Professional Documents
Culture Documents
2, 2019
Sheik Rahman
Petroleum Engineering School,
New South Wales University,
Sydney, Australia
Email: sheik.rahman@unsw.edu.au
Biographical notes: Reda Rabiee Abdel Azim holds a PhD from the
University of New South Wales, Australia 2015, MSc and BSc from the
University of Cairo, Egypt in 2005 and 2010 respectively, all in Petroleum
Engineering. Prior joining Cairo University as an Assistant Professor, he
worked with Schlumberger Company (2006–2009) as a reservoir engineer for
more than three years and then with Technical Petroleum Services (TPS)
Company (2009–2011) as a Senior Consultant Petroleum Engineer. In TPS he
led research programs in the area of integrated reservoir studies and pressure
transient analysis. In 2011, he joined the University of New South Wales’s
1 Introduction
Upscaling of oil and water relative permeability curves have been performed by different
approaches including steady state methods and dynamic upscaling methods. In steady
state methods, the fluids are assumed to be in a steady state conditions, which means that
the fluid saturation distribution within the grid blocks does not change with time and
fractional flow remains constant. These assumptions are valid only over a small scale
(20 cm, or less) in which oil and water phases may come into capillary equilibrium.
There are several steady state method used in upscaling process depending on the balance
between governing forces (Kumar and Jerauld, 1996; Pickup and Stephen, 2000) which
include:
a capillary dominated flow
b viscous dominated flow
c gravity dominated flow.
In capillary dominated flow, the viscous and gravity forces are negligible and fluids are
in capillary equilibrium and saturation distribution is calculated by capillary pressure
curve. The accuracy of this method is tested by Pickup and Stephen (2000) and it has
been concluded that the method is not as accurate as dynamic two phase upscaling, but its
performance is worthy in some cases (e.g., low injection rate).
In viscous dominated flow (for example in area close to the injection well and inside
the fractures), with increase in fluid velocity, the injected fluid moves faster in the high
permeability regions (fractures). This leads to dispersion of the water flood front during
the displacement process (Dongxiao and Tchelepi, 1999). The amount of the dispersion
depends on the mobility ratio between the flowing phases. In this case, capillary forces
become smaller than the viscous forces and Buckley-Leverett formulation and Darcy’s
law can be used in upscaling process. In the case, where viscous force is dominated,
many authors used a single phase upscaling in a two phase system (Christie and Blunt
2001). In this case the dispersion of the flood front due to heterogeneities will not be
taken into account which may result in unreasonable upscaled relative permeability
curve.
190 R.R.A. Azim and S. Rahman
Gravity dominated flow depends on the fluid segregation, in which the lighter fluid
(low density) flows to high side and the high density to the low side. Since this gravity
effect depends on the density difference between fluids, gas floods are more likely to be
gravity dominated than water floods. In this case, vertical equilibrium can be assumed
and 3D simulation may be reduced to 2D.
The dynamic upscaling method is the most difficult one as it is time consuming and
requires two phase flow simulation on a fine grid scale. This method has the advantage of
taking into account of dispersion of the flood front due to heterogeneities or capillary or
gravitational forces. The dynamic upscaling method is divided into two classes:
a weighted pressure method presented by Kyte and Berry (1975)
b total mobility method presented by Stone 1991.
In weighted pressure method, the upscaled relative permeabilities are computed using
Darcy’s law and the sum of flow rates and average of pressure gradient across the grid
blocks. While, in total mobility method the upscaling works on the fractional flows. Kyte
and Berry’s (1975) method has many disadvantages that affect the upscaling results.
They include:
1 negative values of relative permeability if pressure gradient across the grid block has
the same sign as the flow rate
2 infinite relative permeability if the pressure gradient is zero
3 large error occurs in case where displacement is gravity dominated.
In addition, Kyte and Berry (1975) method requires a fine scale simulation to generate
pressure distribution in the studied domain. Fine scale simulations require high
computational resources (Barker and Thibeau, 1997). In addition, generation of fine scale
data (permeability and porosity) requires geostatistical interpolation which may create
uncertainties (Farmer, 2002). To improve the accuracy of this method and avoid
computational problems, Stone (1991) used analytical approach to compute average total
mobility and fractional flow. Darman (2000) improved the Kyte and Berry (1975)
method for cases where the gravity effects are significant. Darman’s (2000) method is
quite similar to Kyte and Berry’s (1975) method except for the transmissibility weighting
which is used to calculate the average pressure across the grid blocks.
