This paper (SPE 175892) was accepted for presentation at the SPE/CSUR Unconventional Methodology
Resources Conference, Calgary, 2022 October 2015, and revised for publication. Original
manuscript received for review 20 April 2016. Revised manuscript received for review 22 This study is conducted in the following six steps: (1) gathering
December 2016. Paper peer approved 4 January 2017. and preparing flowback data; (2) understanding the flow behavior
Horizontal well
Fig. 2Schematic of a fractured horizontal well. The effective fractures include the hydraulic and reactivated natural fractures
communicating with wellbore and contributing to the fluid recovery during flowback.
3. Pick unit-slope region from the log-log plot of RNP vs. tMB
4. Determine the slope m between RNP and tMB (described by Eqs. 1 and 2) in Cartesian plot
5. Estimate Cf-max and Cf-min with the procedure proposed in Estimating Fracture Compressibility by Use of DFIT Data
Fig. 3Flow chart for estimating maximum and minimum effective fracture pore volume.
Yes
Calculate Gclosure
Obtain Pwf-max and Pwf-min Obtain min Calculate Pclosure
using Eq. 7
Fig. 4Flow chart for estimating the maximum and minimum fracture compressibility with DFIT data.
1. Obtain Pclosure for wells with DFIT data. Flowback Rate and Pressure Behavior
2. Obtain the correlation between Gclosure and Gf, using Eq. 7, This section first provides the basic information about the reser-
from wells with DFIT data. voir, wells, and completion parameters. Then, we qualitatively
3. Calculate Pclosure for wells without DFIT data using the cor- interpret the flowback rate and pressure data from the seven multi-
relation in Step 2. fractured horizontal wells.
4. Obtain Pwf-min and Pwf-max during the flow regime of frac-
ture depletion. Well and Completion Information. We study seven multifrac-
5. Calculate Pn-min and Pn-max for each well using Eq. 6. tured horizontal wells completed in Woodford Formation in Ana-
6. Read Cf-min at the point of (Pn-max, 1 /f-max) and Cf-max at darko Basin (Fig. 5). Woodford is primarily a mudstone with a
the point of (Pn-min, 1/f-min) from the graph shown in porosity range of 310% and a permeability range of 10100 nd.
Appendix A. Before hydraulic-fracturing treatment, a small-size fracturing
is performed on the first stage at the toe to estimate the initial res-
Quantification of the Drive Mechanisms. According to Ezulike ervoir pressure (Pi). Fifty to 200 bbl of water is pumped at the
et al. (2016), the drive mechanisms of fracture closure and water rate of 4 to 10 bbl/min into three to four clusters in the first stage.
expansion could be quantified by compaction-drive index (CDI) The first stage in these wells shows breakdown during the pump-
and water-drive index (WDI) separately. Fracture closure repre- ing process. A gauge is then delivered to the first stage before the
sents the reduction of Vef caused by fluid withdrawal during flow- well is shut in. The gauge records the bottomhole pressure during
back, proppant crushing/embedment, and compressibility of rock shut-in for several days, until being pulled out for fracturing treat-
grains. CDI and WDI are defined as ment of the remaining stages. Pi is obtained from a pressure
buildup analysis conducted on the bottomhole pressure recorded
Cf by the gauge. After fracturing, the plugs used for isolating each
CDI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Ct stage are drilled out, and the wells are shut in for 12 hours to set
Cw up the flowback equipment. Flowback for the seven wells is per-
WDI : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 formed through casing. The water/gas/oil rate and casing pressure
Ct
are measured hourly during flowback.
Table 1 provides a summary of completion information and
Pearson Correlation Coefficient. The Pearson correlation coef-
reservoir properties including fluid type, TVD, the initial bottom-
ficient (r) (Pearson 1895) provides a measure of the linear correla-
hole pressure during flowback (Pwfi), initial reservoir pressure,
tion between two series of values, and ranges from 1 to 1. The
and the bubblepoint pressure [Pb, which is obtained from the pres-
positive and negative values of r mean positive and negative cor-
sure/volume/temperature (PVT) analysis performed on the oil
relations, respectively. The larger value of absolute r represents a
samples from these wells]. Wells A, B, D, and E are oil wells. Pi
better correlation between the two parameters. Eq. 11 shows how
for these four oil wells varies from 4,104 to 5,815 psi. Pb for these
to calculate r for two series of data (xi and yi):
four wells ranges from 3,016 to 4,133 psi. Wells C, F, and G are
X gas wells that have a larger TVD and Pi compared with the other
xi xyi y
i
four oil wells.
