Professional Documents
Culture Documents
Abstract
Numerous wells in tight gas reservoirs have long-term performance, which exhibit only
linear flow. The decline curves for these wells do not exhibit pseudo-radial flow when the outer
boundary effects are detected. The objective of this work is focused on decline curve analysis
methods for long-term linear flow in fractured tight gas wells as well as application of linear
analysis, curve fitting by using non-linear optimization and numerical simulation models suitable
for analysis of tight gas reservoirs.
In many tight gas wells, long-term linear flow occurs during depletion. This is not only
caused by a hydraulic fracture, but is probably caused by other factors such as anisotropy, high
permeability streaks (layered reservoirs), and naturally fractured reservoirs. With the several
causes of long-term linear flow in tight gas wells, a wrong idea and misinterpretation of
production and pressure data may originate.
The analysis of the behavior show that the usual transient well tests equations should be
used because they usually apply to constant rate conditions. Long-term production is more nearly
a constant pressure condition (assumed with decline curve analysis). The constant pressure
equations are presented here and compared with the constant rate equations. Type curves are
shown for the two cases.
A simple analysis can be made by finding the slope of a m( pi ) m( pwf ) / qg vs. t plot
and detecting where the outer boundary effect occurs. With this information, values for k xe ,
drainage area, Vp and OGIP can be estimated without knowing permeability, k, or thickness, h.
This direct calculation (without k and h) is seems to be a coincidence of linear flow and is not
possible for other flow geometries. Example calculations are shown in actual field case.
Introduction
Tight gas wells are present in reservoirs where the reservoir rock has the property of low
permeability (k < 0.1 md). Linear flow regime has been reported to dominate in numerous
situations. It is especially an important flow regime associated with production from tight gas
reservoirs. In many tight gas wells, long-term linear flow occurs during depletion1.
Linear flow is characterized by half slope line when pressure or reciprocal of production
data are plotted versus time on log-log plot.
Long-term linear flow in tight gas reservoirs may be caused by anisotropy, naturally
fractures (or “dual-porosity”) reservoirs, linear or elongated reservoirs, high permeability streaks,
and hydraulic fracture (vertical wells intercepted by vertical, horizontal, or diagonal fractures,
and horizontal wells with fractures).
kf w
FCD (1)
kx f
Under these conditions, the flow is linear and is perpendicular to the hydraulic fracture.
The flow geometry of Fig. 1 is simpler than most of the fractured well cases in the
petroleum literature. The technical literature has many complicated cases for fractured wells4-10.
The solutions for these cases are presented as type curves, tables, or sometimes as equations.
Usually the cases involve fractures in infinite reservoirs, which means that flow eventually
leaves the linear regime and then becomes pseudo-radial. Often the fracture is treated as having
limited conductivity. In addition, almost all of the type curves are for constant rate flow. Very
few of these type curves are for constant (wellbore) pressure, which is of interest for long-term
production analysis.
Apparently, the only solution published for the Fig. 1 problem is for a square with
constant flow rate11. This is a special case of a more general set of solutions. No solutions were
found for the rectangular case or the constant pressure case before 1998. The reason for this is
probably that the majority of the pressure analysis literature is directed toward pressure buildup
applications rather than long term performance. However, Wattenbarger et al.1,3 have adapted the
solutions of Miller12 and Nabor and Barham13 for fractured wells in the geometry of Fig. 1. They
included both the constant rate and constant pressure cases for linear flow in a rectangle. Their
equations provide tools for analyzing long term performance of tight gas wells.
The objectives of this work are focused on decline curve analysis methods for long-term
linear flow in fractured tight gas wells as well as application of linear analysis, curve fitting by
using non-linear optimization and numerical simulation models suitable for analysis of hydraulic
fractured tight gas reservoirs.
Linear flow
ye
xe
Fig. 1 - Top view of a rectangular reservoir and a hydraulically fractured well which has only
linear flow into the fracture (xf = xe).
Literature review
The following citations from the literature support the earlier observation of the existence of
long-term linear flow in most of tight gas reservoir basins. Some articles reported that long-term
linear flow was continuous for years in a large amount of wells (50 or 60 % of wells).
Miller12 gave solutions for linear flow in aquifers. These solutions are for the infinite
acting and bounded aquifers for both the constant rate and constant pressure solutions.
