Академический Документы
Профессиональный Документы
Культура Документы
Summary fracture and unsteady-state Darcy flow in the matrix. This facili-
In this paper, a mathematical model has been developed and suc- tates displaying changes in the flux distribution along the fracture,
cessfully applied to accurately determine the fracture conductivity whereas non-Darcy flow in a fracture leads to the apparent fracture
in tight formations with non-Darcy flow behavior. A new non- conductivity being significantly less than its true value. Subse-
Darcy flow number is first defined to account for the effect of quently, Umnuayponwiwat et al. (2000) used a similar model to
characteristic length in a hydraulic fracture. A semianalytical investigate the flux distribution along the fracture and interaction
method is then applied to solve the newly formulated mathemati- between the fracture and matrix and found that an increase in the
cal model by discretizing the fracture into small segments, assum- fracture flow would result in an increase in the non-Darcy flow
ing that there exists unsteady flow between the adjacent segments. effects. Because of its difference from Darcy’s condition, buildup
The newly developed model has been validated by simplifying it tests could not be analyzed as superposition of drawdown tests for
to the traditional Forchheimer (i.e., non-Darcy) model and by per- the nonlinearity of non-Darcy problem (Guppy et al. 1982b).
forming numerical simulation with a reservoir simulator as well. By examining non-Darcy effects in the proppant packs with
The pressure response and its corresponding derivative type laboratory tests, Martins et al. (1990) discovered that pressure-loss
curves have been reproduced to examine non-Darcy flow behavior behavior occurs at a high flow rate. Vincent et al. (1999) pointed
under different fracture conductivities. Both relative minimum out that increasing fracture width or placing high-permeability
permeability and characteristic length are found to impose a nega- proppant could increase conductivity. Settari et al. (2000) used a
tive effect on the fracture conductivity. Compared with relative finite-difference simulator for evaluating performance of hydraulic
minimum permeability, characteristic length is a strong function fracturing of high-permeability gas wells, thereby reducing the
dominating the non-Darcy flow behavior in the fractures. It is associated non-Darcy skin effects. It was found that, in a moder-
obvious that the fracture conductivity can be accurately deter- ate- to high-permeability reservoir, the non-Darcy effect is too
mined when non-Darcy flow behavior in the fracture network is large to be neglected. Gill et al. (2003) examined the effect of non-
taken into account. Darcy flow in a low-permeability gas reservoir and subsequently
developed correlations to limit non-Darcy flow within acceptable
ranges. After examining the effect of non-Darcy flow in nanopore
Introduction shale reservoirs, Swami et al. (2012) concluded that the non-Darcy
Primary recovery factor remains low (approximately 5 to 10% of effect is not only significant, but also dependent on pore radius.
original oil in place) in tight oil formations even after long hori- Smith et al. (2004) found that non-Darcy flow should not be simply
zontal wells have been drilled and massively fractured (Manrique considered as the reduction of the true conductivity, whereas Gup-
et al. 2010). This is ascribed to the fact that fluid flow in such tight py’s correlations (Guppy et al. 1982a) would overestimate the
formations can be completely dependent on the fracture network, non-Darcy flow effects. As for non-Darcy effects at low flow rates,
whereas the matrix plays only a source role (Manrique et al. Miskimins et al. (2005) concluded that non-Darcy effects would
2007). Numerous efforts have been made to describe the non- reduce the flow capacity by 5 to 30% in tight gas reservoirs with
Darcy effect on the basis of the traditional Forchheimer equation, relatively low permeability (0.01 md). By incorporating the
the limits of which have been identified (Andrade et al. 1997; Hill unsteady-state solutions in both the fracture and the matrix, Zeng
et al. 2001; Stanley and Andrade 2001). By replacing the inertial and Zhao (2010) pointed out that this solution can be especially
factor in the Forchheimer model with two new parameters—mini- applied to tight formations. The Barree-Conway model has been
mum permeability plateau and characteristic length—the Barree- used to describe the pressure-transient behavior of non-Darcy flow
Conway model is claimed to be capable of describing non-Darcy in hydraulically fractured wells with the finite-difference method
flow behavior from low to high flow rates (Barree and Conway (Al-Otaibi and Wu 2011). Because limited field data have been
2004; Lai et al. 2012). So far, few attempts have been made to made available in the literature, it is essential that the versatile
quantify the fracture conductivity in such hydraulically fractured Barree-Conway model be modified and used to accurately describe
formations because of both its complexity and the existence of non-Darcy flow behavior in hydraulically fractured wells.
