You are on page 1of 15

DOI: 10.

2118/196164-PA Date: 20-August-20 Stage: Page: 1 Total Pages: 15

A Semianalytical Approach for Analysis


of Wells Exhibiting Multiphase Transient
Linear Flow: Application to Field Data
Hamidreza Hamdi, University of Calgary; Hamid Behmanesh, NCS Multistage;
and Christopher R. Clarkson, University of Calgary

Summary
Rate-transient analysis (RTA) is a useful reservoir/hydraulic fracture characterization method that can be applied to multifractured hori-
zontal wells (MFHWs) producing from low-permeability (tight) and shale reservoirs. In this paper, we applied a recently developed
three-phase RTA technique to the analysis of production data from an MFHW completed in a low-permeability volatile oil reservoir in
the Western Canadian Sedimentary Basin.
This RTA technique is used to analyze the transient linear flow regime for wells operated under constant flowing bottomhole pres-
sure (BHP) conditions. With this method, the slope pffiffiof
ffi the square-root-of-time plot applied to any of the producing phases can be used
to directly calculate the linear flow parameter xf k without defining pseudovariables. The method requires a set of input pressure/
volume/temperature (PVT) data and an estimate of two-phase relative permeability curves. For the field case studied herein, the PVT
model is constructed by tuning an equation of state (EOS) from a set of PVT experiments, while the relative permeability curves are
estimated from numerical model history-matching results.
The subject well, an MFHW completed in 15 stages, produces oil, water, and gas at a nearly constant (measured downhole) flowing
BHP. This well is completed in a low-permeability,
pffiffiffi near-critical volatile oil system. For this field case, application of the recently pro-
posed RTA method leads to an estimate of xf k that is in close agreement (within 7%) with the results of a numerical model history
match performed in parallel. The RTA method also provides pressure–saturation (P–S) relationships for all three phases that are within
2% of those derived from the numerical model. The derived P–S relationships are central to the use of other RTA methods that require
calculation of multiphase pseudovariables.
The three-phase RTA p technique
ffiffiffi developed herein is a simple-yet-rigorous and accurate alternative to numerical model history-
matching for estimating xf k when fluid properties and relative permeability data are available.

Introduction
Economical hydrocarbon production from unconventional tight and shale reservoirs through the use of MFHW technology is designed
to increase the contact area between the wellbore and the reservoir. Production data from MFHWs completed in unconventional tight
and shale reservoirs commonly exhibit a continuous decline in hydrocarbon production rates; flowing pressures often decline rapidly at
early times and transition to a relatively constant value at later times. Due to an extremely low-permeability matrix, the diffusion front
moves very slowly in unconventional reservoirs, which frequently leads to long periods of transient linear flow that can last for months
or even years (Wattenbarger et al. 1998). However, with the current trend in finer fracture and well spacing, the transient linear flow
period is becoming increasingly abbreviated.
RTA is a common engineering tool that is used to analyze the production data from MFHWs to estimate the reservoir and hydraulic
fracture properties, such as permeability and fracture half-length. Assuming that transient linear flow is the dominant transient linear flow
period, production data are commonly plotted on a square-root-of-time plot, in which the reciprocalp offfi rate 1/q (assuming constant flowing
pressure production) or rate-normalized pressure (RNP) is plotted against the square root of time ( t). Transient linear flow appears as a
straight-line
pffiffiffi trend on this plot. The slope of a line fit to the linearized data can then be used to directly estimate the linear flow parameter
xf k. For single-phase oil reservoirs, this estimation is reasonable, provided any fluid property variability with pressure is corrected for in
the analysis. For gas cases, in which fluid property variations with pressure are strong, introduction of single-phase (gas) pseudopressure
and pseudotime transformations to convert the gas case to an equivalent liquid case can linearize the governing flow equations, and liquid
solutions to the diffusivity can be used to perform the analysis (Nobakht and Mattar 2012; Behmanesh et al. 2015a).
Analysis of transient linear flow regime in the presence of multiphase flow requires special treatment because p theffiffiffi relative permeabil-
ities of the producing phases are constantly changing with time and pressure. Significant errors in calculating xf k are expected if the
multiphase flow is not handled correctly. A survey of the literature shows that the majority of published work only considers two-phase
flow (oil þ gas) problems. Different approaches have been used to correct for two-phase flow: the dynamic drainage area method of
Qanbari and Clarkson (2016), semianalytical methods used by multiple authors (Zhang et al. 2016; Tabatabaie and Pooladi-Darvish
2017), and the modified pseudovariable approach (Behmanesh et al. 2015b). These methods have generally successfully been applied to
the analysis of production data from wells producing oil and gas under constant flowing pressure constraints during the transient con-
stant pressure linear flow (CPLF) regime.
However, in practice, there are certain situations in which wells have a high water cut due to flowback of injected water from the hydraulic
fracturing treatment, the existence of mobile water in the reservoir, or a combination of both. An example of high water cuts caused by
mobile formation water is in the Delaware Basin (Upper Wolfcamp Formation), where MFHWs are producing with more than 75% of water
cut (Gannaway et al. 2016). In these situations, ignoring the water phase in the calculations can cause a significant underestimation of reser-
voir parameter values using the square-root-of-time plot when only the dominant hydrocarbon phase is analyzed, or when two-phase
(oil þ gas) analysis is performed without considering three-phase flow conditions. Therefore, additional modifications of the classical RTA
methods are required to improve the estimation of reservoir and fracture properties from recorded pressure and flow-rate data.

Copyright V
C 2020 Society of Petroleum Engineers

This paper (SPE 196164) was accepted for presentation at the SPE Annual Technical Conference and Exhibition, Calgary, Alberta, Canada, 30 September–2 October 2019, and revised for
publication. Original manuscript received for review 13 April 2020. Revised manuscript received for review 26 June 2020. Paper peer approved 6 July 2020.

2020 SPE Journal 1

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 2 Total Pages: 15

For boundary-dominated flow, some authors, such as Behmanesh et al. (2018b) and Shahamat and Clarkson (2018), have introduced
corrections for multiphase flow into pseudovariables used in the flowing material balance method. However, for the CPLF regime, the
published work is extremely limited and is frequently based on the total fluid production and total mobility concept (Eker et al. 2014;
Uzun et al. 2016). The total fluid production/mobility method assumes constant formation volume factors for the fluids, and small (and
constant) isothermal compressibility of the total fluid and rock, that often is not applicable to systems such as the highly volatile oil sys-
tems with large drawdowns. Other work based on modified pseudovariables (Clarkson et al. 2019) require knowledge of unsteady-state
pressure and phase/saturation relationships, which are not readily available or easily determinable. pffiffiffi
Hamdi et al. (2018) recently introduced the theoretical background of a semianalytical method to accurately estimate xf k in the
CPLF regime under three-phase flow conditions. This method is based on the Boltzmann transformation of the modified black oil for-
mulation to reduce the highly nonlinear governing differential equations to a set of ordinary differential equations (ODEs). Other semi-
analytical solutions for the three-phase flow under the context of compositional formulation have also been developed by Zhang and
Ayala (2018). The developed semianalytical method in this paper follows the same procedure as Zhang et al. (2016) for the CPLF
under two-phase (oil þ gas) flow conditions and is inspired by the work of Al-Khalifah et al. (1987) who solved the three-phase flow
equations for constant rate radial flow. This semianalytical method is considered a straight-line analysis method, which can be used to
analyze a portion of production data that fall under the transient linear flow regime.
As previously stated, there are not many reliable RTA techniques (if they exist) other than some semianalytical methods that can be
used to analyze the transient three-phase production data from MFHWs. Semianalytical methods are very useful and can provide
simple and concise solutions to some complex multiphase flow problems. However, because many semianalytical methods are only
applied to synthetic examples, the practicality of the solutions is usually overlooked. A main contribution of this work is to showcase
the usefulness of the semianalytical method for some challenging multiphase flow problems.
In the current work, the semianalytical approach by Hamdi et al. (2018) is used to analyze the three-phase transient production data
from an MFHW well completed in a low-permeability reservoir hosting volatile oil in the Western Canadian Sedimentary Basin. The
results are compared with the 1D numerical history-matching results. In addition, history-matching is conducted using a 2D model to
include the stress-sensitive permeability data and evaluate its effect on the estimated linear flow parameter. In this study, tNavigatorV R

(Rock Flow Dynamics 2019) is used to build the fluid model and to perform the numerical simulations and the history-matching pro-
cess. The validity of employing the constant flowing BHP to analyze the production data from the studied well data is demonstrated by
comparing the results with variable pressure simulations. Further, the effect of uncertainty on the RTA results are also discussed. This
paper provides practical insights into the application of the developed semianalytical model to real field data.