r r
X r
X : . . . . . . . . . . . . . 11 Table 1 summarizes the completion information including
xi x2 yi y2 number of fracture stages, number of clusters, gross perforated
i i interval (GPI), mass of proppant, and total injected-water volume
Woodford Plays
DEWEY
KINGFISHER
DLAINE LOGAN
CUSTER
Meramec Play
CANADIAN
North Area
WASHITYA
CLEVELAND
CADDO
GRADY
MCCLAIN
KIOWA
GARVIN
South Area
STEPHENS
MURR
(TIV) for each well. Fig. 6 schematically illustrates fracture 2. Region 2 happens at the end of Region 1, and shows single-
stages, perforation cluster, and GPI. Average stage spacing repre- phase water flowback with flattening pressure. Pcasing and
sents the average distance between stages, whereas the average calculated Pwf generally flatten out before oil or gas produc-
cluster spacing represents the average distance between perfora- tion at the surface. This region happens at the end of Region
tion clusters. GPI is defined as the distance between the first per- 1. The pressure data flatten approximately 3 to 7 days after
foration cluster in the first stage and the last perforation cluster in starting the flowback and lasts approximately 1 to 6 days.
the last stage. As listed in Table 1, these wells have 9 to 12 frac- This region disappears before the production of oil/gas at
ture stages and 45 to 59 clusters. GPI varies from 1284 to 1490 m. the surface.
TIV ranges from 13 740 m3 to 27 620 m3. Wells A and B show a delay between the end of pres-
sure-flattening and the appearance of oil/gas at the surface.
Field Observations. Fig. 7 shows diagnostic plots of hourly This delay for the two wells is caused by shutting down the
recorded flowback rate and pressure data from Wells A and G, wells, for running gas lift. As shown later in Table 4, the
which produce oil and gas, respectively. Appendix B shows simi- initial Pi for the two wells is relatively low. Gas lift is per-
lar plots for Wells B, C, D, E, and F. From these diagnostic plots, formed on Wells A and B shortly after this region.
we observe three distinct regions: 3. Region 3 shows multiphase flowback, occurring after the
1. Region 1 shows single-phase water production with declin- single-phase region. This region shows a short-term two-
ing pressure. It occurs at the early times and lasts approxi- phase water and gas production and a long-term three-phase
mately 3 to 7 days. In this region, both the casing pressure oil, gas, and water production. The two-phase flow lasts
(Pcasing) and the calculated bottomhole pressure (Pwf) are approximately 1 to 5 days. In this region, water is produced
initially high and quickly drop with time. The water-rate at a relatively low rate, and the casing pressure keeps
plots in this region also show a peak that is followed by a dropping.
decline behavior. Abbasi et al. (2012, 2014) also reported
this similar single-phase region in flowback data from sev- Comparative Analysis of Flowback Data. This section inter-
eral tight gas and oil wells. prets the flattening of bottomhole and casing pressures during
Mass of
Well Number of Number of Proppant TIV
Pwfi (psi) Pi (psi) Pb (psi) 6 3 3
Name Fluid Type TVD (m) Stages Clusters GPI (m) (10 kg) (10 m )
Surface
Wellbore
Perforation cluster
Cluster spacing
Stage spacing
Fig. 6Schematic of a multifractured horizontal well showing four stages with 12 perforation clusters. Gross perforation interval is
defined as the distance between the first perforation cluster in the first stage and the last perforation cluster in the last stage.
early flowback. Also, it investigates how initial reservoir pressure sure drop along the fracture is assumed to be negligible. This
and hydrocarbon type affect the duration of Regions 1 and 2. flattening happens when the average pressure in fractures
Flattening of Bottomhole and Casing Pressures. Fig. 7 drops below the reservoir pressure. Also, it means that this
shows a flattening in the recorded Pcasing and calculated Pwf in point is the onset of hydrocarbon breakthrough from matrix to
Region 2. Jones et al. (2014) explained this phenomenon by the fracture network.
traces of oil/gas entering the wellbore and lightening the fluid col- Pressure-flattening can be investigated further by analyzing
umn. However, this explanation appears inapplicable to the well early-flowback data. Fig. 8 shows the log-log plot of RNP vs. tMB
data in Fig. 7, because no production of oil or gas is observed in and semilog plot of the calculated Pwf vs. tMB for Wells A and G
Region 2. One possible reason for this is that the volume of pro- (see Appendix C for the corresponding plots of the remaining five
duced oil or gas is insignificant compared with the large water wells). Here, we compare the semilog plots of Pwf vs. tMB with the
volume during this period, which makes it challenging to record log-log plots of RNP vs. tMB.
hydrocarbon production. Field experience indicates that the meas- The unit-slope represents water depletion from a closed-tank
urable oil rate begins at 1 to 2 bbl/hr, and the measurable gas rate system, where the fracture network does not receive pressure sup-
begins at 200 to 2,000 Mcf/D. port from matrix. As shown in Fig. 8, the unit-slope (fracture
Pcasing is measured at the surface, and Pwf is calculated from depletion) happens during Region 1, and deviates as Pwf flattens
wellhead data by use of Fanning single-phase correlations. The out in Region 2. The deviation from unit-slope indicates the fluid
main uncertainty in calculating Pwf relates to water density and influx from matrix into fracture network. Therefore, the flattened
pipe friction. Flowback water salinity can be used to estimate pressure in Region 2 can be interpreted as fracture-matrix
water density. Pipe friction is calculated by use of Fanning fric- flow communication. In this study, we apply the tank model
tion for single-phase water flow (IHS 2014). Pwf may be overesti- described in the Model for Estimating Effetive Fracture Pore Vol-
mated if the oil-or-gas volume in the wellbore is neglected for ume subsection to estimate Vef by use of the flowback data during
calculating it with Pcasing. Region 1.
