SPE-170971-MS

A Trilinear Flow Model for a Fractured Horizontal Well in a Fractal

Unconventional Reservoir
O. Ozcan, Colorado School of Mines; H. Sarak, Istanbul Technical University; E. Ozkan, Colorado School of
Mines; R. Raghavan, Phillips Petroleum Company (Retired)

Copyright 2014, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Annual Technical Conference and Exhibition held in Amsterdam, The Netherlands, 27–29 October 2014.

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
Based on the premises of anomalous diffusion models in fractal porous media, an alternative to
dual-porosity based formulations of flow in fractured unconventional reservoirs is presented. The new
formulation is implemented in the trilinear flow idealization for a fractured horizontal well in a tight
formation and verified by using the asymptotic cases. The results of the new model are compared with the
dual-porosity based trilinear flow formulation and the differences and similarities are delineated. A
discussion of the characteristics of the pressure and derivative responses obtained from the trilinear
anomalous-diffusion model is provided and related to the fractal nature of fractured media. Physical
interpretations are also assigned to fractional derivatives and the phenomenological coefficient of the
fractional flux law. It is shown that the anomalous diffusion formulation does not require explicit
references to the intrinsic properties of the matrix and fracture media and thus relaxes the stringent
requirements used in dual-porosity idealizations to couple matrix and fracture flows. The trilinear
anomalous-diffusion model should be useful for performance predictions and pressure- and rate-transient
analysis of fractured horizontal wells in tight unconventional reservoirs.
Introduction
Two common approaches to model naturally fractured media are the discrete fracture network (DFN)
models and dual-porosity idealizations (Fig. 1). In discrete fracture network (DFN) models (Fig. 1A), it
is possible to consider the details of each fracture and the distribution and connectivity of the fracture
network. However, DFN models require extensive characterization studies and also lead to computation-
ally inefficient models. In general, the level of detail that can be utilized from the DFN model is limited
by the capabilities of the flow model, which will use the DFN model. Therefore, despite their potential,
the DFN models are not the tool of choice for most routine engineering applications.
Dual-porosity models (Barenblatt et al., 1960, Warren and Root, 1963, Kazemi, 1969) are based on the
continuity assumption and effective averaging of the naturally fractured medium properties (Fig. 1B).
They are appropriate for systems where a repeated pattern of continuous fractures can be distributed to the
entire flow domain. Considering the large variations of scale, connectivity, and conductivity of fractures,
particularly in tight unconventional reservoirs, dual-porosity assumption is only a first order approxima-
2 SPE-170971-MS

Figure 1—Two common modeling approaches for naturally fractured reservoirs

tion (Kuchuk and Biryukov, 2013 & 2014). Due to the limited data-requirement and computational
conveniences, however, the dual-porosity assumption has been used as a practical approach to model a
fracture network around horizontal wells in tight unconventional reservoirs.
To address the concerns about the applicability of dual-porosity idealizations (Kuchuk and Biryukov,
2013 & 2014) and to present a practical alternative for modeling fractured unconventional reservoirs, this
paper considers the premises of anomalous diffusion in heterogeneous porous media. Specifically, we
propose to model the naturally fractured tight porous media by using fractals instead of the dual-porosity
idealization. As noted by Raghavan and Chen (2013a & b), many forms of probability density have been
considered in the literature to model anomalous (fractional) diffusion in fractal structures. Fundamentally,
the differential equation describing anomalous diffusion “must to be constructed on the basis of empirical
arguments, statistical laws, and numerical experiments.” More importantly, “their suitability to address
transient anomalous diffusion toward fractured wells is yet to be evaluated.” Therefore, the main
objectives of this paper is to demonstrate the potential of anomalous diffusion concept in modeling flow
in unconventional reservoirs and discuss the relations of the model parameters to the physical structure of
the flow domain.
In this paper, anomalous diffusion considerations are incorporated into the trilinear model for a
fractured horizontal well in tight formations (Brown, 2009, Brown et al., 2011, Ozkan et al., 2011). The
choice of the trilinear flow model is because of its relative simplicity and availability for dual-porosity
idealization (Appendix A), which allows direct comparisons. Because dual-porosity idealization explicitly
relates model parameters to the physical properties of the fractured medium, such comparisons provide
suggestions for the physical interpretations of the anomalous diffusion model parameters. Although the
trilinear anomalous-diffusion model is introduced and used in this paper to demonstrate and verify the idea
of concept, it can be readily used for performance predictions and pressure- and rate-transient analysis of
fractured horizontal wells in tight unconventional reservoirs.
The solutions in this paper are given in terms of the scaled variables defined by Eqs. A1 through A10
in the Appendix. Note that the scaled variables correspond to the conventional dimensionless variables for
the trilinear dual-porosity model. For the trilinear anomalous-diffusion model, however, pD and tD (Eqs.
A1 and A2, respectively) are defined based on the phenomenological coefficient, ka, in the constitutive
fractional flux law, which is not in the units of permeability. Therefore, although for simplicity, we use
the same notation for both models, pD and tD are not dimensionless quantities for the trilinear anomalous-
diffusion model.
Anomalous Diffusion in Fractal Porous Media
In the last two decades, non-local, hereditary descriptions of flow and transport have gained notable
popularity among scientists, engineers, and mathematicians focusing on applications in various forms of
nano-porous systems (e.g., Gefen et al. 1983, Le Mehaute and Crepy 1983, Nigmatullin 1984,
O’Shaughnessy and Procaccia, 1985, Chang and Yortsos 1990, Dassas and Duby 1995, Caputo 1998,
Molz III et al. 2002, Flamenco-Lopez and Camacho-Velazquez, 2003, Camacho-Velazquez et al., 2011,
Raghavan 2011, Fomin et al. 2011). Until recently, these efforts have not attracted much attention in the
SPE-170971-MS 3