non-Darcy flow behavior. Therefore, it is of fundamental and In this paper, a new mathematical model has been developed
practical importance to accurately determine the fracture conduc- and successfully applied to determine fracture conductivity
tivity in tight formations with non-Darcy flow behavior. with non-Darcy flow behavior in a hydraulic fracture. After the
The non-Darcy flow in the fracture has been investigated to unsteady-state solution in the fracture is applied to accurately
examine its effect on the transient pressure behavior. After analyz- model the flow in the tight formation, a semianalytical method is
ing the transient behavior of various conductivities with a finite- developed to obtain the solutions for the flow rates in the fracture
difference simulator, Holditch and Morse (1976) concluded that and matrix. As such, the pressure and its derivative type curves can
non-Darcy flow in the fracture reduced the apparent conductivity. be generated, respectively. The non-Darcy effects on fracture con-
Guppy et al. (1982a) proposed a comprehensive semianalytical ductivity by relative minimum permeability and characteristic
model of non-Darcy flow in wells with finite-conductivity frac- length are discussed in tight formations. In addition, a field case is
tures, assuming there exists steady-state non-Darcy flow in the shown to evaluate the actual conductivity under the non-Darcy
effect in tight formations.
Copyright V
C 2014 Society of Petroleum Engineers
This paper (SPE 162548) was accepted for presentation at the SPE Canadian Theoretical Model
Unconventional Resources Conference, Calgary, 30 October–1 November 2012, and
revised for publication. Original manuscript received for review 12 August 2012. Paper peer Mathematical Formulation. In this study, Fig. 1 shows a verti-
approved 18 June 2013. cal fracture that is fully penetrating along the well in a closed
pðx; y; t ¼ 0Þ ¼ pi : . . . . . . . . . . . . . . . . . . . . . . . . . ð7cÞ
and where qN is the flux inside the fracture. The boundary condi- pD ðxD ¼ 0; yD ¼ 0; tD Þ ¼ 0: . . . . . . . . . . . . . . . . . . . ð13cÞ
tions can be written as
kf hwf @pf Qsc Bg Semianalytical Solution. The semianalytical method applied to
d ¼ ; . . . . . . . . . . . . . . . . . . ð5aÞ
l @x x¼0 2 solve the hydraulic-fracture system has been widely discussed in
the literature (Guppy et al. 1982a; Umnuayponwiwat et al. 2000;
@pf
¼ 0: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð5bÞ Zeng and Zhao 2010). In general, the whole system is divided
@x x¼xf into the matrix system and the fracture system, each of which is
solved separately. The fracture system is first discretized into sev-
The pressure drop in the formation can be expressed by eral segments in each of which the flux qfD can be approximated
to its average value within the segment. With such an approxima-
km @ 2 p @ 2 p @p
þ ¼ /m ctm ; . . . . . . . . . . . . . . . . . . ð6Þ tion, equations for the matrix and the fracture systems can be
l @x2 @y2 @t solved in the Laplace domain in real time, respectively. Their
dimensionless parameters (i.e., kmr and FND) with the Barree-Con- that FND imposes a smaller effect on kmr as it is decreased. When
way model. Accordingly, effects of these two parameters on pres- FND is set to be a small value, the non-Darcy effect will not affect
sure response are discussed here because effects of wellbore the pressure response significantly, even at a very small kmr value.