Jones et al. (2014) proposed that the flattening value of calcu- The Duration of Region 1 and Region 2. The duration of
lated Pwf could be approximated as reservoir pressure if the pres- Regions 1 and 2 is related to the water rate, which is influenced
8 10 8 10
Gas Rate (Mcf/D), Oil Rate (BOPD),
Fig. 7Hourly flowback data show three regions. Region 1 shows single-phase water production and declining pressure. Region
2 shows single-phase water production and flattening pressure. Region 3 shows multiphase production. (a) Well A is a tight oil
well; (b) Well G is a tight gas well.
Well A Well G
10 12 10 12
RNP
RNP
6 6
0.1 0.1
Pwf flattens as 3 Unit-slope Pwf flattens as 3
Unit-slope unit-slope deviates unit-slope deviates
0.01 0 0.01 0
0.01 0.1 1 10 100 0.01 0.1 1 10 100
tMB (days) tMB (days)
(a) (b)
Fig. 8Log-log plot of RNP vs. tMB shows a clear unit-slope before the calculated bottomhole pressure flattens out (RNP is in the
unit of psi/STB/D). (a) Well A is a tight oil well; (b) Well G is a tight gas well.
Normalized Time of Region 1
0 0
0 2 4 6 8 10 0 2 4 6 8 10
Initial Reservoir Pressure (1,000 psi) Initial Reservoir Pressure (1,000 psi)
(a) (b)
Fig. 9(a) Crossplot of normalized time of Region 1 vs. Pi; (b) crossplot of normalized time of Region 2 vs. Pi. The red and green
points represent gas and oil wells, respectively.
by choke size. A larger choke size leads to a higher flow rate, and This can be explained by Darcys law: The flow rate of oil
leads further to shorter Regions 1 and 2. Therefore, we normalize increases with the drawdown, which increases with Pi.
the duration of Regions 1 and 2 with choke size to eliminate the Also, from Fig. 9, gas wells generally show shorter Regions 1
effect of variable choke size during early flowback. The normal- and 2 compared with oil wells. Wells C and F are gas wells, and
ized time for Regions 1 and 2 is defined as the product of average they have 5 days of Regions 1 and 2 in total, whereas the duration
choke size and the flowback duration for Regions 1 and of the two regions for oil wells is more than 8 days. This can be
2, respectively. explained by higher mobility of gas compared with oil when flow-
Fig. 9 shows the effect of Pi on the normalized time of Regions ing through matrix pores, fractures, and wellbore.
1 and 2, respectively. Fig. 9a shows a negative correlation between
the normalized time for Region 1 and Pi. The negative correlation Application and Discussions
in Fig. 9a can be explained by the fact that the wells with higher Pi This section is organized into the following five parts: (1) Esti-
show an earlier breakthrough. Fig. 9b shows a negative correlation mating the fracture compressibility for each well; (2) comparing
between the normalized time of Region 2 vs. Pi. The normalized drive mechanisms before breakthrough of oil or gas; (3) estimat-
time of Region 2 indicates the time of hydrocarbon transporting ing Vef and comparing it with TIV and load-recovery volume
through fractures and wellbore to the surface. The negative corre- (LRV); (4) comparing the propped and unpropped-fracture vol-
lation in Fig. 9b suggests that hydrocarbon front moves faster ume; and (5) investigating the correlations between the estimated
through fracture networks and wellbore for wells with higher Pi. Vef and LRV, and completion-design parameters.
Table 2Computed fracture compressibility for each well with DFIT data. Cf-min and Cf-max correspond to (Pn-max, 12 /f-max) and (Pn-min,
12/f-min), respectively. /f-min and /f-max are 47.5 and 100% for all wells.
(1,000 psi)
natural fractures after shut-in. However, the seven wells studied
8
in this paper started their flowback up to 1 day after shut-in. Fur-
4 thermore, DFIT analysis from one offset well shows an FE of
Low water rate approximately 87%. The high value of FE from DFIT analysis
Stage 2 implies low leakoff (Wallace et al. 2014). Therefore, FE from
flowback analysis may approximate the value of FE from DFIT.
