You are on page 1of 20

Journal of Natural Gas Science and Engineering 76 (2020) 103187

Contents lists available at ScienceDirect

Journal of Natural Gas Science and Engineering


journal homepage: http://www.elsevier.com/locate/jngse

Evaluation of the impact of multi-phase flow on reservoir signatures in the


Wolfcamp shale
C.R. Clarkson a, *, B. Yuan a, Z. Zhang a, F. Tabasinejad a, H. Behmanesh b, H. Hamdi a,
D. Anderson b, J. Thompson b, D. Lougheed c
a
Department of Geoscience, University of Calgary, Canada
b
NCS Multistage, Calgary, AB, Canada
c
SAGA Wisdom, AB, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: The dominant transient flow regime for multi-fractured horizontal wells (MFHWs) producing from low-
Multiphase flow permeability and shale (unconventional) reservoirs has historically been interpreted to be transient linear
Flow regime identification flow (TLF) with analysis performed in the framework of classical diffusion (CD). Recently, observed deviations
Transient linear flow
away from this classical behavior for Permian Basin Wolfcamp shale (oil) wells have been attributed to reservoir
Shale
heterogeneity and anomalous diffusion (AD). However, some wells in the basin also produce below bubble point
pressure and exhibit characteristics of multi-phase flow. The objective of the current study is therefore to
investigate if multi-phase flow may also be contributing to observed deviations from classic TLF behavior.
The conventional log-log diagnostics used to identify flow regimes do not account for reservoir complexities
such as multi-phase flow and reservoir heterogeneity. Failure to correct for these effects when they are occurring
may result in misdiagnosis of flow regimes. A new workflow is therefore introduced herein to improve flow
regime identification when reservoir complexities are exhibited, with a specific focus on multi-phase flow. The
workflow involves the correction of log-log diagnostics for multi-phase flow through the use of modified pseudo-
variables (pseudo-pressure and pseudo-time) once characteristics of multi-phase flow are identified (e.g. rapid
increase of gas-oil ratio after bubble point pressure is reached). Although reservoir heterogeneity and AD are
accepted causes of deviations from TLF, the impact of multi-phase flow has not been investigated in detail.
Therefore, corrections to pseudo-variables for multi-phase flow, a known reservoir complexity exhibited by
Wolfcamp shale wells, are presented. Pressure-dependent permeability is also accounted for in pseudo-variable
calculations, although its impact is demonstrated to be relatively minor in this study.
Application of the new workflow to a simulated case and a Wolfcamp shale field case demonstrates the
following: 1) multi-phase flow, and in particular the appearance of a mobile gas phase after two-phase oil and
water production, results in deviations from classical TLF behavior when data is analyzed using conventional
(uncorrected) diagnostics; 2) application of the modified diagnostics to a simulated case that includes multi-
phase flow results in the “true” flow regime signature of TLF being observed; 3) application of the modified
diagnostics to a field case exhibiting evidence of multi-phase flow reduces the deviation from TLF.
This study suggests that multi-phase flow may be impairing our ability to identify flow regimes using con­
ventional flow regime identification methods.

1. Introduction identify flow regime sequences, and specifically to quantify the


power-law exponents for use in forecasting well production. The log-log
Recently, a classic study by Chu et al. (2019) provided evidence of plots utilized included rate-normalized pressure (RNP) and its derivative
non-classical diffusion behavior in the Wolfcamp shale of the Permian (RNP’R ), the integral of rate-normalized pressure (IRNP) and its deriva­
Basin. Those authors focused on three field cases exhibiting power-law tive (IRNP’R ) and the Chow Pressure Group (CPG) (Chow, 1952) versus
decline behavior, and applied log-log diagnostic plots to these cases to

* Corresponding author.
E-mail address: clarksoc@ucalgary.ca (C.R. Clarkson).

https://doi.org/10.1016/j.jngse.2020.103187
Received 10 October 2019; Received in revised form 27 January 2020; Accepted 28 January 2020
Available online 4 February 2020
1875-5100/© 2020 Elsevier B.V. All rights reserved.
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

material balance time. The subscript ‘R’ is used in the derivative to transient linear flow is expected), then modified diagnostics that ac­
denote that the derivative is taken with respect to natural log of time, count for various reservoir complexities are applied to the data to
sometimes referred to as the ‘semi-log’ or ‘radial’ derivative. The pri­ determine the influence that they have on flow regime interpretation.
mary phase (oil) was used in the diagnostic plot calculations. Impor­ The new workflow is applied to a simulated and field case, the latter
tantly, measured flowing bottomhole pressures (FBHP) were also used in including a Wolfcamp shale well for which deviations from classical
the calculations, instead of the more common FBHPs calculated from diffusion flow regimes have been interpreted using the conventional
surface pressures, providing greater confidence in flow regime identi­ diagnostics of Chu et al. (2019). The reservoir complexities accounted
fication. What Chu et al. (2019) observed is that power-law exponents for in this work include multi-phase flow and pressure-dependent
(obtained from the slope of the IRNPIRNP/IRNP’R plot, or stabilization permeability, the former being identifiable from field data, while the
level from the CPG plot), differed from classical diffusion CD behavior latter is inferred from laboratory testing. Simulation sensitivities have
(e.g. ½ slope for transient linear flow on IRNP/IRNP’R plot). This devi­ demonstrated that the effect of pressure-dependent permeability on flow
ation from CD behavior was later used as evidence of AD by Raghavan regimes identified with conventional diagnostics is relatively muted for
and Chen (2018), who suggested that sub-diffusive flow could be the range of values investigated, and therefore the focus of this study is
occurring in the fracture and matrix of the Wolfcamp shale. Key to this on multi-phase flow. The effect of reservoir heterogeneity, while pre­
interpretation is the use of the diagnostics applied by Chu et al. (2019) – viously noted to be significant (and discussed further below), is delib­
while those authors acknowledged that multi-phase flow was occurring erately left out of the diagnostics to isolate how strong an effect that
in some instances, there was no attempt to determine if this was multi-phase flow has on the flow regime signatures. The hypothesis
affecting their flow regime interpretation, or correct for it if it did. that is tested in this work is that multi-phase flow is at least partially
Indeed, while the literature is replete with studies focused on the responsible for deviations from classical diffusion behavior interpreted
development of methods to correct RTA/PTA rate-transient analysis/ for some Permian Basin wells.
pressure-transient analysis model inversion methods for complexities The implications of this study are of great importance to RTA per­
such as multi-phase flow, there appears to be only a few studies on the formed on unconventional reservoirs. If flow regimes are misidentified,
effect that these complexities have on an important early step in RTA/ the incorrect model may be selected, resulting in inaccurate reservoir
PTA: flow regime identification. As examples of the former application, and hydraulic fracture characterization. This in turn could lead to poor
Zhang et al. (2016) and Becker et al. (2016) were the first to apply the reservoir management decisions.
similarity method, coupled with the Runge-Kutta solver, to solve the
two-phase transient linear flow equation for application to RTA; Hamdi 2. Theory and methods
et al. (2018, 2019) used a similar (similarity-based) approach to analyze
constant flowing pressure linear flow under three-phase flow conditions In this section, the conventional diagnostic plots used to identify flow
(see Clarkson et al., 2019a for a more extensive literature review for this regimes for MFHWs completed in unconventional reservoirs are first
topic). reviewed. These include log-log plots of RNP/RNP’R , IRNP/IRNP’R and
In the field of PTA, it has previously been demonstrated that flow CPG versus material balance time. Example applications of these diag­
regime misidentification can occur when the single-phase pseudo-pres­ nostic plots to flow regime identification are then illustrated with
sure is used to analyze well test responses in gas-condensate reservoirs simulated and field examples, including one of the Permian Basin ex­
with strong multi-phase flow behavior. The use of single-phase pseudo- amples studied by Chu et al. (2019). A workflow used herein for eval­
pressure in these situations can frequently cause the appearance of uating the impact of reservoir complexities (in particular, multi-phase
radial composite signatures in homogeneous reservoirs, which can be flow) on flow regime interpretation is then discussed. Finally, modifi­
corrected by implementing two-phase pseudo-pressures (Jones and cation of the diagnostic plots to account for these reservoir complexities
Raghavan, 1988; Jones et al., 1989; Raghavan et al., 1999; Bozorgzadeh using modified pseudo-variables is outlined.
and Gringarten, 2005; and Hamdi et al., 2013). For RTA of coalbed
methane (CBM) reservoirs exhibiting two-phase flow of gas and water,
Clarkson et al. (2012) and Clarkson (2013) incorporated relative 2.1. Conventional diagnostics for flow regime ID
permeability information into derivative calculations to improve flow
regime identification. Later, Zhu et al. (2018) used a log-log plot of RNP, Following the work of Chu et al. (2019), the conventional diagnostics
modified to include pseudo-pressures containing gas relative perme­ used in this work are applied to the primary phase (oil) and are
ability, to identify flow regimes for hydraulically-fractured CBM wells. formulated in terms of pressure and time. With these diagnostics, no
To our knowledge, correction of flow-regime identification tools for the correction for complex reservoir behavior (e.g. multi-phase flow) are
effects of multi-phase flow in shale reservoirs has not been attempted. In applied, hence use of the term ‘conventional’. These diagnostics include
addition to multi-phase flow, several other phenomena observed in log-log plots of RNP/RNP’R , IRNP/IRNP’R and CPG versus material bal­
shales might act to distort flow regime signatures when not accounted ance time, which are defined as follows:
for in the diagnostics, including stress-dependent permeability (e.g.
Clarkson et al., 2013) and reservoir heterogeneity (e.g. Acun ~ a, 2016, RNP ¼
pi pwf ðtÞ
(1)
2017). qo ðtÞ
The purpose of the current study is therefore to investigate the in­
dRNP
fluence that certain shale reservoir complexities1 have on flow regime RNP’R ¼ (2)
dlnðtc Þ’
interpretation. A workflow is developed in which the ‘conventional’
diagnostics used by Chu et al. (2019), which assume single-phase flow Z tc
1 pi pwf ðtc Þ
and do not account for reservoir complexities, are first applied to pro­ IRNP ¼ dtc ; (3)
tc qo ðtc Þ
vide a first pass flow regime identification. If non-classical diffusion flow 0

regimes are observed (e.g. deviations from ½ slope behavior when dIRNP
IRNP ¼ (4)
dlnðtc Þ’
1
For the purpose of this manuscript, we have defined “reservoir complex­ RNP
ities” as any reservoir property or flow characteristic of a reservoir that causes CPG ¼ (5)
2RNP’R
departure from classical diffusion behavior on conventional log-log diagnostics,
as utilized by Chu et al. (2019), and summarized in this work.

2
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Qo ðtÞ time is again equal to a.


tc ¼ (6)
qo ðtÞ Finally, when RNP is a power-law function, then CPG becomes:
1
Where pi is initial reservoir pressure, pwf ðtÞ is the flowing bottomhole CPG ¼ (11)
2a
pressure of the well (assumed to be time-variant), qo ðtÞ is the fluid
production rate and Qo ðtÞ is cumulative oil production. These equations Note that the Bourdet algorithm (Bourdet et al., 1989) is used in this
assume oil with negligible pressure-dependent fluid properties – if the work for derivative calculations.
pressure dependencies of oil properties are significant, pseudo-pressure
for oil may be substituted, as is commonly done for gas cases. Material 2.2. Example applications of conventional diagnostics
balance time (tc ), a superposition time function that is strictly applicable
to boundary-dominated flow, is used in Eqs. (2) and (4) in order to To demonstrate application of the diagnostics provided above to the
observe a constant rate signature (e.g. unit slope for boundary- identification of flow regimes within the framework of classical diffu­
dominated flow) for the case of constant flowing production, or vari­ sion, a simulated example using the Wattenbarger et al. (1998) closed
able rate/flowing pressure production. Further, use of tc helps to reduce boundary constant flowing pressure solution for the
some of noise in the derivative calculations. Chu et al. (2019) found that fracture-to-boundary case is analyzed. A conceptual model for this case
the integral versions of RNP=RNP’R (IRNP=IRNP’R ) considerably is provided in Fig. 1a. The model input values used for this case can be
improved flow regime identification for the Permian Basin wells they found in a previous publication (Clarkson et al., 2019c). Single-phase
studied. Note that CPG may be modified to include IRNP= IRNP’R as flow of undersaturated oil with negligible pressure dependent fluid
follows: properties through a homogeneous isotropic reservoir with negligible
pressure-dependent properties was assumed. For this example, transient
IRNP
CPG ¼ (7) linear flow is identified as a positive half-slope (slope ¼ a ¼ 0.5) on
2IRNP’R
IRNP’R plot, with CPG ¼ 1 (Fig. 1b). After a transition period,
The radial derivative has historically been used in well test analysis boundary-dominated flow is identified as a unit (slope ¼ a ¼ 1) on the
because of the importance of properly identifying radial flow, which IRNP’R plot and a CPG ¼ 0.5.
appears as a zero slope on a log-log plot of RNP’R . While radial flow is The above example of classical diffusion behavior was selected
commonly observed in conventional reservoirs, it is less commonly because the sequence of transient linear flow to boundary-dominated
observed for MFHWs completed in low permeability (and shale) reser­ flow has been reported to be quite common for multi-fractured hori­
voirs. For the latter scenario, power-law behavior (straight line on a log- zontal wells completed in shales (e.g., Anderson et al., 2010). An addi­
log plot of RNP versus time) is commonly observed (Chu et al., 2019). tional flow regime that may be observed at early time, caused by the
Writing RNP as a power-law function of material balance time yields: presence of finite conductivity fractures, is bilinear flow, which is
identified as having a slope (a) ¼ 0:25 on the log-log RNP=RNP’R or
RNP ¼ Atc a (8)
IRNP=IRNP’R plot, or CPG ¼ 2.
RNP’R may then be written as: A field example of a MFHW exhibiting transient linear flow to
boundary-dominated flow behavior was studied by various authors
RNP’R ¼ aAtc a (9) (Clarkson and Pedersen, 2010, 2011; Clarkson and Qanbari, 2015a,b)
From Eqs. (8) and (9) it is apparent that a log-log plot of RNP and and is provided in Fig. 2. This well was completed in a low permeability
region of the Cardium Formation in Western Canada, referred to as a
RNP’R versus tc has a slope ¼ a when RNP is a power-law function.
“Halo” play by Clarkson and Pedersen (2010). This play is characterized
Similarly, IRNP’R is calculated as follows when RNP ¼ At ac :
by heterogeneities at the core scale caused primarily by bioturbation (e.
a g. Solano et al., 2016); however, the permeability is relatively high
IRNP’R ¼ Ata (10)
1þa c (typically 0.01 to > 0.1 md). Inter-fracture flow was determined by
Clarkson and Pedersen (2011) to be over within the first two weeks of
The slope of a log-log plot of IRNP and IRNP’R versus material balance
production, and therefore the linear flow and boundary-dominated flow

