Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269204125

Discrete Fracture Network Simulation of Production Data from Unconventional


Wells

Conference Paper · January 2014


DOI: 10.15530/urtec-2014-1923802

CITATIONS READS
5 674

5 authors, including:

Thomas Doe Clifford Knitter


University of Washington Seattle Golder Associates
76 PUBLICATIONS 798 CITATIONS 6 PUBLICATIONS 78 CITATIONS

SEE PROFILE SEE PROFILE

Cristian Enachescu
MOL
23 PUBLICATIONS 198 CITATIONS

SEE PROFILE

All content following this page was uploaded by Cristian Enachescu on 16 March 2016.

The user has requested enhancement of the downloaded file.


URTeC: 1923802

Discrete Fracture Network Simulation of Production Data from


Unconventional Wells
Thomas Doe*, Chunmei Shi, Golder Associates Inc., and Cristian Enachescu, Golder
Associates GmbH
Copyright 2014, Unconventional Resources Technology Conference (URTeC)

This paper was prepared for presentation at the Unconventional Resources Technology Conference held in Denver, Colorado, USA, 25-27 August 2014.

The URTeC Technical Program Committee accepted this presentation on the basis of information contained in an abstract submitted by the author(s). The contents of this paper
have not been reviewed by URTeC and URTeC does not warrant the accuracy, reliability, or timeliness of any information herein. All information is the responsibility of, and, is
subject to corrections by the author(s). Any person or entity that relies on any information obtained from this paper does so at their own risk. The information herein does not
necessarily reflect any position of URTeC. Any reproduction, distribution, or storage of any part of this paper without the written consent of URTeC is prohibited.

Summary

Pressure transient analyses using discrete fracture network models can provide insights to fracture length, stimulated
reservoir volume, and fracture-matrix interaction. These questions are critical to understanding the production
behaviors of unconventional reservoirs. This paper presents the results of Discrete Fracture Network (DFN)
simulations using (1) finite-conductivity fractures with constant spacing and size and (2) more realistic fracture
networks based on stimulated natural fracture networks. These are compared with production data from Eagle Ford
wells that are deconvoluted to produce equivalent constant-rate pressure derivatives.
Simulations of finite-conductivity hydraulic fractures with constant length and spacing produce distinct flow
regimes. When the matrix has low permeability, the pressure derivative has half-slope, linear-flow periods
corresponding to flow from the fractures and then the flow from the matrix blocks to the fractures. The transition
between fracture and matrix flow has a steeper, near unit slope, which represents the depletion of the fracture before
significant matrix flow occurs. When the matrix is relatively permeable, the effect of the matrix appears before the
end of the fracture-only flow and the result is quarter-slope bilinear flow. Intermediate matrix permeabilities
produce transitional behaviors between trilinear and bilinear flow. The transitions between fracture and matrix flow
provide a basis for assessing fracture length, provided the matrix permeability is known; however, the transitions
occur very early in production, within a day or much less time.
A DFN model provides a more realistic simulation of production from the Eagle Ford shale in southwest Texas.
The model uses fracture orientations and intensities from seismic and outcrop data. The fractures are assumed to be
non-conductive prior to stimulation. The stimulated network model is calibrated to microseismic data. Unlike the
simple simulations with constant matrix block sizes, realistic networks have a distribution of block sizes that
produce derivative slopes between on half and one.. Production data from Eagle Ford wells were analyzed using
pressure deconvolution methods to obtain equivalent constant-rate pressure derivatives. Late time derivatives show
similar slopes to the complex DFN fracture simulations.
The quality of the early-time deconvolution transformations of normal production data is not adequate for assessing
fracture length due to primarily to low early-time data density. Understanding fracture geometries and behaviors
may require several days of build-up data from wells that are shut in after production has started.