Total mobility method are more robust than Kyte and Berry method in the sense that
infinite values can be avoided and negative values occur less frequently, but it can still
occur if the net flows of the two phase are in the opposite directions. Both methods, Kyte
and Berry and Stone’s total mobility are considered computationally expensive, but at
least Stone’s total mobility tend to be more robust than Kyte and Berry method.
The multiphase upscaling process is complicated compared to single phase and still
not well developed (Carlson, 2003; Yang et al., 2013). Existence of fractures in the
domain adds further complexity. In addition, in coarse grid, the sweep profile is
homogenous and the up-scaled relative permeability does not represent flow behaviour
appropriately. Although numerous studies have been carried out using conventional
upscaling techniques (dynamic upscaling methods) in fractured reservoir simulation
(Rossen and Shen, 1989; Talukdar et al., 2000; Ding et al., 2006) until now no
appropriate relative permeability upscaling methodology is documented using 3D flow
simulation in discrete fractures.
3D numerical and experimental study 191
For the purpose of multi-phase flow simulation the laboratory acquired relative
permeability is up-scaled to grid block scale. Fractured reservoirs are highly
heterogeneous in nature and therefore properties, such as the density, azimuth and dip of
fractures vary significantly from grid block to grid block which make the flow simulation
computationally exhaustive. In order to overcome this difficulty all sub-domains (grid
blocks) are grouped into a number fracture patterns based on azimuth and dip. Then the
sub-domain of each specified fracture pattern is discretised into elements of size same as
Fahad et al.’s (2017) laboratory model, 20 cm × 10 m × 2 cm. Size of these elements is
controlled using the 3D mesh generator. Then the relative permeability is defined for
each element (tetrahedral element) using Fahad et al. (2017) correlation. Then fine scale
flow simulation is carried out to up-scale the lab scale relative permeability to subdomain
scale. Once the relative permeability curve is up-scaled to each of the characteristic
subdomains, the up-scaled relative permeability are subsequently distributed to the entire
domains (1875) by matching the fracture pattern with corresponding sub-domain.
Core scale displacement experiments (both steady and unsteady state) are carried out
to derive oil and water relative permeability curves and the corresponding production
data which are then used to upscale to sub-domain reservoir scale.
Detailed derivation of the equations is presented in Abdel Azim (2016) and in
Appendix A (Abdel Azim, 2016). The poro-elastic aspects of the model, such as changes
in stresses with production draw-down or vice versa are validated against appropriate
analytical solutions and finite element tool box (Abdel Azim, 2016).
In general, behaviour of two phase fluid flow system through fractures and matrix porous
media is governed by generalised Darcy’s law and continuity equation.
The Darcy’s law is expressed as:
kij krπ
φ sπ u πs = [ − pπ + ρπ gi ] π = w, nw (1)
μπ
General continuity equation for wetting phase incorporating the concept of effective
stress can be expressed as follow:
D ∂ε Dc
1− 3K ∂t + 3K +
k k ∂ ρ s s m
∇( pw + ρw gh) + φ w w + ρw w
ij rw m
−∇T = + ρwQw = 0 (2)
μ
w wβ ∂t β
w β w
1 − φ D ∂p
− 2
−
m ( 3Km )
K ∂t
General continuity equation for non-wetting phase can be expressed as follow:
D ∂ε Dc
1 − 3K ∂t + 3K +
k k ∂ ρ s s m
∇ ( po + ρo gh) + φ o o + ρo o + ρ Q = 0 (3)
ij ro m
−∇T =
μ
o oβ ∂t β
o o
1 − φ D ∂p
o o
− 2
−
m ( 3Km )
K ∂t
3D numerical and experimental study 193
where φ is the porosity of the media, Sπ is the saturation for each phase, uπs is the relative
velocity vector between fluid phase and solid phase, kij is the permeability tensor, krπ is
the relative permeability for each fluid phase π, μπ, ρπ and pπ are dynamic viscosity,
density of fluid, and fluid pressure for each phase respectively, gi is the gravity
acceleration vector, βπ is the fluid formation volume factor, Km is the bulk modulus of
solid grain, D is the elastic stiffness matrix, Qπ and represents external sources or sinks.
The computational procedure for solving fully coupled two phase fluid flow equations is
described in details in Figure 1.