Fig. 1. (a) Conceptualization of a constant flowing pressure (CP) drawdown test for a hydraulically fractured vertical well. A cross-sectional view (top) and plan view
(bottom) are provided. The grey bar at the center of the plan view depiction is the hydraulic fracture. The dashed green lines are isobars (lines of constant pressure)
that, in plan view in these diagrams, track the propagation of the pressure transient out into the reservoir at the different timesteps 1–3 until the boundaries of the
reservoir are reached (timestep 4). In cross-section view, the green dashed lines illustrate the pressure drop at the wellbore and pressure gradient into the reservoir.
After the boundaries have been reached, boundary-dominated flow occurs, represented by the red solid lines. (b) Oil RNP=RNP’R and CPG plot (‘conventional’ di­
agnostics) for a simulated case of transient linear to boundary-dominated flow (CP case), corresponding to the conceptual model in (a).

3
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

suggested that sub-linear flow (0:5 < a < 1) may be caused by fracture
networks with a complex fracture geometry. He then developed an
analytical method that accounted for this behavior (as well as sub-radial
flow). Acun~ a (2016) attributed geometric non-uniformity of the fracture
network alone to be the cause of sub-linear flow behavior. In a later, but
related work, Yuan et al. (2019a) developed an analytical method for
RTA of unconventional reservoirs exhibiting non-linear square-root of
time behavior. They interpreted this behavior to be caused by reservoir
heterogeneity and stress-dependent reservoir properties (approximated
as a function of pressure), and corrected the linear flow solution for
these effects. Their method assumed a non-uniform distribution of nat­
ural fractures away from the primary hydraulic fracture (see Fig. 3), and
that the resulting rock properties are fractally distributed. One of the
field examples studied by Yuan et al. (2019a,b) is provided in Fig. 4. This
example (shale gas, single-phase gas system) was originally provided by
Stalgorova and Mattar (2012) to demonstrate application of their
analytical enhanced fracture region (EFR) model; the EFR model
developed by those authors accounted for variability in reservoir
permeability caused by fracture complexity by assuming a
Fig. 2. Oil IRNP=IRNP’R and CPG plot (‘conventional’ diagnostics) for MFHW composite-type reservoir system where each reservoir region has a
completed in a low permeability region of the Cardium Formation in Western
Canada. Data affected by inaccurate flowing pressure estimates are evident in
the IRNP’R plot.

was interpreted to occur between horizontal laterals. Early time data ( tc


< 20 days) is affected by inaccurate flowing pressure estimates caused
by rapid fluid level changes in the wellbore during pump off. This
example highlights an issue with the use of the CPG for flow regime
identification. While IRNP’R exhibits a ½ slope for tc > 20 days until
boundaries are reached ( tc > 300 days), IRNP exhibits a slope < ½ for
most of the transient flow period. It is well-known in well-test literature
that RNP is affected by skin, but its derivative is not. This was discussed
in the context of RTA by Nobakht and Mattar (2012). In this example, it
appears that a positive skin effect (either real, or caused by inaccurate
flowing pressure estimates) is affecting IRNP but not IRNP’R . This causes
the CPG, for which RNP or IRNP appears in the numerator, to be > 1 (not
shown), which we believe is not the true flow regime signature. In this
case, and likely in many cases where skin is affecting early time data, the
CPG may provide an erroneous flow regime signature. At the time of
writing a new version of the CPG is being developed to improve flow
regime identification. In the meantime, we recommend the use of RNP’R Fig. 4. IRNPpg =IRNP’pg R and CPGpg plot (‘conventional’ diagnostics) for shale
or IRNP’R in these scenarios. gas example studied by Yuan et al. (2019a,b). Note that gas pseudo-variables
have been substituted for pressure and time, indicated by subscript ‘pg’ in the
In recent years, deviations from a ¼ 0:5 (linear flow) for MFHWs
diagnostics. A positive ½ slope line is drawn for reference to show deviation
completed in unconventional reservoirs have been rationalized to be
away from linear flow on IRNP’pg R plot.
caused primarily by reservoir heterogeneity. For example, Acun ~ a (2016)

Fig. 3. Conceptual diagram showing reservoir quality degradation away from the primary hydraulic fracture. After Yuan et al. (2019a).

4
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

constant porosity and permeability, but with variability from region to 2.3. Workflow to improve flow regime ID
region. The conventional log-log diagnostics, using IRNP= IRNP’R and
CPG, without correction for reservoir complexities, applied to the Stal­ In order to test the hypothesis that complex reservoir behavior, such
gorova and Mattar (2012) field example, is shown in Fig. 4. However, as multi-phase flow, may be affecting flow regime identification, the
gas pseudo-pressure and material balance pseudo-time are substituted following workflow (illustrated in Fig. 6) is used:
for pressure and time in IRNP=IRNP’R and CPG to account for gas prop­
erty changes with pressure (see Eqs. (12)–(15) below); these variables 1. Apply conventional diagnostic plots for data quality assessment and
have been renamed IRNPpg =IRNP’pg R and CPGpg to indicate that this flow regime identification. Using these results, determine if the
conventional log-log diagnostic plots used for flow regime ID need to
substitution has been made. Deviations from classical linear flow char­
be modified to account for complex reservoir behavior
acteristics (CPGpg <1, IRNPpg =IRNP’pg R slope > 0.5) occur between ma­
2. Modify the log-log diagnostic plots for flow regime ID to account for
terial balance pseudo-time values of 10 and 200 days. At late time, false the complex reservoir behavior identified in step 1, if necessary
boundary effects appear to be occurring, possibly caused by liquid 3. Re-assess flow regimes
loading. This example provides evidence that, for single-phase shale gas
cases, reservoir heterogeneity can cause deviations from classic linear For step 1, in addition to the application of the conventional log-log
flow behavior. Acun ~ a (2016, 2017) provided additional simulated and
diagnostic plots for flow regime ID discussed above (log-log RNP=RNP’R ,
field examples.
IRNP=IRNP’R and CPG plots), it is recommended that the data evaluation
The last example of the application of conventional diagnostics is
steps and diagnostic plots suggested by Nobakht and Mattar (2009) and
field Case 1 from Chu et al. (2019), which is a Wolfcamp Shale well
Ilk et al. (2010) be used to assess data quality and identify potential
(Fig. 5). Note that this plot, re-created from Chu et al. (2019) differs
problems that may impact analysis. In fact, it is further recommended
somewhat from those authors because data for the first day were not
that the data assessment procedures precede any attempt at flow regime
available, data frequency is daily in the current work, and likely
identification and subsequent RTA, as suggested by Ilk et al. (2010) and
differing amounts of smoothing were applied in the derivative calcula­
Clarkson (2013).
tions. However, similar trends are observed. At early time, the CPG is
After compilation and review of all data necessary to perform a
close to 1, but transitions to CPG ~0.8, translating to a > 1= 2 – these
model-based analysis, including surveillance data, data assessment
values are consistent with the two CPG stabilizations reported by Chu
should be performed for the purpose of 1) determining if the data quality
et al. (2019). Using the Chu et al. (2019) conventional (uncorrected)
is sufficient for model-based analysis 2) diagnosing non-reservoir effects
diagnostics, Raghavan and Chen (2018) interpreted deviations from
that could influence analysis (and filtering out associated data if
classic linear flow behavior as evidence that anomalous diffusion is
necessary) and 3) diagnosing reservoir effects (multi-phase flow, drive
occurring (particularly sub-diffusion in the fractures and matrix).
mechanism) that may impact analysis, amongst other purposes. Partic­
Reservoir heterogeneity (combined with classical diffusion) is another
ularly relevant to the current study is 3), for which basic plots of flowing
plausible cause (as interpreted for the example in Fig. 4). However there
pressure, production data and producing fluid ratios are recommended
may be other factors affecting the flow regime interpretation, using the
to help with the diagnosis of multi-phase flow.
Chu et al. (2019) diagnostics. For example, Chu et al. (2019) noted that
These analyses are then used to determine if complex reservoir
for Case 1 (Fig. 5), gas lift was applied within the first week to assist with
behavior may be impacting flow regime identification when conven­
unloading fracturing fluid and reservoir oil, and that FBHP was below
tional diagnostics are applied. Examples of complexities that could
bubble point pressure within the first two months of production. Our
impact flow regime ID include:
hypothesis tested herein is that complexities such as the occurrence of
multi-phase flow, not accounted for in the conventional log-log diag­
� Change from single-phase flow to multi-phase flow, or multi-phase
nostic analysis applied above, may be affecting flow regime
flow from the start of production
interpretation.
� Stress sensitivity of fractures and matrix properties (porosity and
permeability)
� Fracture heterogeneity (fracture-network density, primary fracture
conductivity, spacing, and unequal lengths)
� Matrix heterogeneity (permeability)
� Some combination of the above

Non-reservoir or operational effects, such as the following, may also


impact flow regime identification:

� Errors in the calculation of downhole pressures using surface pres­


sures when multi-phase flow occurs in wellbore
� Liquid loading (if surface pressures are used for downhole pressure
calculation)
� Artificial lift (gas-lift, downhole pumps) (if surface pressures are used
for downhole pressure calculation)
� Well re-stimulation
� Frac hits, and inter-well communication

Note that the focus of the current work is the correction of log-log
diagnostics used in flow regime identification for complex reservoir
behavior, and in particular multi-phase flow (step 2), which is discussed
further below. Application if step 3 is illustrated with a simulated and
Fig. 5. Conventional log-log diagnostics applied to field Case 1 of Chu et al. field example.
(2019). A positive ½ slope line is drawn for reference to show deviation away
from linear flow on the IRNP’R plot at late time (tc > 20 days).

5
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Fig. 6. Workflow for improved flow regime identification when reservoir complexities are apparent.

