Spe 175941 MS PDF

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

SPE-175941-MS

Parameters Affecting Load Recovery and Oil Breakthrough Time after


Hydraulic Fracturing in Tight Oil Wells
B. Zanganeh and M. Soroush, University of Calgary; J. D. Williams-Kovacs, University of Calgary and TAQA
North Ltd.; C. R. Clarkson, University of Calgary

Copyright 2015, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE/CSUR Unconventional Resources Conference held in Calgary, Alberta, Canada, 20 –22 October 2015.

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
After hydraulic fracture treatment, wells typically produce significant volumes of water, which is
primarily recovered fracturing fluid. The amount of fracturing fluid produced back is called load recovery.
Single phase water production may continue for several days before oil breakthrough from the formation
occurs. A wide range of load recoveries (15%– 80%) are reported in the literature, depending on various
reservoir, fracture, pumping and operational parameters. This study conducts a comprehensive numerical
and statistical analysis to identify the most important parameters affecting load recovery. The sensitivity
of oil breakthrough time to these parameters is also investigated.
A fit-for-purpose model was created using a commercial coupled reservoir flow/geomechanics simu-
lator to simulate fracturing treatment and flowback operations. Water and oil production trends, and flow
regimes observed in the model, are verified using rate-transient analysis (RTA). The effect of twenty
parameters including pumping rate, shut-in time, producing bottomhole pressure, production rate con-
straint, matrix porosity and permeability, matrix and hydraulic fracture compressibility, relative perme-
ability end points and capillary pressure on load recovery and oil breakthrough time is investigated by
utilizing Design of Experiment (DOE). The developed model considers non-uniform leak-off during
pumping, providing a better understating of water retention, while the application of the DOE method
allows the effect of various parameters and their interaction to be studied rigorously.
A critical finding is that the majority of total load recovery occurs within the first three months of
production. Furthermore, matrix permeability, shut-in time, connate water saturation and fracture com-
pressibility, were found to be the most dominant factors, which can impact the load recovery by up to
22%, 15%, 14%, and 13%, respectively. Matrix permeability, production rate constraint, fracture residual
oil saturation and shut-in time are four of the prominent parameters affecting oil breakthrough time.
Understanding the impact of operational parameters such as shut-in time and production rate can help
industry to implement recycling/disposal options and ultimately optimize load recovery. This can also
create significant environmental compliance benefits for operators in addition to benefitting operational
efficiency.
2 SPE-175941-MS