Nabor and Barham13 generalized Miller’s solutions in dimensionless variables and
derived solutions for constant pressure outer boundary case. In linear heat conduction, Carslaw
and Jaeger14 present the mathematics for these solutions.
Bagnall and Ryan15 described some field cases that show linear flow trend in the
production data of Devonian shale cases. They concluded that Devonian shale gas production is
controlled largely by natural fracturing where the gas rate from such wells also is function on the
density and width of these natural fractures.
Boardman and Knutson16 recognized three kinds of behavior: linear, intermediate, and
near radial from the plots of cumulative gas production versus time of many Uinta basin wells.
However, most wells exhibited near linear flow for at least five years of production. They
reported that under conditions of radial flow, slopes are expected to be 1.0 and under conditions
of linear flow, 0.5.
Hale and Evers17 used numerical simulation to achieve type curves for vertically
fractured wells producing at constant rate or constant bottom-hole flowing pressure. They
3
Complete Solutions. For linear flow into fractured wells, the following equations apply to all
times. Eq. 2 is the solution for constant rate production from a closed linear reservoir.
1 ye xe 1 ye 1 2
2 2 xe
pwD t Dx e 2 exp n t Dx e (2)
2 3 xe ye xe n 1 n ye
Eq. 3 shows the solution for constant pwf production from a closed linear reservoir.
1 ye 1
(3)
qD 4 xe n 2 2 x 2
exp
e
t
4 ye e
Dx
n odd
k h ( pi pwf )
pwD (4)
. qo Bo o
1412
1 k h ( pi pwf )
(5)
qD 1412. qo Bo o
0.00633k t
t D xe (6)
o ct xe 2
The units used are k in md, h in ft, p in psi, qo in STB/D, Bo in rb/STB, t in days, o in cp, ct in
psi-1, and xe in ft. We observe that the definitions of pwD and 1/qD seem to be equal. In the pwD
case pwf varies while in the 1/qD case qo varies.
The following solutions are simplified forms for short and late time flow.
pwD t Dx e (7)
Eq. 7 is the linear flow (half slope) solution for tDxe < 0.01 with flow into both sides of the
fracture. Eq. 8 is the infinite-acting linear solution for constant pressure from a closed linear
reservoir.
1
t Dxe (8)
qD 2
Where the dimensionless variables are shown from Eqs. 4 to 6. Both constant rate and constant
pwf cases have different infinite-acting solutions that differ by a factor of /2. However, the
equivalent two equations are identical for the pseudo radial flow solution in the semi-log
straight-line region defined by Eq. 9.
1 1
pwD ln(t D ) 0.4045 (9)
qD 2
Long-term or late time linear flow approximations. The outer boundary at ye is reach at long
times. Later, the well behaves as either boundary dominated flow for the constant rate (pseudo-
steady state) or outer boundary-dominated flow for the constant pwf case. Eqs. 10 and 11 are the
approximations for closed reservoir for constant rate and constant pressure respectively, in
dimensionless forms, and referenced the initial reservoir pressure.
xe y
pwD t Dxe e (10)
2 ye 6 xe
and
1 ye 2 xe
2
exp t Dx e (11)
qD 4 xe 4 ye
Both Eqs. 10 and 11 can be expressed as stabilized flow and a productivity index
equation would be developed. The stabilized production rate, in terms of average pressure, p ,
for constant rate in filed units is expressed as:
kh( p pwf )
qo (12)
y
141.2 Bo o e
6 xe
6
kh( p pwf )
qo (13)
2 y
141.2 Bo o e
xe
The most important use of the long-term approximations is in the form of productivity index
equations. The productivity index equation is expressed as:
q
J (14)
( p pwf )
If the drainage volumes are known, average pressure, p , can be estimated by material balance
calculations.
Expressions 12 and 13 would be represented in different equations for productivity index.
Substituting Eqs. 12 and 13 into Eq. 14 give us Eqs. 15 and 16 showing the expressions for
constant rate and constant pressure production, respectively.
kh
J CR (15)
y
141.2 Bo o e
6 xe
and
kh
J CP (16)
2 y
141.2 Bo o e
xe
The pressure profiles versus distance, x, for Eqs. 12 and 13 are not equal. The constant
rate case (Eq. 12) has a parabola shape for these late times. At various long times these profiles
are parallel, typical of pseudo-steady state. On the other hand, the constant pwf case (Eq. 13) has a
shape that is proportional to sin(x/2ye) function at late times14. These different shapes result in
different relationships between ( p -pwf) and q, as pointed by the difference in Eqs. 12 and 13.