2.4. Modification of diagnostics for complex reservoir behavior modified pseudo-variables were then used in modified log-log diag­
nostic plots for flow regime identification, and in modified straight-line
The conventional approach for modifying both log-log diagnostics analysis methods for RTA. In this way, analysis becomes an iterative
and analytical solutions for pressure-dependent fluid properties is to process, whereby flow regime identification and analysis occur simul­
replace pressure and time with pseudo-pressure and pseudo-time (i.e. taneously, converging on a consistent, if not unique, solution. This
using pseudo-variables). For example, in the analysis of gas reser­ approach is applied herein to correct log-log diagnostic plots for the
voirs, RNP and RNP’R may be redefined using gas pseudo-variables as effects of multi-phase flow and stress-dependent (approximated as
follows: pressure-dependent) permeability in the Wolfcamp shale.
� The calculation of pseudo-variables for the Wolfcamp shale is
RNPpg ¼
ppg ðpi Þ ppg pwf
� ; (12) particularly complex because of the possibility of simultaneous flow of
qg tcag oil, gas and water phases in the reservoir, and the variable operating
conditions that are encountered. In the field cases studied by Chu et al.
RNP’pg ¼
dRNPpg
� (13) (2019), oil, water and gas production is common to all cases. However,
the development of a mobile gas phase in the reservoir, evident from
R
dln tcag
large increases in GOR over time, depends on when (or if) the flowing
Where ppg ðpi Þ is gas phase pseudo-pressure calculated at initial pressure, pressure drops below bubble point pressure and critical gas saturation.
and ppg ðpwf Þ is gas phase pseudo-pressure calculated at flowing pressure, Once a mobile gas phase is formed, the oil phase (effective) permeability
with ppg defined as: is significantly reduced. When a mobile water phase also exists in the
Z p reservoir, flow of each phase is governed by three-phase relative
p
ppg ¼ 2 dp (14) permeability. As a result, log-log diagnostics that only rely on primary
0 μg ðpÞZðpÞ
phase production, and without correction for multi-phase flow, will be
in significant error.
And tcag is material balance pseudo-time defined as:
A key problem with pseudo-variable calculations, for the case of
Z
μgi cti t qg ðtÞ multi-phase flow, is the requirement for pressure-saturation (P–S) re­
tcag ¼ dt (15)
qg ðtÞ 0 μg ðpÞct ðpÞ lationships. While Hamdi et al. (2018, 2019) were able to derive P–S
relationships using a semi-analytical method for the case of constant
p is the reference pressure, typically taken as the pore volume average flowing pressure and multi-phase transient linear flow, this is not
pressure in the reservoir for conventional, high permeability reservoirs. currently possible for variable operating condition cases. This has
For low permeability reservoirs, with long transient flow periods, it is necessitated the use of numerical simulation output to derive saturations
more appropriate to evaluate this pressure as the average pressure in the and pressures (and the P–S relationships).
distance-of-investigation (DOI), as demonstrated by Anderson and In the following, the modification of pseudo-pressure and pseudo-
Mattar (2007) for radial flow, and by Nobakht and Clarkson (2012a,b) time variables for multi-phase flow and pressure-dependent perme­
for transient linear flow. ability is discussed, along with how these modified pseudo-variables are
In recent years, it has become commonplace to modify pseudo- used to correct the log-log diagnostic plots (RNP=RNP’R , IRNP=IRNP’R
variables for complex reservoir behavior, such as stress-dependent and CPG).
permeability (approximated as pressure-dependent permeability) and Pseudo-pressure modification for multi-phase flow and pressure-
two-phase or multi-phase flow (e.g. Behmanesh et al., 2018a,b). How­ dependent permeability. As described in our previous work (Behmanesh
ever, the use of pseudo-variables corrected for complex reservoir char­ et al., 2018b), and as outlined by Fevang and Whitson (1996) for gas
acteristics are most commonly applied to analytical solutions that have condensate reservoirs, for cases where different flow regions develop
been derived for liquid flow through reservoirs with static, and simple, away from the hydraulic fracture due to the presence of different mobile
reservoir properties. This technique is less commonly applied to correct phases, the pseudo-pressure difference integral must be subdivided
log-log diagnostic plots (for flow regime ID) for complex reservoir accordingly, from the well FBHP to the initial reservoir pressure. For the
behavior, although exceptions were noted in the Introduction. The Wolfcamp shale case studied herein (Fig. 5), the flow regions provided in
reason is one of practicality: some properties, such as relative perme­ Fig. 7 need to be represented. A three-phase region occurs near the
ability or pressure-dependent permeability, are estimated only after primary hydraulic fracture (Region 1), corresponding to transient flow
some analysis is performed, which conventionally occurs after di­ of oil/gas/water; further out in the formation, where pressure is below
agnostics are applied. Clarkson (2013) suggested that, in the analysis of bubble point pressure (pb ), but greater than that for which critical gas
two-phase CBM reservoirs, pressure- and saturation-dependent (e.g. saturation (Sgc ) is achieved (pgc ), a transient two-phase (oil/water) flow
relative permeability) variables for use in pseudo-variable calculations region (Region 2) occurs where gas mobility is negligible; still further
may be derived by using the outputs of analytical or numerical simu­ out in the reservoir, where pressure is greater than pb , only oil and water
lation models, used in turn to history-match well production data. The

6
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

time-varying systems with simpler steady-state solutions. Lee et al.


(1998) extended the method for modeling long-term production per­
formance of various reservoir types, but with a focus on single-phase
flow. In the current work, a new application of this concept is used to
derive the multi-phase material balance pseudo-time. For Eq. (17), the
integration is performed over time, and therefore a reference pressure is
required for calculation of the pressure-dependent reservoir and fluid
properties. The reference pressure used herein is derived by using a
pore-volume-weighted-averaging technique within the distance of
investigation at each timestep. Specifically, to provide a rigorous esti­
mate of α and β in Eq. (17), pressure and saturation of each grid block
from a calibrated (history-matched) simulation model are used to esti­
mate the average pressure/saturation for each region in Fig. 7.
Z tao
∅i μoi cti Soi 1 ∅i μoi cti Soi 1
tcao ¼ qo ðτao Þdτao ¼
ki kroi qo ðtao Þ 0 ki kroi qo ðtÞ
Z t R y3
Fig. 7. Illustration of three flow regions, corresponding to three- and two-phase
qo ðtÞ 0
αðyÞ dy
! dt (17)
flow, that develop with distance from a hydraulic fracture for an unconven­ 0 R y3 dβðyÞ
tional light oil reservoir. Modified from Behmanesh et al. (2018b). 0 dpðyÞ
dy

phases exist in the reservoir, and transient flow of oil/water is occurring αðyÞ is calculated for either three-phase (αOGW ) or two-phase (αOW )
(Region 3). flow. βðyÞ can also be calculated for either three-phase or two-phase
The modified pseudo-pressure difference, accounting for the three conditions, depending upon the pressure/saturation conditions in the
flow regions in Fig. 7, is provided as follows: nth grid block of the (simulated) reservoir. The integral term in Eq. (17)
Z pgc Z pb Z pi ! can be further expressed as:
� μ Boi
ppo ðpi Þ ppo pwf ¼ oi αOGW dp þ αOW dp þ αOW dp R y3 R y1 Ry Ry
ki kroi αi dy αOGW dy þ y12 αOW dy þ y23 αOW dy
(18)
pwf pgc pb 0 ! ¼ � 0 � � � � �
(16) R y3 dβðyÞ R y1 dβOGW Ry
dy þ y12 dβOG W
Ry
dy þ y23 dβdpOW dy
0 dpðyÞ
dy 0 dp dp

Where:
� where:
kðpÞkro So ; Sw ; Sg kðpÞkro ðSo ; Sw Þ
αOGW ¼ ; αOW ¼ ; �
μo ðpÞBo ðpÞ μo ðpÞBo ðpÞ ∅ðpÞSo Sg ; Sw ∅ðpÞSo ðSw Þ
​ βOGW ¼ ; βOW ¼
Bo ðpÞ Bo ðpÞ
and, ppo ðpÞ is oil pseudopressure, kðpÞ is pressure-dependent absolute
permeability of the reservoir, modeled with the Pedrosa (1986) equa­ As noted above, for modified pseudo-pressure and material balance
tion, k ¼ ki e γk ðpi pÞ , where γk is the permeability modulus, pseudo-time calculations, numerical simulation outputs are required to
kro ðSo ; Sg ; Sw Þ is relative permeability to oil modeled using Stone’s first derive the saturation and pressure for use in pseudo-variable evaluation.
three phase relative permeability model (Stone, 1973), and μo ðpÞ and For this purpose, a 1-D simulation model was used for both simulated
Bo ðpÞ are oil viscosity and formation volume factor, respectively. Note (verification) case and field case history-match (discussed in Results
that by inclusion of kðpÞ, pressure-dependent permeability effects can be section), with a base geometry shown in Fig. 8. This conceptual model
accounted for along with multi-phase flow. Importantly, it should be represents an element of symmetry (one fracture) that is scaled up to the
noted that pressure-dependent permeability is used herein to approxi­ number of fractures in multi-fractured horizontal well.
mate the impact of changing effective stress on permeability during Log-log diagnostics modification for multi-phase flow and pressure-
depletion. The impact of stress is more accurately accounted for using dependent permeability. Using the modified pseudo-variables defined
coupled flow-geomechanical models. Use of flow simulation alone, with Eqs. (16) and (17), the log-log diagnostics can then be modified for
however, requires that permeability change be incorporated as a func­ multi-phase flow and pressure-dependent permeability. For example,
tion pressure, not stress. As such, the decoupling of pressure and stress IRNP=IRNP’R and CPG are re-defined as follows:
required by flow simulation will introduce errors in the representation Z tcao �
1 ppo ðpi Þ ppo pwf
of stress-dependent properties. A detailed derivation of this version of IRNPpo ¼ dτcao ; (19)
pseudo-pressure difference is presented in Appendix A. In order to tcao 0 qo ðτcao Þ
calculate the integral of Eq. (16) for cases with multi-phase flow, a
dIRNPpo
relationship between saturation and pressure is required. In this work, IRNP’po R ¼ ; (20)
dlnðtcao Þ
this relationship is developed by reading pressure and saturation of each
grid block over distance from a numerical model calibrated to field data. IRNPpo
Material balance pseudo-time modification for multi-phase flow and CPGpo ¼ (21)
2IRNP’po R
pressure-dependent permeability. As is well-known in RTA literature,
pseudo-time, in addition to pseudo-pressure, is required to (at least Application of the workflow described above will now be demon­
approximately) linearize the flow equation in the multi-phase flow case, strated with a simulation (verification) case, and a field case (Case 1,
and is further required to correct the log-log diagnostics in this work. Eq. Fig. 5) previously evaluated by Chu et al. (2019).
(17) provides the multi-phase material balance pseudo-time utilized in
this work, which references the three flow regions provided in Fig. 7. 3. Results
The detailed derivation of multi-phase material balance pseudo-time is
provided in Appendix B. Note that in this work, the concept of “suc­ In this section, a simulation case, with reservoir and fluid properties
cession of pseudo-steady states” is applied to approximate the transient believed to be reasonably representative of the Wolfcamp shale, is used
flow period. This method originated with Muskat (1937) for modeling to demonstrate the impact of multi-phase and pressure-dependent

7
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Table 1
Summary of numerical model inputs for simulation and field Case 1.
Parameter Matrix Hydraulic
Fracture

Initial pressure (psia) 3405 Same as for


matrix
Reservoir temperature (� F) 165 Same as for
matrix
Initial permeability (md) 4.6 � 100
10 5
Initial porosity (fraction) 0.096 0.12
6
Rock compressibility (1/psia) 5 � 10 4 � 10 5
5
Permeability modulus (1/psia) 4 � 10 9 � 10 4
Oil API gravity 39.6 Same as for
matrix
Gas specific gravity 0.917 Same as for
matrix
Bubble point pressure (psia) 2335 Same as for
matrix
Undersaturated oil compressibility (1/psia) 9.8 � Same as for
10 6 matrix
Number of fractures – 115
Thickness (ft) 530 Same as for
matrix
Fig. 8. Base geometry used for numerical model applied to a simulated (veri­ Initial water saturation (fraction) 0.55 0.6
fication) and field case provided in the Results section. Model output (satura­ SWCON - endpoint saturation: connate water 0.22 0
tions and pressures) is used in pseudo-variable calculations. Ye is fracture SWCRIT - endpoint saturation: critical water 0.22 0
spacing (ft), wf is fracture width (ft) and xf is fracture half-length (ft). The base SOIRW - endpoint saturation: irreducible oil for 0.05 0.19
simulation model has the following number of grids: 7*200*1 (X*Y*Z). Grid water-oil table
SORW - endpoint saturation: residual oil for water- 0.05 0.19
resolution in the Y-direction is deemed sufficient for accurate estimation of
oil table
transient saturations and pressures.
SOIRG - endpoint saturation: irreducible oil for 0.23 0.05
gas-liquid table
permeability on flow regime signatures interpreted with conventional SORG - endpoint saturation: residual oil for gas- 0.23 0.05
liquid table
log-log diagnostics. This case also serves as a verification of the methods
SGCON - endpoint saturation: connate gas 0.01 0
provided in the previous section to correct the diagnostics for these SGCRIT - endpoint saturation: critical gas 0.01 0
complex reservoir behaviors. Next, the workflow is applied to field Case KROCW - kro at connate water 1 1
1 of Chu et al. (2019). This field case, previously interpreted to exem­ KRWIRO - krw at irreducible oil 1 1
plify strong evidence of anomalous diffusion, is re-interpreted with the KRGCL - krg at connate liquid 1 1
KROGCG - krog at connate gas 1 1
modified log-log diagnostics developed herein. The purpose of this
Exponent for calculating krw from KRWIRO 6.5 7
re-analysis is to determine if multi-phase flow may be affecting our Exponent for calculating krow from KROCW 2.9 1.5
ability to properly diagnose flow regimes and therefore our ability to Exponent for calculating krog from KROGCG 6.5 1.5
determine if anomalous diffusion is strongly affecting flow characteris­ Exponent for calculating krg from KRGCL 1.15 1.5
tics in the Wolfcamp shale.
plots of the simulated production data, producing fluid ratios and
flowing pressure data. From these plots, it observed that GOR remains
3.1. Simulation (verification) case
flat for only a very short period of time (<10 days, Fig. 9b), then ramps
steeply up as FBHP drops below bubble point pressure (2335 psia) and
For this simulation case, the base geometry in Fig. 8 is assumed. For
after critical gas saturation is reached. Gas production increases for the
both simulated and field cases, 200 grid blocks were used in the y-di­
first 65 days due to the increase in relative permeability to gas during
rection (linear flow direction) with a grid size of only 0.3 ft. In addition,
this time. WOR decreases continuously throughout the production
to minimize the effects of cross-flow in other directions in the reservoir,
period, flattening after about 400 days.
coarse grids (7 grids and 1 grid, respectively) were used for y- and z-
directions. Grid-size-sensitivity studies have demonstrated that this fine Fig. 10 illustrates application of the conventional log-log IRNP=
grid setting results in negligible errors in production performance IRNP’R and CPG diagnostics (Eqs. (3), (4) and (7)) to the simulated case.
simulation (Clarkson and Qanbari, 2016). The hydraulic fractures are Note that the integral forms are used to provide direct comparison with
assumed to be identical for every stage. An element of symmetry (single the field case below. From this plot it is evident that a very brief period
fracture, full fracture simulation) is used to speed up simulation run (<4 days, material balance time) of CPG > 1 occurs, which is interpreted
times, but is then scaled up to the number of fractures for the MFHW. to be bilinear flow. Transient linear or near-transient linear flow
The numerical model was populated with the data provided in behavior (1/2 slope on IRNP’R plot, CPG ¼ 1) is exhibited for material
Table 1. The simulated well length is 6842 ft. In order to ensure that the balance time values between 7 and 30 days, after which the CPG tran­
simulation case is realistic, all the inputs are identical to those used later sitions to a value of ~0.7 (sub-linear flow). The cause of deviation from
to history-match field Case 1, although the real flowing pressures are classical diffusion behavior is evident when GOR is superimposed on the
approximated with a power-law regression to provide smoother data for plot. We see that the transition from the initial transient linear flow to
log-log diagnostic analysis. Note that a moderate amount of pressure- apparent sub-linear flow approximately corresponds to a transition from
dependent permeability is assumed for the matrix (reservoir), as re­ constant GOR (two-phase flow oil and water in the formation when
flected in the permeability modulus (4 � 10 5 psia 1), whereas it is FBHP above bubble point pressure) to an increase in GOR (formation of
assumed to be strong for the hydraulic fracture (9 � 10 4 psia 1). mobile gas phase as FBHP drops below bubble point pressure, and Sg >
Conventional diagnostics. The first step in the workflow described in Sgc). As a result, if this case were analyzed without referencing the fluid
the Theory and Methods section is to apply conventional diagnostics for production ratios and flowing pressures (and their relationship), the
data quality assessment and flow regime identification. Fig. 9 provides flow regime could have been mis-diagnosed. The numerical simulator