Introduction
Horizontal drilling coupled with multistage hydraulic fracturing has proven to be an effective solution for
producing oil and natural gas at economic rates from ultra-low permeability shale/tight oil and gas
formations. According to the U.S. Energy Information Administration (EIA), 4.2 million barrels per day
of crude oil, or about 49% of total U.S. crude oil production in 2014, were produced directly from tight
oil resources. In Canada, tight oil production doubled between 2011 and 2014, from 0.2 to 0.4 million
barrels per day, with most coming from Alberta and Saskatchewan (EIA, 2015).
Hydraulic fracturing is a method of well stimulation in which large volumes of fracturing fluid are
injected into rock formations at very high pressure to fracture the rock and create flow paths for stored
hydrocarbons (King, 2012). The majority of hydraulic fracture (HF) treatments currently pumped use
water-based fracturing fluids, which contain nearly 99% water, as well as chemical additives and
proppant. After HF treatment, the well is flowed back, typically producing large volumes of load fluid
over the first several weeks of production. Single phase water production is often observed at the onset
of flowback and may continue for several days before oil breakthrough from the formation occurs. The
amount of fracturing fluid produced back is called load recovery. The unrecovered fracturing fluid is
stored inside the HF or leaks off into the rock matrix during treatment and subsequent shut-in period. A
wide range of load recoveries (15– 80%) have been reported in the literature depending on a variety of
parameters mentioned above (EPA Research Study, 2010; King, 2012).
Flowback water may contain chemicals, salts and pollutants, and its proper treatment as well as
disposal/recycling is of fundamental importance in hydrocarbon production from tight and ultra-tight
formations. Trapped fracturing fluid inside the fractures and formation may also affect oil/gas recovery.
As a result, identifying parameters affecting load recovery can create significant environmental compli-
ance benefits for operators, in addition to benefitting operational efficiency.
Several authors have used numerical simulation to model flowback, and demonstrated that load
recovery is a function of reservoir permeability, relative permeability, capillary pressure, shut-in time,
drawdown rate, HF conductivity, producing bottomhole flowing pressure (BHFP) and fluid properties
(Tannich, 1975; Wang et al., 2010; Gdanski and Walters, 2010; Zanganeh et al., 2015; Wang and Leung,
2015). One common issue in these studies is that they assume that HF dimensions and properties (length,
width and conductivity) are constant. However, HF characteristics are strong functions of stress, geome-
chanical parameters such as Young’s modules and Poisson’s ratio, leak-off rate and pumping schedule.
Another issue encountered in previous studies is fracturing fluid saturation initialization in the model.
Because a significant volume of injected fracturing fluid leaks off into the rock matrix during HF
propagation and subsequent shut-in, it is critical to initialize the simulation model with an accurate
saturation distribution around HF, which can be done by accounting for non-uniform fracturing fluid
leak-off during propagation. It is expected that more leak-off will occur near the wellbore than the tip of
the HF. Proper modeling of the fracturing fluid leak-off and saturation distribution around the HF results
in a better estimate of oil breakthrough time (OBT) as well.
In this study, a fit-for-purpose model was developed using a commercial coupled reservoir flow/
geomechanics simulator that can model HF initiation and propagation (and associated changing fracture
dimensions), non-uniform water leak-off into the formation during treatment and shut-in, and the effect
of stress on HF properties. The effect of twenty parameters (factors) including pumping rate, shut-in time,
BHFP, liquid production rate constraint, HF conductivity, matrix porosity and permeability, matrix and
HF compressibility, relative permeability end points and capillary pressure on load recovery and OBT
(responses) was investigated by utilizing factorial DOE in the sensitivity analysis. The objective of this
paper is to identify the impact of these factors on responses, and investigate if these factors interact. In
other words, the application of the DOE method allows one to study the effect of various pumping,
production and reservoir parameters and their interaction effects simultaneously.
SPE-175941-MS 3

Methods

In this section coupled reservoir flow/geomechanics simulation, the approach for hydraulic fracture
modeling, the fit-for-purpose model used for this study and simulation parameters are discussed. The DOE
method as well as its corresponding terms and concepts are also reviewed.
Coupled Reservoir Flow/Geomechanics Simulation
In this study GEM®, a numerical reservoir simulator developed and commercialized by Computer
Modeling Group (CMG), and its coupled geomechanics module, was used. The coupling between
reservoir fluid flow and geomechanical deformation using this model is described in detail in Tran et al.
(2005). As illustrated in Fig. 1 (Tran et al., 2009), pressure and temperature equations are solved first in
the reservoir simulator, then the solution is passed to the geomechanics module where these variables
contribute as source terms in the deformation equation. Displacement, strain and stress are calculated
within this module, and then passed to the reservoir simulator via coefficients for porosity and permea-
bility. These two variables are used to calculate new pressure and temperature values which are re-sent
to geomechanics module to recalculate deformation, strain and stress. This process is iterative and
continues until convergence criteria are satisfied for a given time step.

Figure 1—Workflow for coupling between reservoir simulator and geomechanics module in GEM (Tran et al., 2009)

Hydraulic Fracture (HF) Modeling


Tran et al. (2009) used the modified Barton-Bandis (1990) model for evaluation of CO2 leakage during
sequestration operations due to rock failure and opening of fractures in the cap rock. Padmakar (2013)
used the same approach for modeling of diagnostic fracture injection tests (DFIT). The modified
Barton-Bandis model is used in this study for modeling of HF initiation and propagation during high
pressure injection of fracturing fluid. In this approach, a dual permeability model consisting of matrix
blocks and natural fractures is defined; the fractures are initially inactive. For certain stress conditions,
4 SPE-175941-MS

