Académique Documents
Professionnel Documents
Culture Documents
Summary reserves accurately for these resource plays have become more
Traditional decline methods such as Arps’ rate/time relations and urgent and important than ever before.
their variations do not work for wells producing from supertight Mattar et al. (2008) have discussed various techniques for
or shale reservoirs in which fracture flow is dominant. Most of the production analysis and forecasting of shale-gas reservoirs, while
production data from these wells exhibit fracture-dominated flow Lee and Sidle (2010) have analyzed some of the commonly used
regimes and rarely reach late-time flow regimes, even over several procedures for forecasting and identifying key strengths and
years of production. Without the presence of pseudoradial and limitations of these techniques. The commonly used method for
boundary-dominated flows (BDFs), neither matrix permeability decline-curve analysis is Arps’ hyperbolic rate decline (Arps 1945),
nor drainage area can be established. This indicates that matrix if adequate production data are available. Fetkovich type curves
contribution is negligible compared with fracture contribution, (1980) are not appropriate because the matched b value is usually
and the expected ultimate recovery (EUR) cannot be based on a greater than unity. For horizontal wells with multiple transverse
traditional concept of drainage area. fractures, numerical-simulation modeling is a highly preferred
An alternative approach is proposed to estimate EUR from option. There are, however, some validity and applicability issues
wells in which fracture flow is dominant and matrix contribu- with these techniques.
tion is negligible. To support these fracture flows, the connected The hyperbolic decline equation is conveniently used because
fracture density of the fractured area must increase over time. it can have a “best fit” for the long transient linear-flow regime
This increase is possible because of local stress changes under observed in shale-gas wells with b values greater than unity. Lee
fracture depletion. Pressure depletion within fracture networks and Sidle (2010) showed that values of b equal to or greater than
would reactivate the existing faults or fractures, which may breach unity can cause the reserves to have physically unreasonable
the hydraulic integrity of the shale that seals these features. If properties. To avoid this drawback, “stretched-exponential” models
these faults or fractures are reactivated, their permeabilities will have been proposed recently (Valkó 2009; Ilk et al. 2008, 2009).
increase, facilitating enhanced fluid migration. For fracture flows The current simulation models that are available in the mar-
at a constant flowing bottomhole pressure, a log-log plot of rate ket are still using conventional technologies for unconventional
over cumulative production vs. time will yield a straight line with reservoirs. Some of the assumptions used in simulation modeling
a unity slope regardless of fracture types. In practice, a slope of are inconsistent with the field-data observations such as pressure
greater than unity is normally observed because of actual field initialization and radial (or elliptical) transient flow. Field data
operations, data approximation, and flow-regime changes. A rate/ indicate that the pressure transition through a shale-gas zone is at
time or cumulative production/time relationship can be established a disequilibrium state, while pressure initialization in simulation
on the basis of the intercept and slope values of this log-log plot modeling is based on an equilibrium state with its fluid gradient.
and initial gas rate. Newsham and Rushing (2002) showed that pressure gradients
Field examples from several supertight and shale gas plays for measured in the Lower Bossier can be as high as 63 kPa/m
both dry and high-liquid gas production, and for oil production (3 psi/ft) or more than 25 times its saturated gas gradient or three
were used to test the new model. All display the predicted straight- times the lithostatic gradient. Spencer (1987) presents some pres-
line trend, with its slope and intercept related to reservoir types. sure gradients in overpressured and gas-bearing shale plays in the
In other words, a certain fractured flow regime or a combination Rocky Mountain region. The interpreted pressure profile across
of flow types that dominate a given area or play because of its the active zone for wells in the Piceance basin shows a pressure
reservoir-rock characteristics and/or fracture-stimulation practices gradient of approximately 36 kPa/m.