oil-field applications due to the dominance of advective (Darcy) flow in conventional reservoirs. In
unconventional shale-gas reservoirs, on the other hand, diffusive flow mechanisms have been recently
incorporated into flow models due to their considerable contribution to flow in shale matrix (Javadpour
et al., 2007, Ozkan et al., 2010, Apaydin, 2012). In these works, the advective and diffusive mechanisms
were assumed to be independent of each other and locally defined based on the corresponding gradients
of the process variables (pressure and concentration); an assumption that presumes linearly additive fluxes
and permits the use of the classical diffusion equation. There has been enough evidence in the literature
(Fomin et al., 2011) that these are crude assumptions and classical diffusion is a special case, not a norm,
for flow and transport in heterogeneous nano-porous medium.
In statistical physics, diffusion is the result of the random Brownian motion of individual particles.
Classical (normal) diffusion is usually associated with homogeneous porous media. It is a special case
where the random Brownian motion of the diffusing particles is governed by a Gaussian probability
density whose variance is proportional to the first power of time; that is the mean square displacement of
a particle is a linear function of time (␴r2 ~ t). However, a convincing number of works have indicated
anomalous diffusion in which the mean square variance grows faster (superdiffusion) or slower (subdif-
fusion) than that in a Gaussian diffusion process. Thus, a general relationship between the mean square
variance and time is given by
(1)

The disordered structure of unconventional, nano-porous media is more in the realm of anomalous
diffusion models where the variance of the evolution equations is proportional to the fractional power of
time. In addition, transport pathways created by the natural and induced fractures have been shown to be
fractals (Sahimi and Yortsos, 1976). Moreover, transport in disordered systems often involves long-range
correlations, which is another fundamental characteristic of fractal systems, and the local gradients of the
mean diffusion process variables depend on the global pressure field. These aspects of flow and transport
are comprised in non-local anomalous diffusion processes.
Because nonlinear modeling of anomalous diffusion is computationally expensive, fractal and frac-
tional derivatives have been introduced. In petroleum engineering, fractal diffusion has been used to
account for the stochastic nature of heterogeneity, mostly in the form of natural fractures. The phenom-
enological descriptions used in fractional diffusion models are different from the conventional ones and
have not yet been commonly implemented in petroleum engineering. The limited applications by Chang
and Yortsos (1990), Beier (1994), Flamenco-Lopez and Camacho-Velazquez (2003), and Camacho-
Velazquez et al., (2011) focused on fractal modeling of production from fractured media using vertical
wells. It is worthwhile to note that the essential difference of the fractal and the fractional derivatives lies
in the former being a local operator, while the latter is a global operator. Recently, Raghavan (2011) and
Raghavan and Chen (2013a & b) extended the fractional diffusion models to semi-infinite fractal media
produced through a hydraulic fracture.
As suggested by Raghavan and Chen (2013a & b), in this work, we use the following constitutive
relation (flux law) to describe flow in naturally fractured nano-porous media:
(2)

where 0 ⱕ ␣ ⱕ 1 and ␭␣ is a phenomenological coefficient. In this work, we assume that the anomalous
diffusion is related to the petrophysical heterogeneity of the medium and express the phenomenological
coefficient in the following form:
(3)
4 SPE-170971-MS

The temporal fractional derivative in Eq. 2 is

defined in the Caputo (1967) sense:
(4)

The convolution integral in Eq. 4 signifies the

hereditary nature of anomalous diffusion on a het-
erogeneous velocity field. In addition, k␣ in Eq. 3 is
a dynamic property and different from the conven-
tional Darcy permeability (it has the units of L2T1-
␣). Physical interpretation of k␣ is not straightfor- Figure 2—Schematic of the trilinear flow model (Brown, 2009, Brown et
ward and, based on Eqs. 2 through 4, static al., 2011, Ozkan et al., 2011)

measurements are not suitable to determine k␣. Cur-

rently, the only viable technique to determine k␣ is Table 1—WELL, RESERVOIR, AND FLUID DATA (Intrinsic Prop-
to match the dynamic (transient pressure or flow erties)

rate) data with an appropriate model. Formation thickness, h, ft 250

Wellbore radius, rw, ft 0.25
Using the flux law (Eq. 2) with the mass conser-
Horizontal well length, Lh, ft 3000
vation equation yields the following 2D, temporal- Number of hydraulic fractures, nF 15
fractional (anomalous) diffusion equation: Distance between hydraulic fractures, dF, ft 200
Distance to boundary parallel to well (1/2 well spacing), 250
(5) xe, ft
Inner reservoir size, ye, ft 100
where Viscosity, ␮, cp 0.3
(6) Hydraulic fracture porosity, ␾F, fraction 0.38
Hydraulic fracture permeability, kF, md 5.0E⫹04
Hydraulic fracture total compressibility, ctF, psi-1 1.0E-04
Subject to the appropriate boundary conditions, Hydraulic fracture half-length, xF, ft 250
the temporal-fractional-diffusion equation given by Hydraulic fracture width, wF, ft 0.01
Eq. 5 is solved in Appendix B to obtain the solution Outer reservoir permeability, kO, md 1.0E-04
for a tight, naturally fractured, unconventional res- Outer reservoir porosity, ␾O 0.05
Outer reservoir compressibility, ctO, psi-1 1.0E-05
ervoir.
Constant flow rate, q, stb/day 150