Theory
The three-phase
rffiffiffiffi semianalytical method uses the modified black oil formulation and is based on application of the Boltzmann variable
/
g¼x . The Boltzmann transform reduces the highly nonlinear partial differential equations, which are the function of time and
kt
space, to a set of ODEs as a function of the Boltzmann variable. These new ODEs describe the gradients of pressure and phase satura-
tions as a function of reservoir fluid properties, relative permeability, and the corresponding partial derivative of the properties with
respect to pressure and phase saturations. Because oil saturation is obtained by calculating gas and water saturations (i.e.,
So ¼ 1  Sw  Sg ), the final saturation ODEs are obtained by eliminating the oil-phase partial differential equation using the equations
for the other two phases. As shown in Hamdi et al. (2018):
2 3
dp g dp g
ðaa  aa Þ þ ðab  ab Þ ð aa 0
 aa0
Þ þ ðab 0
 ab0
Þ
dSw dSg dg 2 dp 6dg 2 7
¼
g _   6 4 dp g  7 5; . . . . . . . . . . . . . . . . . . . . . . . . . . . ð1Þ
dg dg dp dg
ðaa_  aa_ Þ þ ab  ab_ ðaa_  aa_ Þ þ ab_  ab_
dg 2 dg 2
dp g 0 0 dp g
ðac0  ca0 Þ þ ðan  cb Þ ðaa0  aa0 Þ þ ðab0  ab0 Þ
dg 2 dg 2
dp g _   dp g  
ða_c  ca_ Þ þ na  cb_ ðaa_  aa_ Þ þ ab_  ab_
dSg dp dg 2 dg 2
¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ð2Þ
dg dg dp g   dp g
 
ðac  ca Þ þ ðan  cb Þ ðaa  aa Þ þ ðab  ab Þ
 
dg 2 dg 2
  þ dp
dp g _ g  
ða_c  ca_ Þ þ na  cb_ ðaa_  aa_ Þ þ ab_  ab_
dg 2 dg 2
where the derivative operators h*i, h’i, and hi represent the partial derivatives with respect to gas saturation Sg, pressure p, and water
saturation Sw. All the equations are presented in consistent units. The phase mobilities (a, a, c) and accumulation terms (b, b, n) are
defined as follows:
1 Rs
a¼ krg þ kro ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð3Þ
lg Bg lo Bo
1 Rs
b¼ Sg þ So ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð4Þ
Bg Bo
1 Rv
a¼ kro þ krg ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð5Þ
lo Bo lg Bg
1 Rv
b ¼ So þ Sg ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð6Þ
Bo Bg
1
c¼ krw ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð7Þ
lw B w
1
n¼ Sw : . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð8Þ
Bw

2 2020 SPE Journal

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 3 Total Pages: 15

It should be noted that for three-phase flow conditions, the gas and water relative permeabilities, that is, krg (Sg) and krw (Sw), are
obtained from the two-phase oil/water and gas/oil curves, whereas the oil relative permeability kro (Sw, Sg) is calculated from three-
phase relative permeability models. In this study, Stone’s first model (Stone 1973) is used to estimate the oil relative permeability from
the corresponding values read from the two-phase relative permeability curves, that is, krow (Sw) and krog (Sg).
Some of the key assumptions in developing this semianalytical method are listed below:
• Capillary and gravity forces are not included in the mathematical development.
• Constant flowing BHP is assumed.
• Only 1D transient linear flow in the reservoir is considered.
• Darcy flow is assumed.
• Infinite conductivity for primary hydraulic fractures is assumed.
• The impact of stress-sensitive phenomena, adsorption, and pore confinement are ignored.
• Uniform initial saturation and pressure are assumed.
• One set of relative permeability curves is used.
• The reservoir properties (porosity and permeability) are uniformly distributed within the reservoir.
In a recent study conducted by Zhang and Ayala (2017), a semianalytical method is introduced to include the capillary pressure data
for the two-phase gas-condensate problems. A similar approach can also be adapted here for the three-phase flow problems, but this
requires additional development to include the capillary terms in the equations.

A Solution Scheme for the Semianalytical Model. The purpose of the semianalytical method is to find the solution of pressure and
phase saturations as a function of the Boltzmann variable from the ODEs describing dp/dg, dSw/dg (Eq.1) and dSg/dg (Eq. 2). The solu-
tion to these equations is obtained by introducing an auxiliary variable y ¼ adp/dg. Using this variable serves two purposes: it creates a
theoretical basis to obtain an ODE for dp/dg and allows Darcy’s equation to directly backcalculate the linear flow parameter without
defining any pseudovariables.
Therefore, the modified black oil partial differential equations are reduced to a set of four ODEs (Eq. 9) with different boundary con-
ditions, which can be solved simultaneously to obtain p, Sw, Sg, and y as a function of g:
8 
>
> dp 1 pð1Þ ¼ pi
>
> ða Þ ¼ y;
>
> dg a pð0Þ ¼ pwf
>
>
>
> dSw
>
> ðbÞ ¼ Eq:1; Sw ð1Þ ¼ Swi
<
dg
   
>
> dy g 0 dp _ dSw  dSg dp
>
> ðc Þ ¼  b þ b þ b y ð1 Þ ¼ a
>
> dg 2 dg dg dg dg 1
>
>
>
>
>
> dS g
: ðdÞ ¼ Eq:2 Sg ð1Þ ¼ Sgi :                                        ð9Þ
dg

The solution to Eq. 9 is obtained using numerical methods, such as the Runge-Kutta method (Chapra and Canale 2014). However, it
should be noted that Eq. 9 is a classical boundary value problem, in which we have two boundary conditions for the pressure equation
(Eq. 9a). This requires the shooting method strategy (Chapra and Canale 2014), where p(g ¼ 1) is used as the only boundary condition,
and y(1) is adjusted such that the solution to the system of ODEs satisfies the other boundary condition for pressure, that is, p(g ¼ 0) ¼ pwf.

Application of the Semianalytical Method in RTA. The benefit of using the semianalytical method is twofold:
• The solution of the ODEs provides p, Sw, and Sg as a function of g. Therefore, one can simply plot phase saturations as a function
of pressure to obtain the unsteady-state P–S paths. The P–S relationships serve as critical inputs for RTA models that require mul-
tiphase pseudovariables. An example of this application is given in Behmanesh et al. (2018a).
• The solution of ODEs can also provide y (the auxiliary variable) as a function of g. As a result, one can easily estimate the gradi-
ent term adp/dg (and subsequently dp/dg) at the wellbore location (i.e., at g ¼ 0). This value can be used in the Darcy equation
(for any of the production phases) to provide a theoretical basis for calculating the linear flow parameter as follows:
  rffiffiffiffi    pffiffiffiffiffiffi
dp   / dp xf h k/
qo;sc ¼ kAa ¼ k: xf h :a p; Sw ; Sg : : ¼ pffi yjg¼0 ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð10Þ
dx x¼0 kt dg g¼0 t
  rffiffiffiffi  
dp   / dp
qg;sc ¼ kAa ¼ k: xf h :a p; Sw ; Sg : : ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð11Þ
dx x¼0 kt dg g¼0
  rffiffiffiffi  
dp   / dp
qw;sc ¼ kAc ¼ k: xf h :c p; Sw ; Sg : : ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð12Þ
dx x¼0 kt dg g¼0

where k is matrix permeability, / is porosity, A is the contact area (i.e., xfh for the quarter fracture model), xf is the fracture half-length, and t
is time. All thepequations
ffi are specified with consistent units. These equations demonstrate that, for the pure transient linear flow regime, a
plot of 1/q vs. p t isffi a straight line. Therefore, one can simply plot the reciprocal of the measured rates [or RNP (Dp/q) for variable operating
 pffiffiffiffiffiffivs. t to estimate the slope of the line fit to the linear flow period of the data. For oil flow rates, the slope of this line
conditions] pffiffiisffi
xf h  k/  yjg¼0 . Knowing yjg¼0 from the solution of the semianalytical method, one can directly estimate the linear flow parameter xf k.