storage (CD), fracture conductivity (CfD), and fracture diffusivity
(Cg) have been extensively analyzed and discussed (Cinco-Ley. Non-Darcy Number (FND). Figs. 7 and 8 present drawdown
et al. 1978; Guppy et al. 1982a; Lee and Holditch 1982). type curves for FND ¼ 1, 10, and 100 at kmr ¼ 0.01 and 0.10,
respectively. As shown in Fig. 7, the pressure drop and its
Relative Minimum Permeability (kmr). In the Barree-Conway
model, kmr is first introduced, which is caused by streamlining or
flow heterogeneity. The drawdown type curves for kmr ¼ 0.01,
0.10, and 0.50 at FND ¼ 100 are depicted in Fig. 5. The typical
one-fourth slope at early times of both pressure and pressure-de-
rivative curves is exhibited to show a bilinear-flow region for
non-Darcy effect. Then, a linear flow with a slope of 0.5 follows
at the end of the bilinear-flow period. Finally, a pseudoradial flow
is attained as the flux in the fracture has been stabilized. As the Well-1
value of kmr decreases, the pressure drop and its derivative curves
increase because of strong non-Darcy effect. It is found that the
non-Darcy effect becomes significant when kmr becomes less than
0.50. As kmr approaches 1.00, however, the non-Darcy effect
tends to disappear and the pressure response is found to be very
similar to the Darcy case.
Fig. 6 presents the type curves for kmr ¼ 0.01, 0.10, and 0.50 at
FND ¼5. When FND is set to be a smaller value, the pressure and its
derivative curves also show a trend similar to that of the case at a (a)
higher FND value. However, the non-Darcy effect becomes less
3
significant compared with that at a high FND value. This means 10
CMG
1 This work
10 2
10
This study
Forchheimer equation
dp (psi)
1
10
pwD, dpwD/dlntD
0
10
0
10
–1
10
–1
10 –5 –4 –3 –2 –1 0 1 2 3 4
10 10 10 10 10 10 10 10 10 10
t (hr)
–2
10 –5 –4 –3 –2 –1 0 1 2 3 (b)
10 10 10 10 10 10 10 10 10
tD Fig. 4—(a) Top view of the 3D grid system and (b) model valida-
tion between this study and the Computer Modelling Group
Fig. 3—Model validation for the Forchheimer equation. simulation results.
1 1
10 10
Linear
Flow
0 0
10 10
pwD, dpwD/dlntD
pwD, dpwD/dlntD
Bilinear
Flow
Pseudo-
–1 Radial Flow –1
10 10
–3 –3
10 –5 –4 –3 –2 –1 0 1 2 3
10
–5 –4 –3 –2 –1 0 1 2 3
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
tD tD
Fig. 5—Type curves for non-Darcy effect of relative minimum Fig. 6—Type curves for non-Darcy effect of relative minimum
permeability kmr 5 0.01, 0.10, 0.50, and 1.00 at FND 5 100, permeability kmr 5 0.01, 0.10, 0.50, and 1.00 at FND 5 5,
respectively. respectively.
derivative curves increase as the non-Darcy number FND increased to 1.00, the Barree-Conway model can be converted
increases. This is attributed to the fact that the non-Darcy effect is into Darcy’s model so that the flow will approach the Darcy flow
increased as FND is increased. Obviously, as relative minimum condition. As can be seen in Figs. 5 and 7, the non-Darcy effect
permeability kmr is decreased, the non-Darcy effect becomes will become more evident at a smaller relative minimum perme-
much stronger. Meanwhile, the non-Darcy number FND becomes ability (kmr < 0.5) and a larger non-Darcy number (FND > 1).
more sensitive than that at a larger kmr. This is because, as kmr is Comparing Fig. 6 with Fig. 8, it can be found that both relative
1 1
10 10
0 0
10 10
pwD, dpwD/dlntD
pwD, dpwD/dlntD
–1 –1
10 10
–3 –3
10 –5 –4 –3 –2 –1 0 1 2 3
10 –5 –4 –3 –2 –1 0 1 2 3
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
tD tD
Fig. 7—Type curves for non-Darcy effect with non-Darcy num- Fig. 8—Type curves for non-Darcy effect with non-Darcy num-
ber FND 5 1, 10, and 100 at kmr 5 0.01, respectively. ber FND 51, 10, and 100 at kmr 5 0.10, respectively.
minimum permeability kmr and non-Darcy number FND can affect Darcy flow on flow performance in a tight oil and gas formation,
the pressure response. However, non-Darcy number FND, which the basic data set (Cinco-Ley et al. 1978; Guppy et al. 1982a;
usually results in a large pressure drop at a large kmr, is more sen- Umnuayponwiwat et al. 2000) used in this study is tabulated in
sitive than the relative minimum permeability kmr. Similarly, Fig. Tables 3 and 4, respectively. It is assumed that fracture diffusiv-
8 shows that pressure drop and its derivative curves increase as ity (Cg) and relative minimum permeability (kmr ) are to be 105
the non-Darcy number is increased. and 0.01, respectively.
