You are on page 1of 14

SPE 144031

Rate-Normalized Pressure Analysis for Determination of Shale Gas Well


Performance
Bo Song, Christine Ehlig-Economides/Texas A&M University

Copyright 2011, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE North American Unconventional Gas Conference and Exhibition held in The Woodlands, Texas, USA, 1416 June 2011.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
The horizontal well with multiple transverse fractures has proven to be an effective strategy for shale gas reservoir
exploitation. However, to improve the well design it is important to quantify the reservoir permeability and the extent of the
stimulated reservoir volume (SRV) defined by the created hydraulic fractures.
Rate-normalized pressure (RNP) is easily computed from production data and is more straightforward to interpret than
pressure normalized rate. Pseudolinear flow before pressure interference happens between two adjacent transverse fractures,
seen as a slope trend in the RNP derivative, and pseudo pseudosteady state, seen as a straight trend with slightly less than
unit slope, are key to the analysis. When only pseudolinear flow appears, a minimum extent for the SRV is indicated as well
as a maximum value for the shale permeability. The appearance of pseudo pseudosteady state flow provides an estimate for
the pore volume of the SRV and a minimum value for the shale permeability.When both pseudolinear and pseudo
pseudosteady state flow regimes appear, permeability and the average half-length and spacing of created fractures can be
estimated.
We illustrate the analysis procedure with field data from key shale gas plays. Estimates for both the expected ultimate
recovery and recovery efficiency as a fraction of the gas in place in the SRV are possible using this interpretation technique.
Sensitivity of the analysis to uncertainties in shale thickness, porosity, and adsorption parameters are explored.
Introduction
The horizontal well with multiple transverse fractures (MTFW) has proven to be an effective way to exploit shale gas
reservoirs economically. The MTFW strategy has been widely used in the Barnett Shale and the Marcellus Shale (Arthur,
Bohm and Layne, 2008). Several authors have studied the behavior of MTFWs: Clarkson et al (2009), Freeman et al (2009)
and Al-Kobashi et al (2006) described flow regimes for the MTFW including pseudolinear flow, radial and coumpound linear
flow before infinite-acting, pseudosteady state or steady state flow related to the extent of the well drainage area. Song,
Economides and Ehlig-Economides (2011) did pressure drawdown behavior and rate decline analyses for MTFW, and
provided a MTFW design strategy for shale gas exploitation that emphasized the use of permeability for optimizing the
fracture spacing. They introduced the term pseudo pseudosteady state for the flow regime following pseudolinear flow to
closedly spaced parallel fractures and explained its relationship to the pore volume of the stimulated reservoir volume (SRV)
described by Meyer et al (2010).
The characterization of gas adsorption in shale gas reservoirs has also been studied. Kuuskraa et al (1985) indicated the
importance of gas adsorption to gas recovery and behavior of shale gas wells in Ohio, West Virginia, and Kentucky. Lane,
Watson and Lancaster (1989) indicated that the gas in shale reservoirs is stored both as free gas in matrix pores and fractures
and as adsorbed gas on the surface of matrix particles. Zuber et al (2002) provided a conceptual model for New Albany
Shale, and Schepers et al (2009) formulated a triple porosity/dual permeability model including the consideration of both free
gas and adsorbed gas. Song, Economides and Ehlig-Economides (2011) defined the adsorption index to quantify the
2 SPE 144031

adsorption effect on pressure penetration and production, and developed an empirical correlation to compute the adsorption
index according to adsorption parameters and reservoir properties.
This paper shows how to use rate-normalized pressure (RNP) computed from production data to quantify the SRV pore
and shale permeability from two key flow regimes: pseudolinear (PL) and pseudo pseudosteady state (PPSS). When only PL
flow is visible on the RNP and RNP derivative plot, the maximum shale permeability and a minimum SRV pore volume can
be estimated. The appearance of PPSS gives a direct estimate for SRV pore volume and a minimum value for shale
permeability. If both flow regimes apprear, both the SRV pore volume and shale permeability can be determined. The SRV
pore volume, in turn, enables estimation of the average half-length of the created hydraulic fractures.
Stimulated Reservoir Volume (SRV)
Operators commonly execute the MTFW in a series of hydraulic fracture stages, starting from the toe of the horizontal well.
After one stage is completed a bridge plug is installed before starting the next stage. Before fracturing a segment of the well
a series of perforation clusters, typically 5 or more, are shot with high shot density, each approximately 2 to 4 ft long and
spaced evenly within the designed stage length. Typical spacing between perforation clusters may be 50 to 100 ft or more.
For our study none of the wells had direct confirmation that a hydraulic fracture is created from each perforation cluster. For
our analysis, we assumed that on average there is one fracture per stage or one fracture per cluster. Figure 1 shows a diagram
of the fractures created in one stage. The model used for this study idealizes the stimulated reservoir volume (SRV) for the
shale gas well as parallel transverse fractures evnly spaced along the well and each with the same half-length and
conductivity.