Trilinear Model for Fractured

Horizontal Wells in Tight
Unconventional Reservoirs
A schematic of the trilinear model is shown in Fig. 2. The basic premise of the trilinear model is that the
flows in the three contiguous flow regions, outer reservoir, inner reservoir, and hydraulic fracture, are
linear and can be coupled by the continuity of pressure and flux at the interfaces of the regions. The details
of the trilinear model are given by Brown (2009), Brown et al. (2011), and Ozkan et al. (2011) and will
not be repeated here. For our discussions, it is important to note that the inner reservoir of the trilinear
model consists of a naturally fractured porous medium and was originally modeled by the dual-porosity
idealization. A summary of the trilinear dual-porosity model is given in Appendix A.
The trilinear dual-porosity (TDP) model will be used, in this work, for verification, comparison, and
physical interpretations. To develop the new trilinear model, the naturally fractured inner reservoir will be
interpreted as a fractal medium and modeled by anomalous diffusion. The derivation of the trilinear model
with anomalous diffusion in the inner reservoir follows the lines similar to those in the TDP model
derivation. The main difference is the use of the fractional diffusion equation for the solution of the inner
reservoir problem. The details of the derivation of the trilinear anomalous-diffusion (TAD) model are
given in Appendix B and by Ozcan (2014).
SPE-170971-MS 5

Table 2—INNER RESERVOIR DATA

TDP (Intrinsic Properties) TAD

Matrix permeability, km, md 1.0E-4 Phenomenological coefficient, k␣, md-day1-␣ 1.2

Matrix porosity, ␾m 0.05 Porosity compressibility product, (␾ct)␣, psi-1 4.62E-4
Matrix total compressibility, ctm, psi-1 1.0E-5
Natural fracture permeability, kf, md 1.0E⫹3
Natural fracture porosity, ␾f, fraction 0.7
Natural fracture total compressibility, ctf, psi-1 5.5E-1
Natural fracture width, hf, ft 3.0E-3

Figure 3—Verification of the TAD solution for normal diffusion (␣ ⴝ 1) in a homogeneous reservoir.

In the following discussions, we will use the subscripts, I and O for the inner and outer reservoir
properties, respectively, i to indicate the initial value of the property, F and f for the hydraulic and natural
fracture properties, respectively, m to refer to the property of the matrix, t to indicate total property, e to
refer to the external boundary, and D for scaled variables. When we refer to the TDP model, I will stand
for f. For the TAD model, on the other hand, I will correspond to ␣.
Verification of the TAD Model
Here, we will present the verification of the TAD model, discuss the results, and provide interpretations.
For the results discussed in this paper, we have used the data in Tables 1 and 2. Other case-specific data
are given in additional tables and also shown on the figures.
To verify the TAD model, we first use the asymptotic case of ␣ ⫽ 1. This case corresponds to normal
diffusion in a homogeneous reservoir, which can be obtained from the TDP solution for f(s) ⫽ 1
(Appendix A) when kf and (␾ct)f are chosen equal to k␣ and (␾ct)␣. The results in Fig. 3 show excellent
agreement between the TAD and TDP solutions.
As another means of verification, in Fig. 4, we match the results obtained from the TDP model for kf
⫽ 106 md and km ⫽ 10-4 md with the TAD model (Table 2). As shown by Fig. 4, the TAD model for ␣
⫽ 0.8 and k␣ provides a reasonable match with the TDP model. This example has been provided to show
that the TAD model captures the naturally fractured reservoir behavior idealized by the TDP models. It
does not, however, imply a general correspondence of the TAD and TDP models. As discussed in the
following section, the TAD model displays flow characteristics not observed for the TDP model.
Discussion of Results
Figure 5 shows the pressure and derivative responses of the TAD model for 0.1 ⱕ ␣ ⱕ 1 and a fixed value
of k␣ ⫽ 1.2. All pressure and derivative responses in Fig. 5 display straight-line trends at early,
6 SPE-170971-MS

Figure 4 —Matching TDP model results with the TAD solution.