these fracture blocks are activated and act as HF blocks. Fig. 2 illustrates the path of fracture block
permeability (kf) during hydraulic fracturing and flowback versus normal effective stress, ␴=n, which is the
difference between minimum stress and pore pressure. The process starts at point A, where the fracture
block is inactive and has a very low value of permeability. ␴=n decreases during injection of the fracturing
fluid (due to increasing pore pressure). When ␴=n reaches the fracture opening stress (frs), the grid block
is converted to a HF block, and its permeability steps up to HF permeability (khf). As long as ␴=n is less
than zero, fracture permeability remains at khf (path DCE). After fracturing treatment, if ␴=n becomes
greater than zero and frs, fracture permeability steps from khf to fracture closure permeability (kccf), and
then follows the Barton-Bandis model (curve FG) to a residual value of closure permeability (krcf).
Fracture width is also calculated internally within the module (GEM User’s Guide).

Figure 2—Diagram illustrating fracture permeability pathway as a function of stress (GEM User’s Guide)

Fit-for-Purpose Simulation Model Setup


Fig. 3 is a plan view schematic of the model used in this paper. In order to reduce the simulation time,
only a 2-D single stage HF and its surrounding matrix are modeled. The model width (fracture spacing)
is 100 ft and model length is 800 ft. It is assumed that the HF can propagate in a single plane perpendicular
to the minimum horizontal stress (black line) leading to the formation of a bi-wing planar fracture. To
further simplify the model, an element of symmetry (yellow outline) was considered. Throughout this
paper, reported production/injection rates and load recovery pertain to this outlined region (i.e., ¼ of a
single HF stage). An estimate of a total well’s production/injection rates can be obtained by multiplying
the reported values by 4 and the total number of HF stages.
SPE-175941-MS 5

Figure 3—Plan view schematic of 2-D simulation model

Water, as the main component of slickwater fracturing fluid, is injected at a specified flow rate into the
perforation. After being shut-in for some time, the simulation model is run for one year.
Simulation model input parameters which are used for all cases (and are not sensitized to) are
summarized in Table 1. There is no free gas in the reservoir; BHFP is chosen such that it is always above
the bubble point and therefore no gas is liberated in the formation or fractures during production.

Table 1—Simulation parameters used for all cases


Input Parameter Value

Depth (ft) 8000


Reservoir temperature (oF) 200
Young’s modulus (psi) 3 ⫻ 106
Poisson’s ratio 0.25
Initial reservoir pressure (psi) 5000
Injection time (minutes) 72
Fracture opening stress (psi) ⫺500
Shmin (psi) 5600
Shmax (psi) 6500
Sv (psi) 8000

Design of Experiment (DOE)


Twenty parameters (factors) that impact load recovery and OBT, and their corresponding high and low
levels, are listed in Table 2. These levels are based on a review of literature and field conditions, as well
as engineering judgment. It must be noted again that injection/production rates and constraints are
downscaled for a single stage of HF and its element of symmetry. The factors include fracturing fluid
(water) injection rate, shut-in time, BHFP, liquid production rate constraint, matrix and HF compress-
6 SPE-175941-MS

ibility, matrix porosity and permeability, HF permeability, matrix and HF relative permeability end points
and exponents.

Table 2—Factors and their levels used for sensitivity analysis in CMOST
Symbol Factor Low Level High Level

F1 Water injection rate (bbl/d) 108 360


F2 Shut-in time (days) 1 10
F3 BHFP (psi) 500 3000
F4 Liquid production rate constraint (bbl/d) 1 100
F5 Matrix compressibility (psi⫺1) 1⫻10⫺6 3⫻10⫺5
F6 HF compressibility (psi⫺1) 1⫻10⫺6 1⫻10⫺4
F7 Matrix porosity (%) 4 10
F8 Matrix permeability (md) 0.0001 0.01
F9 HF permeability (md) 100 10000
F10 Matrix Swcon 0.1 0.4
F11 Matrix Soirw 0.1 0.4
F12 Matrix Krwiro 0.3 0.8
F13 Matrix Krocw 0.3 0.8
F14 Matrix Nw 3 5
F15 Matrix No 3 5
F16 HF Swcon 0.01 0.2
F17 HF Soirw 0.01 0.2
F18 HF Krwiro 0.7 1
F19 HF Krocw 0.7 1
F20 HF Nw 1 2