0 6
0 1 2 3 4 5 6 7 The values of FEmax in Table 5 for Wells F and G are higher
than 100%. This may be a result of overestimating their Vef-max.
Days From Start of Flowback The value of FE from DFIT analysis is close to the values of FEmax
calculated from flowback analysis for the remaining five wells,
Fig. 12The early-flowback pressure and rate data in Region 1
suggesting that the estimated Vef-max is more-representative than
for Well G generally show two stages: (1) Stage 1 shows high
pressure and relatively low water rate; (2) Stage 2 shows Vef-min for the fracture network. Therefore, in this paper, Vef-max is
quickly dropping pressure and relatively high water rate. The used as effective fracture pore volume for the sake of consistency.
Pwfi for Well G is 9,962 psi, which is higher than its Pclosure
(9,602 psi).
Propped- and Unpropped-Fracture Volume. Fig. 14 compares
the volume of proppant, propped-fracture volume, unpropped-frac-
TIV, and FEmin is defined as the ratio of Vef-min to TIV. Vef is gen- ture volume, and Vef-max for each well. The volume of proppant is
erally less than TIV, and Vef-max is larger than LRV. A more- estimated by dividing the proppant mass by its density. The
detailed analysis of Vef, TIV, and LRV is presented in the Correla- propped-fracture volume represents the effective fracture volume
tion Analysis subsection. with proppants, and the unpropped-fracture volume represents the
Vef-max and Vef-min correspond to the Cf-min and Cf-max, respec- effective fracture volume without proppants. We estimate the
tively. Fig. 13 shows that the estimated Vef is very sensitive to Cf. propped-fracture volume by dividing the volume of proppant by
Vef decreases by more than 50% as Cf changes from Cf-min to the mineralization ratio, which is defined in Eq. 8. The mineraliza-
Cf-max. As mentioned in the Estimating Fracture Compressibility tion ratio is approximately 52.5% when assuming cubic packings
by Use of DFIT Data subsection, estimating Cf by use of Appen- of proppant in fractures. The unpropped-fracture volume is esti-
dix A requires Pn and /f. Pn for each well is calculated by use of mated by deducting the propped fracture volume from Vef-max.
Eq. 5. However, /f is an uncertain value assumed to range from For example, the type of proppant used for Well C is Santrol
47.5 to 100% for each well. Therefore, the uncertainty of estimat- SDC, and the size of the proppant is 40/70 mesh. This type
ing Vef is mainly because of the uncertainty in fracture porosity. of proppant has a density of 2.57 g/cm3. Well C pumps
Table 3Comparing the drive mechanisms before hydrocarbon breakthrough (Stage 2).
Table 5Summary of estimated values for effective fracture pore volume and fracturing parameters.
Effective Fracture Pore Volume (103 m3)
40
20 Volume of proppant
E
Estimated maximum effective fracture pore volume
P
Propped-fracture volume
32
U
Unpropped-fracture volume
10
16
8
Vef-min
Cf-min Cf-max 0
0 Well A Well B Well C Well D Well E
1105 1104 1103 1102
Fig. 14Comparing the volume of proppant, Vef-max, propped-
Fracture Compressibility (psi1) fracture volume, unpropped-fracture volume, and TIV shows
that the unpropped-fracture volume accounts for a large per-
Fig. 13Log-log plot of Vef vs. Cf by use of Eq. 4 shows that Vef centage of Vef-max and TIV (Wells F and G are not used for this
is sensitive to Cf. comparison).
approximately 1360 t of proppant. The calculated volume of prop- of the total fracture volume. It is expected that large proppants
pant for Well C is approximately 529 m3, the propped-fracture cannot reach into the unpropped fractures.
volume is approximately 1008 m3, and the unpropped-fracture Fig. 14 also compares the unpropped-fracture volume with
volume is estimated at 18 482 m3. TIV. It shows that the unpropped-fracture volume accounts for
Fig. 14 shows that the propped-fracture volume is relatively more than 60% of the TIV. As mentioned in the Comparative
small compared with Vef-max. The percentage of propped-fracture Analysis of Flowback Data subsection, we estimate Vef by use of
volume in Vef-max may be even smaller, because the proppant may the flowback data in Region 1 when the fractures are filled with
shrink, crush, and embed into the fracture walls under high-stress water. The unpropped-fracture volume can be approximated as
conditions (Warpinski 2010; Raysoni and Weaver 2013). The the volume of fracturing water in unpropped fractures. The results
result suggests that most of the fracture volume is unpropped. indicate that the unpropped fractures host most of the fracturing
McKenna (2014) modeled the proppant distribution by use of the water. Sharma and Manchanda (2015) also suggested that 90% of
discrete-fracture-network (DFN) approach. Their results also indi- fracturing water is in unpropped fractures, by a simple calculation
cate that the propped-fracture volume accounts for a small portion on the basis of the volumetric balance of the fracturing water.