Figure 5—Pressure and derivative responses obtained from the TAD solution for various ␣.

intermediate, and late times. As the straight-line trends are the diagnostic features of flow regimes, Table
3 presents the asymptotic approximations of the TAD model, which suggest the possible flow regimes for
fractured horizontal wells. The constants used in the expressions in Table 3 are given in Table A1 in the
Appendix. Because the derivations are lengthy, they will not be presented here. The details can be found
in Ozcan (2014). Multiple expressions given for each flow period in Table 3 are because of the different
combinations of the parameters in the model and not all expressions will apply to a given system.
However, the existence of multiple expressions indicates the versatility of the TAD model to cover a large
variety of flow conditions.
In Fig. 5, the pressure responses for all ␣ intersect at an early time. Similarly, another intersection point
exists at early-times for the derivative responses. For times larger than the intersection time, the pressure
drop increases as the value of ␣ decreases. This is consistent with the expectation that ␣ ⬍ 1 corresponds
to subdiffusion; in other words, as ␣ becomes smaller, the velocity field becomes more heterogeneous and
the movement of the fluid is interrupted more often.
At early times, two straight lines with the slopes of 1/4, corresponding to ␣ ⫽ 1, and 1/2, corresponding
to ␣ ⫽ 0, bound the pressure and derivative responses from below and above, respectively. Similarly, at
late times, two unit-slope straight lines for ␣ ⫽ 1 and ␣ ⫽ 0 bound the pressure and derivative responses
from below and above, respectively. Based on the trends of the data observed in Fig. 5, and the asymptotic
relations given in Table 3, all pressure and derivative responses, except for ␣ ⫽ 0, collapse into the same
unit-slope straight line at late times. For the time ranges used in Fig. 5, this behavior is evident for ␣ ⱖ
0.6, but only implied by the trends of the data for 0 ⬍ ␣ ⬍ 0.6. Because ⌬p ⬎1E⫹04 psi for all ␣ after
SPE-170971-MS 7

Table 3—INTERMEDIATE- AND LATE-TIME ASYMPTOTIC APPROXIMATIONS OF THE TAD SOLUTION

t ⬎1E⫹08 hr, from a practical perspective, the late-time, unit-slope trend will not be observed for ␣ ⬍
0.6 for the cases in Fig. 5.
To further comment on the flow-regime characteristics, we scrutinize the derivative responses as a
function of ␣ in Fig. 6. As expected from the asymptotic relations in Table 3, the derivative responses in
Fig. 6 display straight lines with a variety of slopes at early, intermediate, and late times. As also noted
in Fig. 5, the late-time derivative responses are bounded by two unit-slope straight lines for ␣ ⫽ 1 and
␣ ⫽ 0, which indicate the depletion of the system (boundary-dominated flow). Theoretically, all derivative
responses for 0 ⬍ ␣ ⬍ 1 are expected to merge with ␣ ⫽ 1 (normal diffusion) case and display a unit-slope
8 SPE-170971-MS

Figure 6 —Slopes of the straight lines observed from the results of the TAD model

line at late times. Although the results in Fig. 6 indicate this trend, the merger does not happen in practical
times for ␣ ⬍ 0.6. This is because of slowing diffusion in matrix as ␣ ¡ 0. In general, ␣ ¡ 0 indicates
longer interruptions of the fracture flow by the matrix elements, which can be physically caused by a
sparsely fractured, tight-matrix and loosely connected fractures. As a consequence, effective depletion of
the matrix, and thus the total system, takes longer when ␣ becomes smaller. At the limit of ␣ ⫽ 0, the
delay approaches infinity and the matrix is never depleted. On the other hand, as ␣ ¡ 1, flow in the
fracture network is not much hindered by the interruptions of the matrix; that is, the system is densely
fractured and the fractures are effectively connected. Therefore, the system is depleted faster and more
efficiently.
Slowing diffusion in matrix causes the slopes of the derivative responses decrease from 1 for ␣ ⫽ 0
to 3/4 for ␣ ⫽ 0.5 at intermediate times and an approximate straight-line may be fitted through the
derivative responses for 0 ⬍ ␣ ⱕ 0.5 at late-intermediate times. For ␣ ⬎ 0.5, the derivative responses
display a transitional behavior at intermediate times with a shallower slope than 3/4, which is not
reasonably constant to fit a straight line through the data, before merging with the derivative responses for
␣ ⫽ 1 and displaying a unit-slope straight line at late times.
Straight lines with slopes from 1/4 for ␣ ⫽ 1 to 1/2 for ␣ ⫽ 0 characterize the early-time derivative
responses in Fig. 6. These results are consistent with the interpretation that for ␣ ¡ 1, natural fractures
dominate the flow in the reservoir and the flow from the reservoir to finite-conductivity hydraulic fractures
begin early (while there is still linear flow in hydraulic fractures) to cause a bilinear flow behavior. In the
case of ␣ ¡ 0, the reservoir response is very weak due to the dominance of the matrix and the early-time
behavior is governed by the flow in hydraulic fractures.
To interpret the observation from Fig. 6 within the context of dual-porosity idealization, in Fig. 7, we
consider the TDP responses for the data in Table 4 (we have obtained the ␻ and ␭ values in Table 4 by
changing the thickness of the matrix elements). Fig. 7 indicates that a longer bilinear flow (1/4 slope) is
observed on derivative responses when ␭ is smaller; that is, when matrix blocks are smaller (more
fractures). Similarly, linear flow (1/2-slope) is observed for the larger values of ␭, which correspond to
larger matrix blocks. These results are consistent with the observations made from Fig. 6 for the TAD
model.
In the results discussed thus far, only the effect of the anomalous diffusion exponent, ␣, on pressure
and derivative characteristics have been considered. In Fig. 8, we consider the combined effects of the
phenomenological coefficient, k␣, and the anomalous diffusion exponent, ␣, on pressure and derivative
characteristics. The results in Fig. 8 indicate that an increase in k␣ for constant ␣ decreases both the
SPE-170971-MS 9