all produce a narrow range of intercepts and slopes. An individual- The concept of stimulated rock volume (SRV) that is supported
well performance or EUR can be derived that is based on this range by microseismic monitoring of hydraulic-fracture treatments is not
if the best 3-month average or the initial production rate of the well confirmed by the observed radial flow from a numerical simulator,
is already known or estimated. The results show that this alterna- as pointed out by Mattar et al. (2008). These authors claim that
tive approach is easier to use, gives a reliable EUR, and can be the radial flow of fractured shale wells observed in a numerical
used to replace the traditional decline methods for unconventional simulator is more likely to be a false radial flow and may not be
reservoirs. The new approach is also able to provide statistical representative of the enhanced permeability of the SRV.
methods to analyze production forecasts of resource plays and This paper introduces an empirically derived decline model that
to establish a range of results of these forecasts, including prob- is based on a long-term linear flow in a large number of wells in
ability distributions of reserves in terms of P90 (lower side) to tight and shale-gas reservoirs. On the basis of this model, a new
P10 (higher side). method has been developed for production analysis and forecasting
of unconventional reservoirs. This method also uses probability
Introduction distributions of reserves in forecasting resource plays, to represent
Unconventional reservoirs, especially wet shale gas and oil, are any uncertainty in reserves estimation.
currently being pursued aggressively for new development in
both the US and Canada. Forecasting production and estimating Method Development
Long-Term Linear Flow. Fig. 1 shows the production history of
a vertical well producing from the Barnett shale. Production data
show a half-slope line, indicating linear flow lasting for more than
Copyright © 2011 Society of Petroleum Engineers
5 years of production.
This paper (SPE 137748) was accepted for presentation at the Canadian Unconventional Long-term linear flow in a large number of wells in tight or
Resources and International Petroleum Conference, Calgary, 19–21 October 2010 and
revised for publication. Original manuscript received for review 29 September 2010. Revised
shale gas has been observed as early as 1976 (Bagnall and Ryan
manuscript received for review 27 January 2011. Paper peer approved 23 March 2011. 1976). Since then, many researchers have tried to explain what
WHP, MPa
10 10
1 1
1 10 100 1,000 10,000
Days
Fig. 1—A log-log plot of field production and pressure data for a shale-gas well. Production data show a half-slope line, indicat-
ing linear flow lasting for more than 5 years of production.
causes this phenomenon. Several authors have shown that this The gas cumulative will be
linear flow exists in any type of fractures whether they are infi-
t 1− n
G p = ∫0 qdt = q1
nite- (Agarwal et al. 1979) or finite-conductivity (Cinco-Ley et al. t
. . . . . . . . . . . . . . . . . . . . . . . . . . . (2)
1978; Cinco-Ley and Samaniego 1981a, 1981b), single or multi- (1 − n)
stage (Bello and Wattenbarger 2010), hydraulic (Arévalo-Villagrán
et al. 2001), or natural fractures (Arévalo-Villagrán et al. 2006). This gives
The difference among these models is the length of this transient
linear-flow regime.
To prolong the linear-flow regime in any of those models, q (1− n) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)
=
all one needs to do is enlarge the term xf2/km in which xf is the Gp t
effective fracture half-length of the fracture system and km is the
matrix permeability. For any given shale-gas reservoir, matrix
permeability is fixed. Thus, the connected fracture density of the Field Applications. Field data from several shale-gas plays were
fractured area must be increasing over time to support this fracture used to test Eq. 3. These field data were operated under the actual
flow over the life of a producing well. This increase is achievable field conditions that may violate some of the ideal assumptions
because of local stress changes under fracture depletion (Warpinski used to derive Eq. 3. Some of the results are shown in Fig. 2 in
and Branagan 1989). Pressure depletion within fracture networks which log-log plots for q/Gp vs. time in days were constructed.
would reactivate the existing faults or fractures, which may breach These plots all give a log-log straight line with a negative slope,
the hydraulic integrity of the shale that seals these features. If −m, and an intercept of a. That is,
these faults or fractures are reactivated, their permeabilities will
increase, facilitating enhanced fluid migration. A fractured shale q
model that imitates this long-term linear flow and nonequivalent = at − m. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4)
pressure distribution will be briefly discussed later and is the Gp
subject of another paper.