Introduction

The production of oil from unconventional reservoirs like Texas’s Eagle Ford shale has significantly boosted the
domestic oil production of the United States. Despite the success in developing these resources, there may be room
URTeC 1923802 2

for significant improvement in optimizing the recovery from these formations. This improvement can be achieved
from a better understanding of the geometry of stimulated fracture networks.
The main motivation of this work is the determination of fracture size, especially length. Length is critical for
designing the well spacing. Well spacing that is too large may leave unrecovered oil, while spacing that is too close
may result in duplicate stimulation of the same rock volume. In principle, microseismic monitoring tracks the paths,
both vertical and lateral, of hydraulic fracture growth. While microseismic events are associated with the fracture
development, it is not clear that the fracture surface defined by microseismic events is the same as the productive
surface of the hydraulic fractures. Fractures may close after the end of treatments if they are not propped, as
proppants likely do not reach small aperture regions of the fracture near its advancing tip.
Pressure transient analysis may provide an independent method for obtaining hydraulic fracture length. The
approach uses the changes in rate and pressure over time either in production data or in well tests to evaluate
formation properties as well as to assess the geometry of the flow system. Determining the length of hydraulic
fractures from pressure and flow transient data is a well-established method of pressure transient analysis in
conventional reservoirs.
This paper addresses the question of whether pressure transient methods can be applied to unconventional reservoirs
using discrete fracture network (DFN) models. The main topics are the following:
• Hydraulic fractures with constant spacing and constant length.
• Realistic stimulated natural fractures with variable size and matrix-block sizes
• Comparison of simulations to deconvoluted, equivalent-rate production data
The examples in this paper are drawn from the Eagle Ford shale of southwest Texas.

Background

The classic theories of pressure transient behavior in hydraulically fractured wells involve either fractures with
infinite conductivity (Gringarten et al., 1975) or a finite conductivity (Cinco-Ley and Samaniego, 1981). The
former produces a half-slope, linear flow derivative that changes into radial flow as the pressure influence moves
beyond the tip of the fracture. Finite conductivity fractures classically produce quarter-slope bilinear flow where
there is a pressure transient in both the fracture and the matrix.
Unconventional reservoirs differ from the classic cases by using horizontal wells with multiple hydraulic fracture
stages. This produces additional transient flow regimes that are associated with the depletion of matrix volumes
between hydraulic fractures and ultimately flow from outside the footprint of the well’s hydraulic fractures.
The trilinear flow model is an emerging concept for pressure transient behavior of unconventional wells (Song and
Ehlig-Economides, 2011; Ehlig-Economides et al., 2012; Apiwathanasorn and Ehlig-Economides, 2012; Brown et
al., 2009; Ozkan et al., 2011, Medeiros et al., 2010; Moghadam et al., 2010). The three linear flow periods of this
conceptual model include the following:
1. Linear flow in either finite-conductivity hydraulic fractures or linear flow from natural fractures to infinite
conductivity hydraulic fractures (Fracture Linear Flow)
2. Linear flow from matrix to either the natural fractures or the hydraulic fracture (Matrix Linear Flow)
3. Linear flow from outside the hydraulic fracture footprint of the entire well (Formation Linear Flow)
In a reservoir with a low permeability matrix, each linear flow period is separated by a steeper-slope depletion
before the next linear flow period starts. The first depletion occurs in the hydraulic fractures, the second in the
matrix between the fractures, and the final depletion is the matrix between wells. Figure 1 illustrates these linear
flow periods along with DFN model visualizations of the pressure. Due the proprietary nature of the data used in
this paper, not all units or pressure scales are given. They are based, however, on actual production data from Eagle
Ford wells.
URTeC 1923802 3

(a)

(b)

Figure 1. (a) Trilinear flow in an unconventional reservoir. The fracture linear flow may represent either finite-
conductivity hydraulic fractures or natural fracture system with infinite conductivity hydraulic fractures. (b) DFN
visualization of pressure.

DFN Simulations on Regularly-Spaced, Constant-Length Fractures

The simulations reported in this paper use Golder Associates DFN code, FracMan® (Dershowitz et al., 2010; Cottrell
et al., 2013). The FracMan simulations assign a permeability and aperture to the hydraulic fracture; hence, these are
finite-conductivity rather than infinite-conductivity fractures. If the fractures have infinite conductivity then there is
no fracture-flow regime and no end time for fracture flow that would indicate fracture length. For simple, finite-
conductivity hydraulic fractures, the fracture half-length expresses itself in two ways:
• It controls the transitions from early-time fracture-dominated flow to matrix-dominated flow.
• It affects the matrix-dominated flow through the fracture area, which is the product of fracture half-length
and fracture height. However, making use of the matrix-dominated flow period to get fracture half-length
does require knowing both the fracture height and the matrix permeability.
The results presented here focus on using a simple model of a single hydraulic fracture to investigate the
interrelationships of fracture half-length and matrix permeability. The matrix is assumed to be a block with a
thickness that extends 20 meters (m) from the hydraulic fracture surface. This block thickness is equivalent to a
hydraulic facture spacing of 40 m.
URTeC 1923802 4