Figure 1: Plane View of Stimulated Reservoir Volume (SRV) Created by one stage of the MTFW

The dimensions of such an idealized rectangle SRV are:

hSRV h: the maximum height of the SRV is the reservoir thickness


wSRV =2xf + xs /2: the width of the SRV is the fracture length (twice fracture half-length plus an estimated distance
the pressure penetrates beyond the fracture tips given by of the fracture spacing
LSRV =Lw; the length of the SRV is the length of horizontal well

Therefore, the bulk volume of SRV is VSRV = L h (2 xf + xs/2). Considering both free gas and adsorbed gas, the original gas in
place is:

VL pi (1)
Gi = Gi , free +Gi , ads = Lh(2 x f + x s / 2) / B gi + rock
p L + pi

where in percentage is the reservoir porosity, Bgi is intial gas formation volume factor in scf/rcf, rock is reservoir bulk
density in ton/scf, VL is the Langmuir volume in scf/ton defined in the Langmuir Mode,l and pL is Langmuir pressure in psi
also as defined in the Langmuir model.
SPE 144031 3

Song, Economides and Ehlig-Economides (2011) noted that the impact of adsorption on the transient response is seen as a
parallel shift in the log-log plot for pressure change and its derivative versus time. They introduced the adsorption index (Iads)
to quantify this shift and developed a correlation for Iads given by

6875.34 1 log 2 ( p L / pi ) (2)


I ads = ( + 2.4298 10 4 pi 0.1992) 1.0215 exp[ ] surf rockVL 0.30483 + 1
pi 0.6644 2 0 .8829
The correlation applies for 0.01 < < 0.1, 0.01 g/cc < ads (= surfrock VL) < 0.1g/cc, 1000 psi < pi <10000 psi and pi /10 < pL
<10 pi. When parameters are outside the above ranges, simulation is recommended to determine the adsorption index.
Rate-Normalized Pressure (RNP)
The well production is seldom at a constant rate; nor is it at a constant flowing pressure. Computed as
pi p wf (t )
RNP =
q (t ) (3)
dRNP
RNP ' =
d ln t e
rate-normalized pressure (RNP) and its derivative (RNP) represent the production behavior that would be observed if the
well were produced at a constant reference rate. In the above equations

te = Q(t ) / q(t ) (4)


is material balance time as defined by Palacio and Blasingame (1993).
Pressure Normalized Rate (PNR)
Palacio and Blasingame (1993) developed type curve analysis based on the pressure-normalized rate (PNR) computed as

Pressure Nomalized Rate:


q(t ) (5)
PNR(t ) =
pi pwf (t )
Pressure Normalized Rate Integral:
t t
1 1 q ( )
te 0 p
PNR Int = PNR ( )d = d (6)
te 0 i p wf ( )

Pressure Normalized Rate Integral derivative:


'dPNRInt (7)
PNRInt =
d ln(te )
They plot each of the above quantities versus material balance time.
Drawdown Flow Regimes of the MTFW
Figure 2 (Song, Economides and Ehlig-Economides, 2011) shows the typical flow regimes during MTFW production include
fracture storage, pseudolinear flow (PL) normal to the transverse fractures, pseudo pseudosteady state flow (PPSS),
compound linear flow, pseudoradial flow, and boundary behavior. Figure 3 shows these flow regimes as they appear on a
log-log plot of RNP and RNP derivative versus material balance time.

Equations for pseudolinear flow, the time at the end of pseudolinear flow, and pseudo pseudosteady state provide
relationships among important parameters. During PL every transverse fracture produces independently, and the total fracture
length and reservoir permeability are related by:
1/ 2
724q g B g g (8)
nf xf k =
mlf I ads h c
t
where nf is the the number of transverse fractures. During PL flow the RNP derivative slope is .
The time at the end of PL flow occurs when flow to adjacent transverse fractures interferes. The time of this occurrence is
given by
948 g ct x s2
t eplf = I ads (9)
16k
4 SPE 144031

Figure 2: Potential Flow Regimes during MTFW Production

Figure 3: Drawdown Behavoir of the MTFW Expressed by Rate Normalized Pressure and Its Derivative

where xs is fracture spacing in ft.


