Fuel 142 (2015) 189198

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Numerical study of the effect of uneven proppant distribution between

multiple fractures on shale gas well performance
Wei Yu a,, Tiantian Zhang a, Song Du b, Kamy Sepehrnoori a
a
b

Department of Petroleum and Geosystems Engineering, The University of Texas at Austin, Austin, TX, United States
Department of Petroleum Engineering, Texas A&M University, College Station, TX, United States

h i g h l i g h t s
 Effect of uneven proppant distribution on well performance is investigated.
 Reservoir model with hydraulic fractures is validated using eld production data.
 Sensitivity studies are performed to quantify the key parameters.
 The range for gas production due to proppant distribution is obtained.

a r t i c l e

i n f o

Article history:
Received 30 August 2014
Received in revised form 20 October 2014
Accepted 27 October 2014
Available online 17 November 2014
Keywords:
Proppant distribution
Marcellus shale
Sensitivity study
Gas desorption
Geomechanics

a b s t r a c t
Uniform proppant distribution in multiple perforation clusters after hydraulic fracturing plays an
important role in the commercial production of shale gas. However, it is very challenging to achieve a
uniform proppant distribution during operation. In some cases, proppant distribution is uneven in different clusters within the same hydraulic fracturing stage. The effect of the uneven proppant distribution on
well performance is not well understood and has been largely neglected in most reservoir simulations.
Hence, it is paramount to develop a reservoir simulation approach to properly examine the relationship
between proppant distribution and well performance for shale gas reservoirs. In this paper, we use
numerical reservoir simulation to model the proppant distribution. The reservoir model with multiple
hydraulic fractures is validated by eld production data from Marcellus shale. Effects of gas desorption
and stress-dependent fracture conductivity are considered in the simulation model. We perform sensitivity studies to quantify the key parameters affecting the well performance between uniform and nonuniform proppant distribution. The six variables, which are cluster spacing, initial reservoir pressure,
fracture conductivity, fracture half-length, fracture height, and matrix permeability, are investigated.
The fracture conductivity ratio of 1:1.5:2.5:4 for four clusters in the same fracturing stage is investigated
for the uneven proppant distribution scenario. This work provides insights into a better understanding of
the effect of proppant distribution on well performance.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
Development of unconventional resources such as shale gas and
tight oil has been boosted by the advancements in two key technologies: horizontal drilling and multi-stage hydraulic fracturing.
Most horizontal wells are drilled in the direction of minimum horizontal stress with the purpose of creating multiple transverse
hydraulic fractures. The plug-and-perf operation is one common
completion technique, which is widely used to create multiple
fractures through spaced perforation clusters within one isolated
Corresponding author. Tel.: +1 512 574 0080; fax: +1 512 471 9678.
E-mail address: weiyu@utexas.edu (W. Yu).
http://dx.doi.org/10.1016/j.fuel.2014.10.074
0016-2361/ 2014 Elsevier Ltd. All rights reserved.

fracturing stage. Proppants such as sand and ceramic with small

size are injected with the uid into the fractures, which can hold
fractures open to provide a conductive path for uid ow from reservoir to wellbore. Multiple long hydraulic fractures with uniform
proppant distribution and sufcient fracture conductivity play an
important role in achieving effective well stimulation and economic production of shale reservoirs [17]. Cipolla et al. [8] studied
the effect of proppant distribution in the fracture network on well
performance and showed that proppant distribution signicantly
affects the fracture network conductivity and treatment design.
Gu et al. [9] presented that proppant transport in natural fractures
has an important impact on critical fracture conductivity required
for stimulation of shale reservoirs. However, it is very challenging

190

W. Yu et al. / Fuel 142 (2015) 189198

Nomenclature
CMG
CFD
EUR
Fcd
LGR
MMSCF
SRV

Computer Modeling Group

computational uid dynamics
estimated ultimate recovery
fracture conductivity, md-ft
local grid renement
106 standard cubic feet, ft3
stimulated reservoir volume