8
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Fig. 9. Simulation output for model corresponding to the base geometry in Fig. 8. (a) Semi-log plot of gas, oil and water production rates and Cartesian plot of FBHP.
(b) Cartesian plot of GOR and FBHP. (c) Cartesian plot of WOR and FBHP.

discussed below therefore correct for these effects.


Modified diagnostics. The simulated data were re-analyzed using the
modified diagnostics (including pseudo-variables modified for multi-
phase flow and pressure-dependent permeability) described in the
Theory and Methods section. As noted in that section, output (satura­
tions and pressures) from the simulation model are required for modi­
fied pseudo-variable calculations. Specifically, these outputs are used to
derive the relationship between oil phase saturation and pressure (P–S
relationship) presented in Fig. 11. Important values to note include the
bubble point pressure (pb ¼ 2335 psia), initial oil saturation (Soi ¼ 0.45),
and the critical gas saturation (Sgc ¼ 0.01), with the corresponding

Fig. 10. Conventional log-log diagnostics applied to simulation case. A positive


½ slope line is drawn for reference to show deviation away from linear flow on
the IRNP’R plot at late time (tc > 30 days). GOR is also shown to illustrate the
relationship between GOR and flow regime changes.

used to generate the synthetic data for this case utilizes approximate
(numerical) solutions to the classical diffusivity equation. Therefore,
interpretations of reservoir heterogeneity, or AD behavior, would have
been erroneous for this case.
From the results of Figs. 9–10, it is evident that multi-phase flow, and
in particular the strong influence of the gas phase on oil relative
permeability, and system compressibility, are the primary causes of
Fig. 11. Pressure-oil saturation relationship derived from numerical simulation
deviations from classical behavior in this case. The modified diagnostics model output for simulated case.

9
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

pressure, pgc ¼ 2250 psia. The P–S relationship in Fig. 11 is then with GOR now included, is provided in Fig. 14. Note that <700 days of
implemented to calculate the modified pseudo-variables. production history in Fig. 13, consistent with the analysis time period in
Utilizing the corrected pseudo-variables, the modified log-log Chu et al. (2019), are analyzed with flow-regime diagnostics. This is to
IRNPpo =IRNP’po R and CPGpo diagnostics (Eqs. (19)–(21)) are plotted in ensure that the well is always in the transient flow regime (this is
Fig. 12. Comparing Fig. 12 (modified diagnostics) with Fig. 10 (con­ confirmed through evaluation of the pressure change at the reservoir
ventional diagnostics), we see that, after a brief period of bilinear flow boundary at different times in the simulation model discussed below).
(CPGpo >1), the corrected pseudo-variables result in a ½ slope on the While a ½ slope in Fig. 14 is observed on the IRNP’R plot, and the CPG
IRNP’po R and a stabilized CPGpo very close to 1, with some deviation is close to 1 at early time (material balance time < 20 days), the slope on
likely caused by using numerical model output in the pseudo-variable the IRNP’R plot increases, and the CPG shifts to values < 1 after the
calculations. With these corrections in place, the true flow regime strong ramp up in GOR, which in turn is caused by FBHP dropping below
signature of transient linear flow (classical diffusion), as expected from bubble point pressure and gas saturation exceeding critical gas satura­
the flow simulator, and simulated flow geometry (Fig. 8), is observed. tion. This suggests that, as with the simulated case above, multi-phase
flow, and in particular the appearance of a mobile gas phase in the
reservoir, may in part be responsible for the apparent AD behavior
3.2. Field case 1 (from Chu et al., 2019) observed late in time. With the observation that multi-phase flow (and in
particular the presence of a mobile gas phase) is affecting the flow
This field case corresponds to Case 1 presented by Chu et al. (2019). regime interpretation, this effect (plus pressure-dependent perme­
The conventional log-log diagnostics had subsequently been interpreted ability), is accounted for in the modified diagnostics discussed below.
by Raghavan and Chen (2018) to support the conclusion that anomalous Modified diagnostics. An additional step that must be taken for the
diffusion was occurring in the Wolfcamp shale reservoir, as discussed field case, in order to modify the log-log diagnostics for multi-phase
above. flow, is to history-match the well production data with a numerical
The workflow provided in the Theory and Methods section was again simulator. This is required to obtain reasonable saturation and pressure
applied to this case, but with the additional step of using a numerical estimates for use in modified pseudo-variable calculations. The history-
model to history-match the production rates, with the accurately- match was achieved by using a model with the base geometry shown in
measured downhole flowing pressure data applied as a constraint. Fig. 8 (element of symmetry, corresponding to one hydraulic fracture)
This additional step was necessary to provide saturations and pressures and then scaling up the results to the number of fractures. Note that
for use in modified pseudo-variable calculations. reservoir heterogeneity was deliberately ignored in the history-
Conventional diagnostics. Fig. 13 provides plots of the field production matching to 1) determine if a homogeneous model, with reasonable
data, producing fluid ratios and flowing pressure data for this case. From estimates of fracture, reservoir and fluid properties, could match the
these plots, it observed that, while WOR decreases continuously through data and 2) isolate the effects of multi-phase flow and pressure-
the production period, GOR significantly increases. A closer look at early dependent permeability on the diagnostics.
time (Fig. 13d) reveals that GOR is relatively suppressed and flat for the Model input values are provided in Table 1; the actual measured
first two weeks of production, but ramps up continuously thereafter, flowing bottomhole pressures (Fig. 13) were used as a constraint and the
consistent with FBHP dropping below the estimated bubble point pres­ flow rates for all three phases were matched. The values of matrix
sure (2335 psia). Notice that, although bubble point pressure is reached permeability and porosity are identical to those provided by Chu et al.
early in the life of this well, there is no rapid drop in well performance, (2019) for this case. The primary parameters adjusted to achieve the
just the expected and predictable well production characteristics caused history-match were relative permeability, pressure-dependent perme­
by the appearance of gas phase in the reservoir, as seen in conventional ability, and hydraulic fracture properties. The resulting history-match is
oil reservoirs. provided in Fig. 15 – the match is reasonable, except for the early-time
Fig. 5 provided the conventional log-log IRNP=IRNP’R and CPG di­ gas rate, for which the reported values are affected by gas-lift operations
agnostics (Eqs. (3), (4) and (7)) for this field case. This same plot, but (Chu et al., 2019). In order to improve the match of oil production at
early time (50 < t < 100) days), a very narrow zone (0.3 ft) was added
adjacent to the primary hydraulic fracture to simulate a fracture face
skin effect; fracture face skin may have several possible causes
including: imbibition of fracturing fluids, causing a change in fluid
saturations, effective permeability to each phase, and possibly alteration
of wettability and relative permeability; and mechanical damage of the
reservoir matrix near the fracture face, resulting in a loss of absolute
permeability. A more thorough discussion of damage mechanisms was
provided by Barati et al. (2009). The ‘damage’ zone was assigned a
permeability smaller than the matrix, and relative permeability different
from the primary hydraulic fracture and the matrix (Table 2). The
addition of this narrow damage zone does not have a significant effect on
the late time match.
Several different physical models were trialed in an attempt to
improve the early time oil production match, including an enhanced
fracture region (EFR) model (composite reservoir where the region
nearest to the primary hydraulic fractures has elevated permeability and
porosity), a reservoir heterogeneity model with permeability continu­
ously decreasing away from the primary hydraulic fracture (analogous
Fig. 12. Modified log-log diagnostics applied to simulation (verification) case. to Fig. 3), and a homogeneous model (Fig. 8). However, all of them
GOR is also plotted. After a short period of bilinear flow (tcao < 5 days), a failed to match the rapid decline in oil rates at early time unless the
transient linear flow period develops, as identified with a slope of ½ on damage zone was added. We therefore selected the simplest (homoge­
IRNP’po R plot and CPGpo ¼ 1.0. This is the expected flow regime sequence for neous) model, with the damage zone, because of the fewer number of
the simulated case. The modified diagnostics successfully correct for effects of adjustable parameters, and the desire to isolate the effects of reservoir
multi-phase flow and pressure-dependent permeability.

10
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Fig. 13. Field data for Case 1 of Chu et al. (2019) used to diagnose complex reservoir behavior. (a) Semi-log plot of gas, oil and water production rates and Cartesian
plot of flowing bottomhole pressure. (b) Cartesian plot of GOR and flowing bottomhole pressure. The strong ramp up of GOR suggests the development of a mobile
gas phase in reservoir, and the onset of multi-phase flow (oil þ gas þ water). (c) Cartesian plot of WOR and flowing bottomhole pressure. (d) Plot of FBHP and GOR
illustrating the ramp up in GOR as FBHP drops below bubble point pressure.

simulation case presented above, differing from it only because the field
rates/flowing pressures are used instead of the simulator rates and
flowing pressures as in the previous case.
Utilizing the corrected pseudo-variables, the modified log-log
IRNPpo =IRNP’po R and CPGpo diagnostics (Eqs. (19)–(21)) are plotted in
Fig. 17. Comparing Fig. 17 (modified diagnostics) with Fig. 14 (con­
ventional diagnostics), we see that application of the corrected pseudo-
variables in the diagnostics generally shifts the data towards a ½ slope
on the IRNP’po R plot and CPGpo ¼ 1, even causing some CPGpo values (e.
g. between material balance pseudo-times of 30 and 200 days) to be
greater than 1. However, it should be noted, that the result is highly
sensitive to the degree of smoothing applied in the derivative. It can be
said that, while application of the modified diagnostics tends to correct
the flow signature back towards the classical diffusion case, for some
regions of the plot (for example at early time, for material balance
pseudo-time values < 50 days), the result is not as convincing as the
simulated case provided above. A possible cause could be the imperfect
history-match (Fig. 15) achieved with this simple model (e.g., the data of
gas-lift operations at early time was missed), which in turn would result
in errors in extracted saturations and pressures used in modified pseudo-
Fig. 14. Conventional diagnostics applied to field Case 1 of Chu et al. (2019).
GOR is shown to illustrate the relationship between GOR and flow regime variable calculations. Nonetheless, from this analysis it is clear that
changes. A positive ½ slope line is drawn for reference to show deviation away multi-phase flow is affecting the flow regime signature, and must be
from linear flow on IRNP’R plot at late time (tc > 20 days). corrected for to obtain more of a correct flow regime signature.

heterogeneity and multi-phase flow. 4. Discussion


The history match in Fig. 15 is not unique, because of the number of
uncertain parameters that were adjusted to achieve the match. However, In this work, it has been demonstrated, using a simulated and field
the match (with model parameters within acceptable ranges) is deemed example (Wolfcamp shale), that flow regime signatures, as observed
sufficient to yield reasonable estimates of saturations and pressures. using conventional log-log diagnostic plots, may be distorted in the
The P–S relationship, derived from simulation model output, is presence of complex reservoir behavior, such as multi-phase flow. This
presented in Fig. 16. The relationship looks very similar to the observation has the practical importance that, in such cases, conven­
tional diagnostics cannot be relied upon to select a flow model that

11
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Fig. 15. Numerical simulation model history-match of (a) oil, (b) gas and (c) water production rates for field Case 1 of Chu et al. (2019).