We can evaluate the ratio of productivity index for constant rate production, JCR, to the
productivity index for constant pressure, JCP. This ratio of productivity is estimated as:
J CR 12
1.216 (17)
J CP 2
Type curve analysis. Wattenbarger et al.3 developed a practical way to use these solutions
through log-log type curves. If in complete solutions for constant rate (Eq. 2) and for constant
pressure (Eq. 3) we use tDye rather than tDxe we can develop, the following solutions for closed
reservoirs for constant and constant pressure, respectively.
xe 1
pwD t Dy e 1
1
2 3
n 2
exp n 2 2t Dy e (18)
ye n 1
and
xe 1 1
(19)
y q 4
n 2 2
e D
exp t Dye
n odd 4
2
0.00633k t xe
tD ye t Dxe (20)
ct ye 2 ye
2
0.00633k t ye
t D xe t Dye (21)
ct xe 2 xe
Fig. 2 shows a type curve for both constant rate and constant pressure cases for all times
of significance (complete solution). This plot is done by using Eqs. 18 and 19, in which we can
plot tDye versus (xe / ye)pwD and (xe / ye)(1/qD) rather than pwD and 1/qD. This procedure gives us
only one curve for each case, for any rectangular geometry, rather than series of curves.
In the constant rate case, Eq. 21 can be substituted in both the early infinite-acting
solution (Eq. 7) and the later outer boundary-dominated solution (Eq. 10) to obtain Eqs. 22 and
23, respectively.
xe
pwD t Dy e (22)
ye
1000
Exponential
100
(xe/ye) pwD or (xe/ye)(1/qD)
Constant Pressure
Drawdown Slope = 1
10
(tDye)ehs = 1/4
Constant Rate
Drawdown
1
(tDye)ehs =1/2
0.1
Slopes = 1/2
0.01
0.001 0.01 0.1 1 10 100
tDye
Fig. 2 – This plot shows linear flow type curve for constant production rate and constant pressure
for closed reservoirs.
Similarly, for constant pressure case Eq. 21 is substituted in both the early infinite-acting
solution (Eq. 8) and the later outer boundary-dominated solution (Eq. 11) to obtain Eqs. 24 and
25, respectively
xe 1
t Dy e (24)
e D
y q 2
xe 1 2
exp t Dye (25)
ye q D 4 4
We can see that Fig. 2 shows the equation forms of the short and long term solutions
(Eqs. 22 to 25). Figs. 3 and 4 show type curves for both constant rate (Eqs. 2, 7, and 10) and
constant pressure drawdown (Eqs. 3, 8, and 11), respectively. These type curves are drawn for
tDxe versus selected values of (xe / ye) for the two cases.
In the log-log plot of difference in pressure or reciprocal of production rate versus time
(type curves in Figs. 2 to 4) we can see an important particularity that they all bend upward when
the outer boundary is reached. This behavior is the contrary of the pseudo-radial flow when the
9
100
10
pwD
Slope = 1/2
0.1
0.1 1 10 100 1000
tDxe
Fig. 3 – Log-log plot that shows linear flow type curve for constant production rate and closed
reservoirs.
End of the Half-Slope. A significant particularity of the type curves in Fig. 2 is the end of the
half-slope straight line or end of the infinite acting linear flow. In analysis of field data, this end
of the half slope would be used to evaluate reservoir magnitude. For the constant rate case the
end of the half slope is (tDye)ehs = ½ and for the constant pressure case is (tDye)ehs = ¼. Both
numerical values are taken to be values where the curves visibly leave from the half slope
straight line. These values are useful for the objective of founding minimum drainage areas from
actual data.
Decline curve. Fig. 5 shows a log-log decline curve for the constant pressure production solution
evaluated by using Eq. 19. These values of qD / xe ye are reciprocal of the values that were
shown in Fig. 2. Easily we can use this type curve for long-term performance where pwf is
constant throughout the observed history. In comparison with other plots (Figs. 2 to 4), this plot
is more useful as a qualitative tool.