to achieve uniform proppant distribution and maintaining such

fracture conductivity because of proppant settlement, proppant
nes generation and migration in the fracture [10], proppant diagenesis [11], proppant embedment in softer rock, and proppant
crushing in harder rock [1213]. In addition, Fredd et al. [14] performed a series of experimental studies of hydraulic fracture conductivity and found that fracture displacement, size and
distribution of asperities, and rock mechanical properties signicantly impact fracture conductivity. Vincent [15] suggested that
the assumption of uniform proppant distribution within narrower
fractures with hundreds or thousands of feet in length may not be
reasonable. It is important to investigate the impact of proppant
distribution on well performance of shale gas reservoirs.
Although the plug-and-perf operation method has been extensively applied in shale reservoirs, achieving uniform proppant distribution among all perforation clusters for each fracturing stage is
challenging [16]. Cipolla et al. [17] have shown that about 40% of
the perforation clusters were not contributing to gas recovery.
Also, Miller et al. [18] reported that only 1/3 of perforation clusters
contributes to 2/3 of gas production in some shale basins based on
production logs more than 100 horizontal shale wells from six
shale basins. Daneshy [19] investigated the proppant distribution
between different perforation clusters within single stage in the
plug-and-perf operation and reported that there is uneven proppant distribution among perforation clusters and most proppant
likely enter the last perforation cluster. The amount of proppant
entered the last cluster near the toe is almost four times the proppant amount in the rst cluster toward the heel. This is because
proppant grains with higher density and larger size and mass than
fracturing uid cannot easily change direction and enter the perforations uniformly. In addition, Crespo et al. [16] observed the phenomenon of uneven proppant distribution within three perforation
clusters through conducting a large-scale experimental study to
mimic the plug-and-perf operation, and it will be more severe in
cases with higher proppant density and smaller ow rates. Bokane
et al. [20] used computational uid dynamics (CFD) technique to
simulate proppant transport in different perforation clusters
within a single stage and to understand the phenomena of uneven
proppant distribution within perforations. However, the impact of
uneven proppant distribution between different clusters within a
single stage on ultimate gas recovery has not been evaluated quantitatively. Additionally, most reservoir modeling works in the literature assume uniform proppant distribution among perforation
clusters within a single stage. Hence, a detailed study for investigation of the impact of uneven proppant distribution between different clusters on well performance in shale gas production is still
signicantly necessary.
In this paper, numerical reservoir simulation approach was validated by eld production data from Marcellus shale using a
numerical reservoir simulator considering gas desorption and
stress-dependent fracture conductivity effects. The simulator was

SI metric conversion factors

ft  0.3048 M
ft3  0.02832 m3
(F 32)/1.8 C
cp  0.001 Pa s
psi  6.895 kPa
md  1e 15 m2

used to model gas production with various proppant distribution

proles. In our previous work [21], we had set up a reservoir model
assuming four clusters per stage with fracture conductivity ratio of
1:1.5:2.5:4 and found that the impact of uneven proppant distribution on well performance is signicant. In this paper, we extended
the original work to perform a thorough sensitivity study to investigate the effect of uneven proppant distribution between different
clusters within one stage on well performance based on typical eld
data from Marcellus shale. Six uncertainty parameters such as
matrix permeability, fracture height, fracture half-length, fracture
conductivity, cluster spacing and initial reservoir pressure, are considered. The results of this work can provide a better understanding
of the impact of proppant distribution on shale gas production.
2. Methodology
2.1. Numerical modeling of shale gas reservoir
In this work, a numerical reservoir simulator of CMG-IMEX [22]
is used to model multiple hydraulic fractures and gas production in
Marcellus shale reservoirs. In our simulation model, bi-wing
hydraulic fractures are explicitly modeled using local grid renement (LGR) with logarithmic cell spacing, which can accurately
model gas ow from shale matrix to hydraulic fractures. A no-ow
boundary condition is used for the reservoir model. Non-Darcy
ow is considered for which the non-Darcy Beta factor, used in
the Forchheimer number, is determined using a correlation proposed by Evans and Civan [23]. This approach has been extensively
used to model transient gas ow in hydraulically fractured shale
gas reservoirs [2428]. In the simulation model, the gas desorption
effect is modeled using the classic Langmuir isotherm with two tting parameters of Langmuir pressure and Langmuir volume [29],
which is based on the assumption that there is a dynamic equilibrium at constant temperature and pressure between the adsorbed
and free gas. When modeling geomechanics in hydraulic fractures,
i.e., stress-dependent fracture conductivity, a specic compaction
table is used to account for decreasing conductivities of propped
fractures with the increase in closure stress or decrease in pressure. The compaction table is assigned in the simulator to cells
describing propped hydraulic fractures. The stress-dependent fracture conductivity curves used in the following simulation studies
are generated based on experimental measurements for stiff shale
samples by Alramahi and Sundberg [30], which were discussed in
our previous work [31].
2.2. Validation of numerical model
Once a numerical reservoir model including multiple hydraulic
fractures is built, it requires validation with eld production data
to ensure the reliability of simulation results. After validation, it