Hydraulic fractures for the simulations presented in this paper use a height of 60 m and the simulations use constant-
rate production with separate realizations for five fracture half-lengths: 50, 100, 150, 200, and 400 m. For each
fracture half-length, we perform a simulation for each of six matrix permeability values: 2.00E-06, 1.00E-05, 5.00E-
05, 2.50E-04, 1.25E-03, and 6.25E-03 millidarcys, as well as a simulation with no matrix. The range of matrix
values is based on an assumed Eagle Ford permeability of 5.00E-05 millidarcys bracketed by a range of higher and
lower values (Orangi et al., 2011). The reservoir properties of the hydraulic fractures are based on early-time PI
values from production data.

Simulation of Matrix Flow in a DFN Model

As a DFN model, FracMan simulates flow in fractures only. That said, there are two methods for including matrix
in DFN simulations (Figure 2). One is called a “DFN dual porosity” and the other is “matrix fracture”. For the
simulations in this paper, DFN dual porosity is used for simple hydraulic fractures, and the matrix fracture is used
for complex hydraulic fractures.
DFN dual porosity associates a one-dimensional flow solution with each finite element on each fracture face
(Dershowitz and Miller, 1995). This creates a matrix “block” attached to both faces of the fracture (Figure 2). The
approach is very efficient but it works mainly if the flow from the matrix is only to the closest fracture face. It does
not account for possible overlap in space of matrix blocks from different fractures. The DFN dual porosity approach
has the numerical advantage of allowing fine discretization of the matrix near the fracture face while leaving the
fracture with a relatively coarse mesh. The fine discretization of the matrix solution improves the accuracy of the
matrix flow solution especially in early times. We defined the DFN dual porosity using a slab with a thickness of
20 m that represented the approximate 40-m spacing of the hydraulic fractures. The flow symmetry of the matrix
between adjacent hydraulic fractures has the effect of a no-flow boundary on the matrix-flow behavior. As the
matrix block is the same length as a hydraulic fracture, the model has no matrix volume outside the SRV.
Therefore, the model does not produce the last of the trilinear flow model’s flow regimes, that is, flow from outside
the SRV.
The matrix fracture approach (Figure 2) represents the matrix as an equivalent fracture with an aperture and
permeability that reproduce the pore volume and flow capacities of the matrix layer (Doe, et at. 2013). The aperture
of this “matrix fracture” is the matrix thickness times the porosity. This gives the matrix fracture the same pore
volume as the matrix it represents. The permeability of the matrix fracture is the matrix permeability divided by the
porosity, which preserves flow capacity of the matrix, or its kh. Simulations of complex fracture networks described
later in this paper use this method.
A major challenge of any simulator of unconventional reservoirs is a large contrast of properties between the
fractures and the matrix. This requires care in discretizing the matrix fracture near the fracture intersections.
FracMan includes a mesh editing capability that performs this refinement, but even with refinements to meter scale
or less, the solution will have some errors in the early time data.

Figure 2. Representing matrix in a DFN model – DFN dual porosity and matrix fracture
URTeC 1923802 5