After the departure from PL flow, the RNP derivative slope increases to nearly unity. Song, Economides, and Ehlig-
Economides (2011) described this behavior as pseudo pseudosteady state because during this time each fracture produces
from its own drainage volume bounded parallel to the fracture plane by the virtual closed boundary resulting from
interference between adjacent fractures, but unbounded beyond the fracture tips unless similar adjacent wells exist. As
indicated previously, the location of the missing end boundaries is approximated as xf + xs/4. The SRV pore volume, Vp (ft3),
can be estimated from the PPSS RNP derivative trend as

dRNP RNP ' 41.7 q ref B g (10)


= =
dte te V p I ads ct

Futhermore, assuming the payzone thickness and reservoir porosity are known, the fracture half-length for a fully penetrating
fracture can be estimated as:
SPE 144031 5

Vp xs 41.7q ref B g x (11)


xf = = / I ads s
2 Lh 4 2 L h( dRNP ) c 4
w t
dt e

The transformation of Eq.9 will give a way to estimate the formation permeability, shown as Eq.12:

948 g ct x s2
k= I ads (12)
16t eplf

Equation 8 provides an additional relationship between permeability and the total fracture half-length, nfxf. Using this
equation to estimate the number of fractures enables confirmation of the effective spacing between fractures, xs = Lw/nf.
Finally, the match with rate decline and cumulative production using pressure as the input variable provides a means to
determine the adsorption index. When the adsorption is greater than unity, the corresponding fracture half-length is less, and
permeability is more. The net result is an increase in cumulative production. Hence, if cumulative production is less than the
observed amount when the match assuming unit adsorption index shows a fit with the RNP and RNP, this is a sign that the
adsorption index is greater than unity. For the increased adsorption index, decreasing the fracture half-length to xf/Iads, and
increasing the value for permeability to kIads results in the match with cumulative production. Quantifying the adsorption
index does not enable estimation of the Langmuir pressure, pL, and volume, VL, parameters, but when Iads >1, significant
reserves may be present as adsorbed gas, and core measurements on a shale sample can provide estimates for these values.
Alternatively, microseismic measurements may provide an independent measure for the SRV extent. If the SRV area
implied by the match with production data appears to be larger than that indicated by microseismic, this could be an
indication that adsorbed gas is contributing to the production. If the SRV area implied by the match with production data
appears to be smaller than that indicated by the microseismic analysis, this could be an indication that not all microseismic
events indicate effectively stimulated pore volume.
Field Examples
Three field examples illustrate the application of the RNP production data analysis. The first example from a well in the
Fayetteville Shale shows both PL and PPSS trends in the RNP derivative. This case illustrates estimation of the SRV pore
volume, the volume of adsorbed gas, average fracture half-length, the shale permeability, and the expected well recovery
factor for an assumed economic limit rate. The second example from the Haynesville Shale exhibits only PPSS trends in the
derivative. In this case, the SRV pore volume estimated from the PPSS trend gives an implied value for the average fracture
half-length, and we estimate a minimum value for permeability assuming a value for the fracture spacing and a unit
adsorption index. The third example from New Albany Shale exhibits only the PL trend. This case yields an estimate for the
maximum value for the shale permeability and for the product, n f x f k , which, in turn, gives a corresponding minimum
value for the average fracture half-length.
In all three examples, casing pressure was recorded on at least a daily basis. Because the production is predominantly
single-phase gas, the casing pressure trends mimic the bottomhole pressure closely.
Fayetteville Shale Example
Table 1 shows the well, fluid, and formation data for this example. Figure 4 shows the well production rate and pressure data,
and Figure 5 shows the RNP plot. The RNP plot shows both pesudolinear flow indicated by a half-slope trend in the RNP
derivative and PPSS indicated by a unit slope trend in the RNP derivative.
First, without considering adsorption, using Eq. 10 for RNP =
Table 1. Well, Fluid, and Formation Data for the 12.38 psi and te = 51,926 hr, we estimate the SRV pore volume from
Fayetteville Shale Example the PPSS trend as 10.2.106 ft3. Equation 11 implies that the average
Horizontal Well Length L 4300 ft fracture half-length is 51 ft. However, this assumes the fracture
Number of cluster nc 72 height is equal to the formation thickness of 322 ft. Instead, it seems
Total Number of fractures nf 72
more likely that the fracture height would be less, and that the
Effective Fracture Spacing xs 49 ft
created fractures may have a radial geometry. Assuming the same
Reservoir Depth d 2344 ft
Reservoir Thickness h 322 ft
area for the fracture face plus the additional length of half the
Initial Reservoir Pressure pi 1050 psi fracture spacing leads to a square transverse fracture with height 180
Reservoir Temperature Tr 125 F ft and half-length 90 ft.
Porosity 0.08
Assuming a fracture is created from each cluster, Eq 12 estimates
the permeability at the time at the end of PL, teplf = 3583 hr, at
1.56.10-4 md = 156 nd. To check this, we solve Eq. 8 for the number
6 SPE 144031