Table 2
Summary of numerical model inputs for damage zone used in simulation (field
Case 1).
Parameter Damage Zone

Initial pressure (psia) 3405


Reservoir temperature (� F) 165
5
Initial permeability (md) 1.6 � 10
Initial porosity (fraction) 0.09
Rock compressibility (1/psia) 3 � 10 5
Permeability modulus (1/psia) 3 � 10 4
Oil API gravity 39.6
Gas specific gravity 0.917
Bubble point pressure (psia) 2335
6
Undersaturated oil compressibility (1/psia) 9.8 � 10
Number of fractures –
Thickness (ft) 530
Initial water saturation (fraction) 0.57
SWCON - endpoint saturation: connate water 0.22
SWCRIT - endpoint saturation: critical water 0.22
SOIRW - endpoint saturation: irreducible oil for water-oil table 0.05
SORW - endpoint saturation: residual oil for water-oil table 0.05 Fig. 16. Pressure-oil saturation relationship derived from calibrated (through
SOIRG - endpoint saturation: irreducible oil for gas-liquid table 0.23 history-matching field Case 1 data, shown in Fig. 15) numerical simulation
SORG - endpoint saturation: residual oil for gas-liquid table 0.23 model output.
SGCON - endpoint saturation: connate gas 0.01
SGCRIT - endpoint saturation: critical gas 0.01
KROCW - kro at connate water 1 corresponds to the flow regime sequence, which is the basis of quanti­
KRWIRO - krw at irreducible oil 1 tative rate- and pressure-transient analysis. Further, the distortion of
KRGCL - krg at connate liquid 1 flow regime signatures using these methods also hampers our ability to
KROGCG - krog at connate gas 1 distinguish classical and anomalous diffusion flow mechanisms. It is
Exponent for calculating krw from KRWIRO 7
important to note that we do not deny that AD behavior is occurring in
Exponent for calculating krow from KROCW 3
Exponent for calculating krog from KROGCG 2.5 the field cases studied, just that our ability to distinguish AD from CD
Exponent for calculating krg from KRGCL 2.8 behavior may be obscured by incorrect diagnostics.
A method has been presented to “correct” the log-log diagnostics
using pseudo-variables that account for reservoir complexities. The
focus of this work has been on correction for multi-phase flow. The

12
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

analytical methods may be used to calculate the modified pseudo-


variables (e.g. Behmanesh et al., 2018a) by applying P–S relationships
developed using the approach of Hamdi et al. (2018, 2019). Interest­
ingly, for this case (Hamid et al., 2019), the flow regime signatures for
each phase, as identified with conventional (uncorrected diagnostics)
are not distorted by multi-phase flow (and show the transient linear flow
signature) – it is therefore the combination of multi-phase flow and
variable operating conditions that cause the deviations, as demonstrated
with the simulation case and field Case 1 provided herein. However, the
semi-analytical method of Hamdi et al. (2018, 2019) will not work for
variable operating conditions (necessitating the approach used in this
work), and still requires information highly uncertain (e.g. relative
permeability). In future work, we will simplify log-log diagnostic plot
corrections (including pseudo-variable calculations) to make them more
practical.

4.2. Investigation of the relative impact of reservoir complexities on flow


regime signatures
Fig. 17. Modified log-log diagnostics applied to field Case 1 from Chu et al.
As noted above, in this work, only the impact of multi-phase flow þ
(2019). A slope of ½ on IRNP’po R plot and CPGpo ¼ 1.0 are shown for reference.
pressure-dependent permeability on flow regime signatures identified
The modified diagnostics, which account for the effects of multi-phase flow and
pressure-dependent permeability, shift the data closer to transient linear flow.
with conventional log-log diagnostics was investigated. Of these two
phenomena, multi-phase flow, and in particular the presence of a mobile
gas phase, is thought to be the greatest control on flow regime distortion.
reason for this is that: 1) multi-phase flow is easily diagnosed in the field
This interpretation is also supported through the analysis of an addi­
cases studied by plotting the fluid ratios (in particular GOR); 2) relative
tional field case (Case 3) provided by Chu et al. (2019), for which
permeability is quantifiable in the lab, although its scale-up to field cases
“flowing BHP was above bubble point pressure for two years”. Fig. 18
remains an area of continued research; and 3) other known causes of
provides the conventional log-log diagnostics for this case – only the first
flow regime signature distortion, such as reservoir heterogeneity, are
~150 days (real time) of data were analyzed because later data were
much more difficult to verify and quantify. For example, although recent
affected by frac hits. Again noting the difference in data frequency, and
progress has been made to quantify reservoir heterogeneity using field
possibly degree of derivative smoothing compared with the Chu et al.
data specifically for the Wolfcamp shale (see Ciezobka et al., 2018 for
(2019) results for this case, it can be seen that the CPG stabilizes at
summary, and additional references therein), and the Eagle Ford shale
values between 0.9 and 1.0 for material balance time > 10 days, which is
(Raterman et al., 2018, 2019), in particular fracture complexity, these
consistent with that observed by Chu et al. (2019). The constant GOR
data are not likely applicable to all areas of the field. Further, even for
observed for this case supports the conclusion of Chu et al. (2019) that
areas where the data are available, due to very sparse reservoir sam­
the well is flowing above bubble point. The stabilized CPG value is
pling, the results are difficult to represent properly in a flow simulator.
significantly closer to 1.0 (classical diffusion) than that for Case 1
It is important to note that, although we have attempted to isolate the
analyzed above (CPG ~0.8), which again suggests that gas phase
impacts of multi-phase flow and reservoir heterogeneity in the analysis
mobility is a significant cause of apparent sub-linear flow behavior. The
of field Case 1, it is possible that the effect of reservoir heterogeneity is
slight sub-linear behavior observed for Case 3 of Chu et al. (2019) may
included in the shape of the relative permeability curves used for
be caused by a combination of two-phase (oil and water),
history-matching. It is not yet clear how much reservoir heterogeneity
stress-dependent permeability, reservoir heterogeneity and AD.
has impacted the shape of the relative permeability curves, because, as
Some additional numerical simulation sensitivities have been
stated above, reservoir heterogeneity is difficult to quantify using
existing field data. Because the topic of diagnostic plot modification for
improved flow regime interpretation is a relatively unexplored concept,
further discussion is provided below on topics that we feel require
further investigation.

4.1. Addressing the practicality of the proposed diagnostics modification


method

While the proposed method for log-log diagnostics correction (in­


clusion of modified pseudo-variables) is believed to be a rigorous
approach to account for complex reservoir behavior, it is currently not
practical, prohibiting routine field application. For field cases, the
approach requires the use of numerical simulation, calibrated through
history-matching of field data, to provide pressures and saturations for
pseudo-variable calculations. However, diagnostics are supposed to be
used in advance of, not coincident with, modeling, in order to identify
the correct model to be used for simulation.
The approach demonstrated herein was used to prove theoretically
that diagnostics may be corrected to account for some reservoir com­
plexities, even if the methods used for pseudo-variable calculation are Fig. 18. Conventional log-log diagnostics applied to field Case 3 of Chu et al.
impractical. For the specific case of multi-phase, transient linear flow (2019) with GOR plotted. A slope of ½ on the IRNP’po R plot and CPGpo ¼ 0.9 are
during constant-pressure production, we have demonstrated that semi- shown for reference, suggesting near-transient linear flow for tc > 20 days.

13
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

performed to evaluate the relative impact of pressure-dependent 4.3. Proposed approach for evaluating the impact of complex reservoir
permeability, water production and gas production, and reservoir het­ properties at the lab scale
erogeneity on flow regimes identified with conventional diagnostics.
While the results of this sensitivity will be provided in a future paper, It should be apparent from this study that determining which
they may be summarized as follows: mechanism is causing deviations from classical behavior using field data
(conventional diagnostics) is a difficult task. One approach that can be
� Pressure-dependent permeability (for single-phase flow case) has a used to identify reservoir-specific flow mechanisms is the analysis of gas
relatively small effect on conventional diagnostics, even for cases flow through core-scale samples. A promising approach that may be
where flowing pressure is changing strongly; the impact only starts to used to distinguish these mechanisms at the lab-scale was proposed by
become significant with permeability moduli on the order of 10 3 Clarkson et al. (2019b) and Vahedian et al. (2018). Those authors
(psia 1) demonstrated that rate-transient analysis can be performed on gas
� Water production, while leading to a reduction in oil relative produced from core plugs after gas injection for the purpose of
permeability, and delaying the start of boundary-dominated flow, extracting permeability and pore volume information from the core. A
has a relatively small effect on the conventional diagnostics during conceptual model used to illustrate injection and production from one
the transient linear flow period, even for cases with high producing end of a core plug is provided in Fig. 20. Clarkson et al. (2019b) and
WOR Vahedian et al. (2018) described the experimental setup and procedures
� Gas flow in the reservoir has a strong impact on conventional di­ required to perform core tests using RTA theory.
agnostics, and of the factors sensitized to, causes the greatest de­ An example application of log-log diagnostics for gas
viations from classical behavior, especially for cases with large (IRNPpg =IRNP’pg R and CPGpg ) to the analysis of methane gas production
increases in producing GOR over time from a core plug extracted from a low-permeability siltstone reservoir
� For cases where there is a non-uniform distribution of natural frac­ (Montney Formation) in Western Canada is provided in Fig. 21. Note
tures away from the primary hydraulic fracture (see Fig. 3), and the Qg ðtÞ
that gas material balance time, tc ¼ qg ðtÞ , is used in this example, not
resulting rock properties are fractally distributed, a deviation from
linear flow may be observed material balance pseudo-time. In this example, the sequence of transient
linear flow followed by boundary-dominated flow is confidently iden­
To illustrate the latter point, a fine-grid simulation case generated by tified from the log-log diagnostics, meaning that classical diffusion ap­
Yuan et al. (2019b) was used. Rock properties were fractally distributed pears to be operating for this sample. Classical diffusion was observed in
in a manner represented conceptually in Fig. 3; single phase flow of oil this case, possibly due to a lack of strong heterogeneities in the core.
was assumed. The data input for the model are provided in Tablea 3 of However, this core testing approach could be used to evaluate flow
Yuan et al. (2019b). Application of the conventional log-log diagnostics mechanisms for both natural (shale core plugs) and synthesized mate­
to this case is shown in Fig. 19. For this case, only a slight deviation from rials in order to gain a better understanding of the controls on AD at the
linear flow is observed (CPG ~0.9), but further sensitivities are required. core scale. For example, AD parameters (e.g. α’f , α’m ) derived from
In future work, a more thorough investigation of the relative impact modeling this kind of core production data could be related to other
of other complexities listed in the Theory and Methods section, petrophysical properties that can be directly measured for the core, such
including reservoir/fracture heterogeneity and operational effects, and as pore size distributions, or fractal dimension of the fracture system. If
the uncertainty analysis of our history matching will be investigated. A deviations from linear flow are observed using the core-derived rate-
future paper will address the inclusion of reservoir heterogeneity, along transient signatures, the cause could be isolated because the sample can
with multi-phase flow and pressure-dependent permeability, into the be fully characterized, unlike a subsurface reservoir. However, these
modified log-log diagnostics for flow regime identification. observations would be difficult to scale up to the field, unless scaling
laws such as fractals can be applied. For example, if a fractal-type
fracture geometry is observed in core, and it affects the flow regime
signature, then one may be able to infer this relationship for field data.
This is an important topic of future research.

5. Conclusions

Flow regime identification is a critical step used in rate-transient and


pressure-transient analysis (RTA/PTA) to determine what model is
appropriate for analyzing the data. This process involves the use of log-
log diagnostic plots on which flow regimes may be identified as char­
acteristic slopes. These slopes are conventionally interpreted in the

Fig. 19. Conventional log-log diagnostics applied to simulation case of Yuan


et al. (2019b) in which a fractal distribution of reservoir properties was
assumed CPG ¼ 0.9 is shown for reference, suggesting near-transient linear flow Fig. 20. A conceptual model used to illustrate injection and production from
for tc > 400 days. one end of a core plug. Modified from Clarkson et al. (2019b).