10
p
p
m( p ) 2 dp (26)
po
z
To obtain the gas solutions pwD is replaced by mwD in the liquid solutions. The mwD parameter is
defined as follows.
k h [m( pi ) m( pwf )]
mwD (27)
1424 q g T
100
10
1/qD
Slope = 1/2
0.1
0.01 0.1 1 10 100
tDxe
Fig. 4 – This plot shows linear flow type curve for constant pressure drawdown and closed
reservoirs.
1 k h [m( pi ) m( pwf )]
(28)
qD 1424 qgT
11
0.00633k t
t D xe (29)
(ct )i xe
2
0.00633k t
tD ye (30)
( ct )i ye
2
In gas fluid for transient flow, the liquid formulations can be applied very accurately.
These liquid solutions have been previously shown and are plotted in the type curves. This has
been determined for a number of well flow problems over the years3.
100
Slopes = 1/2
10
Constant Pressure
Drawdown
1
qD/ (xe/ye)
(tDye)ehs = 1/4
0.1
0.01
Exponential
0.001
0.001 0.01 0.1 1 10
tDye
Fig. 5 – This plot shows a log-log decline curve for constant pressure production and closed
linear reservoirs.
Therefore, for the infinite-acting t performance Eqs. 26 to 30 behave very well. However,
these solutions do not match very well after the outer boundary effects have been felt for a long-
term period due to changes in gas properties. Fig. 6 shows the comparison of simulating gas well
behavior and the analytic liquid solutions for a long-term period. In this plot we can see that due
of fluid property variation the gas case begins to depart from the liquid solution at a
dimensionless time near of 5.0. This departure occurs at about ten times the time when the outer
boundary effects where first seen. Wattenbarger2 explains that the actual time of departure is
function on the gas properties of the case studied. If these solutions (Eqs. 26 to 30) are used in
12
100
10
(xf /ye) pD
0.1
0.01 0.1 1 10 100
tDye
Linear analysis.
Log-log decline curve plot. The first stage is to make a log-log plot of either for
constant rate production a plot of pseudo-pressure difference divided by rate,
m( pi ) m( pwf ) / qg , versus time, t, or for constant pressure production a plot of decline curve of
gas rate, qg, or m( p ) m( p )/ q
i wf versus time, t. This plot is recommended particularly if a
g
number of wells are being investigated. This is a simple plot and will show indications of linear
flow if the plot has a “half slope”. Of course, the analysis of this plot assumes that pwf is constant
for the constant pressure production case, and that qg is constant for the constant rate production.
Many wells may have periods where the well pressure varies or the well is shut in
because of market contracts, curtailment, etc. For these arguments, the half-slope may not
13
Fig. 7 - Semi-log decline curve that shows gas rate versus time in years for well J-7.
We can see for this well we need additional analysis and finally we need to use numerical
reservoir simulation to confirm linear flow behavior. This is necessary because the variation of
production rates and possible variation of pressures must have some effect on the decline curve.
In this case, nonetheless, the log-log decline curve seems to be a good and easy analysis tool for
well J-7.
Specialized plot of m( pi ) m( pwf ) / qg vs. t . The square-root of time, t , plot is a
specialized and more definitive plot to identify the linear flow behavior. A straight line of the
14
Fig. 8 - Log-log diagnostic plot that shows gas production data versus time for well J-7 (decline
curve).
200.8107 T
mCR (31)
(ct )i k xe h
315.4327 T
mCP (32)
(ct )i k xe h
Fig. 9 shows a plot of qg and m( pi ) m( pwf ) / qg versus t and Fig. 10 is a plot of both
m( p ) m( p )/ q
i wf g and Gp versus t , both plots are for a period of 24 years. We can see in
15
1.E+04 3.E+07
2
1.E+02 1.E+07
1.E+01 0.E+00
0 10 20 30 40 50 60 70 80 90 100
1/2 1/2
time (days )
Fig. 9 – Plot of qg and m( pi ) m( pwf ) / qg vs. t for well J-7 that gives a slope
mCP = 2.2x105 D1/2psia2/Mscf-cp
200.8107 T
k xe (33)
(ct )i mCR h
2
40,324.94 T
kx 2
ct i mCRh
e (34)
803.2427 T
k AC (35)
(ct )i mCR
16
3.E+07 2,000
[m(pi)-m(pwf)]/qg
Gp
[m(pi)-m(pwf)]/qg (psia -D/Mscf-cp)
1,500
2.E+07
Gp (MMscf)
2
1,000
1.E+07
6 2 1/2
mcp = 22x10 (psia -D )/(Mscf-cp) 500
0.E+00 0
0 10 20 30 40 50 60 70 80 90 100
1/2 1/2
time (days )
Fig. 10 – Specialized plot that shows Gp and m( pi ) m( pwf ) / qg vs. t for well J-7.
