in any cases where capillary forces are large and pressure
gradients are small. Holditch noted that the combination of
relative permeability and capillary forces in a low
permeability reservoir system, with small pore-size
distribution, can result in a severe reduction in post-frac
deliverability.14 The tendency for tight, water-wet reservoirs to
hold high water saturation along the fracture face, resulting in
zero effective gas permeability over a large part of the created
fracture surface area, may be a noteworthy source of damage
to post-frac production in some cases.
A similar study to that presented by Holditch was
conducted for this paper using a multiphase 3-D numerical
reservoir simulator, including the effects of gravity and
capillary forces with a discontinuity at the fracture-formation
face. Representative relative permeability and capillary
pressure curves were used for the reservoir and proppant pack.
The model was used to predict gas and water saturation
distributions in and around the fracture during cleanup. The
model was initialized by injecting water into the fracture and
surrounding reservoir, then putting the well on production.
70

Fracture Height, feet

60
50

Initial
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6

40
30
20
10
0
0

0.2

0.4

0.6

0.8

1.2

Water Saturation, fraction

Figure 11: Vertical water saturation distribution in a fracture


during clean-up

Figure 11 shows the computed water saturation


distribution for a point approximately mid-way along the
length of the fracture from the wellbore to the frac tip. The
various curves show the progression of desaturation as a
function of producing time. In this case the fracture was filled
with water (1 cp) assuming no polymer residue, filter-cake, or
residual gel viscosity. The saturation profile curves reach a
near-equilibrium state after about one week of production and
the saturation distribution becomes static.
The model results show a tendency for some gravity
segregation within the fracture. The water saturation in the top
15 feet of the fracture decreases to 40-50% (50-60% gas

saturation). The bulk of the fracture height retains a water


saturation of about 55% while the bottom few feet of the
fracture remain almost completely water saturated. Superimposing the gas-water relative permeability relationship on
this saturation distribution clearly shows that the gas flow
distribution will be non-uniform across the height of the
fracture. The expected gas relative permeability may approach
75% in the top few feet of the fracture, yet stay as low as 25%
over 80% of the fracture height. The saturation and resulting
velocity profile will lead to a distribution of non-Darcy effects
across the fracture height and can make any accurate estimate
of overall effective conductivity difficult.
The behavior of the water saturation distribution along the
fracture face during this clean-up period is also interesting to
observe. During the first week of gas flow the water saturation
along the face of the fracture drops from nearly 100% to about
70%. The saturation distribution along the height of the
fracture is effectively constant. Some of the water contained in
this invaded zone around the fracture is produced. A
significant amount of the water is imbibed further back into
the reservoir, although this water flow direction is countercurrent to production. The spontaneous imbibition of water
results in an increase in water saturation from about 40% to
more than 50% in the reservoir 2-5 feet from the fracture face.
This counter-current imbibition can only occur when the
capillary forces pulling water into the matrix exceed the
viscous forces driving the water into the fracture.
Impact of reservoir deliverability
In many cases the conductivity of a proppant pack has been
described in the absence of the effect of reservoir behavior.
The preceding review of the impact of various mechanisms on
proppant pack conductivity shows clearly that it is not possible
to fully separate reservoir deliverability from proppant
performance. It is the reservoir deliverability, including
transmissibility (kh/), available drawdown and saturation
state that determines the rate of flow of each phase in the
fracture, hence the velocity and density of the fluid stream.
This sets the value of the pseudo-Reynolds number at any time
during fracture cleanup and production, and controls the
relationship of viscous to gravity and capillary gradients.
The available pseudo-Reynolds number controls the
degree of fracture clean-up, both for bulk gel residue
displacement and filer-cake removal. For these mechanisms a
large value of Rep improves clean-up and conductivity. At the
same time a large value of Rep causes an increase in non-Darcy
pressure drop, resulting in a drop in effective conductivity.
Balancing these competing factors requires knowledge of the
dominant fracture damage mechanism. In some cases the
composition and characteristics of the fracturing fluid impact
the stimulation efficiency as much (or more) than the choice
of proppant.
The impact of reservoir deliverability and clean-up rate
makes establishment of high conductivity very difficult in a
low permeability or depleted reservoir. It also points out why
the same fluid and proppant combination can perform very
differently in different wells.

SPE 84306

Gel yield-point and critical FCD


The combined effect of all the damage mechanisms discussed
up to this point results in an adjusted proppant pack
permeability and effective pack width. Together these
parameters determine the effective proppant pack
conductivity, kfwf. To characterize the performance of the
fracture in a specific reservoir, the relative flow capacity of the
fracture must be compared to the flow capacity of the
formation. The relative conductivity of the fracture and
formation can be described by the dimensionless fracture flow
capacity, FCD, given by equation (5).