If a fracture flow regime (either linear or bilinear) is prolonged Note that the slope is negative but m is always positive. As shown
over the life of a well, the gas-flow rate q will be later, m is always greater than unity for shale reservoirs. If m is
less than unity, it may indicate a conventional tight well.
q = q1t − n , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (1) Equations for q and Gp can be derived as shown in Appendix A:
a 1− m
where n is one-half for linear flow, n is one-quarter for bilinear q (t −1)
= t − m e 1− m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (5)
flow, and q1 is the flow rate at Day 1. q1
1 1 1
q/Gp , 1/Day
0.1
q/Gp, 1/Day
Fig. 2—A log-log plot of q/Gp vs. t for various shale-gas wells with more than 5 years of production. The data all form a straight
line with a slope m and an intercept a. A slope of higher than unity is normally obtained, to account for any deviations from the
ideal cases.
Tight/Shallow Gas
2.5
Dry Shale Gas
2 Wet Shale Gas
1.5
a
1 a = 0.7364 m 4.7954
R 2 = 0.9417
0.5
0
0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3
m
10
q max
1.50 1 year 2 years
5 years
1
1.00
q/q 1
m~1.00
m=1.05 0.50
0.1 m=1.10 Dots - 25 Barnett Samples
m=1.15 Solid lines - Eq. B-5 (P50)
m=1.20
m=1.25 0.00
m=1.3 0 5 10 15 20 25 30
0.01 6 3
Cum Gas, 10 m
1 10 100 1,000 10,000
Days Fig. 5—Best 3-month averages for a sample of 25 Barnett wells
were plotted against their cumulative productions of 1, 2, and
Fig. 4—Type curves for q/q1 vs. time in days for m>1. 5 years.
m=1.30
0.01
m=1.96 m=1.60
0.001
1 10 100 1,000 10,000
tm or tD
Fig. 6—A comparison between Arps’ and the new method’s type curves for dimensionless rate and time.
For comparison, the Arps rate/time relations (Arps 1945) were gas-rate equivalent and finally were normalized on the basis of the
reconstructed in the same plot using dimensionless time tD=Dit highest rate from these wells, as shown in Fig. 7. Rates on Wells 2
and dimensionless rate q1D for their normal range of b from 0 to 1. and 3 were reinitialized because of some significant downtimes.
Note that q1D is defined as the dimensionless rate at tD=1, not the Fig. 7a shows that the flow regime varies from bilinear to linear
same as normally based on qi. The results show that Arps’ curve of flow. Generally, wells with bilinear flow regime will have lower
b=0.5 is equivalent to the new curve of m=1.96 at late time (tD>10), maximum flow rate compared to wells with linear flow. The lower
as shown in Fig. 6. For b>0.5, Arps curves give more-optimistic deliverability rates and/or bilinear flow may be caused by liquid
forecasts compared to the new approach. holdup in fractures or by fracture choking.
Values of a and m obtained from a retrograde shale gas have
Field Examples covered a large range of values; these vary from 1.02 to 1.67
In this section, several field examples from different unconven- and from 1.04 to 1.21, respectively, as shown in Fig. 7b for this
tional gas plays were used to test the new approach. Examples were example.
selected to illustrate the decline behaviors in various completion
and production scenarios such as the cases of retrograde gas, newly Bakken Oil. An abandoned vertical well that produced from the
developing or developed plays, and vertical or horizontal wells. For Bakken formation in the Viewfield pool, Saskatchewan, Canada,
changes in flow regime, a and m values will be highlighted for each for more than 26 years is shown here as an example for oil pro-
example in the following discussion. A comparison of results is duction from a fractured shale or tight formation. Fig. 8a shows
also made with other methods. the well was under a transient linear flow (half slope) for all of its
production life except for a few months at the start of production.