14
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

dominant transient flow regime for the model flow geometry is


identified (transient linear flow); failure to correct the diagnostics
results in a misdiagnosis of sub-linear flow
� For one Wolfcamp shale case studied, where multi-phase flow is
evident, the modified diagnostics result in a flow regime signature
that is closer to, but not exactly equal, that expected for classical
diffusion, while the conventional (uncorrected) diagnostics clearly
show deviations from classical behavior
� For another Wolfcamp shale case, where only two-phase flow is
evident, the conventional (uncorrected) diagnostics result in a flow
regime signature that is very close to that expected for classical
diffusion
� The CPG diagnostic may lead to erroneous flow regime signatures
when skin effects are present – in these cases, use of IRNP’R is
preferred
� The proposed modified diagnostics, while theoretically correct and
rigorous, are not practical for routine field application

Fig. 21. Conventional log-log diagnostics applied to core gas production data Future work will endeavor to improve the practicality of the pro­
from Clarkson et al. (2019b). Transient linear flow is followed by
posed log-log diagnostics modification, investigate the impact of other
boundary-dominated flow.
reservoir complexities (e.g. reservoir heterogeneity) on flow regime
identification, improve the usefulness of the CPG diagnostic, and
context of classical diffusion, and may be determined theoretically from continue to develop experimental methods to evaluate the impact of
analytical solutions. Multi-fractured horizontal wells completed in shale various reservoir complexities on flow regime signatures. Further, a
reservoirs often exhibit power-law decline behavior, with transient future paper will address the inclusion of reservoir heterogeneity, along
linear flow (classical diffusion) being a commonly observed flow regime. with multi-phase flow and pressure-dependent permeability, into the
Deviations away from this classical (linear flow) regime have recently modified log-log diagnostics for improved flow regime identification.
been observed for Permian Wolfcamp shale wells and used as evidence The relative impact of these effects on flow regime interpretation will
that the anomalous diffusion transport mechanism is operating in this therefore be quantified.
reservoir. However, besides anomalous diffusion and reservoir hetero­
geneity, which are both plausible causes of deviations from classical CRediT authorship contribution statement
behavior, it is argued herein that other reservoir complexities may be
contributing to these deviations including multi-phase flow. Because C.R. Clarkson: Conceptualization, Methodology, Writing - original
conventional log-log diagnostic plots do not account for these effects, draft, Supervision. B. Yuan: Software, Formal analysis, Writing - review
the flow regime signatures may be distorted, hampering our ability to & editing. Z. Zhang: Software, Formal analysis, Writing - review &
properly identify flow regimes and select an appropriate model for editing. F. Tabasinejad: Software. H. Behmanesh: Formal analysis. H.
analysis of the data. Incorrect model selection will of course result in Hamdi: Formal analysis, Writing - review & editing. D. Anderson:
errors in reservoir/hydraulic fracture property evaluation. Project administration, Writing - review & editing. J. Thompson:
In this work, a workflow is introduced to “correct” log-log diagnostic Project administration, Writing - review & editing. D. Lougheed: Proj­
plots for multi-phase flow and stress-dependent (approximated as ect administration, Writing - review & editing.
pressure-dependent) permeability, the former believed to be particu­
larly important for some Wolfcamp shale wells. Specifically, pseudo- Acknowledgements
pressure and pseudo-time variables are corrected for these effects and
used to create modified log-log diagnostics. The authors are grateful to the sponsors of Tight Oil Consortium
Through the analysis of simulated and field cases, the following (TOC), hosted at University of Calgary, and would like to acknowledge
conclusions can be drawn: financial support provided through an NSERC Collaborative Research
and Development (CRDPJ 501135–16) grant to Clarkson. Chris Clarkson
� Multi-phase flow can cause deviations away from classical linear would like to thank Encana and Shell for support of his Chair position in
flow behavior when conventional (uncorrected) log-log diagnostics Unconventional Gas and Light Oil Research at the University of Calgary,
are used for flow regime identification Department of Geoscience. An early version of this work was presented
� The appearance of a mobile gas in the reservoir after the production at the SPE/AAPG/SEG Unconventional Resources Technology Confer­
of oil and water phases only, and consequent changes in relative ence (URTeC) held in Denver, Colorado, USA (22-24 July, 2019) and
permeability and fluid compressibility, cause the greatest deviations published as URTEC-2019-85-MS.
from linear flow
� When the modified diagnostics are applied to a simulated case with
multi-phase flow and pressure-dependent permeability, the expected

Appendix A. Derivation of Pseudo-pressure and Pseudo-time to Account for Multiphase Flow and Stress-Sensitivity

In shale reservoirs hosting light oil, the governing black oil diffusivity equations for oil, gas and water components in a linear flow system can be
written in Cartesian coordinates as follows:
� � � �
∂ kro ∂p 1 ∂ ∅So ∂p
k ¼ (A-1)
∂y μo Bo ∂y 0:00633 ∂p Bo ∂t

15
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

� � � � � �
∂ krg kro ∂p 1 ∂ ∅So ∅Sg ∂p
k þ Rs ¼ Rs þ (A-2)
∂y μg B g μo B o ∂y 0:00633 ∂p Bo Bg ∂t
� � � �
∂ krw ∂p 1 ∂ ∅Sw ∂p
k ¼ (A-3)
∂y μw Bw ∂y 0:00633 ∂p Bw ∂t
Capillary pressure and gravity effects are ignored. In order to make the equations more compact, the following notations are used:
� �
k ∅S k k ∅S ∅S k ∅Sw
α ¼ k ro ; β ¼ o ; ε ¼ k rg þ Rs ro ; b ¼ Rs o þ g ; γ ¼ k rw ; ω ¼
μo B o Bo μg B g μo B o Bo Bg μw B w Bw

The introduction of pseudo-pressure (Al-Hussainy et al., 1966) and pseudo-time (Agarwal, 1980) is a popular approach to approximately linearize
the non-linear diffusivity equation with variable coefficients. Hence, in this work, to incorporate the effect of both pressure-dependent reservoir rock
(e.g. geomechanical effects) and fluid properties (e.g. compressibility), pseudo-time and pseudo-pressure are applied. Although application of
pseudo-pressure and pseudo-time do not serve to completely linearize the diffusivity equation, different authors have studied the usefulness of these
pseudo-variables for both pressure and rate transient analysis with some modifications. For instance, in unconventional reservoirs with transient
linear flow as the dominant flow regime, the definition of pseudo-time in terms of the average pressure within dynamic-drainage-area should be
applied (Anderson and Mattar, 2007); this is the case in our derivations, presented as follows.
The oil diffusivity equation (Eq. A-1) can be partially linearized by introducing the pseudo-pressure transformation as follows:
Z
μ Boi p
ppo ¼ oi αðp; So Þdp (A-4)
ki kroi pref

where μoi is oil viscosity at the initial reservoir pressure, cp; Boi is the initial oil formation volume factor, bbl/STB; ki is the initial reservoir perme­
ability, md; kroi is the initial oil relative permeability, fraction.
The effects of three-phase flow and reservoir permeability stress-sensitivity (approximated as pressure-sensitivity) have been incorporated into the
calculations of α in Eq. A-4 by including oil relative permeability as a function of saturation of the three phases, kro and intrinsic permeability as a
function of pressure, kðpÞ, respectively.
In order to calculate the pseudo-pressure in Eq. A-4, a relationship between saturation and pressure (P–S relationship) is required to integrate
relative permeability over pressure. In this work, P–S is obtained directly from numerical simulation by reading the pressure and saturation of each
grid block over distance. It is concluded that the spatial coordinate has negligible effects on the general P–S relationship through the outputs of each
grid block in numerical simulation, which is consistent with the similarity theory using Boltzmann’s variable (Stewart, 2011). For field cases, the
numerical model must be calibrated to field data through history-matching to ensure that the saturation and pressures obtained are accurate.
Additional methods for estimating the P–S relationship for two- and three-phase cases have been presented in the literature, including: steady-steady
path (Evinger and Muskat, 1942), unsteady-state path (Behmanesh et al., 2015; Zhang et al., 2016; Hamid et al., 2018),
constant-composition-expansion (CCE) (Walsh and Lake, 2003), constant-volume-depletion (CVD) tests (Fevang and Whitson, 1996), and the tank
model approach (Walsh and Lake, 2003).
To more completely linearize the oil diffusivity equation (Eq. A-1), a pseudo-time transformation is also required to linearize the temporal part of
the equation. The pseudo-time is defined using an integration over time, but not pressure, yet, the reservoir and fluid properties are provided as a
function of pressure. Hence, as suggested by Anderson and Mattar (2007), it is appropriate to evaluate the integrand (α=ðdβ=dpÞÞ in terms of the
average pressure within dynamic-drainage-area, while calculating the pseudo-time using Eq. A-5.
0 1
Z t
∅i μoi cti Soi BαC
tao ¼ B C dt
@dβA (A-5)
ki kroi 0 dp
pinv

where ∅i is initial reservoir porosity, fraction; and cti is the total formation compressibility at initial pressure, 1/psia.
The resulting linearized form of oil diffusivity equation for three-phase flow conditions, using three-phase pseudo-pressure and pseudo-time, can
be obtained as follows:
∂2 ppo 1 ∅i μoi cti Soi ∂ppo
¼ (A-6)
∂x2 0:00633 ki kroi ∂tao
Eq. A-6 is the classical linear diffusivity equation with constant coefficients, which can be easily solved to obtain both the transient linear flow
solution and the boundary-dominated flowing solution (pseudo-steady state solution or steady-state solution).

Appendix B. Derivation of Material-Balance Pseudo-time to Account for Variable Rates

Following the linearization of the multiphase diffusivity equation with the use of pseudo-variables, one can easily obtain the solution to Eq. A-6 in
terms of pseudo-pressure and pseudo-time, given the initial and boundary conditions. Duhamel’s theorem (Thompson and Reynolds, 1986) can then
be introduced to the constant-rate solution to obtain a solution that is applicable to more general cases with variable flowing rates/pressure production
data. For the cases with discrete rate changes, superposition of the transient flow solution can be written as follows:
� �� � �� � ��
� ppo ðpi Þ ppo pwf � ppo ðpi Þ ppo pwf � ppo ðpi Þ ppo pwf
ppo ðpi Þ ppo pwf ¼ qo;1 þ qo;2 qo;1 þ … þ qo;n qo;n 1 (B-1)
qo tao qo tao tao;1 qo tao tao;n 1

In integral form, Eq. B-1 can be written as,

16
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Z tao �
��
� ppo ðpi Þ ppo pwf ðtao τao Þ
ppo ðpi Þ ppo pwf ¼ dqo ðtao τao Þ (B-2)
0 qo
Inspired by the work of Liang, 2015, boundary-dominated flow superposition time can preserve the reservoir signature for other flow regimes
including transient linear flow. Moreover, the “dynamic drainage area” (DDA) approach has been developed to evaluate transient flow using the
succession of pseudo-steady states solution with a moving boundary (Muskat, 1937; Lee et al., 1998). This approximation method enables the use of
material balance pseudo-time to analyze the production data during transient flow with acceptable accuracy, which has been examined by various
authors (e.g., Moghanloo et al., 2015; Behmanesh, 2016; Clarkson and Qanbari, 2016; Yuan et al., 2017). Hence, by applying the concept of “suc­
cession of pseudo-steady states” with the assumption that the DOI ðyinv Þ at each timestep is known as a constant, the solution of Eq. (A-6) can be
substituted into Eq. (B-2) yielding:
Z ta Z tao

ppo ðpi Þ ppo pwf ¼ ðb þ mðtao τao ÞÞdqo ðtao τao Þ ¼ bqo ðtao Þ þ m qo ðtao τao Þdτao (B-3)
0 0

which can be rearranged as:


� R tao
ppo ðpi Þ ppo pwf qo ðtao τao Þdτao
¼bþm 0 ¼ b þ mtcao (B-4)
qo ðtao Þ qo ðtao Þ

where,
� �
π � 0:00633 141:2Boi μoi π yinv h
m ¼ 141:2Boi ;b¼ þ Sf
2∅i cti xf yinv hSoi ki kroi h 6xf 2xf

For the transient flow period, yinv is distance of investigation (DOI), propagating into the reservoir over time, ft; h is the reservoir height, ft; xf is the
effective fracture half-length, ft; and tcao is oil material balance pseudo-time. The DOI equation is useful to determine the expanding boundaries at each
timestep needed in the succession of pseudo-steady states. tcao can be represented in terms of real time, t. To achieve this, first, the general material
balance equation in integral form, from the well location (0) to the DOI (yinv ) at each timestep is derived as:
Z yinv
4xf h �
Qo ¼ βi βðyÞ dy (B-5)
5:615 0

Where, Qo is the cumulative oil production, STB; βi is the initial reservoir condition of β; and βðyÞ is the in-situ conditions of β at a given location (nth
grid block in the presented numerical simulation model in this work) in reservoir. Note that, as the value of the integrand is evaluated only within the
distance of investigation (a constant value at each timestep), the integration variable y is independent of time t, within the known distance of
investigation. This equation can be re-arranged in the form of production rate, as follows:
Z yinv !
dQo 4xf h dβðyÞ dpðyÞ
qo ðtÞ ¼ ¼ dy � (B-6)
dt 5:615 0 dpðyÞ dt