Figure 7—Bilinear and linear flow behaviors observed from the results of the TDP model

Table 4 —DATA USED IN FIGURE 7

␭ⴝ20 ␭ⴝ50 ␭ⴝ250 ␭ⴝ500
Property ␻ⴝ5.4E-4 ␻ⴝ2.2E-4 ␻ⴝ4.3E-5 ␻ⴝ2.2E-5

Matrix block dimension, hm, ft 1.25 0.5 0.1 0.05

Natural fracture density, pf, n/ft 0.8 2 10 20
Number of natural fractures, nf 200 500 2500 5000

Figure 8 —Combined effect of the permeability, k, and ␣ on pressure and derivative characteristics of TDP model

pressure drop and the derivative values. Variations of for constant k␣, on the other hand, cause a change
in both the magnitude of the pressure drop and the flow regime characteristics (indicated by the changing
slopes of the derivative responses). It should be also noted that the variation of k␣ for constant ␣ causes
a parallel shift in the pressure and derivative responses for all practical times for ␣ ⱕ 0.5 and at early and
intermediate times for ␣ ⬎ 0.5. The pressure and derivative responses become independent of at late times
for ␣ ⬎ 0.5.
10 SPE-170971-MS

Figure 9 —Rate declines obtained from the TAD solution for various ␣.

Figure 10 —Matching the Barnett field data with the TAD and TDP models.

For completeness, we also present the rate decline characteristics of the TAD model as a function of
␣ in Fig. 9. The physical interpretations presented for the pressure and derivative responses in Figs. 5 and
6 are also applicable to the rate-transient responses shown in Fig. 9. The early-time rate responses, after
the intersection time, display straight lines with slopes between 1/4 for ␣ ⫽ 1 and 1/2 for ␣ ⫽ 0. The
early-time straight lines for ␣ ⫽ 1 and 0 are followed by sharp exponential-decline periods, which are the
terminal flow regimes for these cases. For ␣ ⫽ 0, exponential decline corresponds to the depletion of a
system consisting mostly of a tight-matrix. On the other hand, for ␣ ⫽ 1, the system is dominated by
natural-fractures and their depletion causes the exponential decline behavior.
For 0 ⬍ ␣ ⬍ 1, the flow rates in Fig. 9 display straight lines with slopes less than or equal to 1 (␣ ⫽
0.1) and greater than or equal to 1/2 (␣ ⫽ 0.9) at intermediate times for ␣ ⬎ 0.5 and late times for ␣ ⱕ
0.5 For ␣ ⱕ 0.5, the delay of flow by the tight matrix causes a sharper drop in the flow rates at
intermediate times before the display of the late-time straight lines. For ␣ ⬎ 0.5, the higher decline rates
follow the intermediate-time straight lines.
SPE-170971-MS 11

Barnett Field Example Table 5—DATA USED TO MATCH THE BARNETT WELL DATA
WITH THE TAD MODEL
This example is provided to demonstrate the Formation thickness, h, ft 300
viability of field data analysis by the TAD Reservoir temperature, T, R 565.67
model. We consider the Barnett field data an- Distance to boundary parallel to well (1/2 well spacing), 275
xe, ft
alyzed by Brown et al. (2011) by using the Inner reservoir size, ye, ft 90.3
TDP model. The details of the data are given Viscosity, ␮, cp 0.02
in Brown et al. (2011). Fig. 10 shows the The order of fractional derivative of time, ␣ 0.8
matching of the rate-normalized pseudopres- Phenomenological coefficient of anomalous diffusion, 0.13
k␣, md-day1-␣
sure [⌬m(p)/q] data by the TAD model and the Porosity – compressibility product of inner reservoir, 2.00E-04
results obtained from the match are given in (␾ct)␣,psi-1
Table 5. For comparison, the TDP-model Hydraulic fracture porosity, ␾F 0.38
Hydraulic fracture permeability, kF, md 1.00E⫹03
match obtained by Brown et al. (2011) is also
Hydraulic fracture total compressibility, ctF, 1.00E-04
shown in Fig. 10. Both the TAD and TDP psi-1
models yield a reasonable match and it is not Hydraulic fracture half-length, xF, ft 275
possible to choose one over the other. Because Hydraulic fracture width, wF, ft 0.01
Outer reservoir permeability, kO, md 1.00E-06
the TAD model does not require explicit ref-
Outer reservoir porosity, ␾O 0.04
erences to the intrinsic properties of the matrix Outer reservoir compressibility, ctO, 3.00E-04
and natural fractures, the TAD model requires psi-1
fewer regression parameters than the TDP
model.