Relative permeability curves are generated based on the input endpoints and exponents using the Corey
(1954) model. Capillary pressure is calculated according to the following correlation (Gdanski et al.,
2009):
(1)

Where, Pc is capillary pressure in psi, ␴ ⫽ 72 (surface tension, dynes/cm), Sw is water saturation in


fraction, ␾ is porosity in fraction, k is absolute permeability in md and b1 ⫽ 1.86, b2 ⫽ 6.42, and b3 ⫽
0.5 (Gdanski et al. 2009).
The selected simulation outputs (responses) are load recovery after 3 months (R1), load recovery after
1 year (R2) and OBT (R3). A total number of 220 simulation runs are required to investigate all possible
combinations of the factors and their levels. This number is reduced to 128 runs (28) by utilizing fractional
factorial design within CMOST® (CMOST User’s Guide).
Main and Interaction Effects
In DOE, a Main Effect (ME) is the effect of an independent variable (factor) on a dependent variable
(response) averaging across the levels of any other factors (Awoleke et al., 2012). For example, the main
effect of matrix permeability ME (F8) with low and high levels of 0.0001 and 0.01 md on load recovery
is calculated as follows:
(2)

Where, Ravg(F8⫹) is the average of the observed load recoveries when matrix permeability is 0.01 md,
and Ravg(F8-) is the average of the observed load recoveries when matrix permeability is 0.0001 md. A
SPE-175941-MS 7

positive value for ME indicates that there is a direct relationship between the factor and the response, and
a negative value indicates an inverse relationship.
If the effect of one factor on the response changes due to the level of another factor, these factors are
said to interact. For example, if the effect of matrix permeability (F8) on load recovery changes depending
on the BHFP (F3), these two factors interact and the interaction effect (INT) between matrix permeability
and BHFP is calculated as follows (Awoleke et al., 2012):
(3)

Where,Ravg(F3⫹ⱍF8⫹) is the average of the observed load recoveries at the high level of BHFP given
that matrix permeability is at the high level.

Response Surface
The response surface methodology explores the relationships between factors and responses. The main
idea is to use the set of simulation runs in fractional factorial design to build a proxy (approximation)
model representing the original complex model. The most common proxy model takes a linear form of
a polynomial function (CMOST User’s Guide):
(4)

Where, R is the response (i.e., load recovery or OBT), a0 is the intercept, F1, F2, . . ., Fk are the factors
and a1, a2, . . ., ak are the coefficients (main effects) of the proxy model.
The magnitude of the coefficient is related to the importance of the parameter, with a large value
indicating an important parameter. The removal of statistically insignificant factors from the linear proxy
model results in the reduced linear model. A full description of the method to identify statistical
significance of each factor is given in CMOST User’s Guide. In this paper, a reduced linear proxy model,
with significance probability (␣) of 0.1 (default value of the software), correlating load recovery to
significant factors, is developed.

Results
In this section, typical hydraulic fracture propagation and saturation profiles during fracturing fluid
pumping using the fit-for-purpose model are presented. The sequence of flow regimes observed during
production is discussed. The completed DOE table containing response values is also summarized. The
main and interaction effects as well as possible linear proxy models are presented for load recovery and
OBT, respectively.

HF Propagation and Saturation Profiles


Typical HF propagation and water saturation profiles during injection, and at the end of pumping, are
shown in Fig. 4.As expected, more fracturing fluid leak-off occurs at the wellbore compared to the tip of
HF. Fracture half-length, fracture width and water distribution around HF are functions of input
parameters according to levels in Table 2.
8 SPE-175941-MS

Figure 4 —Hydraulic fracture propagation (HF blocks shown in red) and water saturation profile during injection

Flow Regimes
As determined using water rate normalized pressure (RNP) and its derivative (RNP’) (Fig. 5), the typical
sequence of flow regimes observed during production (flowback) is as follows:

Figure 5—Water RNP and RNP’ plot showing the sequence of flow regimes during the flowback operation
SPE-175941-MS 9

● Linear flow (0.5 slope) inside HF at the start of production. For a 3D model it is likely that radial
flow would be observed in the HF unless the HF length to height aspect ratio is very large.
● HF depletion (boundary dominated flow, unit slope) once the pressure response reaches the tip of
HF.
● Matrix linear flow (0.5 slope) at the start of matrix inflow. A non-linear trend is observed before
this flow regime indicating transition from HF boundary flow to fully developed matrix linear
flow. This transition can be very short depending on the HF conductivity and half-length.
● Matrix boundary dominated flow (unit slope) when the pressure response reaches the model
boundary (fracture spacing).
Observed flow regimes are consistent with field observations and analytical models for flowback
analysis (Williams-Kovacs and Clarkson 2014; Alkouh et al. 2014). The first two flow regimes occur at
very early production time, and depending on the model parameters and time-step size may not be
observed in all cases. In a future paper, numerical simulation with the same approach will be used to
history match field flowback data.

DOE Table
The completed design table with the calculated values for the responses, i.e. load recovery after 3 months
(R1), load recovery after 1 year (R2) and OBT (R3), is presented in Table 3. Note that ‘⫺1’ and ‘1’
represents the low and high level of a factor, respectively.
10 SPE-175941-MS

Table 3—Completed design table

Using the values provided in Table 3, the main effect of all factors on responses can be calculated
according to Eq. 2. For example, the main effect of matrix permeability on R1 is calculated as follows:
SPE-175941-MS 11

Load Recovery
Comparing the results for load recovery after 3 months and 1 year, it is observed that most of the load
recovery occurs during the first 3 months of production. Therefore, load recovery after 3 months is a good
estimate of the total load recovery, and only R1 responses are further analyzed. The range of variation in
load recovery after 3 months is between 7.06% and 93.52%.
Fig. 6 illustrates the calculated main effects of all twenty factors on R1. For example, shut-in time has
a negative effect; as it increases from 1 day to 10 days, load recovery is reduced from 61% to 46%. Similar
interpretations can be made for other factors.

Figure 6 —Main effects of all twenty factors on R1

The most significant factors impacting load recovery are matrix permeability, shut-in time, matrix
Swcon, HF compressibility, matrix porosity, HF permeability, BHFP, production rate constraint and matrix
Krocw, respectively. The effect of other factors including injection rate, matrix compressibility, HF relative
permeability end points and exponents are negligible.
No noticeable interaction effect is observed for R1. This indicates that a linear proxy model similar to
Eq. 4 can be a reasonable representative of the response surface. Eq. 5 (below) is the reduced linear proxy
12 SPE-175941-MS

model correlating 3 months load recovery (R1) to significant factors with R-squared⫽0.9. Based on ␣
value of 0.1, only 10 out of 20 factors are considered as significant effects.
(5)

The above correlation can be used to estimate load recovery in the field under different scenarios. For
example, if matrix permeability ⫽ 0.001 md, Swcon⫽ 0.3, Krocw⫽ 0.5, No⫽ 3, HF compressibility ⫽
1⫻10⫺5, porosity ⫽ 7%, HF permeability ⫽ 1000 md, rate constraint ⫽ 50 bbl/d, BHFP ⫽ 1000 psi and
shut-in time ⫽ 5 days, one can calculate load recovery by substituting these values in Eq. 5 as follows:

The tornado plot in Fig. 7, which was generated based on the reduced linear proxy model (Eq. 5),
shows the relative importance of different factors on load recovery.

Figure 7—Tornado plot based on the reduced linear proxy model (Eq. 5) showing the relative importance of different factors on R1.
Maximum and minimum are the maximum and minimum response (R1) values according to Eq. 5

Oil Breakthrough Time


Response values of OBT (R3) in Table 3 are analyzed in a similar manner to load recovery. Oil
breakthrough can occur as soon as production starts or may be delayed by several days. In this study, OBT
values varied from 0 to 273 hours (greater than 11 days).
Fig. 8 illustrates the main effect of all twenty factors on OBT (R3) and indicates that matrix
permeability and production rate constraint are the most significant factors, both of which have inverse
SPE-175941-MS 13

effects on OBT. For example, as matrix permeability increases from 0.0001 to 0.01 md, mean OBT
reduces from 49 hours to 5 hours. Similar interpretations can be made for other factors.