Note that, dpðyÞ
ðyÞ
is the “total derivative” of βðyÞ with respect to pðyÞ at each location in reservoir. As we are utilizing the concept of the succession of the
dp
pseudo-steady states to approximate transient flow at any time, the common observation for pseudo-steady state flow that the pressure drop dtðyÞ is
constant and independent of reservoir location is also approximately valid. Hence, we can move the pressure drop term outside of the integral.
Furthermore, according to the definition of pseudo-pressure in Eq. (A-4), we can obtain the in-situ pseudo-pressure at a given reservoir location, as
follows:
μoi Boi
dppoðyÞ ¼ αðyÞ dpðyÞ (B-7)
ki kroi
Combining Eq. B-7 with Eq. B-6 yields the change in pseudo-pressure over time at a given location within the distance of investigation, yinv :
μoi Boi 5:615qo ðτÞαðyÞ
dppoðyÞ ¼ ! dτ (B-8)
ki kroi R yinv dβðyÞ
4xf h 0 dpðyÞ
dy

By applying the integral for Eq. B-8 over time, the following is obtained:
Z
� μ Boi t 5:615qo ðτÞαðyÞ
ppo ðpi Þ ppo pðyÞ ¼ oi ! dτ (B-9)
ki kroi 0 R yinv dβðyÞ
4xf h 0 dpðyÞ
dy

Now the changes in average reservoir pseudo-pressure over time can be obtained by using the volumetric average of Eq. B-9 as follows:
R yinv R t 5:615qo ðτÞαðyÞ
0 0
! dτdy
R yinv dβðyÞ
4xf h dpðyÞ dy
0
μoi Boi
ppo ðpi Þ ppo ðpÞ ¼ (B-10)
ki kroi yinv
The calculation of multiple integrals over both distance and time is independent of the order of integrands, and therefore, Eq. B-10 can be further
simplified as:

17
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Z t
R yinv
μoi Boi 5:615 αðyÞ dy
ppo ðpi Þ ppo ðpÞ ¼ qo ðτÞ 0 ! dτ (B-11)
ki kroi 4xf hyinv 0 R yinv dβðyÞ
0 dpðyÞ
dy

In addition, the oil production rate can be expressed using the productivity index and the difference between pseudo-pressure at average reservoir
pressure and pseudo-pressure at well bottomhole flowing pressure (Ramey and Cobb, 1971), as follows:
� �
� 141:2Boi μoi πyinv h
ppo ðpÞ ppo pwf ¼ qo ðtÞ þ Sf (B-12)
ki kroi h 6xf 2xf

As suggested in the work of Walsh and Lake (2003) and Barrios et al. (2003), for small spatial gradients we can assume that average
pseudo-pressure is almost equal to pseudopressure at average pressure, i.e. ppo ðpÞ � ppo ðpÞ. The impact of their difference on the forecasting of well
productivity index will be investigated in future work for more general cases with varying flowing conditions, pressure-dependent rock and fluid
properties, and multiphase relative permeability curves.
Then, combining Eq. B-12 with Eq. B-11, we obtain,
� Z t R yinv � �
ppo ðpi Þ ppo pwf μ Boi 5:615 1 αðyÞ dy 141:2Boi μoi πyinv h
¼ oi qo ðτÞ 0 ! dτ þ þ Sf (B-13)
qo ðtÞ ki kroi 4xf hyinv qo ðtÞ 0 R yinv dβðyÞ ki kroi h 6xf 2xf
0 dpðyÞ
dy

Lastly, by comparing Eq. B-13 with Eq. B-4, we can obtain the general form of material balance pseudo-time, as follows:
Z tao Z t R yinv
∅i μoi cti Soi 1 ∅i μoi cti Soi 1 αðyÞ dy
tcao ¼ qo ðτao Þdτao ¼ qo ðτÞ 0 ! dτ (B-14)
ki kroi qo ðtao Þ 0 ki kroi qo ðtÞ 0 R yinv dβðyÞ
0 dpðyÞ
dy

Appendix C. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.jngse.2020.103187.

Nomenclature

a Exponent of power-law function for RNP, see Eq. (8)


A Constant of power-law equation for RNP, see Eq. (8)
Bo Oil formation volume factor, bbl/STB
Bw Water formation volume factor, bbl/STB
cf Rock compressibility, 1/psi
cg Gas compressibility, 1 psi
co Oil compressibility, 1/psi
ct Total compressibility, 1/psi
h Formation thickness, ft
k Formation permeability, md
krg Gas relative permeability as a function of gas saturation
kro Oil relative permeability as a function of oil saturation
krw Water relative permeability as a function of water saturation
pb Bubblepoint pressure, psi
pgc Pressure for which critical gas saturation is achieved, psi
pi Initial reservoir pressure, psi
pwf Well flowing pressure, psi
pinv Average reservoir pressure within dynamic-drainage-area, psi
qo Oil production rate, STB/D
qg Gas production rate, Mscf/D
Qo ¼ Cumulative oil production, STB
Qg Cumulative gas production, Mscf
ppg Gas pseudo-pressure, psi2/cp
ppo Oil pseudo-pressure, psi
Rs Solution gas-oil ratio, scf/STB
Sf Skin factor of a well with a fracture
Sg Gas saturation, fraction
Sgc Critical gas saturation, fraction
So Oil saturation, fraction
Soi Initial oil saturation, fraction
Sorg Residual oil saturation in oil/gas system, fraction
Sorw Residual oil saturation in oil/water system, fraction

18
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Sw Water saturation, fraction


Swc Critical water saturation, fraction
Swi Initial water saturation, fraction
t Production time, days
tao Oil pseudo-time, days
tc Material balance time, days
tcag Gas material balance pseudo-time, days
tcao Oil material balance oil pseudo-time, days
xf Fracture half-length, ft
yinv Distance of investigation, ft
Z Gas compressibility factor
α Oil flowing efficiency parameter, see Eq. (16)
β Oil storage parameter, see Eq. (18)
γ Water flowing efficiency parameter, see Eq. A-3
γk Permeability modulus, 1/psi
ε Gas flowing efficiency parameter, see Eq. (A-2)
μg Gas viscosity, cp
μo Oil viscosity, cp
μw Water viscosity, cp
∅ Formation porosity, fraction
CPG ¼ Chow pressure group, see Eq. (5)
CPGpo Chow oil pseudo-pressure group, see Eq. (21)
RNP Rate-normalized pressure, psi/(STB/D), see Eq. (1)
RNPpg Rate-normalized gas pseudo-pressure, psi2/cp/scf/D, see Eq. (12)
RNP’R Radial derivative of RNP (for liquids, with respect to lntc ), see Eq. (2)
RNP’pg R Radial derivative of RNPpg (for gases, with respect to lntcag ), see Eq. (13)
IRNP Integral of RNP, psi/(STB/D), see Eq. (3)
IRNPpo Integral of rate-normalized oil pseudo-pressure, psi/(STB/D), see Eq. (19)
IRNP’R Radial derivative of IRNP (for liquids, with respect to lntc ), see Eq. (7)
IRNP’po R Radial derivative of IRNPpo (for oil, with respect to lntcao ), see Eq. (20)
Subscript i Values at initial reservoir pressure
Subscript OGW Oil, gas, water
Subscript OW Oil, water

References Behmanesh, H., Hamdi, H., Clarkson, C.R., 2015. Analysis of transient linear flow
associated with hydraulically-fractured Tight oil wells exhibiting multi-phase flow.
In: Presented at the SPE Middle East Unconventional Gas Conference and Exhibition,
Acu~na, J.A., 2016. Analytical pressure and rate transient models for analysis of complex
Muscat, Oman, 26-28 January. https://doi.org/10.2118/172928-MS. SPE-172928-
fracture networks in Tight reservoirs. In: Presented at SPE/AAPG/SEG
MS.
Unconventional Resources Technology Conference, San Antonio, Texas, 1-3 August.
Behmanesh, H., Hamdi, H., Clarkson, C.R., Thompson, J.M., Anderson, D.M., Chin, A.,
https://doi.org/10.15530/URTEC-2016-2429710. URTEC-2429710-MS.
2018a. Transient linear flow analysis of multi-fractured horizontal wells considering
Acu~na, J.A., 2017. Pressure and rate transient analysis in fracture networks in Tight
three-phase flow and pressure-dependent rock properties. In: Presented at the SPE/
reservoirs using characteristic flow volume. In: Presented at SPE/AAPG/SEG
AAPG/SEG Unconventional Resources Technology Conference, Houston, Texas, 23-
Unconventional Resources Technology Conference, Austin, Texas, 24-26 July.
25 July. https://doi.org/10.15530/URTEC-2018-2884255. URTEC-2884255-MS.
https://doi.org/10.15530/URTEC-2017-2667753. URTEC-2667753-MS.
Behmanesh, H., Mattar, L., Thompson, J.M., Anderson, D.M., Nakaska, D.W., Clarkson, C.
Agarwal, R.G., 1980. A new method to account for producing time effects when
R., 2018b. Treatment of rate-transient analysis during boundary-dominated flow.
drawdown type curves are used to analyze pressure buildup and other test data. In:
SPE J. 23 (4), 1145–1165. https://doi.org/10.2118/189967-PA. SPE-189967-PA.
Presented at SPE Annual Technical Conference and Exhibition. Society of Petroleum
Bourdet, D., Ayoub, J.A., Pirard, Y.M., 1989. Use of pressure derivative in well test
Engineers, Dallas, Texas, 21-24 September. SPE-9289-MS.
interpretation. SPE Form. Eval. 4 (2), 293–302. https://doi.org/10.2118/12777-PA.
Al-Hussainy, R., Ramey Jr., H.J., Crawford, P.B., 1966. The flow of real gases through
SPE-12777-PA.
porous media. J. Petrol. Technol. 18 (5), 624–636.
Bozorgzadeh, M., Gringarten, A.C., 2005. Application of buildup transient pressure
Anderson, D.M., Mattar, L., 2007. An improved pseudo-time for gas reservoirs with
analysis to well deliverability forecasting in gas condensate reservoirs using single-
significant transient flow. J. Can. Pet. Technol. 46 (7), 49–54. https://doi.org/
phase and two-phase pseudopressures. In: Presented at the SPE Europec/EAGE
10.2118/07-07-05. PETSOC-07-07-05.
Annual Conference, Madrid, Spain, 13-16 June. https://doi.org/10.2118/94018-MS.
Anderson, D.M., Nobakht, M., Moghadam, S., Mattar, L., 2010. Analysis of production
SPE-94018-MS.
data from fractures shale gas wells. In: Presented at SPE Unconventional Gas
Chow, V.T., 1952. On the determination of transmissibility and storage coefficients from
Conference, Pittsburgh, Pennsylvania, 23-25 February. https://doi.org/10.2118/
pumping test data. Trans. Am. Geophys. Union 33, 397–404. https://doi.org/
131787-MS. SPE-131787-MS.
10.1029/TR033i003p00397.
Barati, R., Hutchins, R.D., Friedel, T., Ayoub, J.A., Dessinges, M.-N., England, K.W.,
Chu, W.-C., Pandya, N.D., Flumerfelt, R.W., Chen, C., 2019. Rate-transient analysis based
2009. Fracture impact of yield stress and fracture-face damage on production with a
on the power-law behavior for permian wells. SPE Res. Eval. Eng. https://doi.org/
three-phase 2D model. SPE Prod. Oper. 24 (2), 336–345. https://doi.org/10.2118/
10.2118/1687180-PA preprint. SPE-187180-PA.
111457-PA. SPE-111457-PA.
Ciezobka, J., Courtier, J., Wicker, J., 2018. Hydraulic fracturing test site (HFTS) – Project
Barrios, K., Stewart, G., Davies, D., 2003. A novel methodology for the analysis of well
overview and summary of results. In: Presented at the SPE/AAPG/SEG
test responses in gas condensate reservoirs. In: Presented at SPE Latin American and
Unconventional Resources Technology Conference, Houston, Texas, 23-25 July.
Caribbean Petroleum Engineering Conference, Port of Spain, Trinidad, West Indies,
https://doi.org/10.15530/URTEC-2018-2937168. URTEC-2937168-MS.
27-30 April. SPE-81039-MS.
Clarkson, C.R., 2013. Production data analysis of unconventional gas wells: review of
Becker, M.D., Zhang, M., Ayala, L.F., 2016. On the pressure-saturation path in infinite-
theory and best practices. Int. J. Coal Geol. 109–110, 101–146. https://doi.org/
acting unconventional liquid-rich gas reservoirs. J. Nat. Gas Sci. Eng. 35, 97–113.
10.1016/j.coal.2013.01.002.
https://doi.org/10.1016/j.jngse.2016.08.007.
Clarkson, C.R., Pedersen, P.K., 2010. Tight oil production analysis: adaptation of existing
Behmanesh, H., 2016. Rate-Transient Analysis of Tight Gas Condensate and Black Oil
rate-transient analysis techniques. In: Presented at the Canadian Unconventional
Wells Exhibiting Two-phase Flow. PhD dissertation. University of Calgary, Calgary
Resources and International Petroleum Conference, Calgary, Alberta, 19-21 October.
(January 2016).
https://doi.org/10.2118/137352-MS. SPE-137352.