FCD =

k f wf

(5)

kXf
In general, an FCD of 30 and above is taken to represent
an infinite conductivity fracture, where the effective length is
equal to the physically propped length. For any value of FCD
the infinite conductivity half-length can be determined from
equation (6) or the plot in Figure 12. The equation presented
here is based on data originally published by Prats and CincoLey.15,16 It has been adapted to give a consistent representation
of effective fracture half-length compared to created or
propped half-length.

X eff

X created

1 + FCD

1.01

1 .7

..........(6)

the gel mass. If the available pressure differential at some


point along the fracture length is too small, the gel mass
remains as a solid plug and cannot be displaced. The
remaining fracture length, from that point to the propped
fracture tip, remains gel-plugged and unavailable for flow.
The pseudo yield-point, or minimum pressure gradient to
initiate flow, is difficult to measure in the laboratory. Ideally
the test should be done under slowly varying shear stress, until
flow begins. Current procedures give a maximum observed
pressure drop at a constant low flow rate. The values obtained
by these two measurements are different, but the initiation
pressure obtained from the constant-rate test can be correlated
to gel stiffness and can be used to describe yield-point
behavior.
This effect can be empirically represented using a fluiddependent model input parameter, FCDcrit. The input value
specifies the minimum dimensionless flow capacity (FCD)
that will effectively clean-up and contribute to production. The
smaller FCDcrit becomes the more fracture length will
contribute to flow and the longer the effective half-length will
be.
This concept is illustrated in Figure 13. With increasing
distance from the wellbore the fracture conductivity, kfwf,
decreases as the fluid velocity in the fracture drops. The
decrease in velocity also causes a decrease in the local
pressure gradient available to displace gel. The computed
incremental value of FCD depends on both the conductivity
and the overall flow length, which also increases with distance
from the well. These combined factors cause a rapid decline in
apparent FCD with length, as shown in the figure. When FCD
drops below the critical FCD for clean-up, the remaining
fracture length remains gel-plugged.

Effective Length Ratio,


Xfeff /Xfcreated

0.1

0.01

0.001
0.01

0.1

10

100

1000

Fracture Conductivity, FCD

Figure 12: Effective fracture length ratio from FCD

In Figure 12 and equation (6) the created or propped


fracture half-length is Xcreated and the effective infiniteconductivity half-length is Xeff. This relation assumes that the
entire created fracture length is contributing to flow with a
diminishing flow rate, pressure gradient, and effective FCD as
length increases away from the wellbore. In reality, there is
probably a cut-off point at which the fracture ceases to
contribute flow. This cut-off is caused by the development of a
gel pseudo-yield-point.
Like a drilling mud, a concentrated polymer gel tends to
behave more like a solid than a liquid, once it becomes static
in a proppant pack. A minimum pressure differential is
required to break circulation and induce shear and flow in

Figure 13: Illustration of critical FCD for clean-up

The apparent producing fracture half-length, including the


effect of gel plugging, is calculated from the effective infiniteconductivity fracture half-length given by equation (6). The
final apparent producing half length, given by equation (7), is
based on the assumption that an infinite conductivity fracture
is represented by FCD=30. The impact of FCDcrit is shown by
the ratio of the critical clean-up value to an infiniteconductivity fracture response.

10

SPE 84306

X eff

30

1+
1.7 FCD
crit

1.01

... (7)

In equation (7) the apparent fracture length approaches the


infinite-conductivity length as FCDcrit approaches zero. This is
equivalent to having no loss of fracture length to gel plugging
and no minimum cut-off value for FCD. The final apparent
fracture half-length, Xapp, is the fracture half-length that
represents stimulation performance in an infinite-conductivity,
Darcy-flow reservoir model. While FCDcrit is a fluiddependent parameter, a value near 30 appears to represent
many systems. A fluid that leaves a very light gel residue, or a
residue that is easily dispersed, may have a value as low as 10.
Development of a laboratory procedure to quantify this
parameter and correlate it to specific fluid properties is one of
the most important and challenging tasks left in the process of
optimizing a stimulation design.
Integration of damage, cleanup and reservoir
deliverability
To effectively evaluate the combined dynamic interactions of
reservoir transient flow, gel clean-up, multiphase and nonDarcy flow, a complex model is required that can account for
the often competing effects of all these processes. The PredictK simulator combines reservoir transient production
forecasting for an arbitrary shape, rectangular, bounded
reservoir with a damaged hydraulic fracture. The effects of
closure stress, embedment, spalling, filter-cake deposition and
erosion, bulk gel damage, multiphase flow, and non-Darcy
flow are accounted for.
Because of the impact of fracture conductivity on well
production and the effect of reservoir and well flow rate on
conductivity, the solution requires multiple iterations to reach
a stable converged solution for fracture conductivity at each
producing time. The model allows the selection of proppant
and fluid system based on a complete economic analysis of
material costs and generated discounted cash-flow
from production.
The correlations used to drive predictions of conductivity
versus stress, multiphase flow, non-Darcy flow, and clean-up
are based on more than 15 years of laboratory measurements
conducted by the Stim-Lab proppant consortium. The model is
available to members of the consortium.
Field application: Predicted post-frac performance
An example of predicting post-frac production by accounting
for all damage mechanisms is shown in the following field
example. The well is a tight-gas producer with an initial
reservoir pressure of 10,700 psi. Production analysis indicates
an average reservoir permeability of 0.069 md for a net pay
thickness of 18 feet. The zone was hydraulically fractured
with 300,000 lbs of high-strength 20/40-mesh ceramic
proppant and a zirconate crosslinked premium-grade
fracturing fluid. A post-frac history match of the job using a
planar 3-dimensional fracture geometry model showed good
height containment and a very long created length (about 1500