Figure 8 —Main effects of all twenty factors on R3

Interaction effects of the significant factors for OBT are illustrated in Fig. 9.Each panel inside Fig. 9
plots the mean OBT versus one factor at different levels of another factor. Parallel lines indicate no
interaction, while non-parallel lines demonstrate interaction between two factors. For instance, matrix
permeability interacts with all other factors, having the most significant interaction with production rate
constraint.
14 SPE-175941-MS

Figure 9 —Interaction effects for R3

Referring to Fig. 9 – Panel 1, it is observed that for a high value of matrix permeability (0.01 md), the
effect of production rate constraint on OBT is negligible, and its increase from 1 bbl/d to 100 bbl/d causes
only a small decrease (8 hours) in mean OBT. Conversely, when the matrix permeability is low (0.0001
md), changing production rate constraint from 1 bbl/d to 100 bbl/d results in a large drop of mean OBT
(69 hours). Similar behaviour is observed for shut-in time (Fig. 9 – Panel 2).
Considering the large number of interactions between different factors, it is not possible to fit a linear
proxy model through OBT responses with an acceptable R-squared value.
Discussion and Recommendations
The following key observations are made for load recovery based on Figs 6 and 7 and Eq. 5:
● Fracturing fluid leak-off is the main mechanism controlling load recovery. In other words, load
recovery is more a function of fluid flow inside the matrix rather than storage and fluid flow inside
the HF.
● Matrix permeability and shut-in time have a negative effect on load recovery. Larger values of
matrix permeability or shut-in time result in higher fracturing fluid leak-off into the matrix, making
it more difficult to flowback during production.
● Flowback of the fracturing fluid stored inside the HF is easier than the fluid leaked off into the
matrix. Based on the direct relationship between porosity and compressibility (␾1 ⫽ ␾1eC(P ⫺P1))
2

during injection (P2 – P1 ⬎ 0), higher HF compressibility results in larger fracture storage, and
subsequently more stored fluid inside HF and better load recovery.
● Matrix Swcon and matrix porosity have positive and negative effects on load recovery, respectively.
Higher Swcon (or lower matrix porosity) results in lower leak-off of fracturing fluid into the matrix
and better load recovery.
SPE-175941-MS 15

● More drawdown inside the HF results in higher load recovery. This can be accomplished by
increasing HF permeability or production rate constraint as well as decreasing BHFP.
● Increasing matrix No or decreasing matrix Krocw results in lower relative permeability of oil and
subsequently higher relative permeability of water at each specific water saturation. Therefore,
production of leaked off water is easier.
From the above results, it can be observed that operational parameters including shut-in time, BHFP
and production rate constraint have a large impact on load recovery. Contrary to matrix and hydraulic
fracture properties, these parameters can be controlled in the field to optimize the load recovery. If lower
load recovery is preferable due to recycling/disposal limitations, the following actions are recommended:
● Increase shut-in time
● Use higher BHFP during flowback operation
● Produce at a low rate during flowback operation
In some cases, higher load recovery is desirable. The unrecovered fracturing fluid trapped inside the
HF or matrix may have a negative effect on long term production performance of the well. Also, the
recovered fracture fluid can be recycled and utilized in future fracture treatments creating significant
environmental benefits (minimized demand for potable water). In this case it is suggested to:
● Start the flowback operation immediately after the fracturing treatment
● Create higher drawdown inside the HF by using a lower BHFP
● Produce at the maximum possible rate during the flowback operation
Based on the aforementioned interaction effects for OBT (Fig. 9), it can be controlled in very tight
(nano-darcy permeability) formations by changing operational parameters, shut-in time and production
rate. However, there is considerable interaction between these two parameters. When the production rate
is high (100 bbl/d), variation of shut-in time does not affect OBT. On the other hand, when a low
production rate is used, increasing shut-in time from 1 day to 10 days, speeds up OBT from 60 hours to
30 hours (Fig. 9 – Panel 3).