W. Yu et al. / Fuel 142 (2015) 189198

can be used to perform sensitivity studies and production forecasting for a long-time period. In general, owing bottom hole pressure
and gas production data of shale wells are used to perform history
matching to validate the reservoir model. The eld production data
of a Marcellus shale well is chosen to perform history matching,
and the well was located between the lower section and upper section of the Marcellus formation with a true vertical depth ranging
from 8054 ft to 8066 ft [32]. In the eld case study, the horizontal
well was stimulated with seven hydraulic fracturing stages, and
each stage has four active clusters with cluster spacing of 50 ft.
The total number of hydraulic fractures is 28. The numerical model
has a dimension of 5000 ft  1600 ft  300 ft, with a bi-wing fracture system, as shown in Fig. 1. The horizontal well is located along
the middle of fracture planes. The detailed shale and fracture properties of the Marcellus shale well are listed in Table 1. The owing
bottom hole pressure is calculated based on the measured surface
tubing pressure by considering the gas and water ow rates, as
shown in Fig. 2 [32]. It decreases from 1600 psi initially to about
400 psi in a nine-month production period. The inner diameter of
wellbore is 0.7292 ft, and more details about the well can be found
in the work by Yeager and Meyer [32]. The bottom hole pressure is
an input in the reservoir simulator as a history-matching constraint, and cumulative gas production is a history-matching variable. In addition, matrix permeability, fracture half-length and
fracture conductivity are tuned manually to perform history
matching and the other parameters are kept constant, which are
based on the work by Yeager and Meyer [32]. Langmuir pressure
of 500 psi, Langmuir volume of 200 scf/ton, and shale density of
2.46 g/cm3 [26] are used in the gas desorption model. The stressdependent fracture conductivity used in the history matching is
shown in Fig. 3.
Fig. 4 presents the history matching results with and without considering gas desorption and geomechanics effects. As shown, a reasonable match between the numerical simulation results and the
actual eld data is obtained by considering both gas desorption
and geomechanics effects. The gas desorption plays a positive role
on the gas production rate, but has a smaller effect during early period of production. Meanwhile, the geomechanics plays a negative
impact on the gas ow rate with a larger effect at early time of production. The reason is that although a smaller reduction in fracture
conductivity occurs in the early time, it results in a larger decrease
in the gas ow rate due to higher gas ow rate at the early time of
production [26]. Fig. 5 shows the pressure distribution after

191

9 months of eld production, illustrating clearly the stimulated reservoir volume (SRV) in this horizontal fractured Marcellus shale well.
After history matching, we performed production forecasting
for a 30-year period and compared the impacts of gas desorption
and geomechanics on the estimated ultimate recovery (EUR).
Fig. 6 shows the owing bottom hole pressure used for the period
of production forecasting. Comparisons of EUR for a 30-year period
by considering only the gas desorption effect or only the geomechanics effect or both effects together or neither of them are shown
in Fig. 7. As shown, the gas desorption contributes to 21% of EUR at
30 years of gas production when neither the gas desorption nor the
geomechanics effect was considered. The geomechanics model
decreases the EUR by 32% in 30 years of gas production. The EUR
drops by 19% in 30 years when combining gas desorption and geomechanics effects.

3. Uneven proppant distribution between multiple fractures

Uniform proppant distribution within multiple perforation
clusters plays an important role in achieving high fracturing-treatment effectiveness in shale gas reservoirs. Although the technology
of hydraulic fracturing has made signicant progress over the last
several years, proppant transport and distribution in multiple
hydraulic fractures is still unclear. A numerical simulation study
by Daneshy [19] indicated that an uneven proppant distribution
exists between four perforation clusters in one stage using plugand-perf fracturing technique, as shown in Fig. 8. During his simulation study, many factors, such as uid viscosity, injection rate,
and proppant size and density were considered. Among all scenarios, a highly uneven proppant distribution within different clusters
was observed and most proppant entered the last perforation cluster; the proppant concentration of the last cluster was almost four
times the proppant amount in the rst cluster [19]. The average
ratio of proppant concentration within four clusters was approximately 1:1.5:2.5:4.
In this paper, we evaluated three hypothetical scenarios of
proppant distribution in four clusters within one fracturing stage,
as shown in Fig. 9 [21]. Scenario 1 represents uniform distribution
with high proppant concentration, which is the optimal design in
the eld operation; Scenario 2 represents uneven distribution with
proppant concentration ratio of 1:1.5:2.5:4, which might be the
real case in the eld; Scenario 3 represents uniform distribution

Fig. 1. A numerical reservoir model including 28 hydraulic fractures for the Marcellus shale well.