Results of the Simple Hydraulic Fracture Model

Simple hydraulic fracture simulations reproduce the trilinear flow conceptual model for low matrix permeabilities.
The simulation cases included a set of fracture lengths and matrix permeabilities. Figure 3 shows the simulations for
a 400-m fracture half-length and three of the matrix permeability values. The 2 nanodarcy case in Figure 3 shows
fracture depletion prior to the onset of significant matrix flow. The fracture flow and the matrix flow both produce
half slope derivatives separated by a steeper derivative representing the effects of the end of the fracture. At the
other extreme, simulations using 1.2 microdarcys show significant matrix flow, which masks the end of the fracture.
The derivative slope is the characteristic one quarter of bilinear flow. An intermediate permeability case of
50 nanodarcys shows intermediate behavior between the trilinear flow of the low permeability case and the bilinear
flow of the high permeability case.
Further simulations with more permeability cases show a range of slopes between one half and one quarter. The
results of the simple model demonstrate that there is a continuum between trilinear and bilinear flow cases that is
controlled by the fracture–matrix permeability contrast. Trilinear flow occurs in low-permeability matrix situations
where the effects of the end of the fracture appear before there is significant flow from the matrix. With higher
matrix permeabilities, matrix flow commences prior to pressure perturbations reaching the end of the fracture. The
combination of fracture flow and matrix flow produces the well-known bilinear quarter slope. The likelihood of
bilinear flow is enhanced also by having longer fracture lengths or lower permeability hydraulic fractures (results
not presented here).

Derivative, Fracture Half Length 400-m


1.E+05
Matrix Linear Flow (MLF)
Fracture Boundary Flow

1.E+04
Derivative Normalized Pressure, psi/stb/day

Matrix k
Fracture Linear Flow (FLF)
1.E+03 mdarcys

No Matrix
1.E+02
2.00E-06
5.00E-05
Fracture Radial Flow (FRF) 1.25E-03
1.E+01

Matrix Boundary Flow

1.E+00

Fracture-Matrix Bilinear Flow

1.E-01
1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03
Time, days

Figure 3. Flow regimes from simple hydraulic fracture model.


URTeC 1923802 6

A major goal for the simulations was assessing whether pressure transient methods could yield useful information
on the fracture length. For each permeability value we could make a plot of the derivative curves with a range of
fracture lengths from 50 to 400 m (Figure 4). The end of the influence of the fracture was picked visually from the
deviation of each derivative curve from that of the derivative with the next larger hydraulic fracture length. Figure 4
plots the time of the end of fracture flow versus fracture length with a series of curves for different matrix
permeability values. The end of fracture flow appears to have a power law relationship with fracture size (Figure 5).
For high matrix permeabilities this power exponent is 0.25, which corresponds to the quarter slope of bilinear flow.
At low permeabilities the power exponent approaches 0.5, which is the exponent associated with fracture linear
flow. As we observed in the plots of Figure 4, intermediate permeabilities have power exponents between 0.25 and
0.5. In all the simulations, the end of fracture flow occurs very early in the well’s production. Only for the 400-m
fracture length case and the highest permeability value does the end of fracture flow occur after one day. Most of
the cases see the end of fracture effects occurring much earlier in the range of minutes to hours. The early cessation
of fracture flow effects suggests that it will be difficult to determine fracture length from standard production data,
which are not only very noisy in the first few days of production but also are generally not taken with frequencies
much less than a day. Alternatives for obtaining the early time data that reflect fracture length are discussed later in
this paper.

Matrix Permeability 5.0E-5 mdarcys, Derivative


1.E+05

1.E+04
Derivative Normalized Pressure, psi/stb/day

1.E+03 Fracture
Half Size

50-m
1.E+02 100-m
End of Fracture 150-m
Flow 200-m
400-m
1.E+01
Xf=100 m

1.E+00

1.E-01
1.E-05 1.E-04 1.E-03 1.E-02 1.E-01 1.E+00 1.E+01 1.E+02 1.E+03
Time, days

Figure 4. End of fracture flow for fracture lengths from 50 to 400 meters.
URTeC 1923802 7

End of Fracture Flow versus Fracture Size with Power Fits


300

250
y = 520x0.38
y = 820x0.45
200 y = 170x0.26
Fracture Half Length, m

y = 280x0.30

150 y = 1100x0.51

100

50

0
1.E-03 1.E-02 1.E-01 1.E+00 1.E+01
Time, Days

Figure 5. The end of fracture flow versus fracture size for different matrix permeability values with power-law fits.