Conclusions
Anomalous diffusion formulation is a viable alternative to dual-porosity idealization of tight, fractured
unconventional reservoirs. The versatility of the anomalous diffusion model to account for the complex
forms of the reservoir and velocity-field heterogeneity can be inferred from the wide variety of the flow
behaviors described by the asymptotic approximations of the model. Physical interpretations of the key
parameters of the anomalous diffusion model, such as the fractional diffusion exponent and the coefficient
of the flux law, can be deduced from comparisons with the models depending explicitly on the intrinsic
properties of the heterogeneous features of the reservoir, such as the dual-porosity models. The interpre-
tations of the pressure and flow rate behaviors predicted by the anomalous diffusion model are consistent
with the physical expectations and the results of the alternate models. The fact that the anomalous
diffusion formulation does not require explicit references to the intrinsic properties of the matrix and
fracture media relaxes the stringent requirements used in dual-porosity idealizations to couple matrix and
fracture flows. The trilinear anomalous-diffusion model presented in this paper is useful for performance
predictions and pressure- and rate-transient analysis of fractured horizontal wells in tight unconventional
reservoirs.

Acknowledgements
Portions of this work have been performed to meet the MS degree requirements of Ozlem Ozcan at the
Colorado School of Mines. The Turkish Petroleum Corporation (TPAO) has funded Ozlem Ozcan’s MS
studies. The work has been completed under Unconventional Reservoir Engineering Project (UREP). The
support of the member companies of UREP is also acknowledged.

Nomenclature
B Formation volume factor, rb/stb
CFD Hydraulic fracture conductivity, dimensionless
CRD Reservoir conductivity, dimensionless
12 SPE-170971-MS

ct Total compressibility, psi-1

dF Distance between two adjacent hydraulic fractures, ft
f Dual porosity transfer function
h Reservoir thickness, ft
hm Matrix block dimension, ft
k Permeability, md
kI Permeability of the inner reservoir, md
kf Natural fracture permeability, md
kF Hydraulic fracture permeability, md
kO Permeability of the outer reservoir, md
km Matrix permeability, md
k␣ Phenomenological coefficient of anomalous diffusion, md-day1-␣
Lh Horizontal well length, ft
m Pseudopressure, psi2/cp
nf Number of natural fractures
nF Number of hydraulic fractures
p Pressure, psi
q Volumetric rate, stb/day
rw Wellbore radius, ft
s Laplace parameter
t Time, hrs
T Temperature, oR
wF Hydraulic fracture width, ft
x Distance in x-direction, ft
xe Reservoir size, x-direction, ft
xF Hydraulic fracture half-length, ft
y Distance in y-direction, ft
ye Reservoir size, y-direction, ft

Greek
␣ Order of fractional derivative of time
␣O,F Parameter defined in the model
␤O,F Parameter defined in the model
⌬ Difference operator
␩ Diffusivity, ft2/hr
␭ Flow capacity ratio (in TDP model)
␭␣ Phenomenological coefficient (k␣/␮, in TAD model)
␮ Viscosity, cp
␲ Pi constant
␾ Porosity
␻ Storativity ratio

References
Apaydin, O. G. 2012. New Coupling Considerations between Matrix and Multiscale Natural Frac-
tures in Unconventional Resource Reservoirs, PhD Dissertation, Colorado School of Mines, Golden, CO
Barenblatt, G.I., Zeltow, Y.P., and Kochina, I., 1960. Basic Concepts in the Theory of Seepage of
Homogeneous Liquids in Fissured Rocks. J. Appl. Math. Mech. 24(5): 1286 –1303
SPE-170971-MS 13