Conclusions
In this work, a coupled reservoir flow/geomechanics model was used to simulate oil and water production
during flowback operation. Observed production trends and the sequence of flow regimes were consistent
with field data and analytical methods presented by several authors. The model was then used to estimate
load recovery and OBT, and the effects of 20 parameters on these responses were examined by utilizing
fractional factorial design.
For all simulation models, most of the load recovery occurred during the first 3 months of production.
The most important parameters affecting load recovery are matrix permeability, shut-in time, matrix
Swcon, HF compressibility, matrix porosity, HF permeability, BHFP and production rate constraint. Strong
negative effects of matrix permeability and shut-in time, implies the important role of leak-off on load
recovery. A linear correlation was also developed for a quick estimate of load recovery in the field.
OBT was shown to be a strong function of matrix permeability, production rate constraint, shut-in time
and residual oil saturation in HF. However, clear interaction effects were observed between these
parameters. For example, when the matrix permeability is high enough (0.01 md) the effect of other
factors becomes negligible.
The impact of operational parameters such as shut-in time, BHFP and production rate constraint is
great, and modifications can be made in the field to optimize the load recovery or OBT. For example, if
lower load recovery is desirable due to recycling/disposal limitations, it is recommended to increase
shut-in time or use higher BHFP during flowback operation.
16 SPE-175941-MS

In a future paper, numerical simulation, using a similar approach, will be used to history match field
flowback data and examine fracturing fluid properties on load recovery.

Nomenclature
Variables
ak ⫽ Proxy model coefficient
F ⫽ Factor
frs ⫽ Fracture opening stress, psi
INT ⫽ Interaction effect
k ⫽ Absolute permeability, md
kccf ⫽ Hydraulic fracture closure permeability, md
kf ⫽ Fracture block permeability, md
khf ⫽ Hydraulic fracture permeability, md
krcf ⫽ Residual value of closure permeability, md
Krocw ⫽ Oil relative permeability at connate water saturation, fraction
Krwiro ⫽ Water relative permeability at irreducible oil saturation, fraction
m ⫽ Slope, dimensionless
ME ⫽ Main effect
No ⫽ Oil exponent, dimensionless
Nw ⫽ Water exponent, dimensionless
Pc ⫽ Capillary pressure, psi
R ⫽ Response value
Ravg ⫽ Average response value
R1 ⫽ Load recovery after 3 months, %
R2 ⫽ Load recovery after 1 year, %
R3 ⫽ Oil breakthrough time, hours
Shmax ⫽ Maximum horizontal stress, psi
Shmin ⫽ Minimum horizontal stress, psi
Soirw ⫽ Irreducible oil saturation for water-oil table, fraction
Sv ⫽ Vertical stress, psi
Swcon ⫽ Connate water saturation, fraction

Greek Variables
␾ ⫽ Porosity, fraction
␴=n ⫽ Normal effective stress, psi
␴ ⫽ Surface tension, dynes/cm
␣ ⫽ Significance probability, dimensionless

Acknowledgements
The authors would like to thank Computer Modeling Group for providing CMG software and Mr. Varun
Pathak for his valuable advice. The authors would also like to acknowledge the sponsors of the Tight Oil
Consortium (TOC), hosted at the University of Calgary, for their financial support. Funding for the project
was also provided through an NSERC Collaborative Research and Development grant. Clarkson would
like to thank Shell, Encana and AITF for sponsoring his Chair in Unconventional Gas and Light Oil
Research, Department of Geoscience, University of Calgary.
SPE-175941-MS 17