Overview of the Studied Well. The studied well is an MFHW with 15 fracture stages and a horizontal length of 4,000 ft. The subject
well was completed in a low-permeability reservoir within the Western Canadian Sedimentary Basin. The reservoir has an initial pres-
sure of  4,580 psia measured at a depth of 6,600 ft true vertical depth, an average porosity of 4.7%, and an estimated net pay thickness
of 164 ft. Initial reservoir pressure was estimated from a diagnostic fracture injection test, and is consistent with that obtained from
flow/buildup tests performed in an offset well. Porosity was derived from log analysis (calibrated to nearby cored wells) and averaged
over the zone of interest.

2020 SPE Journal 3

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 4 Total Pages: 15

The studied well was landed in the volatile oil window with an API oil gravity of 48, an initial gas/oil ratio of 2,880 scf/STB and a
bubblepoint of  4,300 psia at 152.6 F. This well has a pressure gauge installed downhole (approximately 400 ft above the midpoint
perforations), which records the flowing BHPs. Fig. 1 displays the surface flow rates as well as the recorded flowing BHPs, the latter of
which are of excellent quality. The figure illustrates that after the initial flowback and shut-in period (150 days), there is a rapid
decline in the flowing BHP while the well is producing oil, water, and gas, after which the flowing BHP becomes relatively constant
at  550 psi for almost a year before a second shut-in period. Because of the nearly constant flowing pressure period, and the excellent
quality of pressure and rate data, the subject well is an ideal candidate to demonstrate the applicability of the recently proposed
semianalytical method.

Oil Rate Water Rate


1,200 2,500

Water Rate (STB/D)


Oil Rate (STB/D)

1,000
2,000
800
1,500
600
1,000
400
200 500

0 0
0 200 400 600 800 0 200 400 600 800
Days Days

Gas Rate Measured Flowing BHP


2,500 5,000
Gas Rate (Mscf/D)

2,000

Pressure (psia)
4,000

1,500 3,000

1,000 2,000

500 1,000

0 0
0 200 400 600 800 0 200 400 600 800
Days Days

Fig. 1—Production rates and flowing BHPs (measured downhole) for the studied well.

It should be emphasized that the current semianalytical method is developed for analyzing production data from the transient linear
flow regime with constant flowing BHP. Therefore, the semianalytical solutions cannot fully account for the cases with variable pro-
duction constraints. For single-phase flow problems, superposition principles are readily used to extend the constant flowing pressure/
rate solutions to variable production constraints scenarios. In multiphase flow cases, the application of superposition is not straightfor-
ward. This usually requires the definition of pseudovariables to account for the time-varying pressure/saturation relationship, which is
not the subject of this paper. However, in some cases, without long-term dramatic changes in the recorded pressure and rates, the cur-
rent semianalytical method can still provide an analysis of production data with a relatively constant flowing BHP. Winestock and
Colpitts (1965) and Lee (2016) provide an excellent discussion on when a simple rate-normalized plot can still provide acceptable
results when a constant inner boundary solution is applied instead of the superposition. The use of this simplified approach is discussed
in various publications (Liang et al. 2011; Ilk and Houze 2016; Hasan and Mattar 2017; Poston et al. 2019; Houzé et al. 2020; IHS
Markit 2020). Additionally, application of this method for analyzing some real field cases is presented in Samandarli et al. (2014) and
Li et al. (2018).

Methods
Application of the semianalytical method for accurately extracting the linear flow parameter from the subject well production data
requires prior knowledge of reservoir fluid PVT properties and relative permeability curves. For the studied well, the required black oil
PVT properties were generated using an EOS calibrated to several laboratory experiments. Because laboratory-measured relative per-
meability data were not available for this study, this information was obtained by conducting a numerical history match of the well pro-
duction data. The derived relative permeability data were then used as input to the semianalytical model, which in turn was used to
extract the linear flow parameter from the subject well data. The same derived relative permeability data can also serve as a “best
guess” for the analysis of production data from offset wells completed in the same reservoir (and at similar depths).

Reservoir Fluid Modeling. The reservoir fluid corresponds to a highly volatile oil system. Surface compositions of oil and gas at sepa-
rator conditions were obtained from representative surface samples, which were recombined at a gas/oil ratio of  350 sep.bbl/MMscf.
This resulted in a reservoir fluid with a saturation pressure of 4,853 psig at the reservoir temperature of 152.8 F. A Peng-Robinson EOS
(Robinson and Peng 1978) is used to calibrate the fluid model to several laboratory tests, such as constant composition expansion, con-
stant volume depletion, and separator tests. In this study, PVT Designer (Rock Flow Dynamics 2019) was used for characterizing the
reservoir fluid. The final fluid model has 15 pseudocomponents, which were obtained by lumping the C7þ components. Table 1 lists the
pseudocomponent properties for the final calibrated model. Because production data analysis is performed using modified black oil
properties, which were generated using the Whitson-Torp method (Whitson and Torp 1983), further lumping of the components was
not necessary. Fig. 2 provides the phase envelope and an illustration of the performance of the PVT model when predicting some of the
important measured PVT properties.

4 2020 SPE Journal

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 5 Total Pages: 15

Critical Critical Acentric Molecular Critical


Components Temperature (R) Pressure (psia) Factor Weight Volume (ft3/lbm)
N2 227.16 492.31 0.04 28.013 1.433652
CO2 547.56 1069.86 0.225 44.01 1.505736
CH4 343.08 667.19 0.008 16.043 1.585828
C2H6 549.72 708.34 0.098 30.07 2.370733
C3H8 665.64 615.76 0.152 44.097 3.251748
IC4 734.58 529.05 0.176 58.124 4.212856
nC4 765.36 551.09 0.193 58.124 4.084708
IC5 828.72 490.84 0.227 72.151 4.90165
nC5 845.28 489.37 0.251 72.151 4.869613
C6 925.38 493.88 0.266 85.599 5.404775
C7–C10 1024.01 442.54 0.316 111.446 6.953426
C11–C14 1039.78 375.091 0.451 165.287 10.2221
C15–C19 1203.6 251.86 0.628 233.073 13.92623
C20–C27 1487.84 156.36 0.816 312.007 18.24597
C28–C30þ 1845.89 196.45 1.002 394.58 27.39672

Table 1—Properties of the calibrated Peng-Robinson EOS after matching laboratory PVT tests.

Critical point pressure: 5038.63 Psi 2.0


5,000 Critical point temperature: 292.9 F Oil viscosity—simulated
Cricondenbar: 5213.8 Psi
Cricondentherm: 891.4 F Gas viscosity—simulated
4,000 1.5 Oil viscosity—measured
Viscosity (cp)
Pressure (psi)

3,000
1.0

2,000
0.5

1,000

0
0 2,000 4,000 6,000 8,000
0 200 400 600 800 Pressure (psia)
Temperature (°F)
(a) (b)
50 30
Oil density—measured
45
Oil density—simulated Simulated
25
40 Gas density—simulated
Measured
Relative Volume

35
Density (lb/ft3)

20
30
25 15
20
10
15
10 5
5
0 0
0 2,000 4,000 6,000 8,000 10,000 0 2,000 4,000 6,000 8,000
Pressure (psia) Pressure (psia)
(c) (d)

Fig. 2—(a) Phase envelope and (b, c, and d) results of matching some of the laboratory PVT tests with the fluid model.