Figure 1 Description of the computational procedure of two phase fluid flow calculations
Define a threshold
value of fracture
length
Starting simulation
Calculation of 3D of two phase fluid
permeability tensor for flow
short fractures using single
phase fluid flow
Initialise the reservoir model
using all rock and fluid
Mesh generation for element properties
based permeability tensor and
discrete fractures (long
fractures)
Time step
checking
Increasing No
of time Solve both saturation and relative
step permeability through the relations of
t> Pc-Sw and kr-Sw
t_max
Yes
Solve the water
saturation changes If capillary
End from stabilised pressure effect is
saturation equation ignoring
194 R.R.A. Azim and S. Rahman
(Q − Qisim )
2
J= i
obs
(6)
i =1
where Qiobs is the observed volume of produced fluid Qisim s the simulated volume of
produced fluid, and N is the total number of recovery observations to be history matched.
In step 2, the parameter ‘a’ is modified by a magnitude of ε (ε = 0.001) to estimate the
relative permeability curve which is then used to simulate two phase fluid flow to
calculate new fluid volumes (oil and water). These water and oil volumes are
subsequently used in equation (6) to calculate a new least square function, Ja and ∇Ja
then as follow:
Ja − J
∇J a = (7)
ε
In step 3, the sensitivity coefficient matrix, D is calculated as:
Q obs − Q sim
Da = (8)
ε
Now the first part of the left hand side of the Levenberg-Marquardt algorithm is obtained
and is known as Hessian matrix which is given by:
H ( X ) = DT D (9)
3D numerical and experimental study 195
In step 4, similar procedures are repeated for other controlling parameters b, c, and d until
DTD matrices are complete.
In step 5, the full formulation of left and right hand sides of Levenberg-Marquardt
algorithm are obtained. The algorithm is solved for the improvement term Δxk and the
equation used in this algorithm is as follow:
where λ is the stability factor, Δxk is used to update the control parameters (a, b, c and d)
as:
xk +1 = xk + αΔxk (11)
where α is the step size. Steps from one to five are repeated until Δxk be very small and
error between the observed and calculated volume of produced fluid is minimised.
The first step of the upscaling process is to simulate the unsteady state experiment so that
the history match of the production data can be performed. Core samples, in which
displacement tests are carried out, are presented in Figure 2. These core samples have a
length of 38 mm and diameter of 25.5 mm. The fracture aperture is estimated from the
volume of fluid occupying the fracture space which is 0.05 mm. The reservoir rock and
fluid properties are presented in Table 1.
Figure 2 Core samples with naturally occurring fractures from a typical basement reservoir
(see online version for colours)
196
Table 1
Experiment Length Diameter Porosity Permeability Density Viscosity Interfacial tension Capillary number
R.R.A. Azim and S. Rahman
L D ϕ k Fluids ρ µ σ Nc
Steady state
mm mm % mdarcy gm/cm3 cP mN/m vµ/σ
Corrected gas permeability 73.65 37.84 0.7 0.095
(unfractured sample)
Sample 1 (with 30º fracture) 48.8 25.4 8.28 1.226 Brine 1.063 0.95 38 10–7
Sample 2 (with 45º fracture) 25.3 25.25 16 2.135 Soltrol 0.755 1.46
Sample 3 (with 30º fracture) 30.20 25.30 10.4 1.23
Rock properties data for core samples used in relative permeability experiments
3D numerical and experimental study 197
Figure 3 A schematic representation of un-steady state experimental setup (see online version
for colours)
Prior to the displacement tests, core sample are cleaned in a soxhlet extractor using 100%
Isopropanol solvent. Next, these samples are dried in an oven at temperature of 60°C for
24 hours. Then the dry weight and dimensions (length and diameter) of the samples are
measured. The cleaned samples are placed in a vacuum pump for 12 hours to evacuate
any trapped air until the vacuum pressure is less than 0.1 m bar. De-aired brine (4% KCl)
is then allowed to flow into the sample slowly until the sample is fully submerged in
brine (about six hrs has been taken for the sample to be fully saturated). The sample is
left submerged for another eight hours to allow complete saturation. Finally both steady
and unsteady state displacement tests are carried out to measure production and their
198 R.R.A. Azim and S. Rahman
corresponding pressure. Injection of fluid is adjusted using a dual cylinder pump (Model
260D syringe pump) with high precision (+0.5% of the set rate).