References
Alkouh, A., McKetta, S., Wattenbarger, R.A., 2014. Estimation of Effective-Fracture Volume Using
Water-Flowback and Production Data for Shale-Gas Wells. Journal of Canadian Petroleum
Technology. 53(5):290 –303.
Awoleke, O., Romero, J., Zhu, D., Hill, A.D., 2012. Experimental Investigation of Propped Fracture
Conductivity in Tight Gas Reservoirs Using Factorial Design. Paper SPE 151963 Presented at the
SPE Hydraulic Fracturing Technology Conference, The Woodlands, TX.
Barton, N., Bandis, S., 1990. Review of predictive capabilities of JRC-JCS model in engineering
practice. Proceedings of the International Symposium on Rock Joints, Loen, Norway, pp. 603–
610.
CMOST User’s Guide. Version 2014.10, Computer Modelling Group Ltd., Calgary, AB, Canada.
Corey, A.T, 1954. The Interrelation between Gas and Oil Relative Permeabilities. Production
Monthly, 19 (1): 38.
EIA, 2015. U.S. Energy Information Administration, Annual Energy Outlook, Washington, DC. ⬍
www.eia.gov/forecasts/aeo/⬎
EPA (Environmental Protection Agency), 2010. Hydraulic Fracturing Research Study. ⬍
www.epa.gov/safewater/uic/pdfs/hfresearchstudyfs.pdf⬎
Gdanski, R., Fulton, D., Shen, C., 2009. Fracture-Face-Skin Evolution During Cleanup. SPE Pro-
duction and Operation. 24(1): 22–34.
Gdanski, R., Walters, H., 2010. Impact of Fracture Conductivity and Matrix Relative Permeability on
Load Recovery. Paper SPE 133057 Presented at the SPE Annual Technical Conference and
Exhibition, Florence, Italy.
GEM User’s Guide. Version 2014.10, Computer Modelling Group Ltd., Calgary, AB, Canada.
King, G.E., 2012. Hydraulic Fracturing 101: What Every Representative, Environmentalist, Regulator,
Reporter, Investor, University Researcher, Neighbor and Engineer Should Know About Estimating
Frac Risk and Improving Frac Performance in Unconventional Gas and Oil Wells. Paper SPE
152596 Presented at the SPE Hydraulic Fracturing Technology Conference, The Woodlands, TX.
Padmakar, A.S., 2013. Geomechanics Coupled Reservoir Flow Simulation for Diagnostic Fracture
Injection Test Design and Interpretation in Shale Reservoirs. Paper SPE 166201 Presented at the
SPE Annual Technical Conference and Exhibition, New Orleans, LA.
Tannich, J.D., 1975. Liquid Removal from Hydraulically Fractured Gas Wells. Journal of Petroleum
Technology. 27(11): 1309 –1317.
Tran, D., Nghiem, L., Buchanan, L., 2005. An Overview of Iterative Coupling between Geomechani-
cal Deformation and Reservoir Flow. Paper SPE 97879 Presented at the SPE International Thermal
Operation and Heavy Oil Symposium, Calgary, Alberta, Canada.
Tran, D., Shrivastava, V., Nghiem, L., Kohse, B., 2009. Geomechanical Risk Mitigation for CO2
Sequestration in Saline Aquifers. Paper SPE 125167 Presented at the SPE Annual Technical
Conference and Exhibition, New Orleans, LA.
Wang, J. Y., Holditch, S., McVay, D., 2010. Modeling Fracture-Fluid Cleanup in Tight-Gas Wells.
SPE Journal 15(3): 783–793.
Wang, M., Leung, J., 2015. Numerical Investigation of Fluid-loss Mechanisms During Hydraulic
Fracturing Flow-back Operations in Tight Reservoirs. Journal of Petroleum Science and Engi-
neering 133(1): 85–102.
Williams-Kovacs, J.D, Clarkson, C.R., 2014. Analysis of Multi-Well and Stage-by-Stage Flowback
from Multi-Fractured Horizontal Wells. Paper SPE 171591 Presented at SPE/CSUR Unconven-
tional Resources Conference, Calgary, Alberta, Canada.
18 SPE-175941-MS

Zanganeh, B., Ahmadi, M., Hanks, C., Awoleke, O., 2015. The role of hydraulic fracture geometry
and conductivity profile, unpropped zone conductivity and fracturing fluid flowback on production
performance of shale oil wells. Journal of Unconventional Oil and Gas Resources 9(1):103–113.

You might also like