Retrograde-Gas Case. Wet-shale-gas plays have become more The late-time flow that indicates a big rate drop in oil rate can
desirable lately because of higher liquid contents and attractive be mistaken as BDF, but in reality it is caused by an unexpected
liquid price. Unfortunately, very little is known about their flow increase in water production.
characteristics and how the liquid dropout in the formation and The normalized decline-trend plots for the well are shown in
fracture network behaves. Fig. 8b. If q/Np is based on the total liquid produced, the plot gives
The following example contains production data from four a good fitted straight line (R2 of 0.9824) with a and m values of
horizontal wells from a new emerging play. The production rates 1.7894 and 1.1477, respectively. The values are slightly higher if
from both gas and liquid were recombined and converted into only the oil stream is used in the calculation. Thus, an increase
1
1
¼ slope ½ slope
Normalized Rates
0.1
q/G p , 1/day
y = 1.4112x –1.1121
0.1 Well 1 y = 1.0164x –1.0409
Well 2
Well 3 Well 1
0.01 Well 2
Well 4 Well 3 y = 1.18x–1.1128
Well 4 –1.2139
y = 1.6655x
0.01 0.001
1 10 100 1,000 1 10 100 1,000
Days Days
(a) (b)
Fig. 7—A condensate/shale-gas example. Flow regimes that change from linear to bilinear in this retrograde play may be caused
by liquid dropout in the fracture. There is a wider range for a and m observed in this play compared to other types.
Liquid Rate, m3 /d
0.1
y = 1.8316x –1.1488
10
R 2 = 0.9963
q/N p , 1/day
0.01 y = 1.7894x –1.1477
R 2 = 0.9824
Oil+Water 0.001
1
Oil Only
0.0001 Oil+Water
Oil Only
0 0.00001
1 10 100 1,000 10,000 1 10 100 1,000 10,000
(a) Days (b) Days
Fig. 8—An oil example of fractured shale with 26 years of monthly production data. Fracture-dominated flow regimes lasted for
all of its production life. Rate drop at late time is caused by an increase in water production. (a) Linear flow lasted more than 26
years (9,500 days). (b) Normalized decline-trend plots for different fluid streams.
in water production will cause both a and m to increase in value used to obtain the representative a and m parameters for the well.
(Fig. 8b). Because of the unexpected increase in water production The R2 value is used to determine the best fit of the data. An R2
at the late time, an overprediction of only 8% in the well’s EUR value of more than 0.95 is recommended.
is projected. 3. Step 3: Rate Forecast. q1 Determination. To obtain q1, gas-
flow rate should be plotted against t(a, m). From Eq. 5, we have
Individual-Well Analysis. In this example, a step-by-step pro-
cedure is shown in Fig. 9 of how to perform decline analysis on
an individual-well basis. A vertical well taken from the Barnett q = q1t (a, m), . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (9)
shale has been selected for this illustration. The well has been in
production for the last 5 years and operated with minimum water a
(t1− m −1)
production. where t (a, m) = t − m e 1− m . The q-vs.-t(a,m) plot should give
1. Step 1: Data Check and Correction. The production-data .
a straight line through the origin with slope of q1. However, this
histories such as flowing WHPs, gas rate, and water rate are plot- may not be the best selection in some cases because of the current
ted as shown in Fig. 9a. The well was initially choked back at the wellbore operating conditions. In such cases, Eq. 9 becomes
surface on the basis of the observed flat rate and high WHPs. A
correction was made for the rate on the basis of average operating
pressures. The wet-gas rate should also be corrected to the dry-gas q = q1t (a, m) + q∞, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (10)
rate equivalent if there is a high condensate/gas ratio.
2. Step 2: a and m Determination. A log-log plot of q/Gp (or where q∞ is the rate at infinite time, so it can be zero, positive, or
q/Np for oil) is constructed to determine these values, as shown in negative. In either case, the abandonment rate should be set on the
Fig. 9b. Data are examined to determine which section should be basis of economics.
0.1 R 2 = 0.9744
WHP, MPa
q/G p , 1/day
Skin effect
10 10 0.01 a =1.4105–1 d
m =1.1384
Skin effect and 0.001
Surface choking
1 1 0.0001
1 10 100 1,000 10,000 1 10 100 1,000 10,000
Days Days
(a) (b)
m3/d
Rate Data
or Cum.