Beier, R. A. 1994. Pressure Transient Model for a Vertically Fractured Well in a Fractal Reservoir.
SPE Formation Evaluation 9(2): 122–128. SPE 20582-PA
Brown, M., 2009. Analytical Trilinear Pressure Transient Model for Multiply Fractured Horizontal
Wells in Tight Unconventional Reservoirs. Masters Thesis. Colorado School of Mines. Golden, CO.
Brown, M., Ozkan, E., Raghavan, R., and Kazemi, H. 2011. Practical Solutions for Pressure Transient
Responses of Fractured Horizontal Wells in Unconventional Reservoirs. SPEREE 14(6): 663–676.
Camacho-Velazquez, R., de Swaan-Oliva, A., Vasquez-Cruz, M., 2011. Interference Tests Analysis in
Fractured Formations with a Time Fractional Equation. SPE 153615.
Caputo, M. 1967. Linear Models of Dissipation Whose Q is Almost Frequency Independent-II.
Geophys. J. Int. 13(5): 529 –539
Caputo, M. 1998. 3-dimensional physically consistent diffusion in anisotropic media with memory,
Matematica e Applicazioni Rendiconti Lincei, 9(2): 131–143
Chang, J. and Yortsos, Y. 1990. Pressure Transient Analysis of Fractal Reservoirs. SPE Formation
Evaluation 5(1):
Dassas, Y. and Duby, Y. 1995. Diffusion toward Fractal Interfaces, Potentiostatic, Galvanostatic, and
Linear Sweep Voltammetric Techniques, Journal of The Electrochemical Society, 142(12): 4175–4180
Flamenco-Lopez. F. and Camacho-Velazquez, R. 2003. Determination of Fractal Parameters of
Fracture Networks Using Pressure Transient Data. SPE Reservoir Evaluation & Engineering 6(1). SPE
82607-PA.
Fomin, S., Chugunov, V. and Hashida, T. (2011), Mathematical modeling of Anomalous Diffusion in
Porous Media, Fractional Differential Calculus, 1, 1–28
Gefen, Y., Aharony, A. and Alexander, S. 1983. Anomalous Diffusion on Percolating Clusters, Phys.
Rev. Lett., 50(1): 77–80.
Javadpour, F., Fisher, D., and Unsworth, M. 2007. Nanoscale Gas Flow in Shale Gas Sediments.
Journal of Canadian Petroleum Technology, 46(10): 55–61
Kazemi, H., 1969. Pressure Transient Analysis of Naturally Fractured Reservoirs with Uniform
Fracture Distribution. SPE J. 9(4): 451–462
Kuchuk, F. and Biryukov, D., 2013. Pressure Transient Tests and Flow Regimes in Fractured
Reservoirs. Paper SPE 166296 presented at the SPE Annual Technical Conference and Exhibition, 30
September-2 October, New Orleans, Louisiana, USA
Kuchuk, F. and Biryukov, D., 2014. Pressure Transient Behavior of Continuously and Discretely
Fractured reservoirs. SPE REEE 17(1): 82–97
Le Mehaute, A. and Crepy, G. 1983. Introduction to transfer and motion in fractal media: The
geometry of kinetics, Solid State Ionics, 1(9-10): 17–30
Molz, III, F. J., Fix, III, G. J. and Lu, S. S. 2002. A physical interpretation for the fractional derivative
in Levy diffusion, Applied Mathematics Letters, 15(7): 907–911
Nigmatullin 1984. To the Theoretical Explanation of the Universal Response, Physica Status Solidi B,
Basic Research, 123(2): 739 –745
Ozcan, O., 2014. Fractional Diffusion in Naturally Fractured Unconventional Reservoirs. Master of
Science Thesis, Colorado School of Mines, Golden, CO.
O’Shaughnessy, B. and Procaccia, I. 1985. Analytical solutions for diffusion on fractal objects, Phys.
Rev. Lett. 54(5): 455–458
Ozkan, E., Raghavan, R., and Apaydin, O. G. 2011. Modeling of Fluid Transfer from Shale Matrix to
Fracture Network, paper SPE 134830, presented at the 2010 SPE Annual Technical Conference and
Exhibition, Florence, Italy, Sept. 19 –22, 2010.
Ozkan, E., Brown, M., Raghavan, R., and Kazemi, H. 2011. Comparison of Fractured-Horizontal-
Well Performance in Tight Sand and Shale Reservoirs. SPEREE 14(2): 248 –259
14 SPE-170971-MS

Raghavan, R. 2011. Fractional derivatives: Application to transient flow, Journal of Petroleum

Science and Engineering, 80, 7–13
Raghavan, R. and Chen, C. 2013a. Fractional diffusion in rocks produced by horizontal wells with
multiple, transverse hydraulic fractures of finite conductivity, Journal of Petroleum Science and Engi-
neering, 109, 133–143.
Raghavan, R. and Chen, C. 2013b. Fractured-Well Performance Under Anomalous Diffusion, SPE
Res Eval & Eng, 16(3), 237–245
Sahimi, M. and Yortsos, Y. C. 1990. Applications of Fractal Geometry to Porous Media: A Review,
paper SPE 20476, presented at the SPE Annual Technical Conference and Exhibition, New Orleans, LA,
Sept. 23-26, 1990
Stehfest, H. 1970. Algorithm 368: Numerical inversion of Laplace transforms [D5], Communications
of the ACM, 13(1) 47–49
Warren, J.E. and Root, P.J. 1963. The Behavior of Naturally Fractured Reservoirs. SPE J. 3(3):
245–255
SPE-170971-MS 15

Appendix
Here we present the analytical details of the models used in this paper. Appendix A presents a brief summary of the trilinear
dual-porosity model for completeness. In Appendix B, we present the details of anomalous-diffusion solution for the inner
reservoir component of the trilinear anomalous-diffusion model. Appendix C outlines the derivation of the long-time
asymptotic approximation. For convenience, the derivations are presented in terms of scaled variables. The definitions of scaled
pressure drop and time are given, respectively, as follows:
(A1)

and
(A2)

where ␩I is the inner reservoir diffusivity defined by

(A3)

and
(A4)

Eqs. A1 and A2 correspond to the conventional dimensionless pressure and time for the trilinear dual-porosity model. For
the trilinear anomalous-diffusion model, ␭I (Eq. A4) and ␩I (Eq. A3) refer to ␭␣ (Eq. 3) and ␩␣ (Eq. A6), respectively, where
kI ⫽ k␣ is different from the conventional Darcy permeability and has the units of L2T1-␣ instead of L2. Therefore, pD and tD
are not dimensionless quantities for the trilinear anomalous-diffusion model. We also define the following dimensionless
distance and dimensionless hydraulic-fracture thickness, respectively:
(A5)

where ␰ is x or y, and
(A6)

where Wf is the width of the hydraulic fracture.