30
Well A Well F As shown in Fig. 15 and Table 5, the estimated Vef-max for most
Well C
wells is larger than LRV measured after three years. The load-re-
20 covery ratio for Well F is more than unity, and this suggests that
Well B some water is produced from the formation. In general, LRV after
10 Well D Well E 3 years can be considered as final LRV, because the water-pro-
y = 0.9006x + 7.6238 duction rate is relatively low after 3 years.
R = 0.7461 Vef decreases as fracture depletion progresses (see Flow Re-
0 gime 2 in Fig. 1). Vef-max is assumed to be the upper limit of Vef at
0 10 20 30 the start of fracture depletion. As mentioned in the Results of
LRV (103 m3) Effective Fracture Pore Volume subsection, Vef is taken as the
water volume in fractures when they are filled with water during
Fig. 15Crossplot of Vef-max vs. LRV. The red and green points fracture depletion. Therefore, Vef-max can be approximated as the
represent gas and oil wells, respectively. initial water volume in fractures.
Completion- Gross
Design Number of Number of Perforated Total Injected- Proppant Average
Parameters Stages Clusters Stage Length Interval Water Volume Mass Treatment Rate
Pearson
Correlation
0.395 0.557 0.007 0.857 0.828 0.405 0.204
Coefficient (r)
with Vef-max
Lower values of final LRV compared with Vef-max indicate that clusters may be combined into one term: average cluster spac-
final LRV is still less than the initial water volume in fractures. ing, which is defined as GPI divided by the number of clusters.
This indicates further that there is still some nonrecovered fractur- Detailed discussion on TIV and average cluster spacing are pre-
ing water left in fractures, even a long time after starting produc- sented next.
tion. Experimental studies of Parmar et al. (2014) suggest that a Total Injected-Water Volume. As shown in Fig. 16 and Table
large volume of water can be trapped at the bottom of vertical 5, Vef-max for most wells is approximately 60 to 89% of TIV.
fractures because of gravity effect. This phenomenon is also con- Because Vef-max is the upper limit of Vef, one may conclude that
firmed by the simulation studies of Sharma and Agrawal (2013). Vef for most wells should be less than 60 to 89% of TIV. This
Furthermore, Sharma and Manchanda (2015) showed that fractur- means that more than 11 to 40% of the pumped water may be lost
ing water can also be trapped in the unpropped fractures. As dis- in matrix or ineffective fractures. However, Vef-max for the two
cussed in the Propped- and Unpropped-Fracture Volume gas wells (Wells F and G) is even larger than TIV. This suggests
subsection, most of the fracture volume is unpropped, which may the possibility of Vef-max overestimation as a result of underesti-
host the nonrecovered fracturing water. mating the total compressibility. There might be some initial gas
Effective Fracture Pore Volume and Completion-Design in fractures before flowback. The initial gas in fractures may
Parameters. We use the Pearson correlation coefficient (r) to increase the total compressibility (Xu et al. 2015), which is not
evaluate the possible correlations between Vef-max and comple- considered in our calculations.
tion-design parameters and to explore the key factors controlling Fig. 16 shows a positive correlation between Vef-max and TIV.
the fracture pore volume. The Pearson correlation coefficients are This observation indicates that, in general, Vef-max increases with
obtained by use of Eq. 11 and the values of Vef-max and comple- increasing TIV. For wells with larger TIV, fracturing water with
tion-design parameters. more energy is pumped to fracture the formation. Therefore, a
Correlating Vef-max with completion-design parameters may be larger TIV generally contributes to a larger effective fracture pore
challenging because of the limited number of wells we have. volume. However, there are some wells that have a relatively
Also, there might be interdependency between the design parame- modest Vef-max even with a large TIV pumped. For example, Well
ters. Besides, formation parameters such as permeability may B has a larger TIV than Well C, but Vef-max for Well B is less than
influence the process of leakoff, which may further affect the that for Well C, which is a gas well. One may expect that Vef-max
effective fracture pore volume. Therefore, the correlations shown for Well C is possibly overestimated. As discussed previously,
in this section may need further verifications by use of data from Vef-max for a gas well is possibly overestimated, caused by under-
more wells completed in this play. estimating the total compressibility.