192

W. Yu et al. / Fuel 142 (2015) 189198

Table 1
Parameters used in history matching for the Marcellus shale well.
Parameter

Value(s)

Unit

The model dimension

(length  width  height)
True vertical depth
Pore pressure gradient
Initial reservoir pressure
Closure pressure
Closure pressure gradient
Production time
Reservoir temperature
Initial gas saturation
Gas gravity
Total compressibility
Matrix permeability
Matrix porosity
Fracture conductivity
Fracture half-length
Cluster spacing
Fracture height
Horizontal well length
Number of fractures

5000  1600  300

(1524  487.7  91.4)
8060 (2456.7)
0.62 (1.40  104)
5024 (3.46  107)
5642 (3.89  107)
0.70 (1.58  104)
9 (2.37  107)
169 (76.1)
0.70
0.5726
3  10 6 (4.5  10 10)
0.0001 (1.0  10 19)
0.046
1 (3.0  10 16)
400 (121.9)
50 (15.2)
173 (52.7)
1400 (426.7)
28

ft (m)
ft (m)
psi/ft (Pa/m)
psi (Pa)
psi (Pa)
psi/ft (Pa/m)
month (s)
F (C)
fraction
Dimensionless
psi 1 (Pa 1)
md (m2)
fraction
md-ft (m2 m)
ft (m)
ft (m)
ft (m)
ft (m)
Dimensionless

Fig. 4. History matching results of a Marcellus shale well.

4. Results and discussion

4.1. Base case

Fig. 2. Flowing bottom hole pressure used for history matching (modied from
Yeager and Meyer, [23]).

with low proppant concentration, which is poor design in eld

operations. In addition, proppant distribution in each cluster is
assumed to be uniform laterally and vertically, which may not be
close to the reality but it will simplify the problem in this study
where the uneven proppant distribution between different perforation clusters is our focus.

Fig. 3. Stress-dependent fracture conductivity curve used for history matching.

Based on the reservoir model built for the history matching

study of the Marcellus shale reservoir, we set up a base case model
with well-length of 3500 ft, as shown in Fig. 10. For the base case,
there are 11 fracturing stages and 4 clusters per stage. Therefore,
the total number of hydraulic fractures is 44. The detailed reservoir
properties for this model are the average values based on available
data from published Marcellus shale work [3337] and are listed in
Table 2. In general, there are high uncertainties in reservoir and
fracture properties in shale gas reservoirs. Therefore, we conducted
the sensitivity study of six uncertain parameters on gas production,
such as cluster spacing, initial reservoir pressure, fracture conductivity (Fcd), fracture half-length, fracture height, and matrix
permeability. The range investigated for each uncertain parameter
is listed in Table 3, which is mainly based on the work by Yu and
Sepehrnoori [26,27,33]. Based on history matching results of the
Marcellus shale well, the fracture conductivity is obtained as 1
md-ft. Hence, the range of fracture conductivity is chosen from
0.1 md-ft to 10 md-ft, representing low and high proppant concentration, respectively. The effects of gas desorption and geomechanics are considered for all simulation cases. Langmuir pressure of
500 psi and Langmuir volume of 200 scf/ton are used in the gas
desorption model. For the stress-dependent fracture conductivity
effect, three curves of fracture conductivity changing with pressure
are used for the Marcellus shale corresponding to three differential
initial reservoir pressures (Fig. 11). Each curve starts with the
bottom hole pressure constraint of 500 psi and ends with various
initial reservoir pressures.
In the base case, the cluster spacing is 80 ft, the initial reservoir
pressure is 4000 psi, the minimum fracture conductivity is 1 md-ft,
the fracture half-length is 300 ft, the fracture height is 200 ft, and
the matrix permeability is 0.0001 md. The fracture conductivity
in Scenario 1 is changed to 4 md-ft for all clusters. The fracture conductivity in Scenario 2 has a ratio of 1:1.5:2.5:4 md-ft for cluster
one to four, respectively. In the base case, the fracture conductivity
for Scenario 2 is 1, 1.5, 2.5 and 4 md-ft for cluster one to four within
the single stage. The fracture conductivity of Scenario 3 is constant
at 1 md-ft for all clusters. The comparison of cumulative gas production in a 30-year period between these three scenarios is shown
in Fig. 12. Scenario 1 has the largest cumulative gas production and
Scenario 2 has the smallest. The difference of cumulative gas production between Scenarios 1 and 2 is about 11%, between Scenarios
2 and 3 is about 13%, and between Scenarios 1 and 3 is about 24%. It
suggests that the impact of proppant distribution on gas recovery is
important and should be considered in the eld development in

W. Yu et al. / Fuel 142 (2015) 189198

193

Fig. 5. Pressure distribution after the history matching period.

Fig. 6. Flowing bottom hole pressure used for production forecasting in 30 years.

Fig. 8. Proppant distribution within four perforation clusters in one fracturing stage
(modied from Daneshy, [11]).

smaller than Scenario 1 with uniform proppant distribution and larger fracture conductivity.
In the subsequent simulation studies, we only chose Scenarios 1
and 2 to perform the sensitivity study for the purpose of determining the critical reservoir and fracture parameters controlling the
difference in gas recovery between uneven proppant distribution
and uniform proppant distribution. The difference in gas recovery
is calculated by the absolute difference of cumulative gas production at 30 years between Scenarios 1 and 2 divided by cumulative
gas production of the base case at 30 years with uneven proppant
distribution. Six uncertain parameters are investigated: cluster
spacing, initial reservoir pressure, fracture conductivity, fracture
half-length, fracture height and matrix permeability.