30 Matched Rate
y = 35.18x – 0.5207 Cum Gas
Cal. Cum Gas
20 10
m3/d
10 = 33.7x103
q1y=35.18 m3/d
q =–0.52 103 m3/d
103
0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1
Rate
q/Gp, 1/day
100
a =1.1142
0.01 m =1.0895 y = 140.35x – 0.4779
50
0.001 q 1=140 103 m3d
q∞=–0.48 103
0.0001 0
1 10 100 1,000 10,000 0 0.2 0.4 0.6 0.8 1
(a) Days (b) t(a,m)
Power-Law Exp
1 0.1
1 10 100 1,000 10,000 100,000 1 10 100 1,000 10,000 100,000
(c) Days (d) Days
Fig. 10—Result comparison with power-law-exponential method and Arps hyperbolic with b=1.9.
Once q1 (and q∞) have been determined, gas-rate forecasts can This analog is in the form of type curves (P10, P50, and P90),
be made by using Eq. 9 or 10. which were generated by the new technique.
⎛ q ⎞
4. Step 4: Reserves Estimate. Eq. 4 ⎜ G p = t m ⎟ is used for A set of 25 wells that includes eight horizontal wells and 17
⎝ a ⎠
cumulative calculation with q>qeco, whereas EUR is estimated by vertical wells from the Barnett shale was selected for this study.
Eq. 11: In this study, there is not much difference between the vertical and
horizontal wells except for their initial rate. For this reason, no
q m attempt was made to separate them into two different data sets.
EUR = eco teco, . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (11)
a The following steps are proposed to establish the type curves:
where qeco is the minimum economic rate, and teco is at q= qeco. 1. Establish a and m for each individual well. a and m were
determined for all 25 wells, as shown in Fig. 10b.
Analysis Comparisons. To compare the results, other meth- 2. Construct a probability plot to obtain the P10, P50, and P90
ods such as Arps’ hyperbolic with b greater than unity and the values for m (Fig. 11a).
recently developed power-law-exponential method (Ilk et al. 2009) 3. Establish an a–m relationship such (as shown in Fig. 3) if
were selected. Texas A&M University has developed a rate/time sufficient data are available. Otherwise, use the correlation from
spreadsheet on using the power-law-exponential approach (Ilk that figure.
et al. 2008, 2009). Data from Example 3 from this spreadsheet was 4. Estimate the best 3-month rate averages for both vertical
employed here for the purpose of comparison. The analysis results and horizontal wells. The best 3-month rate averages are 0.51×106
are shown in Fig. 10. The results from the power-law-exponential m3/month (18 MMscf/month) and 1.10×106 m3/month (39 MMscf/
approach and Arps’ hyperbolic with b of 1.9 are included in Fig. 10 month) for the vertical and horizontal wells, respectively. For com-
for comparison. parison, Frantz et al. (2005) obtain the best 3-month gas averages
The reserves that were calculated on the basis of time limit of of 0.54×106 m3/month (19 MMscf/month) and 1.25×106 m3/month
30 years and economical rate of 2.83×103 m3/d (100 Mscf/D) are (44 MMscf/month) that are based on a data sample of 2,842 verti-
summarized in Table 1. cal wells and 246 horizontal wells from the Barnett shale.
Table 1 shows that the new approach provides conservative 5. Use Eq. B-5 to plot q3ma against Gp for various values of a and
reserves estimate compared with the other two methods. m to construct Fig. 12 for new well EUR prediction. Fig. 12a is for
the EUR prediction based on the time limit of 30 years regardless
Analog Type Curves. A field example was used to illustrate how of flow rate, whereas Fig. 12b is based on 30 years of production
a developing play can be used as an analog for other similar plays. with a cutoff rate of 2.83×103 m3/d (100 Mscf/D).
6. Use Eq. B-3 to estimate the q1 values (P10, P50, and P90)
based on the q3ma for both vertical and horizontal wells. The results
for this example are shown in Table 2. The type curves, as shown
TABLE 1—RESERVES COMPARISON BETWEEN METHODS in Fig. 13, were constructed on the basis of these parameters.