Effect of Non-Darcy Flow in Tight Formations Tight Oil Formations. Although the effect of non-Darcy flow on
Because of the low production rate in a tight formation, the non- the transient pressure responses of hydraulically fractured wells
Darcy effect on transient pressure behavior has not yet been con- has been examined, few attempts have been made in tight oil for-
sidered. To respectively analyze and discuss the effect of non- mations at a low production rate. In the tight-oil-formation exam-
ple, the reservoir and fluid data are sourced from Cinco-Ley et al.
(1978). As for a tight oil formation with km ¼ 5.00 md, it is
4
10 assumed that there is a relatively high-conductivity fracture with
CfD ¼ 100.0. As shown in Eq. 14, FND is a rate-dependent factor.
Therefore, oil-production rate is assumed to be constant (195
Derivative, Δpwf, dΔpwf /ln(t)
Pressure Change and Its
10
3 STB/D) so that FND is a function only of s. Barree and Conway
(2004) suggested that s is related to the particle size or size distri-
bution of the porous media. In this study, the value of s is set
2
according to the examples provided by Al-Otaibi and Wu (2011).
10 As can be seen in Fig. 9, when s is increased from 100 to 10,000
ft–1, the value of FND is set to be 22.001, 2.200, and 0.220,
τ = 100, FND = 22.001
τ = 1000, FND = 2.200
whereas responses for two Darcy flow scenarios with CfD ¼ 47.9
10
1
τ = 10000, FND = 0.220 and CfD ¼ 100.0 are included.
kmr = 1.00 (Darcy) Comparing the response between non-Darcy flow and Darcy
kmr = 1.00, CfD = 47.9 flow (CfD ¼ 100.0), non-Darcy flow is found to induce an addi-
0 tional pressure drop. As s is decreased, FND is increased, leading
10 –3 –2 –1 –0 1 2 3 4
10 10 10 10 10 10 10 10 to a strong non-Darcy flow effect. As observed by Guppy et al.
t, Hours (1982a) and Umnuayponwiwat et al. (2000), the additional pres-
sure drop because of the non-Darcy effect can also be expressed
Fig. 9—Type curves for non-Darcy effect in a tight oil formation by a reduction in the apparent conductivity of the fracture. In Fig.
at CfD 5 100.0. 7, the Darcy scenario of CfD ¼ 47.9 can be matched with the non-
qND
qfD
1.0
0.2
0.5
0.1
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
xD xD
Fig. 10—Flux distribution from the matrix to the fracture in a Fig. 11—Flux distribution inside the fracture in a tight oil forma-
tight oil formation at CfD 5 100.0 and FND 5 22.001. tion at CfD 5100.0 and FND 5 22.001.
Darcy scenario of CfD ¼ 100.0 and FND ¼ 2.200 all the time. The bore from the formation through the fracture enter at the fracture
traditional analysis method can result in more than 50% error in the toe.
estimation of conductivity. This is the reason that the type curves For comparison, a relatively low-conductivity fracture is con-
that can be matched with the conventional Darcy model would sidered with CfD ¼ 20.0 in the same formation by changing the
yield an underestimated fracture conductivity if the non-Darcy fracture permeability to a smaller value. As depicted in Fig. 12,
effect exists (Guppy et al. 1982a, 1982b; Umnuayponwiwat et al. FND is still the same as that in a high-conductivity fracture. The
2000). Therefore, the fracture half-length together with other frac- transient pressure and its derivative curves show a trend similar to
ture parameters cannot be accurately determined if the apparent those in a high-conductivity fracture, except that a larger pressure
fracture conductivity is evaluated with the conventional method. drop can be obtained at the early stages. This is ascribed to the fact
To better understand the non-Darcy effect, Fig. 10 presents the that low fracture conductivity can also cause an additional pressure
flux distribution from the matrix to the fracture as a function of drop, which is similar to that of the non-Darcy effect. Fig. 13
time at FND ¼ 22.001. In this case with a strong non-Darcy effect, presents flux distribution along the fracture with a low conductiv-
it is found that the fluid enters from the heel of the fracture at the ity. As shown in Fig. 13, the flux at early time is larger than that in
very beginning, resulting in a large pressure drop until the pseu- the high-conductivity fracture at the same location, directly leading
doradial flow is established. As time proceeds, the flux decreases to a larger pressure drop. In the late stage, the flux distribution is
at the heel of the fracture but increases at its toe. This is ascribed to similar to that of a high-conductivity scenario, indicating that the
the fact that a strong non-Darcy effect causes resistance in the frac- fluid entering the fracture enters less at its heel and more at its toe.