4.2. Effect of cluster spacing

Fig. 7. Production forecasting of gas recovery in a 30-year period.

order to optimize the shale gas production. Fig. 13 shows the pressure distribution after 1 year of production for Scenarios 13. As
shown, the drainage efciency of Scenario 3 with low proppant concentration is the lowest, and the drainage efciency of Scenario 2 is

The effect of cluster spacing on the comparison of well performance between Scenarios 1 and 2 is shown in Fig. 14, while keeping the other reservoir and fracture properties same as those in the
base case. Cluster spacing of 50 ft, 80 ft and 100 ft within each single fracturing stage is investigated under the condition of the same
well length, representing fracturing stage number (4 clusters per
stage) of 18, 11, and 9, respectively. As shown, the difference of

194

W. Yu et al. / Fuel 142 (2015) 189198

Scenario 1: uniform
distribution with high proppant
concentration

Scenario 2: uneven distribution

(proppant concentration ratio
of 1:1.5:2.5:4)

Scenarios 3: uniform
distribution with low proppant
concentration

Fig. 9. Proppant distribution in four perforation clusters within one fracturing stage [21].

Fig. 10. A reservoir simulation model for the Marcellus shale well.

Table 2
Parameters used in simulation study of proppant distribution effect.
Parameter

Value(s)

Unit

The model dimension

(length  width  height)
Production time
Reservoir temperature
Initial gas saturation
Gas gravity
Total compressibility
Matrix porosity
Horizontal well length
Bottom hole pressure

5000  1600  300

(1524  487.7  91.4)
30 (9.47  108)
220 (104.4)
0.70
0.58
3  10 6 (4.5  10 10)
0.06
3500 (1066.8)
500 (3.45  106)

ft (m)

4.3. Effect of initial reservoir pressure

year (s)
F (C)
fraction
Dimensionless
psi 1 (Pa 1)
fraction
ft (m)
psi (Pa)

gas recovery is almost the same for different cluster spacing, which
is approximately 11%. Hence, the effect of cluster spacing based on
the range investigated in this study is negligible.

Fig. 15 shows the impact of initial reservoir pressure on the

comparison of well performance between Scenarios 1 and 2, while
keeping the other reservoir and fracture properties same as those
in the base case. It can be seen that the difference of gas recovery
between Scenarios 1 and 2 is 9%, 11% and 12%, corresponding to
the initial reservoir pressure of 3000 psi, 4000 psi and 5000 psi,
respectively. Since cumulative gas production increases more signicantly for the case with uniform proppant distribution with
the increasing initial reservoir pressure under the same bottom
hole pressure, the difference of gas recovery between two scenarios increases with the increasing initial reservoir pressure. It suggests that in shale reservoirs with high initial reservoir pressure,
it is crucial to take into account the impact of uneven proppant
distribution.

Table 3
Six uncertainty parameters used in simulation study of proppant distribution effect.
Parameter

Base case

Minimum

Maximum

Unit

Cluster spacing
Initial reservoir pressure
Fracture conductivity
Fracture half-length
Fracture height
Matrix permeability

80 (24.4)
4000 (2.76  107)
1 (3.0  10 16)
300 (91.4)
200 (61)
0.0001 (1.0  10 19)

50 (15.2)
3000 (2.07  107)
0.1 (3.0  10 17)
200 (61)
100 (30.5)
0.00001 (1.0  10

100 (30.5)
5000 (3.45  107)
10 (3.0  10 15)
400 (121.9)
300 (91.4)
0.001 (1.0  10 18)

ft (m)
psi (Pa)
md-ft (m2 m)
ft (m)
ft (m)
md (m2)

20

W. Yu et al. / Fuel 142 (2015) 189198

195

Fig. 11. Fracture conductivity changes with pressure corresponding to differential

initial reservoir pressures.

Fig. 12. Comparison of cumulative gas production in a 30-year period between

three scenarios.

4.4. Effect of fracture conductivity

Fig. 16 presents the effect of fracture conductivity on the comparison of well performance between Scenarios 1 and 2, while
keeping the other reservoir and fracture properties same as those
in the base case. It can be observed that the difference of gas recovery is 12%, 11% and 6%, corresponding to the minimum fracture
conductivity of 0.1 md-ft, 1 md-ft and 10 md-ft, respectively. The
increase in cumulative gas production for both scenarios is significant with the increasing fracture conductivity, resulting in the
decrease in difference of gas recovery between two scenarios. Furthermore, the difference between them decreases signicantly
with the minimum fracture conductivity of 10 md-ft, when compared with the base case. It suggests that the negative effect of
uneven proppant distribution on gas recovery could be ignored
when developing shale reservoirs with high fracture conductivity
(over 10 md-ft in this study) for multiple hydraulic fractures.