Table 6 shows positive values of r between Vef-max and the Average Cluster Spacing. Vef-max is normalized by TIV
number of clusters, stage length, TIV, and proppant mass, suggest- (Vef-max/TIV) to eliminate the effect of TIV on Vef-max. As dis-
ing that that these fracturing parameters have a positive effect on cussed previously, Vef-max is possibly overestimated for the gas
Vef-max. It also shows negative values of r between Vef-max and num- wells. In Fig. 17, we plot normalized Vef-max vs. average cluster
ber of stages, GPI, and average treatment rate, suggesting that these spacing only for the oil wells. The limited data in Fig. 17 show that,
fracturing parameters have a negative effect on Vef-max. Zhou et al. in general, the normalized Vef-max decreases by increasing average
(2016) showed the similar effect of number of stages, proppant cluster spacing. In other words, closer cluster spacing generally
mass, and average treatment rate on the load recovery measured leads to a larger effective fracture pore volume. Ingram et al.
within the first 3 weeks, which may be related to Vef-max. (2014) stated that the cluster spacing has a significant effect on
From Table 6, we find that GPI, TIV, and number of clusters fracture initiation and propagation. A close cluster spacing may
have the highest values of coefficient compared with the other cause stress interference between clusters and influence the orien-
parameters. This may suggest that TIV, GPI, and the number of tation of the fractures. However, a close cluster spacing also creates
clusters are among the key design parameters for creating a more interconnected fractures, which enhances the connectivity
larger effective fracture pore volume. GPI and the number of and effectiveness of the fracture network.
40
1
Well D
Vef-max (103 m3)
y = 0.6503x + 2.6583
30 R = 0.7035 0.8
Normalized Vef-max
Well G Well E
Well C Well A
Well F 0.6
20
Well A Well B
Well E 0.4 y = 0.0228x + 1.3866
10 Well B R = 0.6135
Well D
0.2
0
0 10 20 30 0
TIV (103 m3) 10 20 30 40
Average Cluster Spacing (m)
Fig. 16Crossplot of Vef-max vs. TIV. The red and green points
represent gas and oil wells, respectively (Wells F and G are not Fig. 17Normalized Vef-max vs. average cluster spacing for four
used for calculating correlation). oil wells shows a negative correlation.
Technol 27 (1): 2127. SPE-4569-PA. https://doi.org/10.2118/4569- Warpinski, N. R. 2010. Stress Amplification and Arch Dimensions in
PA. Proppant Beds Deposited by Waterfracs. SPE Prod & Oper 25 (4):
Jones, R., Steven Jr., R., Pownall, B. et al. 2014. Estimating Reservoir 461471. SPE-119350-PA. https://doi.org/10.2118/119350-PA.
Pressure From Early Flowback Data. Presented at the Unconventional Williams-Kovacs, J. D. and Clarkson, C. R. 2013. Stochastic Modeling of
Resources Technology Conference, Denver, 2527 August. SPE-2014- Multi-Phase Flowback From Multi-Fractured Horizontal Tight Oil
1934785-MS. https://doi.org/10.15530/urtec-2014-1934785-MS. Wells. Presented at the SPE Unconventional Resources Conference
McCain, W. D. 1991. Reservoir-Fluid Property Correlations-State of the Canada, Calgary, 57 November. SPE-167232-MS. https://doi.org/
Art (includes associated papers 23583 and 23594). SPE J. 6 (2): 10.2118/167232-MS.
266272. SPE-18571-PA. https://doi.org/10.2118/18571-PA. Xu, Y., Adefidipe, O. A. and Dehghanpour, H. 2015. Estimating Fracture
Volume Using Flowback Data From the Horn River Basin: A Material
McKenna, J. P. 2014. Where Did the Proppant Go? Presented at the SPE/
AAPG/SEG Unconventional Resources Technology Conference, Den- Balance Approach. Journal of Natural Gas Science and Engineering
25: 253270. https://doi.org/10.1016/j.jngse.2015.04.036.
ver, 2527 August. SPE-2014-1922843-MS. https://doi.org/10.15530/
Xu, Y., Adefidipe, O. A., and Dehghanpour, H. 2016. A Flowing Material
urtec-2014-1922843-MS.
Balance Equation for Two-Phase Flowback Analysis. Journal of Pe-
Nguyen, P. D., Weaver, J. D., Parker, M. A. et al. 1996. Proppant Flow-
troleum Science and Engineering 142: 170185. https://doi.org/
back Control Additives. Presented at the SPE Annual Technical Con-
10.1016/j.petrol.2016.01.018.
ference and Exhibition, Denver, 69 October. SPE-36689-MS. https://
Zhang, J., Kamenov, A., Hill, A. D. et al. 2014. Laboratory Measurement
doi.org/10.2118/36689-MS.
of Hydraulic-Fracture Conductivities in the Barnett Shale. SPE Prod
Nolte, K. G. 1979. Determination of Fracture Parameters From Fracturing & Oper 29 (3): 216227. SPE-163839-PA. https://doi.org/10.2118/
Pressure Decline. Presented at the SPE Annual Technical Conference 163839-PA.