of fractures using the value for RNP = 1.05 at the time teplf = 3583 hr to compute mlf = 2RNPt-0.5 = 0.414. The result
confirms the fracture spacing used in the permeability estimation from Eq. 12.
However, Fig. 6 shows the match for production rate and cumulative production data using values for xf and k estimated
above, and the model gives a lower cumulative production than observed in the data. A possible explanation may be
adsorption.
We consider the pore volume, average fracture half-length and permeability values estimated without considering
adsorption as apparent values, Vp-app, xf-app and kapp. As noted in Song, Economides, and Ehlig-Economides (2011), the
adsorption effect is seen as a shift the RNP and RNP to the right along the time axis by the adsorption index, Iads. For Iads >
1, we adjust the pore volume, fracture half-length, and permeability as follows:
Vp-real = Vp-app / Iads
xf-real = xf-app / Iads (13)
kreal = kappIads
For this particular well, we determined that an adsorption index of about 1.4 gives the observed cumulative production.
From Eq. 13, the pore volume of SRV is 7.3 106 ft3. In addition, Eq. 13 gives the fracture half-length as 64 ft and
permeability 220 nd.

Figure 4. Fayetteville Shale Well Production Rate and Cumulative Data

Figure 5. Feyetteville Shale Well RNP and RNP Plot

The adsorption index does not resolve the


adsorption parameters, Vp and VL, individually
because different combinations yield the same
adsorption index. Table 2 shows two parameter
combinations yielding the same adsorption index.
Figure 7 shows the match for production rate and
cumulative production for both sets of adsorption
parameters. While the models are identical, what is
important is that the match is an improvement over
that shown in Fig. 6 that did not consider

Table 2: Two Sets of Adsorption Parameters


Parameter Unit Case 1 Case 2
pL psi 5400 2000
ads g/cc 0.014 0.0083
Iads dim. 1.4 1.4
6 3
Figure 6. Fayetteville Shale Well Production Rate and Cumulative Data Match Vpreal 10 ft 7.3 7.3
without Adsorption hSRV ft 160 160
xfreal ft 64 64
kreal nd 220 220
SPE 144031 7

adsorption. Similarly, Figure 8 shows that the match for RNP and RNP is unaffected, and again, the two models are
identical. This shows that transient data cannot resolve these parameters, and it is necessary to measure them from a core
sample. If adsorption parameter values are not available, it is only possible to estimate the volume of free gas in place, Gi,free
= 0.51 BSCF. It is not possible to estimate the volume of adsorbed gas. However, given the adsorption parameters for Case 1
and Case 2, using Eq. 1 we compute the adsorbed gas as

Figure 7: Fayetteville Shale Well Production Rate and Cumulative


Production Match with Adsorption

Figure 8: RNP and RNP Match for Fayettevile Shale Well


0.724 BSCF and 0.779 BSCF, respectively, for Bgi = 0.0142.
In summary, the analysis for this well provides the SRV pore volume 7.3.106 ft3; average fracture half-length of about 64
ft; and formation permeability of about 220 nd, and that the production is supported by adsorption with an adsorption index
of 1.4. The reason all of these estimates are possible is that both PL and PPSS flow regimes are present in the RNP response.
To illustrate the contrast between RNP and PNR analysis, Fig. 9 shows PNR, PNRint and PNRint graphs for the
Fayetteville Shale well data. Of the three graphs, only the PNR shows trends of interest. A negative slope corresponding to
PL flow appears in the PNR until about 30,000 hr material balance time. This is followed by -1 slope corresponding to
pseudo pseudosteady state flow behavior. The departure time from the PL is later than that seen in the RNP response shown
in Fig. 5 and cannot be used in Eq. 9. Neither the PNRint nor its derivative shows useful analysis trends.

Figure 9. PNR, PNRint and Derivative Plot for Fayetteville Well

Figure 10. PNR and Cumulative Production for the Fayettefille Shale
Well
8 SPE 144031