315.4327 T
k xe (36)
(ct )i mCP h
2
99,497.76 T
kx
2
ct i mCPh
e (37)
1261.7308 T
k AC (38)
(ct )i mCP
Where in Eqs. 35 and 38, AC is the cross-section area to flow and it is given as follows:
AC 4 xe h (39)
17
k tehs
ye 0.1125 (40)
(ct )i
k tehs
ye 0.1591 (41)
(ct )i
The boundary distance evaluated with Eqs. 40 and 41 can be considered minimum values
if all the history data is still on the half-slope trend. If our well performance is still on the half-
slope trend, the latest production time is used instead of tehs in Eqs. 40 and 41. This would be the
distance of investigation at the current time. This expression requires that permeability be
known. Of course, this may be not a strong point in estimating ye by this procedure since
permeability may be uncertain. Because in Eqs. 33 to 38 is difficult to separate k form xe and AC.
Thus, the value of ye is function on the value selected for xe and AC.
Calculation of drainage area. We can use Eq. 40 or 41 and substitute for k in Eq. 35 or
38, respectively. The advantage is that we can develop Eqs. 42 and 43 to calculate directly the
drainage area for constant rate and constant pressure, respectively.
90.3648 T
A 4 xe ye tehs (42)
(ct )i mCR h
200.7699 T
A 4 xe ye tehs (43)
(ct )i mCP h
Where in Eqs. 42 and 43, A is the drainage area and it is given as follows:
A 4 xe ye (44)
18
Calculation of pore volume, Vp. We can get a direct determination Vp, if we multiply
Eqs. 42 or 43 by porosity, , and height, h. This may be a good advantage, since and h are
often not known in tight gas reservoirs. The Vp equations for constant rate and constant pressure
are, respectively:
90.3648 T
Vp tehs (45)
( ct )i mCR
200.7699 T
Vp tehs (46)
( ct )i mCP
V p Ah (47)
Calculation of original gas in place, OGIP. Once the pore volume is calculated, then
OGIP can be directly evaluated using Eq. 48.
V p (1 S w )
OGIP (48)
B gi
The evaluation of OGIP is insensitive to the value of water saturation, Sw, used, if gas
compressibility, cg, dominates. We can estimate a precise value of OGIP without actually having
good knowledge of permeability, k, thickness, h, water saturation, Sw, and porosity, .
History matching by using reservoir simulation. The linear analysis and curve fitting analysis
should be accurate if all the assumed conditions are completely met. In the example well
presented here, the reservoir-fracture system seems to behave totally linear flow. This means that
pseudo-radial flow regime was not identified. If either constant rate or constant flowing pressure
19
Example Application
Case 1. This example assumes a well with hydraulic fracture and infinite fracture
conductivity in a rectangular reservoir (Fig. 1). .
The case shows the application of the procedure described above for well J-7, which is a
tight gas well in a field in South Texas. The well has been producing for more than 22 years
without hydraulic fracture. Gas rates in monthly basis are the only production data available
among some reservoir and fluid properties. Reservoir and fluid data are given in Table 1.
Fig. 7 displays a semi-log decline curve of gas rate versus time for well J-7. The
fluctuations in the production rate were caused by shut-ins. Unfortunately, we do not have much
data about those shut-in periods.
Fig. 8 shows a log-log decline curve for this well. The gas production rate follows a
negative “half slope” behavior for most of its life. For this well this negative half-slope is a
confirmation that linear flow existed for long-term. We can see that the data starts bending down
after about fifteen years. The well found the outer boundary at the distance ye at this time. The
estimated value for tehs is 5,625 days. The production data for well J-7 after this time bend
downward due to boundary effects.