ture, forcing the fluids to largely enter from the heel at the begin- In practice, bilinear-flow analysis proposed by Cinco-Ley and
ning of well production. When pseudoradial flow is established, Samaniego-V. (1981) is widely used to estimate fracture conduc-
the flux distribution stabilizes and remains unchanged. Because tivity. It is worthwhile, however, to examine how non-Darcy
the resistance force becomes smaller, most of the fluids enter the behavior will affect bilinear-flow behavior (Umnuayponwiwat
fracture at its toe, causing a smaller pressure drop. It should be et al. 2000). As shown in Fig. 14, the pressure change is increased
noted that the additional pressure drop caused by the non-Darcy following a straight line in the one-fourth of time at a low conduc-
effect is rate dependent and that the flux distribution along the frac- tivity. This is because bilinear flow exists at the beginning of pro-
ture is a function of time until the pseudoradial flow begins. duction. At the same one-fourth of time, the pressure change is
To perform further illustration, flux distribution inside the reduced as s increases. This means that non-Darcy flow leads to
fracture is plotted in Fig. 11. At the beginning of well production, an increase in the pseudopressure drop. Similarly, Fig. 15 shows
the flux inside the fracture generally decreases at each node, the same trend for a scenario with a higher conductivity. Obvi-
though it is decreased sharply in the half of the fracture closer to ously, there exists non-Darcy flow behavior in a tight oil forma-
the wellbore. This is because most of the fluids entering the well- tion even at a low production rate.
3.0
4
10 tD = 1.0E–5
tD = 1.0E–4
2.5
Derivative, Δpwf, dΔpwf / ln(t)
tD = 1.0E–2
Pressure Change and Its
3 tD = 1.0E0
10 tD = 1.0E3
2.0
qfD
2 1.5
10
10 –3
0 0.0
–2 –1 –0 1 2 3 4 0.0 0.2 0.4 0.6 0.8 1.0
10 10 10 10 10 10 10 10
t, Hours xD
Fig. 12—Type curves for non-Darcy effect in a tight oil forma- Fig. 13—Flux distribution from the matrix to the fracture in a
tion at CfD 5 20.0. tight oil formation at CfD 5 20.0 and FND 5 22.001.
60 40
40
20
20
0 0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Δt1/4, Hours Δt1/4, Hours
Fig. 14—Bilinear-flow analysis with non-Darcy effect at Fig. 15—Bilinear-flow analysis with non-Darcy effect at
CfD 5 20.0. CfD 5 100.0.
11 2.5
10
tD = 1.0E–5
Derivative, Δm(pwf), dΔm(pwf) /ln(t)
Pseudo-Pressure Change and Its
tD = 1.0E–4
2.0 tD = 1.0E–2
10 tD = 1.0E0
10
tD = 1.0E3
1.5
qfD
9
10
1.0
τ = 1000, FND = 121.328
8 τ = 5000, FND = 24.266
10 τ = 10000, FND = 12.133
kmr = 1.00 (Darcy)
0.5
kmr = 1.00, CfD = 9.65
7
10 –2 –1 –0 1 2 3 4 5 6 7 0.0
10 10 10 10 10 10 10 10 10 10 0.0 0.2 0.4 0.6 0.8 1.0
t, Hours xD
2.5
2.0
1.5
Field Data
Non-Darcy Flow
1.0 Darcy Flow
0.5
10
0.0 0.1 1 10 100
0.0 0.2 0.4 0.6 0.8 1.0 t, Hours
xD
Fig. 20—Type-curve match of the field data on an oil well in a
Fig. 19—Flux distribution from the matrix to the fracture in a tight formation.
tight gas formation at CfD 5 20.0 and FND 5 121.328.