Figure 10 shows a more useful presentation. Selecting an


economic rate limit of 10 MSCF/d provides a model-
based estimate for the Expected Ultimate Recovery
(EUR) of 0.65 BSCF. Figure 11 shows the same data
plotted against time instead of the material balance time
and indicates that the well can reach the EUR in less than
10 years. Using the gas in place estimate for Case 1
adsorption parameters gives an EUR recovery factor of
nearly 90%.
Interestingly, we note that the model extrapolation
exceeds the cumulative production data points for large
values of the material balance time, as it does in Fig. 7.
Therefore, the EUR of 0.65 BSCF may be somewhat
optimistic.
Figure 11. Production Rate and Cumulative Production for the Hayenesville Shale Example
Fayettefille Shale Well Table 3 shows the well, fluid, and formation data for this
example. Figure 12 shows the production rate and
pressure data, and Figure 13 shows the RNP plot. The
Table 3. Well, Fluid, and Formation Data for the Haynesville Shale RNP plot shows a clear PPSS regime indicated by the unit
Example slope trend in both RNP and RNP. Assuming that the
Horizontal Well Length L 3940 ft adsorption in Haynesville is insignificant, the PPSS trend
Number of Stage ns 15 provides an estimate for the SRV pore volume of 1.35.107
Number of Cluster per Stage ncps 5
ft3 using Eq 10 and average fracture half-length of 150 ft
Total Number of fractures nf 75
Effective Fracture Spacing xs 53 ft
using Eq 11. This also provides an estimate for free gas in
Reservoir Depth d 12228 ft
place of 3.45 BSCF for Bgi = 0.0032.
Reservoir Thickness h 150 ft Unlike the previous example, in this case, the RNP
Initial Reservoir Pressure pi 9000 psi graph only shows PPSS, and the PL flow regime is
Reservoir Temperature Tr 320 F absent. Therefore, it is not possible to estimate the shale
Porosity 0.07 permeability from the RNP graph. However, it is possible
to estimate a minimum value for the shale permeability
by assuming the end of PL flow to be at the latest possible time before the start of the PPSS. This time, shown as 24 hr in Fig.
13, provides the minimum permeability estimate of 0.0014 md. Assuming there is a fracture created by each perforation
cluster provides the match with data shown in Figs. 14 and 15, for permeability of 0.0031 md. Interestingly, Marongiu-Porcu,
Economides, and Ehlig-Economides (2011) estimated the permeability independently from a fracture calibration test in this
well as 0.004 md. Extrapolating PNR using the model match shown in Figs 14 and 15 leads to recovery of 2.5 BSCF in less
than 3 years for an economic rate limit of 10 MSCF/d, giving a recovery factor of about 72%. Because of some problems in
processing the raw pressure data, this estimate is probably pessimistic.

Figure 12. Haynesville Shale Well Production Rate and Cumulative Data
Figure 13. Haynesville Shale Well RNP and RNP Plot
SPE 144031 9

Figure 14. Haynesville Shale Well Production Rate and Cumulative Data Figure 15. Haynesville Shale Well Model Match
Match
New Albany Shale Example
Table 4 lists the well, fluid, and formation data for this example. Figure 16 shows production rate and wellhead pressure data
acquired over time. The RNP plot for this well in Fig. 17 only shows the PL trend.
Without the PPSS trend, it is not possible to quantify directly
Table 4. Well, Fluid, and Formation Data for the New Albany the SRV pore volume. Assuming the departure from PL occurs
Shale Example right after the last recorded data gives an estimate for the
Well Name symbol value unit maximum permeability of 0.045 md. However, the adsorption
Horizontal Well Length L 3500 ft parameters measured from a core sample from New Albany Shale
Number of Stage ns 8 shown in Table 3 give and adsorption index of 1.5 using the
Number of Cluster per Stage ncps 1 correlation in Eq. 2. Therefore, the adjusted maximum
Total Number of fractures nf 8
permeability is approximately 1 microdarcy, or 0.001 md.
Effective Fracture Spacing xs 438 ft
Reservoir Depth d 2380 ft
Microseismic measurements in a well completed in a similar
Reservior Thickness h 56 ft way indicated an average fracture half-length of approximately
Initial Reservoir Pressure pi 714 psi 1200 ft. Using this estimate for xf in Eq. 5 and the adsorption
Reservoir Temperature Tr 89 F index of 1.5 gives an estimate for permeability of about 150 nd =
Porosity 0.06 0.00015 md. Figures 18 and 19 separately show the match for
Langmuir Pressure pL 1044 psi production and the RNP plot. Using data from Table 3 and the
Langmuir Volume VL 125.8 scf/ton assumed average fracture half-length, with Eq. 1 we estimate the
Rock Density rock 2.37 g/cc
OGIP at 2.9 BSCF for Bgi = 0.0217, incuding 45% as free gas and
55% as adsorbed gas.
Using the permeability value estimated from the production data and the fracture spacing for the number of fracture
stages, Eq. 9 indicates that adjacent fractures would not interfere for more than 200 years. If the well remains on production
down to an economic rate of 10 Mscf/d, only 10% of the gas in place will be recovered. To achieve greater recovery
efficiency, the fractures should be more closely spaced. While both the Fayetteville and Haynesville wells injected multiple
perforation clusters for each fracture stage, the New Albany well had only one fracture per stage.
The Importance of Pressure Measurement
Many arbitrary empirical extrapolation methodologies have been developed for determining the estimated ultimate recovery
(EUR) using only reported rate data. A commonly used methodology developed by Valko (2009) is