For this well we need additional analysis and finally we need to use numerical reservoir
simulation to confirm linear flow behavior. This is necessary because the variation of production
20
Or,
k xe 26.801 md 1 / 2 ft
And,
kxe2 718.2927 md ft 2
Or,
A 2,778,270.90 ft 2
This drainage area is equal to 63.78 acres (2,778,270.9/43,560). This is the minimum area of
drainage that was established at 5,625 days. According to this procedure, this means that in 15.4
years, that drainage area has been depleted.
Next, the pore volume is evaluated by using either Eq. 46 or Eq. 47.
21
Or,
However, OGIP can be evaluated directly without knowing k and h by using Eq. 46. This is a
real benefit, since k and h values are often uncertain for tight gas wells.
In function of the above plots, it is evident that well J-7 has been exploiting with the domain of
linear flow.
Case 2. In this case, we assumed a hydraulically fractured well in a square drainage area
of 80 acres (Fig. 11). The well is located in the center of a square. From the center of the square
the xf extends to the boundary in the x direction, xe, where xf = xe. The distance to the outer
boundary in the direction perpendicular to the fracture is ye, so xe = ye the drainage area of the
well is 4 xe xe. The xe = ye distance is evaluated as:
A (80)(43,560)
xe 933.38 ft (48)
4 4
For a square drainage shape with 80 acres spacing, xf = xe = ye = 933.38 ft. The distance
between wells is calculated as 1,866.76 ft. The fracture is assumed to have infinite conductivity.
With these conditions, again the flow is linear and is perpendicular to the fracture.
The k AC product value is estimated by using Eqs. 38 as follows.
Or,
k AC 9,878.31 md 1 / 2 ft 2
AC 4(933.38)(92) 343,483.84 ft 2
22
2
k AC
2
9,878.31
k 0.000827 md (49)
AC 343,483.84
1,866.76 ft xf
Linear flow ye
xe
Fig. 11 - Areal view of a square reservoir geometry and a hydraulically fractured well which has
only linear flow into the fracture (xf = xe = ye ).
Curve fitting procedure. We used the curve fitting analysis by using non-linear optimization
procedure to determine the best value of k AC , yeAC, and OGIP parameters. Table 1 shows the
data of well J-7 that were used to determine gas properties by using correlations. Figs. 12 and 13
are semilog and log-log plots of production data versus time for well J-7, respectively.
Fig. 14 is a log-log plot of 1/qD vs. tD for the same well. All these plots show the
comparison between actual data and analytical results for constant pressure case.
By using curve fitting analysis for case 1 we found a k = 0.0006067 md, AC = 426,876.33
ft2, and L = xe = 652.98 ft. With these information the following calculated parameters were
obtained: k AC 10,514.51 md 1 / 2 ft 2 , yeAC = 278,741,706 ft3, and OGIP 7.055 Bscf .
Similarly, for case 2 with AC = 343,344 ft2 and ye = 933 ft we determined a k = 0.000876 md.
Confirmation with history matching by using reservoir simulation. In all the plots presented
here for well J-7, the tight gas reservoir seems to behave totally linear flow. This means that
pseudo-radial flow regime was not identified. However, in some tight gas wells either constant
rate or constant flowing pressure conditions can be not ideal. In some examples, the actual
producing conditions are not even close to either case constant pressure or constant rate.
23
10,000
Actual
Constant pwf Analytical
1,000
qg (Mscf/D)
100
10
0 5 10 15 20 25
time (years)
Fig. 12 – Semilog plot that shows production data vs. time showing the comparison between
actual data and analytical results for constant pressure case.
For our well J-7 example, daily average rates were input data into GASSIM3B simulator
along with other reservoir and fluid data. Bottom flowing pressure was then estimated for each
timestep. To fit the history data the reservoir-fracture data were changed.
The plot of comparing both the history and simulated [m(pi)-m(pwf)] / qg vs. t is shown
in Fig. 15 for case 1. The modified data which were used for the history match results in Fig. 15
were k = 0.000607 md, xe = 1,124 ft, and ye = 646 ft giving A = 66.7 acres. The values of k and
xe matched value of kxe2 was 766.9 md ft2. These values are slightly different from the linear,
curve fitting analyses, and may require filtering.
Fig. 16 displays history match of qg and m(p)/qg vs. t1/2 by using simulation for case 2.
To fit the history data only the permeability data was changed. The modified permeability data
which was used for the history match process was k = 0.00075 md having xe = ye = 933.38 ft,
and A = 80 acres. This value of k matched values of kxe2 of 654 md ft2 , and k AC of 9,407
md1/2 ft2. These values are different from the linear, curve fitting analyses, and may require
filtering.