Figure 5 Water saturation profile after 1 min of water injection during the unsteady state
experiment (sample with 15° fracture) (see online version for colours)
Sw
Producer
Injector
3D numerical and experimental study 199
The relative permeability curves are generated by history matching of the production data
(the unsteady state experimental data). The history matching is carried out using the
inversion algorithm. For the purpose flow simulation in core scale 3D mesh of the core
sample is generated (see Figure 4). Matrix is represented by triangular prisms while
fracture (cutting the core sample along longitudinal axis) is represented by triangular
elements for the core sample with a single fracture cutting along the axis. The two phase
flow simulator. Figure 5 shows the water saturation profile after one minute of water
injection and Figure 6 shows the comparison between the laboratory data (produced oil
volume) and the simulation data. The relative permeability curves of the core samples
thus obtained are presented in Figures 7(a), 7(b), and 7(c).
Figure 6 Comparison between the volumes of produced oil using laboratory experiment and
simulation model (sample with 15° fracture) (see online version for colours)
The unsteady state experiment run under constant water flow rates of 0.1 cc/min for a
core sample with pore volume of 0.6 cc (see Figure 3). The pressure and cumulative oil
recovery have been recorded versus time then the history matching methodology that has
been presented in this paper used to invert the measured data to relative permeability
curve. All the three samples selected for the relative permeability include a single natural
fracture with dip angle ranging from 30° to 45°. In Figure 2 samples with varying
fracture orientation are presented. Results of these tests are presented in Figures 7(a) to
7(c). Corey type power law (Corey, 1954) is used to parameterise relative permeability
for the three core samples.
200 R.R.A. Azim and S. Rahman
Figure 7 Relative permeability curve of steady state test conducted on sample, (a) with 15°
inclination angle (b) with 45° inclination angle (c) with 30° inclination angle
(see online version for colours)
(a)
(b)
3D numerical and experimental study 201
Figure 7 Relative permeability curve of steady state test conducted on sample, (a) with 15°
inclination angle (b) with 45° inclination angle (c) with 30° inclination angle
(continued) (see online version for colours)
(c)
Residual water saturations obtained during the experiments range from 33–38% and the
oil relative permeability (end point at residual water saturation) range from 0.25 to 0.3.
Both oil and water relative permeability curves have a curvature towards higher water
saturation but near lower water saturations both curves flatten out and relative
permeability changes slowly with water saturation [see Figures 7(a) to 7(c)]. This
behaviour can be explained by the fact that there exist two flood fronts, one inside
fracture and the other inside porous matrix. The flood front sweeps faster though the
fracture than the tight matrix (basement) and as result of this the breakthrough of water
front in the fracture happens first which decreases the relative permeability rapidly with
decreasing water saturation. Once the fracture is swept through, the changes in relative
permeability are mainly due to change in saturation in matrix only. After this point, the
relative permeability of fracture-matrix system is primarily controlled by effective
permeability of the matrix.
Core scale permeability of naturally fractured core samples cut from a typical fractured
basement reservoirs is up-scaled to a reservoir scale of size 500 m × 500 m × 90 m (see
Figure 8) [this sector has been extracted from reservoir with dimensions of (25 km ×
10 km × 300 m)] under poroelastic framework.
202 R.R.A. Azim and S. Rahman
Figure 8 3D fracture map generated using object-based model (see online version for colours)
Note: Fractures are distributed stochastically with different radius, dip and azimuth
angles using fracture intensity value of 0.1 m–1 and fractal dimension value of
(D = 1.25).
Source: Abdelazim and Rahman (2016)
The seepage velocity calculated based on the flow rate integration over fracture surfaces
and matrix porous media and by using total volume of the block.
km
v=− ∇p (11)
μ
3D numerical and experimental study 203
where μ the fluid viscosity and p is the pressure and the continuity equation for local
seepage velocity in the matrix read as:
∇.v = 0 (12)
x=0
In case of the flow is parallel to fracture plane, the seepage velocity normal to the fracture
induces a pressure drop expressed by:
1
v = − ∇p (14)
μ
The effective fracture permeability of fracture can be describes by its aperture b as (in
case the fractures are empty):
b3
keff = − (15)
12
The mass conservation equation for the flow in a fracture is:
∇ s .J = − ( v + −v ) .n (16)
+
where n the unit vector is normal to fracture plane, v is the seepage velocity in the
matrix on the side of n and v − is the seepage velocity on the opposite side.
This transport equation is implemented with the above-mentioned boundary
conditions to calculate the permeability tensors.