Analysis of Production Data for Comparison with DFN models

The examples of simple hydraulic fracture behavior use constant rate production; however, unconventional wells are
usually produced with both variable rates and pressure. One method for calculating constant-rate equivalent
pressure data is the use of deconvolution methods. These methods have seen increasing application since 1998
(Onur and Reynolds, 1998; Levitan, 2003, Von Schroeter et al. , 2004) .
A significant advantage to the deconvolution approach is that it uses all of the rate and pressure data from multiple
production or well-test stages. Rather than focusing on one well test stage, such as a buildup, where the
superposition effects may be uncertain, the deconvolution approach uses all available pressure and rate information.
Figure 6 shows the deconvoluted pressure derivatives for nine Eagle Ford wells. The early time portions of the
deconvoluted plots are less useful than later time portions. The production data were taken with no greater than a
daily frequency. Thus, the accuracy of the derivative for the first several days is relatively poor. The most accurate
portions of the deconvoluted pressures and derivatives are the late time data from approximately 50 days to
1000 days. Most of these derivatives are stabilizing to straight-line logarithmic slopes. It is interesting to note that
these slopes have neither half-slope values nor unit-slope values, but lie somewhere in between. Half-slopes would
indicate linear flow in fractures, or more likely matrix at the late times when the derivatives are most valid. Unit
slopes at late times would likely indicate interference between hydraulic fractures. However, an analysis of the late-
time derivative values yields neither half nor unit slopes (Figure 6). Rather, the late time derivatives are
intermediate between linear flow and depletion effects with derivatives that range from 0.6 to 0.9 with an average of
0.73. The discrepancy between the observed late-time derivative slopes and those of simple models provides a
reason to explore more complex, realistic models.

DFN Simulations of Realistic Stimulated Fractures

Actual hydraulic fractures may be considerably more complex than the simple constant-size and constant-spacing
cases. Using information based on outcrop studies and lineaments based on seismic data, the Eagle Ford may have
natural fractures in two major sets, one striking N40E and the other N70E. Fracture lengths were extrapolated from
power-law plots of interpreted seismic lineaments. The resulting DFN model (Figure 7) was generated in FracMan.
URTeC 1923802 8

The natural fractures were assumed to be non-conductive; however, they may control the development of hydraulic
fracture networks, if they are appropriately oriented with respect to the in situ stress field. FracMan includes a
module for hydraulic fracture generation. The model performs a mass-balance calculation based on injected
volumes and elastic fracture opening (Dershowitz et al., 2010; Cottrell et al., 2013). The model assumes the greatest
horizontal stress is oriented N70oE. The model required very little calibration of the mechanical properties and
stress state to reproduce the form and extent of microseismic data, as shown in Figure 7.

1.E+02

Derivative Deconvoluted Pressure

1.E+01

1.E+00

1.E-01
1.E+00 1.E+01 1.E+02 1.E+03
Time, days

Figure 6. Rate-normalized deconvoluted pressure derivative for six Eagle Ford wells from the same lease.
Reference unit and half slope lines are depicted in gray.

Figure 7. Left: DFN model of Eagle Ford with unstimulated fracture networks. Right: Stimulated fracture network
with comparison to microseismic data.
URTeC 1923802 9

The simulations use a matrix fracture with a porosity of 9%, permeability of 50 nanodarcys, and matrix thickness of
60 m. Figure 8 shows two visualizations of the pressure drawdown one at early time and one at late time. The
early-time visualization shows the initial depletion of the hydraulic fractures themselves along with the depletion of
smaller matrix blocks that lie between closely spaced stimulated natural fractures. The later time visualization shows
the depletion of most of the matrix within the volume stimulated by hydraulic fracture treatment. The stimulated
fracture network then acts as a single sink for production. It is similar to a single simple hydraulic fracture except
the stimulated volume has an irregular outline with a relatively large surface area of pressure perturbation.