and Exhibition, Las Vegas, Nevada, 2326 September. SPE-8341-MS. Zhang, Y. and Ehlig-Economides, C. 2014. Accounting for Remaining
https://doi.org/10.2118/8341-MS. Injected Fracturing Fluid in Shale Gas Wells. Presented at the SPE/
Palacio, J. C. and Blasingame, T. A. 1993. Decline-Curve Analysis With AAPG/SEG Unconventional Resources Technology Conference, Den-
Type Curves Analysis of Gas Well Production Data. Presented at ver, 2527 August. SPE-2014-1892994-MS. https://doi.org/10.15530/
the SPE Rocky Mountain Regional/Low Permeability Reservoirs Sym- urtec-2014-1892994.
posium, Denver, 2628 April. SPE-25909-MS. https://doi.org/ Zhou, Q., Dilmore, R., Kleit, A. et al. 2016. Evaluating Fracture-Fluid
10.2118/25909-MS. Flowback in Marcellus Using Data-Mining Technologies. SPE Prod &
Parmar, J., Dehghanpour, H., and Kuru, E. 2014. Displacement of Oper 31 (2): 133146. SPE-173364-PA. https://doi.org/10.2118/
Water by Gas in Propped Fractures: Combined Effects of Gravity, 173364-PA.
Surface Tension, and Wettability. Journal of Unconventional Oil
and Gas Resources 5: 1021. https://doi.org/10.1016/j.juogr.2013.
11.005. Appendix A: Estimates of Fracture
Pearson, K. 1895. Notes on Regression and Inheritance in the Case of Two
Compressibility
Parents. Proc., the Royal Society of London, Vol. 58, 240242. Please refer to Fig. A-1.
Peters, E. J. 2012. Advanced Petrophysics. In Geology, Porosity, Absolute
Permeability, Heterogeneity, and Geostatistics, Vol. 1. Greenlead K J I HG F E D CB A A. Miner = 0%
Book Group (Reprint). 12,000
A. Miner = 0%
Raysoni, N. and Weaver, J. 2013. Long-Term Hydrothermal Proppant Per- B. Miner = 10%
11,000 C. Miner = 20%
formance. SPE Prod & Oper 28 (4): 414426. SPE-150669-PA. D. Miner = 30%
https://doi.org/10.2118/150669-PA. 10,000 E. Miner = 40%
F. Miner = 50%
Sharma, M. and Agrawal, S. 2013. Impact of Liquid Loading in Hydraulic G. Ratio = 90%
Net Stress on Fracture (psi)
Fractures on Well Productivity. Presented at the SPE Hydraulic Frac- 9,000 H. Ratio = 70%
I. Ratio = 50%
turing Technology Conference, The Woodlands, Texas, 46 February. J. Ratio = 30%
8,000 K. Ratio = 20%
SPE-163837-MS. https://doi.org/10.2118/163837-MS.
L. Ratio = 10%
Sharma, M. M. and Manchanda, R. 2015. The Role of Induced Un- 7,000
propped (IU) Fractures in Unconventional Oil and Gas Wells. Pre-
sented at the SPE Annual Technical Conference and Exhibition, Hous- 6,000
ton, 2830 September. SPE-174946-MS. https://doi.org/10.2118/
5,000
174946-MS.
Song, B. and Ehlig-Economides, C. A. 2011. Rate-Normalized Pressure 4,000
Analysis for Determination of Shale Gas Well Performance. Presented
at the North American Unconventional Gas Conference and Exhibi- 3,000
tion, The Woodlands, Texas, 1416 June. SPE-144031-MS. https://
2,000
doi.org/10.2118/144031-MS.
Tiab, D., Restrepo, D. P., and Igbokoyi, A. O. 2006. Fracture Porosity 1,000
of Naturally Fractured Reservoirs. Presented at the International
Oil Conference and Exhibition in Mexico, Cancun, Mexico, 31 0
August2 September. SPE-104056-MS. https://doi.org/10.2118/ 1106 1105 1104 1103
104056-MS.
Fracture Compressibilty (psi1)
Wallace, J., Kabir, C. S., and Cipolla, C. 2014. Multiphysics Investigation
of Diagnostic Fracture Injection Tests in Unconventional Reservoirs. Fig. A-1Charts for estimating fracture compressibility. Miner
Presented at the SPE Hydraulic Fracturing Technology Conference, is the estimated percentage of secondary mineralization in the
The Woodlands, Texas, 46 February. SPE-168620-MS. https:// natural fractures. Ratio is fracture porosity divided by the sum-
doi.org/10.2118/168620-MS. mation of fracture porosity and vug porosity (Aguilera 1999).
Region 1 Region 3
Gas Lift 3 3
Region 2
0 0 0 0
0 5 10 15 20 0 5 10 15 20
Days From Start of Flowback Days From Start of Flowback
(a) Well B (b) Well C
8 9
Gas Rate (Mcf/D), Oil Rate (BOPD),
0 0 0 0
0 5 10 15 20 0 5 10 15
8 9
Gas Rate (Mcf/D), Oil Rate (BOPD),
Calculated Pwf
Region 1
6
Region 2
4
Region 3
3
0 0
0 5 10 15 20
Days From Start of Flowback
(e) Well F
Fig. B-1Hourly flowback data show three regions. Region 1 shows single-phase of water production and declining pressure.