4.5. Effect of fracture half-length

The effect of fracture half-length on the comparison of well performance between Scenarios 1 and 2 is illustrated in Fig. 17, while
keeping the other reservoir and fracture properties same as those
in the base case. As shown, the difference of gas recovery is 7%,
11% and 15%, corresponding to a fracture half-length of 200 ft,

Fig. 13. Pressure distribution after one year of gas production.

300 ft and 400 ft, respectively. Cumulative gas production increase

more signicantly with an increase in fracture half-length for the
case with uniform proppant distribution, resulting in the increase

196

W. Yu et al. / Fuel 142 (2015) 189198

Fig. 14. Effect of cluster spacing on the comparison of well performance between
Scenarios 1 and 2.

Fig. 16. Effect of fracture conductivity on the comparison of well performance

between Scenarios 1 and 2. (Fracture conductivity of Scenario 2 are 0.1, 0.15, 0.25,
and 0.4 md-ft for cluster one to four within the single stage if the minimum Fcd is
0.1 md-ft. Similar ratio is assumed for the other two cases with the minimum Fcd of
1 md-ft and 10 md-ft.)

Fig. 15. Effect of initial reservoir pressure on the comparison of well performance
between Scenarios 1 and 2.

in difference of gas recovery between two scenarios. Hence, in the

fracturing treatment design with long fracture half-length, the negative effect of uneven proppant distribution on gas recovery should
be considered.
4.6. Effect of fracture height
Fig. 18 shows the impact of fracture height on the comparison
of well performance between Scenarios 1 and 2, while keeping
the other reservoir and fracture properties same as those in the
base case. As shown, the difference of gas recovery is 6%, 11%

Fig. 17. Effect of fracture half-length on the comparison of well performance

between Scenarios 1 and 2.

and 17%, corresponding to a fracture height of 100 ft, 200 ft and

300 ft, respectively. Cumulative gas production increase more
signicantly with the increasing fracture height for the case with
uniform proppant distribution, resulting in the increase in difference of gas recovery between two scenarios. Therefore, for fracturing treatment design with large fracture height, the negative effect
of uneven proppant distribution on gas recovery should be
considered.

W. Yu et al. / Fuel 142 (2015) 189198

197

Fig. 20. Tornado chart of sensitivity study (red color represents high value of
parameter and blue color represents low value of parameter). (For interpretation of
the references to color in this gure legend, the reader is referred to the web version
of this article.)

Fig. 18. Effect of fracture height on the comparison of well performance between
Scenarios 1 and 2.

Finally, the rank of impacts of the above six uncertainty parameters on the difference in gas production for 30 years between Scenarios 1 and 2 are summarized in a Tornado plot, as shown in
Fig. 20. It illustrates that the most sensitive parameter is matrix
permeability, followed by fracture height, fracture half-length,
fracture conductivity, and initial reservoir pressure. The effect of
cluster spacing is less sensitive based on the range investigated
in this study. Also, the overall range of gas recovery difference
between uneven proppant distribution and uniform proppant distribution is between 5% and 17% for the Marcellus shale.
5. Conclusions
We performed the history matching of a gas production well
from Marcellus shale, considering the gas desorption and the geomechanics effect. We compared the difference of gas recovery
between uneven proppant distribution in different clusters within
one fracturing stage and uniform proppant distribution, and performed sensitivity studies over six uncertain parameters. The following conclusions can be drawn from this study:

Fig. 19. Effect of matrix permeability on the comparison of well performance

between Scenarios 1 and 2.

4.7. Effect of matrix permeability

The impact of matrix permeability on the comparison of well
performance between Scenarios 1 and 2 is illustrated in Fig. 19,
while keeping the other reservoir and fracture properties same as
those in the base case. The range of matrix permeability from
0.00001 md to 0.0001 md is considered. It can be seen that the difference of gas recovery is 5%, 11% and 17%, corresponding to matrix
permeability of 0.00001 md, 0.0001 md and 0.001 md, respectively. Therefore, in shale reservoirs with some natural fractures,
resulting in large effective matrix permeability, the negative effect
of uneven proppant distribution on gas recovery should be taken
into account.