Furthermore, dimensionless fracture and reservoir conductivities are defined, respectively, by
(A7)

and
(A8)

and the diffusivity ratios for the hydraulic fracture and the outer reservoir are given, respectively, by
(A9)

and
(A10)

where ␩F and ␩0 are the diffusivities of hydraulic fracture and the outer reservoir, respectively.
The solutions given in below are expressed in terms of Laplace transforms. The overbar symbol is used to indicate the
dimensionless pressure in the Laplace-transform domain and s is the Laplace-transform parameter with respect to dimension-
less time, tD. The numerical evaluations of the Laplace domain solutions require a numerical Laplace inversion algorithm. In
this study, the numerical Laplace inversion algorithm proposed by Stehfest (1970) has been used.
16 SPE-170971-MS

Appendix A
Trilinear Dual-Porosity Model

The trilinear dual-porosity model (Brown, 2009, Brown et al., 2011, and Ozkan et al., 2011) couples the solutions for three
linear flow regions (Fig. 1); the solution for the outer reservoir,
(A11)

for the dual-porosity inner reservoir,

(A12)

and for the finite-conductivity hydraulic fracture,

(A13)

Then, the dimensionless wellbore pressure is obtained from Eq. A11 at xD ⫽ 0 (intersection of the wellbore and the
hydraulic fracture) as follows:
(A14)

Eq. A14 has a nested structure. The ␣F term in Eq A14 carries the properties of all three flow-regions, outer reservoir, inner
reservoir, and the hydraulic fracture, into the solution and is given by
(A15)

where
(A16)

(A17)

(A18)

and
(A19)

In Eq. A19, f(s) is the dual-porosity transfer function representing fluid transfer from matrix to fracture network given by
(A20)

For tight unconventional reservoirs, the dual porosity model considering transient fluid transfer from matrix to fractures
proposed by Kazemi (1969) is more appropriate that the model assuming pseudo-steady fluid-transfer proposed by Barenblatt
et al. (1960) and Warren and Root (1963).
SPE-170971-MS 17

Appendix B
Trilinear Anomalous-Diffusion Model

The outer reservoir solution of the trilinear anomalous-diffusion model is the same as that for the trilinear dual-porosity model
(Brown, 2009, Brown et al., 2011, and Ozkan et al., 2011) and will not be repeated here. To derive the linear flow solution
for the naturally fractured inner reservoir, we start with the temporal-fractional anomalous diffusion equation given by Eq. 5.
Upon integrating the Eq. 5 from 0 to xF and assuming that ⭸pI/⭸y ⫽ f(x) and ⭸pI/⭸t ⫽ f(x), we obtain the following equation
for the flow of a slightly compressible fluid in the inner reservoir in terms of the scaled variables defined in Eqs. A1 through
A10:
(A21)

The continuity of flux at the boundary of the inner and outer reservoirs requires
(A22)

Applying the Laplace transformation to Eq. A21 and A22 with respect to tD yields, respectively,
(A23)

and
(A24)

Substituting Eq. A24 into Eq. A23, and noting from Eq. A11 that
(A25)

gives
(A26)

where we have used (pID)XD ⫽ 1 ⫽ pID by the linear flow assumption and defined
(A27)

The outer and inner boundary conditions for the inner reservoir solution are given, respectively, by
(A28)

and
(A29)

Then, the solution of the boundary value problem defined by Eqs. A26, A28, and A29 is
(A30)

To complete the solution of the trilinear anomalous-diffusion problem, the inner reservoir solution (Eq. A30) must be
coupled with the hydraulic fracture solution. The diffusivity equation for the hydraulic fracture is given, in terms of scaled
variables, by
(A31)

Using the continuity of flux condition at the boundary of the hydraulic fracture and the inner reservoir given by
(A32)

and applying the Laplace transformation to Eq. A31 yields

(A33)
18 SPE-170971-MS

In Eq. A33, we have substituted from Eq. A30,

(A34)

used pFD ⫽ (pFD)yD ⫽ wD/2 by the linear flow assumption in the hydraulic fracture, and defined
(A35)

and
(A36)

Assuming no flow across the fracture tip, the outer boundary condition of the hydraulic fracture is given by
(A37)

and the constant rate passing from the fracture to the horizontal well yields the following inner boundary condition:
(A38)

The solution of the boundary value problem specified by Eq. A33, A37, and A38 gives the following expression for the
scaled pressure of the hydraulic fracture:
(A39)

Assuming that the pressures of the hydraulic fracture and the wellbore are the same at the interface between the hydraulic
fracture and the wellbore (xD ⫽ 0), the wellbore pressure is obtained from Eq. A39 as follows:
(A40)

A flow-choking skin due to the circular intersection between the fracture and the horizontal well (Brown, 2009, Brown et
al., 2011, and Ozkan et al., 2011) can be added to Eq. A40.
SPE-170971-MS 19

Appendix C
Intermediate- and Late-Time Asymptotic Approximation of the TAD Solution

To obtain the asymptotic forms of the TAD solution for intermediate and late times, we consider
(A41)

and apply the following limits:

(A42)

The limiting forms of the terms involved change based on whether ␣ ⫽ 0 or ␣ ⫽ 0 and xeD ⬎ 1 or xeD ⫽ 1. Therefore,
the resulting expressions given in Table 3 are listed with the conditions on ␣ and xeD. The constant used in the expressions in
Table 3 are shown in Table A1. The details of the derivations will not be given here for space limitations. They can be found
in Ozcan (2014).

Table A1—CONSTANTS USED IN THE ASYMPTOTIC APPROXIMATIONS IN TABLE 3