Region 2 shows single-phase of water production and flattening pressure. Region 3 shows multiphase production.
Well B Well C
10 9 10 9
1 6 1 6
RNP
RNP
Unit-slope
0.1 3 0.1 3
Unit-slope deviates Unit-slope deviates
Unit-slope as Pwf flattens out as Pwf flattens out
0.01 0 0.01 0
0.01 0.1 1 10 100 0.01 0.1 1 10 100
tMB (days) tMB (days)
(a) (b)
Well D Well E
10 9 10 9
1 6 1 6
RNP
RNP
0.1 3 0.1 3
Unit-slope Unit-slope deviates Unit-slope Unit-slope deviates
as Pwf flattens out as Pwf flattens out
0.01 0 0.01 0
0.01 0.1 1 10 100 0.01 0.1 1 10 100
tMB (days) tMB (days)
(c) (d)
Well F
10 9
Region 2
Unit-slope deviates
as Pwf flattens out
0.1 3
Unit-slope
0.01 0
0.01 0.1 1 10 100
tMB (days)
(e)
Fig. C-1Log-log plot of RNP vs. tMB shows a clear unit-slope before the calculated bottomhole pressure flattens out (RNP is in
the unit of psi/STB/D).
1 1
Well A Well B
0.8 0.8
RNP
R = 0.9956 R = 0.9974
0.4 0.4
0.2 0.2
0 0
0 2 4 6 8 10 0 2 4 6 8 10
tMB (days) tMB (days)
(a) (b)
1 1
Well C Well D
0.8 0.8
0.6 0.6
RNP
y = 0.1254x + 0.0976
RNP
R = 0.9385
0.4 0.4
y = 0.0752x + 0.0032
0.2 0.2
R = 0.9918
0 0
0 2 4 6 8 10 0 2 4 6 8 10
tMB (days) tMB (days)
(c) (d)
1 1
Well E Well F
0.8 0.8
RNP
0.2 0.2
0 0
0 2 4 6 8 10 0 2 4 6 8 10
0.6
RNP
0.4
y = 0.076x + 0.0527
0.2 R = 0.9783
0
0 2 4 6 8 10
tMB (days)
(g)
Fig. D-1RNP vs. tMB in Cartesian plot and linear fit for each well.
Yingkun Fu is currently a PhD-degree student in the School of Hassan Dehghanpour is an assistant professor in the Depart-
Mining and Petroleum Engineering at the University of Alberta. ment of Civil and Environmental Engineering at the University
His research interests include flowback and production data of Alberta. His primary research interests include measurement
analysis in unconventional reservoirs and software develop- and modeling of multiphase flow in porous media for applica-
ment. Fu holds BSc and MSc degrees in petroleum engineer- tions in enhanced oil recovery and petroleum production
ing from the China University of Geosciences, Beijing. from shale reservoirs. Dehghanpour has authored or
coauthored more than 80 refereed-journal and conference more than seven peer-reviewed-journal papers within the last 3
papers, mainly related to reservoir-engineering aspects of years. He is an SPE member and the First-Prize recipient (PhD
unconventional reservoirs. He is the recipient of the 2014 SPE degree division) in the 2013 SPE Student Paper Contest, Canada
Young Member Outstanding Service Award and the 2015 SPE region. Ezulike holds BSc degrees in geology and physics from the
Distinguished Achievement Award for Petroleum Engineering University of Nigeria and an MSc degree in petroleum engineer-
Faculty, both for the Canada region, and serves on the exec- ing from the African University of Science and Technology.
utive committee for the 2017 SPE Unconventional Resources
R. Steven Jones Jr. is a reservoir-engineering adviser at
Conference. Dehghanpour holds two BS degrees in mechani-
Cimarex Energy Company in Tulsa. Previously, he was with
cal and petroleum engineering from Sharif University of Tech-
nology, Iran, and an MS degree from the University of Alberta Newfield Exploration Company (where the work on this paper
and a PhD degree from the University of Texas at Austin, both was performed) for 15 years. Jones is focused on the reservoir
aspects of appraising and optimizing unconventional
in petroleum engineering.
resource plays. He has authored or coauthored five technical
D. Obinna Ezulike is currently a PhD-degree candidate in the papers. Jones is a registered professional engineer in Okla-
School of Mining and Petroleum Engineering at the University of homa, and is a member of SPE and the American Association
Alberta. His research interests include modeling fluid flow in con- of Petroleum Geologists. He holds a BS degree from the Univer-
ventional and unconventional reservoirs, artificial intelligence, sity of Tulsa and an MS degree from the University of Texas at
and software development. Ezulike has authored or coauthored Austin, both in petroleum engineering.