Numerical Simulation and History Matching. Assuming a planar, symmetric fracture geometry and a 50% fracture efficiency per
stage, a model with 30 identical bi-wing fractures (spaced  130 ft apart) was generated. To reduce the computational cost of the numer-
ical simulations used for history matching, an element of symmetry model was assumed (flow toward one face of a quarter fracture
plane). Consequently, for our analysis, the production rates are divided by 4  nf, where nf ¼ 30 is the number of fractures. To this end,
a 1D numerical model with an infinite conductivity fracture was constructed to extract the relative permeability data using numerical
history matching. The history-matched model also provides an independent estimate of the linear flow parameter for comparison with
the semianalytical model results. To apply the semianalytical model assumptions, the supportive numerical model uses similar relative

2020 SPE Journal 5

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 6 Total Pages: 15

permeability and water saturations in the hydraulic fracture and the matrix. Additionally, the fracture height is assumed to be the same
as the net pay thickness. The parameters adjusted for history matching, and their ranges, are provided in Table 2. The 1D numerical
model has been designed to closely represent the situation in which the semianalytical method is applicable. A schematic of the 1D
model (a quarter fracture model) is shown in Fig. 3. The assumptions of the semianalytical method were stated previously in the
Theory section.

Parameter Minimum Maximum


nw 2 7
now 2 7
nog 3 7
ng 3 7
krg_max 0.1 0.6
krw_max 0.1 0.8
Srow 0.1 0.4
Sw 0.05 0.4
xf (ft) 100 500
k (nd) 10 10,000

Table 2—Ranges of parameters used in the history-matching


process.

A quarter fracture model A


~164 ft

A B

t
5f
~6

Fig. 3—Schematic of the 1D numerical model (a quarter fracture model) used in the history-matching process.

At each iteration of the numerical history-matching process, the two-phase relative permeability curves are generated using the
power-law Corey equations (Christiansen 2007) as follows (Eqs. 13 through 16):
 nw
Sw  Swc
krw ðSw Þ ¼ krw max ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð13Þ
1  Swc  Srow
 now
1  Sw  Srow
krow ðSw Þ ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð14Þ
1  Swc  Srow
 ng
 Sg  Sgc
krg Sg ¼ krg max ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .ð15Þ
1  Sgc  Swc  Srog
 
 1  Sg  Srog  Swc nog
krog Sg ¼ ; . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ð16Þ
1  Swc  Srog

where Swc and Sgc are critical water and gas saturations, respectively, whereas Srow and Srog are the residual oil saturations in the oil/water
and oil/gas systems respectively. History matching was performed using the differential evolution algorithm (Storn and Price 1997).

Results
In this section, the numerical history-matching results are provided and compared with the results of the recently proposed
semianalytical method.

History-Matching Results. To enhance the investigative capability of the history-matching process, the differential evolution algo-
rithm is set up with a relatively large initial population of 30 samples, and the process was repeated for 40 iterations. Table 3 provides
the variables corresponding to history match p with
ffiffiffi the lowest misfit (“best-case”). The history-matching process led to a best-case model
/ 12
with an initial water saturation of 34% and xf k ¼ 6.33 ft-md . The derived relative permeability curves are shown in Fig. 4.
The best-case history match is graphically illustrated in Fig. 5. It can be seen that the model reasonably matches the oil and gas
rates, but the water rates are underpredicted at early time. This is expected, however, because a simplified 1D model is not able to fully
account for all reservoir complexities and the various flow regimes encountered. Nonetheless, the main purpose of this pffiffiffinumerical
model history-matching exercise is to extract the two-phase relative permeability data and to compare the results (i.e., xf k) with the
semianalytical method.

6 2020 SPE Journal

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 7 Total Pages: 15

Parameter Minimum Maximum Matched


nw 2 7 2.6
now 2 7 3.7
nog 3 7 4
ng 3 7 4.4
krg_max 0.1 0.6 0.26
krw_max 0.1 0.8 0.58
Srow 0.1 0.4 0.31
Sw 0.05 0.4 0.16
xf (ft) 100 500 243
k (nd) 10 10,000 680

Table 3—Parameters of the history-matched (calibrated) numeri-


cal model.

1.0 1.0
krw kro krg kro
0.8 0.8

0.6 0.6

kr
kr

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Sw Sg
(a) (b)

Fig. 4—(a) Oil/water and (b) gas/oil relative permeability curves derived from history-matching. These curves are used in the
recently proposed semianalytical model for extracting the linear flow parameter.
Well ‘PROD’ Well ‘PROD’
Gas rate (constant _BHP_1Dmodel.data) Oil rate (constant _BHP_1Dmodel.data)
Gas rate (H) Oil rate (H)
8
20
Liquid Rate (STB/D)
Gas Rate (Mscf/D)

6
15

10 4

5 2

0 100.00 200.00 300.00 400.00 500.00 0 100.00 200.00 300.00 400.00 500.00
Days Days
(a) (b)
6 Well ‘PROD’ Well ‘PROD’
Water rate (constant _BHP_1Dmodel.data) Bottomhole pressure (constant _BHP_1Dmodel.data)
Water rate (H) Bottomhole pressure (H)

5 3,000
Liquid Rate (STB/D)

Pressure (psia)

2,000
3

2
1,000

0 100.00 200.00 300.00 400.00 500.00 0 100.00 200.00 300.00 400.00 500.00
Days Days
(c) (d)

Fig. 5—1D model history-matching results for (a) gas, (b) oil, and (c) water production rates during the online production period.
(d) Flowing BHP is approximated by a constant value of 550 psi after around 200 days. In these plots, “H” denotes the history or
measured data points. The history-matched results are shown by continuous black curves.

2020 SPE Journal 7

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 8 Total Pages: 15

Semianalytical Solution Results. The semianalytical model was initialized using extracted relative permeability data and fluid properties
from the previous steps. Convergence of the semianalytical! model was achieved by selecting y ¼ adp/dg ¼ 6.12102
Scm3 atmðdarcy  sÞ0:5 STB psi ðmd-DÞ0:5
3
 2:95 at g ¼ 100 cm/(darcy  s)0.5 [ 45 ft/(md-D)0.5]. Fig. 6 compares the unsteady-
cm  cp cm bbl-cp ft
state P–S paths, obtained from the semianalytical method, with the results of the numerical simulation (i.e., the history-matched model).
This figure shows that the semianalytical model is able to accurately estimate the transient P–S paths for the CPLF regime within 2% error.

0.9 0.5
Simulation Simulation
0.8 0.4
Semianalytical Semianalytical
0.7 0.15
0.3

Sw
So

Sg
0.6
0.2
0.5
Simulation
0.4 0.1
Semianalytical
0.3 0 0.1
0 1,000 2,000 3,000 4,000 5,000 0 1,000 2,000 3,000 4,000 5,000 0 1,000 2,000 3,000 4,000 5,000
Pressure (psi) Pressure (psi) Pressure (psi)

Fig. 6—Comparison of the unsteady-state P–S relationships obtained from numerical simulation and the semianalytical method.

Fig. 7 provides the square-root-of-time plot for oil (Fig. 7a) and gas (Fig. 7b) phases. RNP (Dp/q) is used to partially account for the
flowing pressure changes in the first 200 days of the online production period. The constant pwf solution can be applied at middle times
ffi pwf becomes approximately constant. Fig. 7 demonstrates that, for either of the hydrocarbon production phases, a plot of Dp/q vs.
when
p
t follows a straight-line trend after initial fluctuations caused by large changes in the flowing BHP from 4,580 to 550 psia. Following
the semianalytical procedure, that is, by fitting a straight line to the linear flow portionpofffiffiffi the data and using the derivative estimation
/ 12
from the semianalytical solution (i.e., adp/dgjg¼0  0.53), the linear flow parameter (xf k) is calculated to be 5.9 and 6.7 ft-md from
the oil and gas phases, respectively. These amount to a 7 and 3% error, respectively, relative to the numerical simulation results of the
best-case model. It should be noted that for this field case, water rates could not be easily used for the linear parameter estimation
because of data noise and uncertainty in the measurement. As shown in Fig. 8, the square-root-of-time plot for the water phase is rela-
tively scattered and the straight line cannot be easily fitted to the data.

18,000 2,500
16,000 Real data Real data
Semianalytical 2,000
Δp/q (psi/STB/D)

14,000 Semianalytical
Δp/q (psi/Mscf/D)

12,000
1,500
10,000
8,000
1,000
6,000
4,000 500
2,000
0 0
0 5 10 15 20 25 0 5 10 15 20 25
√t (Day0.5) √t (Day0.5)
(a) (b)

Fig. 7—Square-root-of-time plots for the (a) oil and (b) gas phases. The semianalytical method (black curve) matches the linear
trend of the measured data points on the square-root-of-time plot. RNPs are used to account for variable flowing conditions
during the first 200 days of online production.