(1) Gas desorption and geomechanics effects should be considered when performing history matching and production
forecasting in the Marcellus shale.
(2) The effect of uneven proppant distribution between different clusters can signicantly reduce the well productivity
when the proppant concentration is low and fracture conductivity is small; while it might be less signicant when
the proppant concentration is high and fracture conductivity
is large, i.e. 10 md-ft in this study.
(3) The most sensitive parameter in the comparison of gas production with uneven and uniform proppant distribution is
matrix permeability, followed by fracture height, fracture
half-length, fracture conductivity, and initial reservoir pressure. The effect of cluster spacing is minimum based on the
range investigated in this study.
(4) The range for the difference of gas recovery between uneven
and uniform proppant distribution in a 30-year period is
obtained as 517% for the Marcellus shale, based on the
range of each uncertain parameter investigated in this study.

Acknowledgements
The authors would like to acknowledge Computer Modeling
Group Ltd. for providing us with usage of CMG software. We would
also like to express our gratitude for reviewers for their careful

198

W. Yu et al. / Fuel 142 (2015) 189198

review of this manuscript. Comments made by Dr. Chowdhury K.

Mamun on this manuscript are greatly appreciated.

[18]

References
[19]
[1] Woodworth TR, Miskimins JL. Extrapolation of laboratory proppant placement
behavior to the eld in slickwater fracturing applications. In: SPE 106089,
presented at SPE hydraulic fracturing technology conference, January 2931,
College Station, TX; 2007.
[2] Warpinski NR, Mayerhofer MJ, Vincent MC, Cipolla CL, Lolon EP. Stimulating
unconventional reservoirs: maximizing network growth while optimizing
fracture conductivity. In: SPE 114173, presented at SPE unconventional
reservoirs conference, February 1012, Keystone, CO; 2008.
[3] Warpinski NR. Stress amplication and arch dimensions in proppant beds
deposited by waterfracs. SPE Prod Oper 2010;25(4):46171.
[4] Cipolla CL, Lolon EP, Mayerhofer MJ, Warpinski NR. The effect of proppant
distribution and un-propped fracture conductivity on well performance in
unconventional gas reservoirs. In: SPE 119368, presented at SPE hydraulic
fracturing technology conference, January 1921, The Woodlands, TX; 2010.
[5] Gaurav A, Dao EK, Mohanty KK. Ultra-lightweight proppants in shale gas
fracturing. In: SPE 138319, presented at SPE tight gas completions conference,
November 23, San Antonio, TX; 2010.
[6] Mahoney RP, Soane D, Kincaid KP, Herring M, Snider PM. Self-suspending
proppant. In: SPE 163818, presented at SPE hydraulic fracturing conference,
February 46, The Woodlands, TX; 2013.
[7] Gu M, Mohanty KK. Effect of foam quality on effectiveness of hydraulic
fracturing in shales. Int J Rock Mech Min Sci 2014;70:27385.
[8] Cipolla CL, Warpinske NR, Mayerhfer MJ, Lolon EP, Vincent MC. The
relationship between fracture complexity, reservoir properties, and fracture
treatment design. SPE Prod Oper 2010;25(4):43852.
[9] Gu M, Kulkarni P, Raee M, Ivarrud E, Mohanty KK. Understanding the
optimum fracture conductivity for naturally fractured shale and tight
reservoirs. In: SPE 171648, presented at SPE/CSUR unconventional resources
conference, September 30October 2, Calgary, Canada; 2014.
[10] Pope C, Benton T, Palisch T. Haynesville shale-one operators approach to well
completions in this evolving play. In: SPE 125079, presented at the annual
technical conference and exhibition, October 47, New Orleans, LA; 2009.
[11] LaFollette RF, Carman PS. Proppant diagenesis: results so far. In: SPE 131782,
presented at the SPE unconventional gas conference, February 2325,
Pittsburgh, PA; 2010.
[12] Darin SR, Huitt JL. Effect of a partial monolayer of propping agent on fracture
ow capacity. AIME-Petrol Trans 1960;219:317.
[13] Fan L, Thompson JW, Robinson JR. Understanding gas production mechanism
and effectiveness of well stimulation in the Haynesville shale through
reservoir simulation. In: SPE 136696, presented at Canadian unconventional
resources and international petroleum conference, October 1921, Calgary,
Canada; 2010.
[14] Fredd CN, MaConnell SB, Boney CL, England KW. Experimental study of
hydraulic fracture conductivity demonstrates the benets of using proppants.
In: SPE 60326, presented at SPE rocky mountain regional/low permeability
reservoirs symposium, March 1215, Denver, CO; 2000.
[15] Vincent MC. The next opportunity to improve hydraulic-fracture stimulation.
JPT 2012;64(3):11827.
[16] Crespo F, Aven NK, Cortez J, Soliman MY, Bokane A, Jain S, Deshpande Y.
Proppant distribution in multistage hydraulic fractured wells: a large-scale
inside-casing investigation. In: SPE 163856, presented at SPE hydraulic
fracturing technology conference, February 46, The Woodlands, TX; 2013.
[17] Cipolla C, Mack M, Maxwell S. Reducing exploration and appraisal risk in low
permeability reservoirs using microseismic fracturing mapping-Part 2. In: SPE