Time limit 30 Years
Rate limit 2.83
3
10 m /d
3 Conventional Tight Gas Wells. All examples illustrated thus far
Time limit Rate limit
result in an m value greater than 1. However, m can be less than 1,
as shown in Fig. 3. The cases in which m is less than unity in Fig.
Gp, 10 m Gp, 10 m
6 3 6 3
Model 3 are from a shallow and tight gas formation. A typical rate-decline
Hyperbolic (b=1.9) 118 222 analysis is shown in Fig. 14.
Power-law 116 188
Notice the slightly concave-up shape of the rate plot, as shown
exponential in Fig. 14d. This concave-up shape, which is caused by a late time
pseudoradial flow is characteristic for m<1. For m>1, a log-log
New approach 106 141
plot of rate vs. time will display a concave-down feature (Fig. 4).
q/Gp, 1/day
P10 1.00 1.075
0.50 0.01
P50
0.001
0.10
P10
0.01 0.0001
1 10 100 1,000 10,000
1 1.05 1.1 1.15 1.2 1.25 1.3 Days
m
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 19 20 21 22 23
25 24 26 27 28
(a) (b)
Fig. 11—A statistical approach to analyze production forecasts for a group of wells or a set of sample wells: (a) a probability
plot for 25 wells and (b) a and m for all 25 wells.
EUR RANGES FOR BARNETT (30 YEARS) EUR RANGES FOR BARNETT (2.83 103 m3/d)
2.0 2.0
1.0 1.0
Rate limit with 30 years
0.5 0.5
Time limit only EUR shifted because
0.0 of rate cutoff
0.0
0 20 40 60 80 100 0 20 40 60 80 100
EUR, E6 m 3 EUR, E6 m3
(a) (b)
Fig. 12—EUR prediction based on time- and rate-limit constraints. Time limit would only give an optimistic prediction for EUR
if the initial best rate is low.
This feature distinguishes flow characteristics of a conventional its fluid gradients. As in the case of a gas-saturated formation, the
(late-time pseudoradial flow with m<1) vs. unconventional (lack pressure gradient should be approximately 2 kPa/m. A hydrauli-
of late-time pseudoradial flow or m>1) reservoir. cally fractured well located in such a formation will be expected to
pass through several distinct flow regimes such as fracture linear,
bilinear, elliptical, and pseudoradial flows, and BDF, as discussed
Discussion by Clarkson and Beierle (2010). An unconventional reservoir may
Absence of Late-Time Pseudoradial Flow. Unsurprisingly, the not be working according to such a model because there is no
m value for all shale-gas wells investigated in this study is greater matrix transient flow here to give rise to radial or elliptical flow.
than unity, or a log-log plot of rate vs. time will display a concave- If the reservoir is shale, the pressure gradient will be overpres-
down feature (Fig. 4). This flow characteristic, which indicates an sured. This can be up to 63 kPa/m (3 psi/ft), as recorded in the
absence of the late-time pseudoradial flow from shale gas reser- Lower Bossier play (Newsham and Rushing 2002) and shown
voirs, can be explained as follows. in Fig. 16. Spencer (1987) shows some pressure gradients in
Fig. 15 was constructed to distinguish between a conventional overpressured and gas-bearing shale plays in the Rocky Mountain
(tight) and shale-gas formation with an assumption that Point A, region. The interpreted pressure profile for the MWX site wells of
the wellbore location, has the same pressure in both conventional the Piceance basin shows that the pressure gradient based on mea-
and unconventional formations. If Well A is located in a tight but sured pore pressures in the active zone is approximately 36 kPa/m
conventional formation, its pressure will be in equilibrium with (1.7 psi/ft). The author of this study has also experienced pressure
3 0.51 1.10
m /month
P90 P50 P10 P90 P50 P10
a 1.88 1.42 1.00 1.88 1.42 1.00
m 1.21 1.14 1.07 1.21 1.14 1.07
q1, 10
3
12.19 19.09 32.58 26.40 41.37 70.59
3
m /d
Fig. 13—Type curves for Barnett shale based on a set of 25 wells, for both vertical and horizontal wells.