50,000
45,000 Real data
Semianalytical
40,000
Δp/q (psi/STB/D)

35,000
30,000
25,000
20,000
15,000
10,000
5,000
0
0 5 10 15 20 25
√t (Day0.5)

Fig. 8—Square-root-of-time plot for the water phase. The semianalytical method (black curve) matches the linear trend of the mea-
sured data points on the square-root-of-time plot.

8 2020 SPE Journal

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 9 Total Pages: 15

It should be noted that in practice, when analyzing multiphase production data, fitting a correct straight line to the data may be chal-
lenging for some important phases (e.g., oil and gas). This is due to the existence of noise in the data or other production reporting
issues. The presence of such errors could lead to significant discrepancies between the linear flow parameter estimation using individual
phases. However, by simultaneous analysis of more production phases (if possible), one can reduce the uncertainty in the line fitting
procedure to have a better estimation ofpthe ffiffiffi linear flow parameter with a smaller error. In other words, this issue can be seen as a
history-matching problem to obtain an xf k that can provide the best fit to the analyzed production phases. The suggested workflow to
conduct the latter task is based on a simple visual adjustment p procedure:
ffi pffiffiffi
1. For the first phase, with better quality data, plot 1/q vs. t, draw a straight line, find the slope, and calculate xf k using the proce-
dure stated earlier (Eqs. 10 through 12). pffiffiffi
2. For the other phase(s), use the estimated xf k from Step 1, but this time calculate a slope (Eqs. 10 through 12) and draw the
straight line to the current data using the calculated slope. If the line
pffiffiffi does not reasonably match the data for the current phase,
pffiffiffi the slope of the line, but this time calculate a new (xf k)new using the adjusted slope.
slightly adjust
3. Using (xf k)new, apply Step 2 to the first phase analyzed,pand ffiffiffi see how the slope changes. If the slopes for all the phases are
reasonably matched, stop the process and record the final xf k.

Discussion
Comparing the Constant Flowing BHP with the Variable Pressure Solutions. As previously demonstrated, the semianalytical
method is developed for the constant flowing BHP constraint. In practice, this approach is only applied to the section of the data corre-
sponding to the transient linear flow regime where the pressure is relatively constant for a long period of time. For these situations, the
semianalytical method can be used as a good approximation to estimate the linear flow parameter. Nonetheless, in the studied field
case, the pressure has shown some variations in early times before it can be reasonably approximated with a constant flowing BHP of
550 psi. To investigate the validity of this constant flowing pressure approximation, the 1D history-matched model is used to compare
the constant BHP solution (i.e., pwf ¼ 550 psi) with the case that the variable pressure profile is used as the constraint in the simulation.
Fig. 9 compares the production rates for both simulation cases and the observed production data. As shown in this figure, after a
short period of time, the production rates obtained from the constant BHP simulation are closely matching the simulated data from the
case with the variable pressure constraint. The data from this period, in which both simulations are nearly identical, correspond to the
dominant transient linear flow regime, which is analyzed by the semianalytical method. This comparison confirms that for the studied
case, the constant BHP solution is a safe approach to estimate the linear flow parameter.

Well ‘PROD’
Oil rate (variableBHP_1Dmodel.data)
Well ‘PROD’
Gas rate (variableBHP_1Dmodel.data)
6 Well ‘PROD’
Water rate (variableBHP_1Dmodel.data)
8 Oil rate (constant _BHP_1Dmodel.data) Gas rate (constant _BHP_1Dmodel.data) Water rate (constant _BHP_1Dmodel.data)

Liquid Rate (STB/D)


Oil rate (H) Gas rate (H) Water rate (H)
20
Liquid Rate (STB/D)

Gas Rate (Mscf/D)

6 4
15

4 10
2
2 5

0 100.00 200.00 300.00 400.00 500.00 0 100.00 200.00 300.00 400.00 500.00 0 100.00 200.00 300.00 400.00 500.00
Days Days Days

Fig. 9—Comparison of the production rates obtained from the constant flowing BHP and the variable pressure solutions for the 1D
history-matched model. The model parameters used for both cases are identical.

Additional Complexities. The semianalytical method is an elegant RTA approach that was developed to analyze the CPLF regime
under three-phase flow conditions. This method was derived under the assumption of a 1D reservoir geometry with an infinite conduc-
tive fracture and uniform reservoir and fracture properties (e.g., porosity, permeability, reservoir thickness). Similar initial (mobile)
water saturations and relative permeability curves were also assumed for both fracture and the reservoir. However, in practice, addi-
tional complexities may exist (e.g., strong reservoir heterogeneities, pore confinement, non-Darcy flow, complex fracture geometries,
and stress-sensitive properties) that violate the assumptions of the semianalytical method. In the following sections, we account for
some complexities by performing numerical model history matching with a 2D numerical model. A schematic of a base 2D model
(a quarter fracture model) used in the simulations is displayed in Fig. 10.
History Matching. In many unconventional reservoirs, large changes in effective stress (or pressure) due to production can strongly
affect the rock’s pore structure. These strong changes in effective stress can also cause embedment or crushing of proppants within the
hydraulic fracture. As a result, the fracture conductivity decreases during production, with a consequent loss in well productivity. Therefore,
it is important to include the effect of such pressure-sensitive parameters within the production data analysis framework. Further develop-
ment is required to include these parameters in the semianalytical method. However, the more robust numerical model can be used to explore
the impact of inclusion of these pressure-dependent parameters on both the history match and the derived reservoir/fracture properties.
Ghanizadeh et al. (2016) recently provided detailed laboratory procedures for measuring variations in the permeability of propped/
unpropped fracture and reservoir matrix in unconventional reservoirs. These measurements result in a set of permeability values (con-
verted to transmissibility multipliers in the simulator) that change exponentially with effective stress or pressure. Fig. 11 shows the var-
iation of the rock and propped fracture transmissibility multipliers with pressure derived from laboratory measurements, which are
included in the updated 2D model history matching. These smooth curves are obtained by fitting an exponential trend to the measured
data. Furthermore, unlike the 1D numerical model and semianalytical models, different relative permeability curves and initial water
saturations are used in the propped fracture and the matrix.
Fig. 12 compares the history-match results with the recorded pressure and rates for the complete production history (i.e., flowback
and online production period) of the studied well. The quality of the match is reasonable, except for the early time water production
rates during the flowback period. However, these discrepancies are expected to occur because many additional complexities (e.g., frac-
ture geometries, reservoir heterogeneities, non-Darcy effects) are not included in the 2D model due to lack of available data. It is noted
that after a short time since the start of the first buildup, the pressure gauge seems to have experienced some issues in which the upward

2020 SPE Journal 9

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 10 Total Pages: 15

trend of pressure data has suddenly reversed, causing an overall downward shift of the recorded pressure data. This downward trend of
recorded pressure data during the buildup period is not a reservoir signature (Mattar and Zaoral 1992). Therefore, the data in this period
are considered unreliable and have been excluded from the matching process.

Permeability ×
(md)
24.20793
PROD

3.21759

0.42767

0.05684

0.00756

Fig. 10—Schematic of the 2D numerical model (a quarter fracture model).

Propped Fracture Matrix


1.2 1.2
Transmissibility Multiplier

Transmissibility Multiplier
1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 1,000 2,000 3,000 4,000 5,000 0 1,000 2,000 3,000 4,000 5,000
Reservoir Pressure (psi) Reservoir Pressure (psi)

Fig. 11—Variation of transmissibility multipliers as a function of pressure for the propped fracture and the matrix. The transmissibility
multipliers are deduced from variations of permeability as a function of effective stress derived from the laboratory experiments.

10 25
9 Measured oil rate Measured gas rate
Gas Rate/120 Mscf/D

Simulation Simulation
Oil Rate/120 STB/D

8 20
7
6 15
5
4 10
3
2 5
1
0 0
0 200 400 600 800 0 200 400 600 800
Days Days
(a) (b)
20 5,000
18 Measured BHP
Water Rate/120 STB/D

Measured water rate


Flowing BHP (psia)

16 Simulation 4,000 Simulation


14
12 3,000
10
8 2,000
6
4 1,000
2
0 0
0 200 400 600 800 0 200 400 600 800
Days Days
(c) (d)

Fig. 12—History-matching results for the 2D model. The measured (historical) production rates and BHP data are shown by col-
ored circles. The simulation results are illustrated with continuous black curves.