Fig. 17 shows a comparison plot among actual data, constant pressure analytical and
simulation results of production data versus time for well J-7. We can see a slightly difference
24
10,000
Actual datal
Constant pwf Analytical
1,000
qg (Mscf/D)
100
10
10 100 1,000 10,000
time (days)
Fig. 13 – Log-log plot that shows the comparison between actual data and constant pressure
analytical for well J-7.
Discussion
The analysis procedure presented in this section is appropriate for many fractured low
permeability or tight gas wells. Both aspects the increase use of long hydraulic fractures and
more close well spacing will tend to achieve the conditions of this linear flow analysis procedure.
The analysis procedure is not complicated but may have comprehensive application when
real data are used. It is important to discuss several situations of its application.
Wattenbarger2 mentioned that in a current study of a field with near of 60 gas wells, long
time linear flow was detected in about one-third of these wells. He explained that large numbers
of the other wells could not be studied under this linear flow method because of critical gas rate
changes caused by gas market constraints. However, none of these wells showed pseudo-radial
flow, which is theoretically anticipated in tight gas wells with hydraulic fractures.
Limited fracture conductivity. The procedure analysis presented here has been for infinite
fracture conductivity. Recalling Eq. 1 we can note that when formation permeability is
extremely low, the fracture conductivity will behave as relatively high in most cases.
25
100
Actual
Constant pwf Analytical
10
1/qD
0
1.E-03 1.E-02 1.E-01 1.E+00
tD
Fig. 14 - Log-log plot of 1/qD vs. tD that shows the comparison between actual data and
analytical results for constant pressure case.
If the conductivity of the fracture is considerably low (FCD less than 50) then this would
look like as a skin in long-term production performance. Agarwal, et al33 analyzed this behavior
in massive hydraulic fracturing of low-permeability gas wells. In a plot of [m(pi) - m(pwf)]/qg vs.
t this type of behavior would be recognizable by a non-zero intercept on the y axis.
In our well J-7 this intercept was not detected. However, it should be accepted that the
short-time behavior was frequently interrupted by the effects of opening a new well on
production. Bilinear flow was detected in early flow of buildup tests on well J-7, implying that
the conductivity of the fracture was reduced2. Still, this bilinear flow was not detected on the
long time performance.
26
1.E+04 3.E+07
qg, History
Slope
[m(pi)-m(pwf)]/qg, History
[m(pi)-m(pwf)]/qg, Simulation 3.E+07
[m(pi)-m(pwf)]/qg (psia2-D/Mscf-cp)
1.E+03 2.E+07
qg (Mscf/D)
2.E+07
1.E+02 1.E+07
5.E+06
1.E+01 0.E+00
0 10 20 30 40 50 60 70 80 90 100
1/2 1/2
time (days )
Fig. 15 – Semilog plot that shows history match of m(p)/qg vs. t1/2 by using simulation for case
1. The matched parameters are k = 0.000607 md, xe = 1,124 ft, ye = 646 ft to give kxe2 = 767 md-
ft2 and A= 66.68 acres.
~ 1/ 4
k ky
~
xe xe xe (50)
kx kx
27
1.E+04 3.E+07
qg, History
Slope
[m(pi)-m(pwf)]/qg, History
[m(pi)-m(pwf)]/qg, Simulation 3.E+07
2
qg (Mscf/D)
2.E+07
1.E+02 1.E+07
5.E+06
1.E+01 0.E+00
0 10 20 30 40 50 60 70 80 90 100
1/2 1/2
time (days )
Fig. 16 – Plot that shows history match of qg and m(p)/qg vs. t1/2 by using simulation for case 2.
The matched permeability is 0.00075 md.
It is likely that anisotropy is sometimes much more severe than this example. The
estimated k xe from either Eqs. 33 or 36 would be the product of the square root of the
equivalent permeability expressed by Eq. 52 ant the transformed dimension ~
x . e
~
k kxk y (52)
However, the drainage area, A, and the resultantly pore volume, Vp, and the OGIP are properly
evaluated even with this anisotropy.