204 R.R.A. Azim and S. Rahman
Therefore, the total seepage velocity over the block is obtained by integrating the
flow rates over fracture surfaces and matrix porous media. Then the results divided by the
total block volume to calculate the block effective permeability tensor.
1 − −
−keff ∂p
vx =
γ γm
vx dv +
sf
J x ds =
μ ∂x
(17)
where sf is the surface for all fractures and γ is the matrix volume.
The details of the subsurface fracture map with detailed characteristic fracture
properties are presented in Abdel Azim et al. (2014). The input data used during the
upscaling process are presented in Table 2. The effective fracture permeability for the
whole reservoir is shown in Figure 10. The two phase flow simulator and the inversion
algorithm that presented in the above sections are used in the upscaling process. Different
steps involved in the upscaling process are presented below:
Step 1 The reservoir is divided into a number of grid blocks (1.75 no of sub-domains)
of sizes 20 m × 20 m × 30 m includes three layers in z-direction. First, four
distinct fracture patterns are identified based on azimuth and dip of fractures that
intersect the subdomains. Note also that (fractures less than the threshold value
of 40 m are considered for discrete flow simulation. These four patterns as
defined in Table 3 are selected to upscale relative permeability from laboratory
scale (core scale) to sub-domain scale. Of these characteristic fracture patters the
fracture pattern 1 is the homogeneous matrix with no intersecting fracture. The
correlation model presented by Fahad et al. (2017) based on a laboratory glass
bead model with dimensions 20 cm × 10 cm × 2 cm, is used to up-scale the core
scale relative permeability to the subdomain scale (20 m ×
20 m × 30 m).
The relative permeability curve for fracture pattern 1 (homogenous) is obtained
by acquiring displacement data for core samples from the basement reservoir
with homogeneous matrix.
Fahad et al. (2017) conducted enormous number of experiments using glass
bead model and based on the results of these experiments, the authors developed
a correlation for relative permeability estimation of fractures porous system. The
correlation as follows:
kro = 1 + ( −0.015 × n2f + 0.155 × n f ) + ( 0.16 × e−4.01(1−sin ψ ) ) − 0.085 × sin θ × kref (18)
nw−ref
S − S wr
krw = krw− max w (19)
1 − Sw
Swr = 1 + ( −0.059 × n2f + 0.355 × n f ) + ( 0.304 × e−3.7(1−sin ψ ) ) − 0.19 × sin θ × Sref (20)
where kro is the oil relative permeability, kref is the reference oil relative
permeability which is kro of the homogenous glass bead pack (Fahad et al.,
2017), nw–ref and Sref are Corey exponent and residual saturation for glass bead
pack respectively, nf is the number of fractures in the system, θ is the fractures
orientation towards the flow direction, ψ and is the angle between the fracture
3D numerical and experimental study 205
Figure 10 3D fracture equivalent permeability the studied reservoir (see online version
for colours)
k ,md
Kxx, mD
Selected Sector
z
y
x
206 R.R.A. Azim and S. Rahman
Step 3 Next, the fine scale numerical simulation run is carried out using the developed
multiphase numerical simulator at constant pressure drop between the injector
and the producer. The oil and water production rates are calculated with time at
the producer for these four characteristic fracture patterns (sub-domains). These
four production data are referred to as original production data. The inversion
technique is then used to match the calculated oil recovery or the water
breakthrough time (time at which water break in to the wellbore) to produce
relative permeability curve that represents the fluid flow behaviour within the
subdomains.
Figure 11 Estimated relative permeability curve for using (a) Fahad et al. (2017) correlation and
(b) simulation work presented in this paper (see online version for colours)
Parameter Value
Reservoir dimensions 500 m × 500 m × 90 m
Matrix permeability 0.1 mD
Matrix porosity 2%
Fracture aperture 7.06 × 10–3 mm
Initial fracture intensity 0.15 m–1
Fractal dimension (D) 1.25
Initial reservoir pressure 5,063 psia (34.9 MPa)
Injection pressure ( injection case) 6,409 psia (44.2 MPa)
Fluid viscosity 1.38 cp
Fluid compressibility 10–5 psi–1
Source: Farag et al. (2010)
3D numerical and experimental study 207
Step 4 Once the relative permeability curve is up-scaled to each of the four
characteristic subdomains as described in step 2, the up-scaled relative
permeability are subsequently distributed to the entire domain comprising of
1,875 subdomains by matching of the fracture pattern with the corresponding
sub-domain. Noteworthy, so far we have now obtained four characteristic
production data and their corresponding relative permeability which describes
two phase flow behaviour in 1,875 sub-domains. This was done in order to save
extensive computation time.