Pressure, MPa
3 Pressure 3 0.1
q/Gp
2 2 0.01
q1 Determination Results
2 10
Gas Rate, 103 m3/d
y = 5.3375x – 0.0424
1.5
y = 5.1312x 1 Actual Rate
1 Cal. Rate 1/4 slope
Actual Cum.
0.1 Cal. Cum
0.5
0 0.01
0 0.1 0.2 0.3 0.4 0.5 1 10 100 1,000 10,000
t(a,m) Days
(c) (d)
Fig. 14—A rate-decline example from a shallow and tight gas formation.
distributions across producing gas shales exceeding the lithostatic from this shale formation, these tiny reservoirs need to be con-
gradient, as in the Montney and other studied plays. nected by a fracture network, without which they will remain
With this high pressure gradient in shale gas, the formation is isolated and unproductive.
thought to be in a dynamical equilibrium state in which hydro- Because of such isolated characteristics of microreservoirs in
carbons have been generated and leaked off the formation over shale-gas formations, capillary pressure or relative permeability
geologic time. Under a normal production time scale, shale-gas curves should be applicable in each individual tiny reservoir, but
formations can be treated as impermeable barriers that contain they may not be applicable applied across the entire shale forma-
millions and millions of tiny isolated reservoirs (such as sandy tion. This can explain why the late-time radial- or elliptical-flow
laminates, microcracks, or isolated sand bodies); to produce gas regimes are absent from all shale-gas wells studied.
3,000 3
DEPTH (X 1,000)
4,000 0.7% RO
1.0
≈ 63 kPa/m pressure
5,000 4 0.8% RO ≈ 36 kPa/m pressure
distribution (acoustic
6,000 distribution (measured
0.8
0 .6
log response)
PS
5 1.5
PS
7,000 pore pressure)
I/F
I/F
T
165°F (74°C)
T
8,000
6 INACTIVE ZONE
9,000
2.0
10,000 200°F (93°C)
4
0.4
11,000 ACTIVE ZONE
33
12,000
PSI/
8 2.5
FT
13,000
10 20 30 40 50 60 kPa
9
2 3 4 5 6 7 8 9 PSI
0 2,000 4,000 6,000 8,000 10,000 12,000 0 120 200 50
Pressure, psig 1,000/div GR Acoustic PRESSURE (X 1,000)
(a) (b)
Fig. 16—Two examples of high pressure-gradient distribution across the producing shale gas: (a) the Lower Bossier play (New-
sham and Rushing 2002) and (b) the Piceance basin (Spencer 1987).
is not considered in the study but can be included in the model. curve-analysis methods. To use q/Gp, both rate and cumulative
Fig. 17 is a summary of the results in the form of normalized production should be initialized with pressure correction. Fig. 7
decline-trend plots. The q/Gp ratio of the proposed fracture model shows the example plots of rate initializations for Wells 2 and 3.
is shown as the blue line. If the stimulated region is much smaller, EUR Constraints. In most cases, without the presence of BDF
the red or BDF curve proposed by Bello and Wattenbarger (2008) in shale reservoirs, drainage area cannot be established for EUR
is obtained. prediction. In this method, the drainage area will be defined only
To compare with the proposed analytical method, the thin when the fracture-expansion process is brought to an end. Thus,
white line with the matched a and m factors of 0.7785 and 1.0481, EUR is not based on the traditional concept of drainage area
respectively, is produced. This line lies exactly on top of the blue (BDF) but on the constraints of the latest trends of (q1, q∞) with
line from the numerical simulation result. both time and economic rate limits. Because of this, results are
The yellow curve shows one of the results of a normal reser- valid only where analog wells are similar in both interwell and
voir model where the virgin area surrounding the SRV is also in fracture spacings.
pressure communication. This yellow curve, which includes a late- In case of spacing differences, a numerical reservoir modeling
time linear flow from the virgin region into the SRV, can also be (see the Numerical Model Comparison subsection) should be used
obtained using the approach from Anderson et al. (2010). Another for any EUR adjustment.