feet), only part of which was effectively propped. Figure 14


shows the predicted proppant concentration distribution at the
end of the job.

Figure 14: Proppant concentration profile for example case

A clean-up and production model was run using the total


propped length for this job (1500 feet) with an average
concentration of 1.3 lb/ft2. The reservoir properties determined
from long-term production analysis were used in the forecast
of post-frac performance. The average proppant concentration
and propped length from the model were input to Predict-K
for the analysis. The actual post-frac production and the rates
determined from the combined fracture and reservoir clean-up
model are shown in Figure 15.
7000

6000

Gas Rate, MSCF/D and WHP, psi

X app =

Actual Rate
Rate-1000 WHP
WHP
Rate-3000 WHP

5000

4000

3000

2000

1000

0
0.00

50.00

100.00

150.00

200.00

250.00

300.00

350.00

Days on Production

Figure 15: Observed and modeled post-frac production rates

The model was run for two cases of WHP, 3000 psi and
1000 psi, while the actual well produced for the first 70 days
with a variable wellhead pressure. In Figure 15 the red line is
the actual WHP over the producing time and the magenta
squares are the actual well flow rate. The dark blue line is the
model predicted flow rate assuming a constant WHP of 1000
psi and the dashed light-blue line is the calculated well rate for
3000 psi WHP. In general the model accurately matches the
observed flow rate.
Even with a created fracture half length of almost 1500
feet, using an average concentration of 1.3 lb/ft2, the model
matches the performance with an effective infiniteconductivity fracture half-length of 140 feet. While the
undamaged baseline conductivity of this proppant was
calculated to be almost 2000 md-ft, the effective conductivity
after accounting for all damage mechanisms in this simulation
was less than 60 md-ft, giving an FCD of 0.578. The overall
final conductivity is computed from equation (8), where the

SPE 84306

11

values of kf and wf are first adjusted for the effects of closure


stress, embedment, and spalling. Equation (6) gives an
infinite-conductivity effective length of 377 feet, ignoring the
impact of gel tip-plugging. The correction for gel plugging,
from equation (7) further reduces the apparent length to
140 feet.

FCD =

k f w f Frp F fc Fnd Fmp

consortium for their support in the development of this


technology. We would also like to acknowledge Marathon Oil
Company for the contribution to the numerical simulation of
multiphase flow. Many model cases were run to provide
insight into gravity and capillary effects which could not be
included here.
Nomenclature

.(8)

k X created
Note that most of the damage factors are dependent on the
flow velocity in the fracture, and its influence on Rep. The final
conductivity determination, therefore, requires an iterative
solution. The initial damage conditions are estimated and the
reservoir deliverability is determined for the given effective
fracture length and conductivity. The new estimate of
deliverability is used to re-calculate Rep and the various
damage factors. The iteration proceeds until the system
converges at a balanced reservoir deliverability and fracture
clean-up state.
The resulting FCD gives the dimensionless fracture
conductivity in the created frac. The effective fracture halflength, Xeff, is calculated from equation (6). The final
adjustment for gel plugging is made by application of equation
(7) and a fluid-dependent value of FCDcrit, here assumed to
be 30.
The relative magnitude of the damage factors for the
example case is shown in Table 5. Each of these damage
mechanisms results from the combined influence of the
selected
proppant,
fluid,
reservoir,
and
well
operating conditions.

fg
Ffc
Fmp
Frp
Fnd
k
kw
kg
L
P
Rep
v
w
Xapp

Xcreated
Xeff
Xf

References
1.
2.