Figure 8. Pressure drawdown in a reservoir with realistic fractures with 50 nanodarcy matrix (50, 250, 500, and
1000 hours)
The pressure and the pressure derivative curves for production from a simulation of all 16 stages together appear in
Figure 9 along with the deconvolution pressure derivative for an Eagle Ford producing well. The derivative curve
has two major differences in comparison with those of simple hydraulic fractures with constant size and spacing.
First of all, the derivative curve lacks a distinctive early-time linear flow associated with the fractures. As the range
of block sizes produces different rates, there is no single distinct transition from fracture linear flow to fracture
depletion or matrix linear flow. A second distinction appears in the late time data where the derivative has a slope
that is neither the one-half value characteristic matrix linear flow nor a unit slope that would be characteristic of
matrix depletion. Rather, with the range of block sizes present there are components of both matrix production and
matrix depletion going on the same time.
Figure 10 shows the results of simulations for each of the 16 stages individually. The curves show a great deal of
variability in the early- and mid-time data, which reflects random variations in the sizes of the stimulated fractures
and of their associated matrix blocks. In late time, however, there is a convergence of the derivative curves for each
stage to a single straight line that has a slope between one half and one, similar to the results of simulations of
production from all 16 stages at the same time.
Realistic hydraulic fractures, as inferred for microseismic monitoring, do not have constant size and spacing. Rather
they define irregular volumes with variable fracture orientations, fracture sizes, and matrix block sizes. Production
from geomechanically generated hydraulic fractures does not follow the clear flow regimes that appear when the
fractures are more regular. The flow regimes associated with fracture linear, bilinear, and matrix linear flow tend to
URTeC 1923802 10

blur due to the variable production rates and depletion times associated with variably-sized matrix blocks. Late-time
behaviors produce derivative slopes that are between one-half and one, which is similar to the actual production
derivative produced by deconvolution of the production data.

100

10
Rate Normalized Derivative

0.1

k=5e-5 mdarcys; h=60m

0.01
1.00E-03 1.00E-02 1.00E-01 1.00E+00 1.00E+01 1.00E+02 1.00E+03
Time, Days

Figure 9. Pressure (blue) and pressure derivative (red) from DFN stimulated fracture model with comparison to
deconvoluted responses from an actual reference production well. Reference lines are unit and half slopes. Note the
simulations have late-time derivative slopes between 0.5 and 1.

100000

10000
Rate Normalized Derivative

1000

100

10

k=5e-5 mdarcys; h=60m

1
1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01 1.0E+00 1.0E+01 1.0E+02 1.0E+03
Time, Days

Figure 10. DFN simulation of 16 stages simulated separately. Reference lines are unit and half slopes.
URTeC 1923802 11

Discussion and Conclusions

Discrete fracture network (DFN) simulations can provide insights into the relative contributions of fracture and
matrix flow over the production life of the well. Simple hydraulic fracture networks of regularly spaced,
constant-sized fractures produce a range of behaviors from trilinear flow where the matrix has a low permeability to
bilinear flow where the matrix has a higher permeability. Simulations show that distinct flow regimes may be
associated with the hydraulic fractures, the flow from matrix to fractures, the interference between fracture stages,
flow from the formation to the entire region drained by the well, and ultimately interference between wells. The
lengths of simple fractures may, in principle, be determined from the timings of the transitions between these
regimes. In practice, however, the transitions from fracture to matrix flow occur within a day or much less.
Production data, especially in the early periods, usually do not have the frequency or resolution for fracture length
analysis. Obtaining data from the periods that are affected by fracture flow may require buildup pressure data taken
when a well is shut in after a period of production.
All of the simulations described in this paper use constant rate production in order to produce analyzable pressure
derivatives. Actual production data from wells in unconventional reservoirs is nearly always variable both in rate
and pressure. Pressure deconvolution methods are useful for converting production data into equivalent rate
pressure derivatives. When this method is applied to production data from some Eagle Ford wells, the late-time
derivative slopes lie between one half and one. Early time data from the deconvolutions are probably not reliable
and would need to be complemented by buildup data. Obtaining matrix permeability from well tests prior to
stimulation would greatly constrain the interpretation of post-stimulation data.
Realistic fracture networks may be constructed from outcrop and seismic data. Subjecting these to simulated
hydraulic fracture stimulation produces fracture networks that reproduce the footprints of microseismic data. These
realistic simulations do not have a single matrix block size, as in the simple simulations. Rather they have a range of
block sizes. This range of block sizes produces transient signatures that have derivative slopes between one half and
one, similar to the deconvoluted production data. These intermediate slopes are likely the result of a distribution of
rates and depletion times from these variably-sized matrix blocks.

Acknowledgements

The authors wish to acknowledge the contributions of our colleagues Christopher May and Stefan Rohs for
assistance with the flow simulations and the pressure deconvolution, and to Cliff Knitter, Bill Dershowitz, and
Emily Drew for their review and comments. We also gratefully acknowledge our clients for their permission to use
data from their studies in this paper.