[20]

[21]

[22]
[23]

[24]

[25]
[26]
[27]

[28]

[29]
[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

138103, presented at SPE Latin American & Caribbean petroleum engineering

conference, December 13, Lima, Peru; 2010.
Miller C, Waters G, Rylander E. Evaluation of production log data from
horizontal wells drilled in organic shales. In: SPE 144326, presented at SPE
North American unconventional gas conference and exhibition, June 1416,
The Woodlands, TX; 2011.
Daneshy A. Uneven distribution of proppants in perf clusters. World Oil
2011;232(4).
Bokane A, Jain S, Deshpande Y, Crespo F. Transport and distribution of
proppant in multistage fractured horizontal wells: a CFD simulation approach.
In: SPE 166096, presented at the SPE annual technical conference and
exhibition, September 30October 2, New Orleans, LA; 2013.
Yu W, Sepehrnoori K. Simulation of proppant distribution effect on well
performance in shale gas reservoirs. In: SPE 167225, presented at SPE
unconventional resources conference, November 57, Calgary, Canada; 2013.
CMG. IMEX users guide. Computer Modeling Group Ltd.; 2012.
Evans RD, Civan F. Characterization of non-darcy multiphase ow in
petroleum bearing formations. Report, U.S. DOE Contract No. DE-AC2290BC14659, School of Petroleum and Geological Engineering, University of
Oklahoma; 1994.
Rubin B. Accurate simulation of non-darcy ow in stimulated fractured shale
reservoirs. In: SPE 132093, presented at the SPE western regional meeting,
May 2729, Anaheim, CA; 2010.
Cipolla CL, Lolon EP, Erdle JC, Rubin B. Reservoir modeling in shale-gas
reservoirs. SPE Res Eval Eng 2010;13(4):63853.
Yu W, Sepehrnoori K. Simulation of gas desorption and geomechanics effects
for unconventional gas reservoirs. Fuel 2014;116:45564.
Yu W, Sepehrnoori K. An efcient reservoir simulation approach to design and
optimize unconventional gas production. J Can Petrol Technol
2014;53(2):10921.
Yu W, Luo Z, Javadpour F, Varavei A, Sepehrnoori K. Sensitivity analysis of
hydraulic fracture geometry in shale gas reservoirs. J Petrol Sci Eng
2014;113:17.
Langmuir I. The adsorption of gases on plane surfaces of glass, mica and
platinum. J Am Chem Soc 1918;40:140361.
Alramahi B, Sundberg MI. Proppant embedment and conductivity of hydraulic
fractures in shales. In: ARMA 12-291, presented at the 46th US rock
mechanics/geomechanics symposium, June 2427, Chicago, IL; 2012.
Yu W, Sepehrnoori K. Optimization of well spacing for Bakken tight oil
reservoirs. In: URTeC 1922108, presented at unconventional resources
technology conference, August 2527, Denver, CO; 2014.
Yeager BB, Meyer BR. Injection/fall-off in the Marcellus Shale: using reservoir
knowledge to improve operational efciency. In: SPE 139067, presented at SPE
eastern regional meeting, October 1214, Morgantown, WV; 2010.
Yu W, Sepehrnoori K. Sensitivity study and history matching and economic
optimization for Marcellus shale. In: URTeC 1923491, presented at unconventional
resources technology conference, August 2527, Denver, CO; 2014.
Ejofodomi E, Baihly J, Malpani R, Altman R, Huchton T, Welch D, Zieche J.
Integrating all available data to improve production in the Marcellus shale. In:
SPE 144321, presented at SPE North American unconventional gas conference
and exhibition, June 1416, The Woodlands, TX; 2011.
Edwards KL, Weissert S., Jackson J., Marcotte D. Marcellus shale hydraulic
fracturing and optimal well spacing to maximize recovery and control costs.
In: SPE 140463, presented at SPE hydraulic fracturing technology conference
and exhibition, January 2426, The Woodlands, TX; 2011.
Jayakumar R, Rai R. Impact of uncertainty in estimation of shale-gas-reservoir
and completion properties on EUR forecast and optimal development
planning: a Marcellus case study. SPE Res Eval Eng 2014;17(1):6073.
Izadi G, Junca JP, Cade R, Rowan T. Multidisciplinary study of hydraulic
fracturing in the Marcellus shale. In: ARMA 146975, presented at the 48th US
rock mechanics/geomechanics symposium, June 14, Minneapolis, MN; 2014.