10 2020 SPE Journal

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 11 Total Pages: 15

pThis
ffiffiffi history-matched model has an initial matrix permeability of 0.00756 md and an average fracture half-length of 160 ft (i.e.,
/
12
xf k ¼ 13.9 ft-md ). As is shown in Fig. 11, the matrix permeability is reduced by 75% when the pressure is reduced to 550 psi. There-
/ 12
fore, the initial linear flow parameter declines to 7.1 ft-md due to a dramatic change in the near fracture permeability. This latter
model-derived value is close to the semianalytical model value, which was fit to the data. The matched fracture conductivity is
25 md-ft. Additionally, estimated initial water saturations in the fracture and the matrix are 0.68 and 0.20, respectively. The history-
matched relative permeability curves for the fracture and matrix are given in Fig. 13.

1.0 1.0
krw kro krg kro
0.8 0.8

Fracture–kr
Fracture–kr

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Sw Sg
(a) (b)
1.0 1.0
krw kro krg kro
0.8 0.8
Matrix–kr

Matrix–kr
0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Sw Sg
(c) (d)

Fig. 13—Two-phase relative permeability curves for the (a and b) propped fracture and the (c and d) matrix derived from the 2D
model history-match.

Relative Permeability. As observed in Fig. 5, for this field case, the 1D history-matched model (or equivalently the semianalytical
method) underpredicts the water production at early times. As a result, the history matching leads to low water relative permeabilities
(0.02) for the range of water saturation variation during the simulation period (Fig. 4). Nonetheless, the existence of a large pressure
drop at the wellbore (pi – pwf  4,300 psi), combined with a large contact area, results in significant water production at later times, and
a decent 1D model can match the water production during the transient linear flow regime.
A closer look at the relative permeability curves derived from both the 1D and 2D model shows that the relative permeability to gas
is restricted to low values for the range of evolved gas saturation estimated in the reservoir during production. This is because, in the
simulation model, the cumulative amount of the free gas in the reservoir, and the dissolved gas in the oil, is larger than the amount of
the measured produced gas. As a result, the history-matching process leads to a gas relative permeability curve with a very small magni-
tude. A possible explanation for this observation could be the effect of pore confinement on the phase behavior, which can cause the
phase envelope to shrink (Pitakbunkate et al. 2016). The shrinkage of the phase envelope would lead to a lower bubblepoint pressure,
and a smaller amount of dissolved gas in the oil. Because we have not included the pore confinement effect in our EOS modeling pro-
cess, the reduced level of gas/oil ratio observed in the measurements is accounted for by using modified gas relative permeability
curves with smaller values. Interestingly, this restricted flow of gas in shales has also been observed by Honarpour et al. (2012), who
derived relative permeability curves using pore scale simulations.

Uncertainty Assessment. The developed semianalytical method provides accurate results when compared with the output of the corre-
sponding 1D numerical simulations. However, the reliability of the estimated linear flow parameter is a direct function of the input vari-
ables, such as relative permeability. pOn ffiffiffi the other hand, the history-matching results are not unique, meaning that by employing
different parameters (e.g., different xf k), equally good history-matching results could be achieved. Fig. 14 provides a range of some
model parameters (combined with a range of relative permeability curves) that can all reasonably match the production history using a
1D model. This plot is obtained by applying a Bayesian history-matching technique using Markov chain Monte Carlo (Gamerman
1997), as described in Hamdi et al. (2017, 2019).
/ 12
Fig. 14 illustrates that, with a linear flow parameter between 5.9 and 10.2 ft-md , a reasonable history match can be achieved. How-
ever, the estimated initial water saturation range (Fig. 14) is larger than the saturation obtained from the deterministic history matching
(Table 3) because no constraint was imposed on the initial water saturation during the matching process. If a different initial water
saturation is assumed, some history-matching parameters such as Swc and/or the Corey exponents (e.g., nw) would be readjusted in the
course of the history-matching process to match the water production. This in turn leads to a different linear flow parameter but within
its expected range of variability
pffiffiffi as shown in/ Fig. 14. For example, Fig. 15 shows an ensemble of relative permeability curves for a 1D
12
model with Swi ¼ 0.35 and xf k ¼ 8.2 ft-md that can match the production data.
It is important to note, however, that this does not mean that the selection of Swi (or any other similar parameters) is completely arbi-
trary in the modeling process. Although some parameters might not show much sensitivity for a specific process (e.g., primary produc-
tion) or a specific flow regime (e.g., transient linear flow regime), their selection could still critically affect decision-making for other
processes (e.g., secondary or tertiary recovery mechanisms) or other flow regimes (e.g., boundary-dominated flow). Therefore, a reliable
model and interpretation should be obtained by integrating all available data to reduce the uncertainty in predicting the future recovery.

2020 SPE Journal 11

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 12 Total Pages: 15

0.20
0.10

0.08 0.15

Frequency

Frequency
0.06
0.10
0.04
0.05
0.02

0 0
0.38 0.40 0.42 0.44 0 2 4 6 8
Sw k (md) ×10–4

0.15 0.15

0.10 0.10
Frequency

Frequency
0.05 0.05

0 0
250 300 350 400 0 5 10
xf (ft) xf√k (ft-md¹⁄₂)

Fig. 14—Probability distributions of the 1D model parameters resulting in a reasonable history match to the production data of the
studied well during the transient linear flow period.

1.0 1.0

0.9 0.9

0.8 0.8

0.7 0.7
krw and krow

krg and krog

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

0.2 0.2
Swc
0.1 0.1

0 0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
Sw Sw
pffiffiffiffi / 12
Fig. 15—Range of two-phase relative permeability curves for a 1D model with Swi 5 0.35 and xf k 5 8.2 ft.md that result in good
history-matching results.

Conclusions
In this work, a newly developed three-phase RTA model is applied to a real field case to estimate the linear flow parameter from pro-
duction data. The production data were obtained from an MFHW completed in a low-permeability reservoir containing a highly volatile
oil in the Western Canadian Sedimentary Basin.
This RTA approach can be considered a straight-line analysis method to fit a line to a portion of data that fall within the common
transient linear flow regime period. The main contribution of this work is to show if the developed semianalytical method can be
applied to some appropriate field data with multiphase flow complexities. A challenging field example from a highly volatile oil reser-
voir is presented, and the results are compared with the numerical simulations.
In this study, modified black oil PVT data for use in the semianalytical model were obtained by matching the Peng-Robinson EOS
to a set of laboratory tests and generating black oil PVT tables using the Whitson-Torps method. However, the relative permeability
data were derived by history matching the studied well data using a representative (1D) model. The calibrated numerical model was
thenpused
ffi to evaluate the accuracy of the semianalytical model results. As demonstrated using the semianalytical method, a plot of RNP
vs. t for any of the measured production phases resulted in a straight line in which slopes were used to independently estimate the
linear flow parameter. Water production rates could not be confidently used for this purpose because of significant scatterness pffiffiin
ffi the
data when displayed on the square-root-of-time plot. Use of either of the hydrocarbon production phases, however, resulted in xf k esti-
mates within 7% (for oil) and 3% (for gas) of the same value derived from the 1D model history match. In addition, the unsteady-state
P–S paths calculated from the semianalytical method were in excellent agreement with the numerical simulation results.
An additional history-matching process using 2D models was also performed to include more complexities when analyzing the pro-
duction data from MFHWs. Some of these additional complexities include the measured stress-dependent properties for the matrix and
hydraulic fractures, nonuniform initial saturation distribution, and different sets of relative permeability curves, including the possibility
of observing other flow regimes through 2D simulations. Interestingly, the results show that considering the stress-dependent properties,

12 2020 SPE Journal

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 13 Total Pages: 15

the linear flow parameter obtained from the more complex 2D numerical history matching is still close to the one obtained from effec-
tive semianalytical and 1D numerical models.
The developed semianalytical method provides accurate results compared with the output of numerical simulations. However, it is
noted that the reliability of the estimated linear flow parameter is a direct function of the input variables and the sensitivity of the input
parameters in the simulations. Therefore, an uncertainty analysis approach is also adapted here to obtain the posterior distributions of
the parameters
pffiffiffi through which a combination can still provide an acceptable history-matching quality. For example, the results show
/12
that with xf k ranging from 5 to 10 ft-md , and a set of relevant relative permeability curves, a reasonable match can still be obtained.
This is not surprising because the existence of noise in the real production data can make the history-matching and/or the line-fitting
process very uncertain.
It should be noted that the developed semianalytical approach is a straight-line analysis method, which is applied to a portion of the
production data in which the transient linear flow regime coincides with a long and relatively constant BHP trend. The straight-line
analysis approaches do not tend to model all the flow regimes with arbitrary variations in p/q. The studied well has shown favorable
characteristics that could be properly analyzed by the developed semianalytical solution.
Although there are many assumptions in developing this semianalytical model [which is an inheriting property of all the (semi-)ana-
lytical methods], the results are encouraging because there are still many real data that can be approximately analyzed using the semian-
alytical methods. Moreover, the field study showcased here can pave the road for additional developments by addressing some of the
limitations in future developments.