Q QD 1 (14)
rp = 1 =1 = [1 / n, ln q D ]
EUR EUR D [1 / n]

where rp is the recovery potential, Q is cumulative production in scf, QD = Q/Qi is dimensionless cumulative production, qD =
qg /qi is dimensionless production rate and n is a dimensionless model parameter. Data are graphed for an assumed value for
n. The value for n is varied from 0 to1 to find the value that produces a straight line with the maximum root mean square
correlation. Once the best straight line is found, it is extrapolated to the economic rate to determine the EUR for the well.
10 SPE 144031

Figure 16. New Albany Shale Well Production Rate and Cumulative
ProductionData Figure 17. New Albany Shale Well RNP and RNP Plot

Figure 18. New Albany Shale Well Production Rate and Cumulative Production
Data Match Figure 19. New Albany Shale Well Model Match

The EUR estimates for the Fayetteville and Haynesville data are shown in Fig. 21 (3, 9 and 18 months) and Fig. 22 (for 3
and 6 months since the production data provided for the Haynesville Well is just for first 6 months). These estimates are more
stable because both wells have reached PPSS flow, and the power law extrapolation is more rigorous in this case. However,
we note that the power law empirical EUR estimate for the Fayettville Shale well is greater than the model-based estimate of
the gas in place, including adsorbed gas.
These examples clearly show the importance of pressure measurement. For dry gas wells wellhead pressure should
suffice. However, we encountered model instabilities when pressure was recorded only once per day. We are investigating
what pressure data rate would avoid such problems. With pressure and rate data it is possible to understand the geometry of
the SRV and how to improve the well design. Empirical extrapolations of rate data without pressure only indicate the EUR.
The SRV pore volume provided from the RNP and PNR analyses using pressure and rate enables a more reliable estimate for
the EUR and, in addition, estimation of the gas recovery factor.
Discussion
Core measurements of porosity and permeability are difficult to achieve and, as for any reservoir, may not be representative
of macro scale properties. The three examples shown in this paper indicate permeability values ranging from 150 nd to 4 d,
and raise questions whether they represent matrix or microfracture permeability.
Uncertainties about the nature of the stimulation in shale gas reservoirs abound. Some authors prefer to characterize
created stimulation as a complex network of fractures, perhaps mainly representing reopened existing filled natural fractures.
Such interpretations are encouraged by microseismic data showing a swarm of events detected while fracturing. However,
such observations must consider the context of known completion strategies. When the operator has shot multiple perforation
clusters spaced uniformly along the horizontal well in a single hydraulic fracture stage, the wide swarm of events observed in
SPE 144031 11

microseismic data should not be surprising, especially when the width of the noise is similar to the length of the stage along
the horizontal well axis. One can question whether microseismic resolution is sufficient to follow multiple parallel hydraulic
fracture propagations spaced as close as 50 ft or less. When significant contrast exists between minimum and maximum
stress and between lateral stresses, we should expect minimal fracture complexity.

Figure 20: Power Law EUR Extrapolation for the New Albany Shale
Well Production Data

Figure 21: Power Law EUR Extrapolation for the Fayetteville Shale Well
Production Data

In this paper we have characterized the set of fractures


numbering from 8 in the New Albany Shale example to
72 in the Fayetteville Shale example as parallel
effectively infinite-conductivity fracture planes each
having an average half-length, height, and effective
spacing. When the fracture half-length is less than the
shale thickness, we assume the fractures are effectively
radial with height equal to twice the fracture half-length
and assume in this case that the SRV height is less than
the shale thickness. The actual dimentions of the SRV
may be different and are dependent on an assumed
porosity value, but the PPSS behavior uniquely
determines the SRV pore volume.
Figure 22: Power Law EUR Extrapolation for the Haynesville Shale Well The permeability determined from this analysis
Production Data represents a property of the bulk shale between created
fractures. Whether matrix or microfracture permeability,
we assume this value is a continuum property. We have seen no evidence in the field data to support dual porosity behavior
that might be expected if the SRV characterization is predominantly that of a random continuum of opened natural fractures.
Instead, in two of the 3 examples, we see evidence of pseudolinear flow characteristic of matrix flow to one or more
effectively infinite-conductivity fracture planes followed by apparent pseudosteady state flow behavior which we have
interpreted as multiple fracture drainage volumes. If operators were to record pressure at a higher data rate, whenever the well
is shut in, buildup transients could be studied. The early time response might indicate dual porosity behavior that is no longer
visible after one day on production.
The appearance of PL flow shows that concerns some operators express about low apparent fracture conductivity appear
to be misplaced. It is essential to recognize that the behavior of the created fractures is based on the relative pressure
gradients representing linear flow in the fracture as kfw and linear flow in the shale to the fracture as kxf. Predominant PL
flow in the long term production data indicates that the magnitude of the ratio between these two products, well known as the
dimensionless fracture conductivity, CfD, is large. We expect this to be the case when the reservoir permeability is in the
nanodarcy range.
By acquiring continuous pressure measurements during long term production, the operators who shared the data used in
the examples in this paper enabled a conceptual interpretation. In contrast, without pressure measurements, rate decline data
supports only empirical statistical analysis. Such trendology enables statistical comparisons, but it does not offer direct
insights about the special nature of the production that can be compared with independent observations provided by
microseismic data. Furthermore, we have shown that recognizing PL and PPSS behavior avoids overestimation of the EUR
12 SPE 144031