result is shown as a green line where both stimulated and virgin Latest Trends. During a well production, many factors can
regions are the same or under a homogeneous state. Pseudoradial contribute to the deviation from a fracture flow to accelerate the
flow will be achievable to make m a value of less than unity. Fig. decline rate such as water production or liquid dropout/holdup in
14 illustrates a field example of a gas well producing from a tight the fracture system and wellbore. The values of q1 and q∞ (Eq. 10
reservoir. and Fig. 9) are designed to account for this acceleration in rate
decline in late time. An increase in water production will also cause
Method Limitations. Downtime Effect. Because of some opera- both a and m to increase in value (Fig. 8).
tional conditions, a well may have to be shut in for some periods Time Limit vs. Rate Limit. Fig. 12 shows a typical EUR predic-
of time during its production life. If the downtime of each shut-in tion for an infill in the Barnett shale for both vertical and horizontal
is significant, rate initialization is required as in traditional decline- wells. The prediction based on a time limit alone would give an
0.01
Anderson et al. 2010
1.E-04
Contribution from
0.001 Undisturbed Shale
Fig. 17—Normalized decline trend plots for various simulated fracture models in shale or tight reservoirs: (a) for all production
data and (b) at late time.
( )
presented at the SPE Hydraulic Fracturing Technology Conference, m
m−a
1− m
College Station, Texas, USA, 29–31 January. doi: 10.2118/106043- qmax = q1 a e 1− m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-2)
MS. m
Appendix A—General Equations for q and Gp Eq. B-1 was obtained by allowing the rate derivative of Eq. 5,
dq/dt, to approach zero.
⎡ ⎤
⎢ q = ε (t ) ⎥⎥ .
General equations are developed on the basis of ⎢
⎢⎣ G p ⎥⎦
Best 3-Month Average, q3ma. Since tmax is less than 45 days (Fig.
q 4), the best 3-month average (q3ma) will be equal to the average of
= G p . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-1)
ε (t ) the first 3 months of production and can be calculated by
Taking a derivative with respect to time, we have
q1 1−am (901− m −1)
q3 ma = e . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (B-3)
d [ q / ε (t ) ] = d G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (A-2) 3a
dt dt p
Eq. B-3 is obtained on the basis of Eq. 6 by calculating the cumula-
Using q’= dq/dt and ε’(t) = dε (t ) , Eq. A-2 becomes tive gas of the first 3 months of production. Gas rate and cumulative
dt gas based on this q3ma would be
q '/ ε (t ) − qε '(t ) / ε 2 (t ) = q , . . . . . . . . . . . . . . . . . . . . . . . . (A-3)
a
(t 1− m
− 901− m ) . . . . . . . . . . . . . . . . . . . . . . . . (B-4)
or q = 3q3 ma at − m e 1− m
and
dq / q = dε (t ) / ε (t ) + ε (t )dt . . . . . . . . . . . . . . . . . . . . . . . . (A-4)
a
(t 1− m
− 901− m ). . . . . . . . . . . . . . . . . . . . . . . . . . . (B-5)
Integrating both sides from t=1 to t, we have: G p = 3q3 mae 1− m
Duong (2011) presents a new forecasting approach using log-log Linear-Flow Plot
rate/cumulative production (q/Gp) vs. time (t) plots for wells in
unconventional reservoirs which often exhibit long-term linear y = 1.2325×104 – 8.0794×104
ðb Di Þðb1Þ=b : . . . . . . . . . . . . . . . . . . . . ð5Þ
Gas Rate, Mscf/D
A¼
ð1 bÞDi EUR = 0.6 Bscf
and
qi
B¼ : . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð6Þ
ð1 bÞDi
Therefore, we can construct a Cartesian plot of Gp against tn
where n ¼ (b–1)/b. This allows easily fitting a straight line which
corresponds to Slope A. For the special case of b ¼ 2, n ¼ 0.5
which is the classic linear-flow plot with Gp plotted against the
square root of time. For bilinear flow with b ¼ 4, n ¼ 0.75. 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7
t 0.091, days
Copyright V
C 2014 Society of Petroleum Engineers Fig. 2—Field example of b51.1 plot (time0.091).