Nomenclature
a, a, c ¼ component mobilities for gas, oil, and water, respectively, 1/(cp  cm3/Scm3)
b, b, n ¼ component accumulation terms for gas, oil, and water, respectively, 1/(cm3/Scm3)
Bg, Bo, Bw ¼ formation volume factors for gas, oil, and water, respectively, cm3/Scm3
h ¼ formation thickness, cm
k ¼ matrix (formation) permeability, darcy
krg, kro, krw ¼ gas, oil, and water relative permeability, respectively, dimensionless
krg_max, krg_max ¼ maximum gas and water relative permeability values, dimensionless
nf ¼ number of symmetrical hydraulic fractures, dimensionless
nog, ng ¼ Corey’s exponents for oil and gas relative permeability curves, dimensionless
now, nw ¼ Corey’s exponents for oil and water relative permeability curves, dimensionless
p ¼ pressure, atm
pi ¼ initial pressure, atm
pwf ¼ flowing BHP, atm
qg, qo, qw ¼ surface flow rate for gas, oil, and water, respectively, cm3/s
Rs ¼ solution gas/oil ratio, Scm3/Scm3
Rv ¼ vaporized oil/gas ratio, Scm3/Scm3
Sg, So, Sw ¼ gas, oil, and water saturations, respectively, dimensionless
Sgc, Swc ¼ critical gas and water saturations, respectively, dimensionless
Srog, Srow ¼ residual oil saturation in the presence of gas or water, dimensionless
t ¼ time, seconds
xf ¼ fracture half-length, cm
g ¼ Boltzmann variable, cm  (darcy  s)–0.5
lg, lo, lw ¼ viscosities of gas, oil, and water, respectively, cp
/ ¼ porosity, dimensionless

Acknowledgments
Hamidreza Hamdi would like to thank Rock Flow Dynamics Inc. for supporting his independent research. Chris Clarkson would like
to acknowledge Ovintiv and Shell for support of his Chair position in Unconventional Gas and Light Oil research at the University of
Calgary, Department of Geoscience. We further gratefully thank the sponsors of the Tight Oil Consortium, hosted at the University of
Calgary, for their support. Partial support for this work was provided through an NSERC CRD grant (CRDPJ 501135–16) held
by Clarkson.

References
Al-Khalifah, A.-J. A., Horne, R. N., and Aziz, K. 1987. In-Place Determination of Reservoir Relative Permeability Using Well Test Analysis. Paper pre-
sented at the SPE Annual Technical Conference and Exhibition, Dallas, Texas, USA, 27–30 September. SPE-16774-MS. https://doi.org/10.2118/
16774-MS.
Behmanesh, H., Clarkson, C. R., Tabatabaie, S. H. et al. 2015a. Impact of Distance-of-Investigation Calculations on Rate-Transient Analysis of Uncon-
ventional Gas and Light-Oil Reservoirs: New Formulations for Linear Flow. J Can Pet Technol 54 (6): 509–519. SPE-178928-PA. https://doi.org/
10.2118/178928-PA.
Behmanesh, H., Hamdi, H., and Clarkson, C. R. 2015b. Production Data Analysis of Tight Gas Condensate Reservoirs. J Nat Gas Sci Eng 22: 22–34.
https://doi.org/10.1016/j.jngse.2014.11.005.
Behmanesh, H., Hamdi, H., Clarkson, C. R. et al. 2018a. Transient Linear Flow Analysis of Multi-Fractured Horizontal Wells Considering Three-Phase
Flow and Pressure-Dependent Rock Properties. Paper presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference, Hous-
ton, Texas, USA, 23–25 July. URTEC-2884255-MS. https://doi.org/10.15530/urtec-2018-2884255.
Behmanesh, H., Mattar, L., Thompson, J. M. et al. 2018b. Treatment of Rate-Transient Analysis during Boundary-Dominated Flow. SPE J. 23 (4):
1145–1165. SPE-189967-PA. https://doi.org/10.2118/189967-PA.
Chapra, S. and Canale, R. P. 2014. Numerical Methods for Engineers, seventh edition, 922. New York, New York, USA: McGraw-Hill Education.
Christiansen, R. L. 2007. Relative Permeability and Capillary Pressure. In Petroleum Engineering Handbook: Vol 1: General Engineering, ed. John R.
Fanchi, Chap. 15, 727–765. Richardson, Texas, USA: Society of Petroleum Engineers.