19
C.R. Clarkson et al. Journal of Natural Gas Science and Engineering 76 (2020) 103187

Clarkson, C.R., Pedersen, P.K., 2011. Production analysis of western Canadian Nobakht, M., Mattar, L., 2012. Analyzing production data from unconventional gas
unconventional light oil plays. In: Presented at the Canadian Unconventional reservoirs with linear flow and apparent skin. J. Can. Pet. Technol. 51 (1), 52–59.
Resources Conference, Calgary, Alberta, 15-17 November. https://doi.org/10.2118/ https://doi.org/10.2118/137454-PA. SPE-137454-PA.
149005-MS. SPE-149005. Nobakht, M., Clarkson, C.R., 2012a. A new analytical method for analyzing linear flow in
Clarkson, C.R., Jordan, C., Ilk, D., Blasingame, T., 2012. Rate-transient analysis of 2- tight/shale gas reservoirs: constant-rate boundary condition. SPE Reservoir Eval.
phase (gas þ water) CBM wells. J. Nat. Gas Sci. Eng. 8, 106–120. https://doi.org/ Eng. 15 (1), 51–59. https://doi.org/10.2118/143990-PA. SPE-143990-PA.
10.1016/j.jngse.2012.01.006. Nobakht, M., Clarkson, C.R., 2012b. A new analytical method for analyzing linear flow in
Clarkson, C.R., Qanbari, F., Nobakht, M., Heffner, L., 2013. Incorporating geomechanical tight/shale gas reservoirs: constant-flowing-pressure boundary condition. SPE
and dynamic hydraulic-fracture-property changes into rate-transient analysis: Reservoir Eval. Eng. 15 (3), 370–384. https://doi.org/10.2118/143989-PA. SPE-
example from the haynesville shale. SPE Reservoir Eval. Eng. 16 (3), 303–316. 143989-PA.
https://doi.org/10.2118/162526-PA. SPE-165526-PA. Pedrosa, O.A., 1986. Pressure transient response in stress-sensitive formations. In:
Clarkson, C.R., Qanbari, F., 2015a. A semianalytical forecasting method for Presented at the SPE California Regional Meeting, Oakland, California, 2-4 April.
unconventional gas and light oil wells: a hybrid approach for addressing the https://doi.org/10.2118/15115-MS. SPE-15115-MS.
limitations of existing empirical and analytical methods. SPE Reservoir Eval. Eng. 18 Raghavan, R., Chu, W.C., Jones, J.R., 1999. Practical considerations in the analysis of
(1), 94–110. https://doi.org/10.2118/170767-PA. SPE-170767-PA. gas-condensate well tests. SPE Reservoir Eval. Eng. 2 (3), 288–295. https://doi.org/
Clarkson, C.R., Qanbari, F., 2015b. An approximate semianalytical multiphase 10.2118/56837-PA. SPE-56837-PA.
forecasting method for multifractured Tight light-oil wells with complex fracture Raghavan, R., Chen, C.C., 2018. A conceptual structure to evaluate wells producing
geometry. J. Can. Pet. Technol. 54 (6), 489–508. https://doi.org/10.2118/178665- fractured rocks of the Permian Basin. In: Presented at the SPE Annual Technical
PA. SPE-178665-PA. Conference and Exhibition, Dallas, Texas, 24-26 September. https://doi.org/
Clarkson, C.R., Qanbari, F., 2016. History matching and forecasting Tight gas condensate 10.2118/191484-MS. SPE-191484-MS.
and oil wells by use of an approximate semi-analytical model derived from the Ramey, H.J., Cobb, W.M., 1971. A general pressure buildup theory for a well in a closed
dynamic-drainage-area concept. SPE Reservoir Eval. Eng. 19 (4), 540–552. https:// drainage area. J. Petrol. Technol. 23 (12), 1493–1505. https://doi.org/10.2118/
doi.org/10.2118/175929-PA. SPE-175929-PA. 3012-PA. SPE-3012-PA.
Clarkson, C., Yuan, B., Zhang, Z., Tabasinejad, F., Behmanesh, H., Hamdi, H., Raterman, K.T., Farrell, H.E., Mora, O.S., et al., 2018. Sampling a stimulated rock
Anderson, D., Thompson, J., Lougheed, D., 2019a. Anomalous diffusion or classical volume: an Eagle Ford example. SPE Reservoir Eval. Eng. 21 (4), 927–941. https://
diffusion in an anomalous reservoir? Evaluation of the impact of multi-phase flow on doi.org/10.2118/191375-PA. SPE-191375-PA.
reservoir signatures in unconventional reservoirs. In: Presented at SPE/AAPG/SEG Raterman, K., Liu, Y., Warrem, L., 2019. Analysis of a drained rock volume: an Eagle
Unconventional Resources Technology Conference, Denver, Colorado, 22-24 July. Ford example. In: Presented at the SPE/AAPG/SEG Unconventional Resources
URTEC-2019-85-MS. Technology Conference, Denver, Colorado, 22-24 July. https://doi.org/10.105530/
Clarkson, C.R., Vahedian, A., Ghanizadeh, A., Song, C., 2019b. A new low-permeability urtec-2019-263. URTEC-2019-263-MS.
reservoir core analysis method based on rate-transient analysis theory. Fuel 235, Solano, N.A., Clarkson, C.R., Krause, F.F., 2016. Characterization of fine-scale rock
1530–1543. https://doi.org/10.1016/j.fuel.2018.07.115. structure and differences in mechanical properties in Tight oil reservoirs: an
Clarkson, C., Yuan, B., Zhang, Z., 2019c. A new straight-line analysis method for evaluation at the scale of elementary lithological components combining
estimating fracture/reservoir properties based on dynamic fluid-in-place photographic and X-ray computed tomographic imaging, profile-permeability and
calculations. In: Presented at SPE Annual Technical Conference and Exhibition, microhardness testing. J. Unconv. Oil Gas Res. 15, 22–42. https://doi.org/10.1016/
Calgary, Canada, 30 September-2 October. SPE-195930-MS. j.juogr.2016.04.003.
Evinger, H.H., Muskat, M., 1942. Calculation of theoretical productivity factor. Trans. Stalgorova, E., Mattar, L., 2012. Practical analytical model to simulate production of
AIME 146 (1), 126–139. https://doi.org/10.2118/942126-G. horizontal wells with branch fractures. In: Presented at the SPE Canadian
Fevang, Ø., Whitson, C.H., 1996. Modeling gas-condensate well deliverability. SPE Unconventional Resources Conference, Calgary, Alberta, 30 October-1 November.
Reservoir Eval. Eng. 11 (4), 221–230. https://doi.org/10.2118/30714-PA. SPE- https://doi.org/10.2118/162515-MS. SPE-162515-MS.
30714-PA. Stewart, G., 2011. Well Test Design & Analysis. PennWell, Tulsa, Oklahoma, p. 1544.
Hamdi, H., Ghahri, P., Sousa, M.C., et al., 2013. Application of two-phase pseudo- Stone, H.L., 1973. Estimation of three-phase relative permeability and residual oil data.
pressure in gas-condensate well-testing with and without positive coupling and J. Can. Pet. Technol. 12 (4), 53–61. https://doi.org/10.2118/73-04-06. PETSOC-73-
inertia. In: Presented at the 75th EAGE Conference & Exhibition Incorporating SPE 04-06.
EUROPEC 2013, London, UK. https://doi.org/10.3997/2214-4609.20130187. Thompson, L.G., Reynolds, A.C., 1986. Analysis of variable-rate well-test pressure data
Hamdi, H., Behmanesh, H., Clarkson, C.R., 2018. A semi-analytical approach for analysis using duhamel’s principle. SPE Form. Eval. 1 (5), 453–469. https://doi.org/
of transient linear flow regime in Tight reservoirs under three-phase flow conditions. 10.2118/13080-PA. SPE-13080-PA.
J. Nat. Gas Sci. Eng. 54, 283–296. https://doi.org/10.1016/j.jngse.2018.04.004. Vahedian, A., Clarkson, C.R., Ghanizadeh, A., Zanganeh, B., Song, C., Hamdi, H., 2018.
Hamdi, H., Behmanesh, H., Clarkson, C.R., 2019. A semi-analytical approach for analysis Use of rate-transient analysis techniques for evaluating experimental core
of wells exhibiting multi-phase transient linear flow: application to field data. In: permeability tests for unconventional reservoirs. In: Presented at the SPE/AAPG/
Presented at the SPE Annual Technical Conference and Exhibition, Calgary, Alberta, SEG Unconventional Resources Technology Conference, Houston, Texas, 23-25 July.
Canada, Sep 30- Oct 2. https://doi.org/10.2118/196164-MS. SPE-196164-MS. https://doi.org/10.15530/URTEC-2018-2902799. URTEC-2902799-MS.
Ilk, D., Anderson, D.M., Stotts, G.W.J., Mattar, L., Blasingame, T.A., 2010. Production- Walsh, M.P., Lake, L.W., 2003. A Generalized Approach to Primary Hydrocarbon
data analysis — challenges, pitfalls, diagnostics. SPE Reservoir Eval. Eng. 13 (3), Recovery. Elsevier, Amsterdam; Boston.
538–552. https://doi.org/10.2118/102048-PA. Wattenbarger, R.A., El-Banbi, A.H., Villegas, M.E., et al., 1998. Production analysis of
Jones, J.R., Raghavan, R., 1988. Interpretation of flowing well response in gas- linear flow into fractured Tight gas wells. In: Presented at the SPE Rocky Mountain
condensate wells. SPE Form. Eval. 3 (3), 578–594. https://doi.org/10.2118/14204- Regional/Low-Permeability Reservoirs Symposium, Denver, Colorado, 5–8 April.
PA. SPE-14204-PA. https://doi.org/10.2118/39931-MS. SPE-39931-MS.
Jones, J.R., Vo, D.T., Raghavan, R., 1989. Interpretation of pressure-buildup responses in Yuan, B., Moghanloo, G.R., Zheng, D., 2017. A novel integrated workflow for evaluation,
gas-condensate wells. SPE Form. Eval. 4 (1), 93–104. https://doi.org/10.2118/ optimization and predication in shale plays. Int. J. Coal Geol. 180, 18–28. https://
15535-PA. SPE-15535-PA. doi.org/10.1016/j.coal.2017.04.014.
Lee, K.S., Miller, M.A., Sepehrnoori, K., 1998. Succession-of-States model for calculating Yuan, B., Zhang, Z., Clarkson, C.R., 2019a. Generalized analytical model of transient
long-time performance of depletion reservoirs. SPE J. 3 (3), 279–284. SPE-51023- linear flow in heterogeneous fractured liquid-rich Tight reservoirs with non-static
PA. properties. Appl. Math. Model. 76, 632–654 j.apm.2019.06.036.
Liang, Y., 2015. A New Method For Production Data Analysis Using Superposition-Rate. Yuan, B., Zhang, Z., Clarkson, C.R., 2019b. Improved distance-of-investigation model for
University of Calgary. Masters Thesis. https://prism.ucalgary.ca/bitstream/handle rate-transient analysis in a heterogeneous unconventional reservoir with nonstatic
/11023/2669/ucalgary_2015_liang_yue.pdf;jsessionid¼F0ADF41E34FDE63A82B89 properties. SPE J. 24 (5), 2362–2377. https://doi.org/10.2118/191698-PA. SPE-
1A36A1C9F7F?sequence¼1. 191698-MS.
Moghanloo, R.G., Yuan, B., Ingrahama, N., et al., 2015. Applying macroscopic material Zhang, M., Becker, M.D., Ayala, L.F., 2016. A similarity method approach for early-
balance to evaluate interplay between dynamic drainage volume and well transient multi-phase flow analysis of liquid-rich unconventional gas reservoirs.
performance in Tight formations. J. Nat. Gas Sci. Eng. 27, 466–478. https://doi.org/ J. Nat. Gas Sci. Eng. 28, 572–586. https://doi.org/10.1016/j.jngse.2015.11.044.
10.1016/j.jngse.2015.07.047. Zhu, S., Salmachi, A., Du, Z., 2018. Two phase rate-transient analysis of a hydraulically
Muskat, M., 1937. The Flow of Homogeneous Fluid through Porous Media. MaGraw-Hill, fractured coal seam gas well: a case study from the ordos basin, China. Int. J. Coal
New York City. Geol. 195, 47–60. https://doi.org/10.1016/j.coal.2018.05.014.
Nobakht, M., Mattar, L., 2009. Diagnostics of data quality for analysis of production data.
In: Presented at the Canadian International Petroleum Conference, Calgary, Alberta,
16-18 June. https://doi.org/10.2118/2009-137. PETSOC-2009-137.

20

You might also like