and enables estimation of not just EUR but an ultimate recovery factor indicating what fraction of the gas in place in the SRV
is recoverable.
In particular, Song, Economides, and Ehlig-Economides (2011) indicated that the same recovery factor achievable with
fewer fractures, but more fractures will achieve nearly full recovery in a shorter time. Creating fewer perforation clusters per
stage with greater half-length will increase the SRV pore volume, thereby yielding greater EUR for the well at the same
recovery factor. Economics would dictate the better choice at a given point in time. With this strategy, provided it is possible
to execute greater fracture half-lengths successfully, fewer wells would be required to effectively drain a lease.
As a final observation, once the PPSS has been observed, the well will show exponential decline behavior for production
under constant flowing pressure.
Conclusions
We have explained simple implications of a conceptual model for the MTFW and applied them to the analysis of long term
production data from wells completed in 3 different shale formations. All three wells recorded casing pressure continuously
since the start of production. The analysis provides the following conclusions:
The stimulated reservoir volume created by MTFW is the key driver to shale gas exploitation.
The presence of both PL and PPSS flow regimes in long-term production data processed as rate-normalized pressure
(RNP) enables estimation of the SRV pore volume, average created fracture half-length, effective fracture spacing, shale
permeability, and the adsorption index.
The presence of PPSS flow without the PL flow regime enables estimation of the SRV pore volume, average created
fracture half-length, and a minimum value for the shale permeability.
The presence of PL flow without the PPSS flow regime enables estimation of a maximum value for the shale
permeability and a minimum SRV pore volume.
Estimation of the adsorption index does not provide quantification of Langmuir adsorption properties, but when this
factor is greater than unity, some of the gas within the SRV is in the adsorbed state. In this case, core measurements can
provide adsorption parameters that will quantify the additional volume of adsorbed gas recoverable from the SRV.
Empirical determination of the EUR using rate only rate data can be erroreous. As long as the well production is in PL
flow, the EUR will continually increase and will overestimate the actual well recovery potential. Once the well reaches
PPSS flow behavior, production at the lowest possible constant wellbore pressure will show an exponential rate decline
that can correctly indicate the EUR for an assumed economic limit rate. With a correct conceptual model RNP analysis
using pressure avoids empirical errors.
Ongoing pressure measurement greatly increases what can be learned from production data analysis leading to insights
on how to improve well design, more accurate quantification for the EUR, and the gas recovery factor.
Nomanclature
Bg = gas formation volume factor, rcf/scf
ct = total compressibility, psi-1
d = reservoir depth, ft
EUR = ultimate estimated recovery, scf
G = original gas in place, scf
h = payzone thickness, ft
Iads = adsorption index
k = reservoir permeability, md
L = horizontal well length, ft
mlf = slope of straight line for linear flow on the coordinate of p versus t1/2, psi/hour1/2
n = Valko power-law model parameter
ncps = number of clusters per stage
nc = number of clusters
nf = number of fractures
ns = number of stages
pi = initial reservoir pressure, psia
pL = Langmuir pressure, psia
PNR = pressure normalized rate, 1/psi
pwf = well bottomhole pressure, psia
qg = gas production rate, mscf/d
qi = initial production rate, mscf/d
qref = reference gas production rate, mscf/d
Q = cumulative production, scf
SPE 144031 13

RNP = rate normalized pressure, psi


t = time, year or hour
te = material balance time, hour
teplf = end of pseudolinear flow, hour
Tr = reservoir temperature, F
Vp = pore volume, scf/ton
VL = Langmuir Volume, scf/ton
VSRV = volume of stimulated reservoir volume, ft3
w = width, ft
xf = transverse fracture half-length, ft
xs = fracture spacing, ft
Greek variables
= porosity, fraction
g = gas viscosity, cp
surfg = density of surface gas, g/cc
rock = rock density, g/cc
= adsorption index correlation coefficient
= power-law model integral variable