High permeability streaks. One more possible reason for linear flow performance other
than the geometry of Figs. 1 or 18 would be massive tight formations, which could be expected
to have layers of higher permeability18. In this case, the drop of pressure in the higher
permeability layers would then cause vertical linear flow into the higher permeability layers.
This would result in long-term linear behavior. Fig. 19 exhibits a sketch of this situation of
vertical linear flow.
28
1,000
qg (Mscf/D)
100
10
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000
time (days)
Fig. 17 – Semilog plot that shows the comparison among actual data, analytical and simulation
results of gas rate vs. time for constant pressure in well J-7.
Layered reservoirs. If well produces from multiple layers, it would still appear as linear
flow if flow were infinite-acting in all layers. The k xe product would be a thickness weighted
average summation of these values for the individual layers (Eq. 4-53).
n
j 1
k j xej
k xe n
(53)
h j 1
j
Determination of fracture length. The fracture length is a continual confuses for a fractured
well. In the fractured well analysis presented here, it is seen that the k xf product is estimated
from the well performance. The normal thought is that permeability must be evaluated from a
pre-frac buildup test. Wattenbarger et al.1 explain that this would apply well in a simple
reservoir, but it is expected that large number of low permeability reservoirs will behave a
multiple-layered reservoir. For these situations, the permeability averaging that is estimated
before the fracture treatment and the k xf averaging that is estimated after the fracture treatment
may not be correlated. Some type curves of complex cases may assist this situation but the
estimated results may not be with unique solution.
29
ye
ye
xe
Fig. 18 – This picture shows the effect of anisotropy from natural fractures on equivalent
reservoir geometry.
Fig. 19 – This sketch shows vertical linear flow into a high permeability layer.
30
qg
J CP
m( p) m( p )
wf
(54)
where the average pressure, p , is estimated from material balance calculations, using actual gas
cumulative production, Gp. The forecasting process is estimated by applying material balance
and gas productivity index equations of selecting future time steps. Forecasting calculations
will be conservative, if the boundary effects have not yet been detected.
Conclusions
1. Actual field data show long-term linear flow behavior for years in a large number of
wells that produces from tight gas reservoirs (low permeability).
2. Several tight gas wells with fracture treatment exhibit only linear flow and they do not
show pseudo-radial flow as sometimes expected with hydraulic fractures.
3. The expressions for analyzing long-term linear flow (constant pwf) are distinct than for
buildup analysis (constant rate).
4. A history match that does properly described the long-term linear identified in some
tight gas reservoirs should be used to forecast gas rates and reserves.
5. If the reservoir outer boundary effect has been detected, drainage area can be directly
estimated. This drainage area would be a minimum estimation if linear flow were still infinite
acting. For this procedure knowledge of permeability, k, is not indispensable.
6. Pore volume, Vp, and original gas in place, OGIP, can be directly evaluated if the outer
boundary effect of the reservoir has been detected. These Vp and OGIP would be minimum
values if linear flow were still infinite acting. Knowledge of permeability, k, and thickness, h,
and porosity, , are not required.
7. The evaluation of OGIP becomes not sensitive to the value used for water saturation,
Sw, if gas compressibility, cg, dominates total compressibility, ct. The evaluation of OGIP may be
determined accurately without having knowledge of permeability, k, porosity, , thickness, h,
and water saturation, Sw.
31
32
Subscripts
CP = constant pressure
CR = constant rate
D = dimensionless
ehs = end of “half-slope” period
g = gas
i = initial conditions
max = maximum value
min = minimum value
o = oil
w = well
x = direction x
y = direction z
33
34
35
Traditional definitions used in this chapter are based on radial flow convention for pD:
1 k h( p i p wf ) 0.00633k t
p wD ; t D xe
qD 141.2qo Bo o o ct xe 2
Then, the solutions are as follows:
Constant q solutions:
Complete solution,
1 ye 1
qD 4 xe n 2 2 xe
2
exp 4 t Dx e
nodd
ye
Early time solution,
1
t Dxe
qD 2
36
1 2 xe 2x 1 k Ac ( p i p wf ) 0.00633k t
p wDLL p wD e t D Le
y e
;
q DLL ye qD 887.2 L q o Bo o o c t L2
Constant q solutions:
Complete solution,
1 1
p DLL t DL
3
n
n 1
2 2
exp n 2 2 t DL
Complete solution,
1 1 1
q DLL 2 n 2 2
exp t DL
nodd 4
37