References

Apiwathanasorn, S. and C. Ehlig-Economides, 2012, Evidence of reopened microfractures in production data


analysis of hydraulically fractured shale gas wells. SPE-162842.
Brown, M., E. Ozkan, R. Raghavan, and H. Kazemi, 2009, Practical solutions for pressure transient responses of
fractured horizontal wells in unconventional reservoirs. SPE 125043.
Cinco-Ley, H., and F. Samaniego, 1981, Transient pressure analysis for fractured wells, Journal of Petroleum
Technology, v. 33, p. 1749-1766.
Cottrell, M., H. Hosseinpour, and W. Dershowitz, 2013. Rapid Discrete Fracture Analysis of Hydraulic Fracture
Development in Naturally Fractured Reservoirs. Paper SPE-168843, URTEC-1582243, Proceedings, First
Unconventional Resources Technology Conference. Society of Petroleum Engineers.
Dershowitz, W. M. Cottrell, D. Lim, and T. Doe, 2010, A discrete fracture network approach for evaluation of
hydraulic fracture stimulation of naturally fractured reservoirs, 44th U.S. Rock Mechanics Symposium and
5th U.S.-Canada Rock Mechanics Symposium, June 27 - 30, 2010 , Salt Lake City, Utah, ARMA paper
10-475.
Dershowitz, W. and I. Miller, 1995. Dual porosity fracture flow and transport, Geophysical Research Letters,
Vol.22, No. 11, p. 1441-1444.
URTeC 1923802 12

Doe, T., A. Lacazette, W. Dershowitz, and C. Knitter, 2013. Evaluating the Effect of Natural Fractures
onProduction from Hydraulically Fractured Wells Using Discrete Fracture Network Models, Paper SPE
168823 /URTeC 1581931, Proceedings, First Unconventional Resources Technology Conference, Society of
Petroleum Engineers.
Ehlig-Economides, C., I. Ahmed, S. Apiwathanasorn, J. Lightner, B. Song, F. Vera, H. Xue, and Y. Zhang,
2012, Stimulate shale volume characterization: multiwall case from the Horn River Shale, II, Flow
Perspective. SPE-159546.
Gringarten, A., H. Ramey, and R. Raghavan, 1975, Applied pressure analysis for fractured wells, Journal of
Petroleum Technology, v. 27, p. 887-892.
Levitan, M, 2003, Practical application of pressure-rate deconvolution to analysis of real well test, SPE Annual
Technical Conference and Exhibition, Denver CO, SPE-84290
Medeiros, F., B. Kurtoglu, E. Ozkan, and H. Kazemi, 2010, Analysis of production data from hydraulically
fractured horizontal wells in shale reservoirs, SPE Res Eval & Eng v. 13, p. 559-568. SPE-110848-PA.
Moghadam,, S., L. Mattar, and M. Pooladi-Darvish, 2010, Dual porosity type curves for shale gas reservoirs,
CSUG/SPE 137535 Canadian Unconventional Resources and International Petroleum Conference, October
19 - 21, 2010, Calgary, Alberta, Canada.
Onur, M. and A. Reynolds, 1998, Numerical Laplace transform transformation of sampled data for well-test
analysis, SPE Reservoir Evaluation and Engineering, v. 1, p. 268-277.
Orangi, A, N. Nagarajan, M. Honarpour, and J. Rozenweig, 2011, Unconventional shale oil and gas-condensate
reservoir production, impact of rock. Fluid, and hydraulic fractures., SPE 140536, SPE Hydraulic Fracturing
Technology Conference, the Woodlands, Texas.
Ozkan E., M. Brown, R. Raghavan, and H. Kazemi, 2011, Comparison of fractured horizontal well performance
in tight sand and shale reservoirs. SPE Reservoir Evaluation & Engineering. SPE-121290.
Song, B., and C. Ehlig-Economides, 2011, Rate-normalized pressure analysis for determination of shale gas well
performance. SPE 144031.
Von Schroeter, T., F. Hollaender, and A. Gringarten, 2004, Deconvolution of well test data as a nonlinear total
least-squares problem, SPE Journal, v. 9, p. 375-390.

View publication stats

You might also like