Superscripts and Subrscripts


app = apparent
D = dimensionless
i = initial
int = integral
real = real
ref = reference
- = average
, = derivative
Acknowledgements
The authors would like to thank Southwestern Energy Company for sharing data used for the Fayetteville Shale example,
Shell-SWEPI and EnCana Oil and Gas for the Haynesville Shale example, and partners in the New Albany Shale project
sponsored by RPSEA for data and funding used for the New Albany Shale example. We would also like to express our
appreciation to Kappa Engineering for use of the Ecrin Topaz software for the production data analysis.
References
Al-Kobaisi, M., Ozkan, E., Kasogi, H., and Ramirez, B. 2006. Pressure-Transient Analysis of Horizontal Wells with
Transverse, Finite-conductivity Fractures. Paper PETSOC 2006-126 presented at the Petroleum Societys 7th Canadian
International Petroleum Conference (57th Annual Technical Meeting), Calgary, Alberta, Canada, 13-15 June.

Arthur, J.D., Bohm, B., and Layne, M. 2008. Hydraulic Fracturing Considerations for Natural Gas Wells of the Marcellus
Shale. Paper presented at the Ground Water Protection Council 2008 Annual Forum, Cincinnati, Ohio, 21-24 September.

Clarkson, C.R., Jordan, C.L., Ilk, D., and Blasingame, T.A. 2009. Production Data Analysis of Fractured and Horizontal
CMB Wells. Paper SPE 125929 presented at the 2009 SPE Eastern Regional Meeting, Charleston, West Virginia, 23-25
September.

Kuuskraa, V.A., Sedwick, K., and Yost II, A.B. 1985. Technically Recoverable Devonian Shale Gas in Ohio, West Virginia
and Kentucky. Paper SPE 14503 presented at the SPE 1985 Eastern Regional Meeting, Morgantown, West Virginia, 6-8
November.

Lane, H.S., Watson, A.T., and Lancaster, D.E. 1989. Identifying and Estimating Desorption from Devonian Shale Gas
Production Data. Paper SPE 19794 presented at the 64th Annual Technical Conference and Exhibition of the Society of
Petroleum Engineers, San Antonio, TX, 8-11 October.
14 SPE 144031

Mayer, B. R., Bazan, L.W., Jacot, R. H., and Lattibeaudiere, M. G. 2010. Optimization of Multiple Transverse Hydraulic
Fractures in Horizontal Wellbores. Paper SPE 131732 presented at the 2010 SPE Unconventional Gas Production
Conference, Pittsburgh, Pennsylvania, 23-25 February.

Mayerhofer, M. J., Lolon, E.P., Warpinski, N. R., Cipolla, C.L., and Walser. D. 2008. What is Stimulated Reservoir Volume
(SRV)? Paper SPE 119890 presented at the 2008 SPE Shale Gas Production Conference, Fort Worth, TX, 16-18
November.

Palacio, J.C., and Blasingame, T.A. Decline-Curve Analysis With Type Curves Analysis of Gas Well Production Data.
Paper SPE 25909 presented at the 1993 SPE Rocky Mountain Regional/Low Permeability Reservoir Symposium,
Denver, Colorabo, 12 - 14 April.

Rushing, J.A., Perego, A.D., and Blasingame, T.A. 2008. Applicability of the Arps Rate-Time Relationships for Evaluating
Decline Behavior and Ultimate Gas Recovery of Coalbed Methane Wells. Paper SPE 114515 presented at the CIPC/SPE
Gas Technology Symposium 2008 Joint Conference, Calgary, Alberta, Canada, 16-19 June.

Song, B., Economides, M.J., and Ehlig-Economides, C.A. 2009. Design of Multiple Transverse Fracture Horizontal Wells in
Shale Gas Reservoirs. Paper SPE 140555 presented at the 2011 SPE Hydraulic Fracturing Technology Conference,
Woodlands, TX, 24 - 26 January.

Schepers, K.C., Gonzalez, R.J., Koperna, G.J., and Oudinot. A.Y. 2009. Reservoir Modeling in Support of Shale Gas
Exploration. Paper SPE 123057 presented at the 2009 SPE Latin American and Caribbean Petroleum Engineering
Conference, Cartagena, Colombia, 31 May- 3 June.

Valko, P.P. 2009. Assigning Value to Stimulation in the Barnett Shale: A Simultaneous Analysis of 7000 Plus Production
Histories and Well Completion Records. Paper SPE 119369 presented at the 2009 SPE Hydraulic Fracturing Technology
Conference, Woodlands, Texas, 19-21 January.

Zuber, M.D., Williamson. J.R., Hill. D.G., Sawyer, W.K., and Frantz. J.H. 2002. A Comprehensive Reservoir Evaluation of a
Shale Reservoir-The New Albany Shale. Paper SPE 77469 presented at the 2002 SPE Annual Technical Conference and
Exhibition, San Antonio, Texas, 29 September-2 October.

You might also like