Damage Factor

Conductivity
Multiplier

Remaining
Conductivity

Ffc

0.63

0.63

Fmp

0.85

0.54

Frp
Fnd

0.69

0.37

0.08

0.03

3.
4.
5.

Table 5: Summary of conductivity damage factors

Conclusions
Clean pack baseline data for proppant pack conductivity are
optimistic. Small variations in proppant properties (less than
15 to 20% difference in conductivity) may not be statistically
significant when applied to field measurements of
performance. Proppant conductivity should be considered with
all damage effects included. Multiple damage mechanisms
affect the actual fracture conductivity under field conditions
and these various damage mechanisms are cumulative.
Effective fracture conductivity is generated from the coupling
of reservoir flow to fracture cleanup, in conjunction with
proppant characteristics. Field conductivity is much lower
than generally believed or expected.
Acknowledgements
The authors would like to acknowledge the support of the
member companies of the Stim-Lab proppant conductivity

= fractional flow of gas


=filter-cake conductivity factor
=multi-phase flow conductivity factor
=regained permeability factor
=non-Darcy flow conductivity factor
=Darcy permeability
= relative permeability to water
= relative permeability to gas
= Flow path length
=Pressure
=pseudo-Reynolds number
=superficial flow velocity
= fracture or pack width
=apparent producing fracture half-length
=created fracture half-length
= effective infinite-conductivity length
= fracture half-length
=Forchheimer coefficient
=viscosity
=density

6.

7.

8.

McGuire, W. J. and Sikora, V. J.: The Effect of Vertical


Fractures on Well Productivity, Trans., AIME, v. 253.
Stim-Lab Proppant Conductivity Consortium Reports,
1986-2003.
Muskat, M.: The Flow of Homogenous Fluids Through
Porous Media, J. W. Edwards, Inc., Ann Arbor,
Michigan, 1946.
Recommended Practices for Evaluating Short Term
Proppant Pack Conductivity, API Recommended Practice
61, American Petroleum Institute, Washington, DC, 1989.
Geng, J., Longhi, E., Behringer, R., Howell, D.: Memory
in Two-dimensional Heap Experiments, Physical Reviews
E, Volume 64, 060301-1.
McGowen, J.M., Vitthal, S., Parker, M.A., Rahimi, A., and
Martch Jr., W.E.: Fluid Selection for Fracturing HighPermeability Formations, paper SPE 26559 presented at
the 68th Annual Technical Conference and Exhibition of
the Society of Petroleum Engineers held in Houston, Texas,
3-6 October 1993.
McGowen, J.M. and Vitthal, S.: Fracturing-Fluid Leakoff
Under Dynamic Conditions Part 1: Development of a
Realistic Laboratory Testing Procedure, paper SPE 36492
presented at the 1996 SPE Annual Technical Conference
and Exhibition held in Denver, Colorado, U.S.A., 6-9
October 1996.
McGowen, J. M., Barree, R. D., and Conway, M. W.:
Incorporating Crossflow and Spurt-Loss Effects in
Filtration Modeling Within a Fully 3D Fracture-Growth
Simulator, paper SPE 56597 presented at the 1999 SPE
Annual Technical Conference and Exhibition held in
Houston, Texas, 3-6 October 1999.

12

SPE 84306

9.
10.

11.
12.
13.
14.
15.
16.

Jones, S. C.: A Rapid Accurate Unsteady-State


Klinkenberg Permeameter, SPEJ, p. 383, 1972.
Jones, S. C.: Using the Inertial Coefficient, , To
Characterize Heterogeneity in Reservoir Rock, paper SPE
16949 presented at the 62nd Annual Technical Conference
and Exhibition of the Society of Petroleum Engineers held
in Dallas, TX September 27-30, 1987.
Perkins, F. M. Jr.: An Investigation of the Role of
Capillary Forces in Laboratory Waterfloods, Trans.,
AIME, v. 210, 1957.
Richardson, J. G., Kerver, J. K., Hafford, J. A., and Osaba,
J. S.: Laboratory Determination of Relative Permeability,
Trans., AIME, v. 195, 1952.
Caudle, B. H., Slobod, R. L., and Brownscombe, E. R.:
Further Developments in the Laboratory Determination of
Relative Permeability, Trans., AIME, v. 192, 1951.
Holditch, S. A.: Factors Affecting Water Blocking and
Gas Flow from Hydraulically Fractured Gas Wells, JPT,
December 1979.
Cinco-Ley, H. and Samaniego-V., F.: Transient Pressure
Analysis for Fractured Gas Wells, JPT (Sept. 1981),
pp 1749-66.
Prats, M.: Effect of Vertical Fractures on Reservoir
Behavior Incompressible Fluid Case, SPEJ (June 1961),
pp 105-118.

Vous aimerez peut-être aussi