2020 SPE Journal 13

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 14 Total Pages: 15

Clarkson, C. R., Yuan, B., Zhang, Z. et al. 2019. Anomalous Diffusion or Classical Diffusion in an Anomalous Reservoir? Evaluation of the Impact of
Multi-Phase Flow on Reservoir Signatures in Unconventional Reservoirs. Paper presented at the SPE/AAPG/SEG Unconventional Resources Tech-
nology Conference, Denver, Colorado, USA, 22–24 July. URTEC-2019-85-MS. https://doi.org/10.15530/urtec-2019-85.
Eker, I., Kurtoglu, B., and Kazemi, H. 2014. Multiphase Rate Transient Analysis in Unconventional Reservoirs: Theory and Applications. Paper pre-
sented at the SPE/CSUR Unconventional Resources Conference, Calgary, Alberta, Canada, 30 September–2 October. SPE-171657-MS. https://
doi.org/10.2118/171657-MS.
Gamerman, D. 1997. Markov Chain Monte Carlo: Stochastic Simulation for Bayesian Inference, 264. Boca Raton, Florida, USA: Chapman & Hall.
Gannaway, G. R., Ibrahim, M., and Pieprzica, C. 2016. Multiphase Analytics in a High Water-Cut Environment. Paper presented at the SPE Annual
Technical Conference and Exhibition, Dubai, UAE, 26–28 September. SPE-181470-MS. https://doi.org/10.2118/181470-MS.
Ghanizadeh, A., Clarkson, C. R., Deglint, H. et al. 2016. Unpropped/Propped Fracture Permeability and Proppant Embedment Evaluation: A Rigorous
Core-Analysis/Imaging Methodology. Paper presented at the SPE/AAPG/SEG Unconventional Resources Technology Conference, San Antonio,
Texas, USA, 1–3 August. URTEC-2459818-MS. https://doi.org/10.15530/URTEC-2016-2459818.
Hamdi, H., Behmanesh, H., and Clarkson, C. R. 2018. A Semi-Analytical Approach for Analysis of the Transient Linear Flow Regime in Tight Reser-
voirs under Three-Phase Flow Conditions. J Nat Gas Sci Eng 54: 283–296. https://doi.org/10.1016/j.jngse.2018.04.004.
Hamdi, H., Clarkson, C. R., Esmail, A. et al. 2019. A Bayesian Approach for Optimizing the Huff-n-Puff Gas Injection Performance in Shale Reservoirs
under Parametric Uncertainty: A Duvernay Shale Example. Paper presented at the SPE Europec Featured at 81st EAGE Conference and Exhibition
London, England, UK, 3–6 June. SPE-195438-MS. https://doi.org/10.2118/195438-MS.
Hamdi, H., Sousa, M. C., and Behmanesh, H. 2017. Bayesian History-Matching and Probabilistic Forecasting for Tight and Shale Wells. Paper
presented at the SPE Unconventional Resources Conference, Calgary, Alberta, Canada, 15–16 February. SPE-185082-MS. https://doi.org/10.2118/
185082-MS.
Hasan, S. S. and Mattar, L. 2017. Does Unit-Slope beyond Maximum Producing Time Always Represent BDF in RTA? Paper presented at the SPE
Unconventional Resources Conference, Calgary, Alberta, Canada, 15–16 February. SPE-185020-MS. https://doi.org/10.2118/185020-MS.
Honarpour, M. M., Nagarajan, N. R., Orangi, A. et al. 2012. Characterization of Critical Fluid PVT, Rock, and Rock-Fluid Properties—Impact on
Reservoir Performance of Liquid Rich Shales. Paper presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA,
8–10 October. SPE-158042-MS. https://doi.org/10.2118/158042-MS.
Houzé, O., Viturat, D., Fjaere, O. S. et al. 2020. Dynamic Data Analysis: The Theory and Practice of Pressure Transient, Production Analysis, Well
Performance Analysis, Production Logging and the Use of Permanent Downhole Gauge Data v5.03.01 (Ebook), 852. Houston, Texas, USA:
KAPPA Engineering.
IHS Markit. 2020. Unconventional Reservoir Theory: IHS Harmony Documentation, https://www.ihsenergy.ca/support/documentation_ca/Harmony/
content/html_files/reference_material/analysis_method_theory/unconventional_reservoir_theory.htm.
Ilk, D. and Houze, O. 2016. Model-Based Well Performance Analysis and Forecasting. In Monograph 4—Estimate Ultimate Recovery of Developed
Wells in Low Permeability Reservoirs, ed. John Seidle, Chap. 7, 155. Houston, Texas, USA: Society of Petroleum Evaluation Engineers.
Lee, J. 2016. Fluid Flow and Alternative Decline Models. In Monograph 4—Estimate Ultimate Recovery of Developed Wells in Low Permeability Reser-
voirs, ed. John Seidle, Chap. 6, 113. Houston, Texas, USA: Society of Petroleum Evaluation Engineers.
Li, H., Luo, H., Zhang, J. et al. 2018. Rate-Transient Analysis in Multifractured Horizontal Wells of Tight Oil Reservoir. SPE Prod & Oper 33 (4):
718–738. SPE-189992-PA. https://doi.org/10.2118/189992-PA.
Liang, P., Mattar, L., and Moghadam, S. 2011. Analyzing Variable Rate/Pressure Data in Transient Linear Flow in Unconventional Gas Reservoirs.
Paper presented at the Canadian Unconventional Resources Conference, Calgary, Alberta, Canada, 15–17 November. SPE-149472-MS. https://
doi.org/10.2118/149472-MS.
Mattar, L. and Zaoral, K. 1992. The Primary Pressure Derivative (PPD) a New Diagnostic Tool in Well Test Interpretation. J Can Pet Technol 31 (4):
63–70. PETSOC-92-04-06. https://doi.org/10.2118/92-04-06.
Nobakht, M., and Mattar, L. 2012. Analyzing Production Data from Unconventional Gas Reservoirs with Linear Flow and Apparent Skin. J Can Pet
Technol 51 (1): 52–59. SPE-137454-PA. https://doi.org/10.2118/137454-PA.
Pitakbunkate, T., Balbuena, P. B., Moridis, G. J. et al. 2016. Effect of Confinement on Pressure/Volume/Temperature Properties of Hydrocarbons in
Shale Reservoirs. SPE J. 21 (2): 621–634. SPE-170685-PA. https://doi.org/10.2118/170685-PA.
Poston, S. W., Laprea-Bigott, M., and Poe, B. D. 2019. Analysis of Oil and Gas Production Performance, 167. Richardson, Texas, USA: Society of
Petroleum Engineers.
Qanbari, F. and Clarkson, C. R. 2016. Rate-Transient Analysis of Liquid-Rich Tight/Shale Reservoirs Using the Dynamic Drainage Area
Concept: Examples from North American Reservoirs. Paper presented at the SPE Low Perm Symposium, Denver, Colorado, USA, 5–6 May.
SPE-180230-MS. https://doi.org/10.2118/180230-MS.
Robinson, D. B. and Peng, D.-Y. 1978. The Characterization of the Heptanes and Heavier Fractions for the GPA Peng-Robinson Programs. Research
Report 28, Gas Processors Association, Tulsa, Oklahoma, USA.
Rock Flow Dynamics. 2019. tNavigator Reservoir Simulator’s User Manual, Version 19.1. Houston, Texas, USA.
Samandarli, O., McDonald, B., Barzola, G. et al. 2014. Understanding Shale Performance: Performance Analysis Workflow with Analytical Models in
Eagle Ford Shale Play. Paper presented at the SPE Unconventional Resources Conference, The Woodlands, Texas, USA, 1–3 April. SPE-169004-
MS. https://doi.org/10.2118/169004-MS.
Shahamat, M. S. and Clarkson, C. R. 2018. Multiwell, Multiphase Flowing Material Balance. SPE Res Eval & Eng 21 (2): 445–461. SPE-185052-PA.
https://doi.org/10.2118/185052-PA.
Stone, H. L. 1973. Estimation of Three-Phase Relative Permeability and Residual Oil Data. J Can Pet Technol 12 (4): 53–61. PETSOC-73-04-06.
https://doi.org/10.2118/73-04-06.
Storn, R. and Price, K. 1997. Differential Evolution – A Simple and Efficient Heuristic for Global Optimization over Continuous Spaces. Journal of
Global Optimization 11 (4): 341–359. https://doi.org/10.1023/A:1008202821328.
Tabatabaie, S. H. and Pooladi-Darvish, M. 2017. Multiphase Linear Flow in Tight Oil Reservoirs. SPE Res Eval & Eng 20 (1): 184–196. SPE-180932-
PA. https://doi.org/10.2118/180932-PA.
Uzun, I., Kurtoglu, B., and Kazemi, H. 2016. Multiphase Rate-Transient Analysis in Unconventional Reservoirs: Theory and Application. SPE Res Eval
& Eng 19 (4): 553–566. SPE-171657-PA. https://doi.org/10.2118/171657-PA.
Wattenbarger, R. A., El-Banbi, A. H., Villegas, M. E. et al. 1998. Production Analysis of Linear Flow into Fractured Tight Gas Wells. Paper presented at
the SPE Rocky Mountain Regional/Low-Permeability Reservoirs Symposium, Denver, Colorado, USA, 5–8 April. SPE-39931-MS. https://doi.org/
10.2118/39931-MS.
Whitson, C. H. and Torp, S. B. 1983. Evaluating Constant-Volume Depletion Data. J Pet Technol 35 (3): 610–620. SPE-10067-PA. https://doi.org/
10.2118/10067-PA.

14 2020 SPE Journal

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119


DOI: 10.2118/196164-PA Date: 20-August-20 Stage: Page: 15 Total Pages: 15

Winestock, A. G. and Colpitts, G. P. 1965. Advances in Estimating Gas Well Deliverability. J Can Pet Technol 4 (3): 111–119. PETSOC-65-03-01.
https://doi.org/10.2118/65-03-01.
Zhang, M. and Ayala, L. F. 2017. Similarity-Based, Semi-Analytical Assessment of Capillary Pressure Effects in Very Tight, Liquid-Rich Gas Plays
during Early-Transient Multiphase Analysis. J Nat Gas Sci Eng 45: 189–206. https://doi.org/10.1016/j.jngse.2017.04.031.
Zhang, M. and Ayala, L. F. 2018. A Semi-Analytical Solution to Compositional Flow in Liquid-Rich Gas Plays. Fuel 212: 274–292. https://doi.org/
10.1016/j.fuel.2017.08.097.
Zhang, M., Becker, M. D., and Ayala, L. F. 2016. A Similarity Method Approach for Early-Transient Multiphase Flow Analysis of Liquid-Rich Uncon-
ventional Gas Reservoirs. J Nat Gas Sci Eng 28: 572–586. https://doi.org/10.1016/j.jngse.2015.11.044.

2020 SPE Journal 15

ID: jaganm Time: 15:25 I Path: //chenas03.cadmus.com/Home$/jaganm$/SA-SPE-J###200119

You might also like