Figure 12 (a) Fracture map (fractures larger than 40 m) from a cross section of the basement
reservoir and (b) fracture map is divided into 625 sub-domains in 2D and 1,825
sub-domains in 3D (see online version for colours)
(a) (b)
Figure 13 Fracture map is divided into 625 sub-domains in 2D and 1,825 sub-domains in 3D
(left) (see online version for colours)
Note: Further, each sub-domain (20 m × 20 m × 30 m) is further divided into lab glass
bead model domain of size 20 cm ×10 cm × 2 cm (right).
208 R.R.A. Azim and S. Rahman
Figure 14 Estimated relative permeability curve for fracture pattern (1) obtained by using
unfractured (homogenous) core sample (see online version for colours)
The relative permeability curve for fracture pattern 1 (homogenous) is obtained based on
homogeneous sample.
The results of the water saturation distribution after 3 hrs of water injection are shown
in Figures 15, 16, and 17. These figures show that water displaces oil in the discrete
fractures first and then the oil in the matrix porous media depending on the permeability
tensor value. The three characteristic oil and water production rates for fractured patterns
are shown in Figures 18, 19, 21, 22, 24, and respectively. From these figures it can be
seen that the water breakthrough takes place after 6 hrs, 3 hrs, and 1 hr of water injection
for fracture pattern 2, 3, and 4 respectively and this breakthrough time estimated by the
upscaling procedure match well with that estimated by using Fahad et al.’s (2017)
correlation. The characteristic water production rate curves between breakthroughs till
end of simulation time match reasonably well with that estimated by Fahad et al. (2017)
correlation. The up-scaled relative permeability curves for the different fracture patterns
(see Table 1) are shown in Figures 20, 23, and 26.
3D numerical and experimental study 209
Figure 15 Water saturation profile for fracture pattern (2) after 0.5 hr of water injection at
Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1 MPa, σh = 33.1 MPa,
and σv = 41.3 MPa (see online version for colours)
SW
Figure 16 Water saturation profile for fracture pattern (3) after 0.5 hr of water injection at
Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1 MPa, σh = 33.1 MPa,
and σv = 41.3 MPa (see online version for colours)
SW
Figure 17 Water saturation profile for fracture pattern (4) after 0.5 hr of water injection at
Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1 MPa, σh = 33.1 MPa,
and σv = 41.3 MPa (see online version for colours)
SW
Figure 18 Comparison between original and upscaled oil production rate for fracture pattern (2)
(see Table 3) with Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1 MPa,
σh = 33.1 MPa, and σv = 41.3 MPa (see online version for colours)
Figure 19 Comparison between original and upscaled water production rate for fracture pattern
(2) (see Table 3) with Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1
MPa, σh = 33.1 MPa, and σv = 41.3 MPa (see online version for colours)
Figure 20 Upscaled relative permeability curves for fracture pattern (2) in a block size (20 m ×
20 m × 30 m) with average fractures dip angle of 76° (see online version for colours)
212 R.R.A. Azim and S. Rahman
Figure 21 Comparison between original and upscaled oil production rate for fracture pattern (3)
(see Table 3) with Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1 MPa,
σh = 33.1 MPa, and σv = 41.3 MPa (see online version for colours)
Figure 22 Comparison between original and upscaled water production rate for fracture pattern
(3) (see Table 3) with Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1
MPa, σh = 33.1 MPa, and σv = 41.3 MPa (see online version for colours)
Figure 23 Upscaled relative permeability curves for fracture pattern (3) in a block size
(20 m × 20 m × 30 m) with average fractures dip angle of 89.5° (see online version
for colours)
Figure 24 Comparison between original and upscaled oil production rate for fracture pattern (4)
(see Table 3) with Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1 MPa,
σh = 33.1 MPa, and σv = 41.3 MPa (see online version for colours)
Figure 25 Comparison between and upscaled water production rate for fracture pattern (4) (see
Table 3) with Pinitial = 34.9 MPa, Pinj = 44.2 MPa, ∆P = 3.1 MPa, σH = 33.1 MPa, σh =
33.1 MPa, and σv = 41.3 MPa (see online version for colours)
8 Conclusions
References
Abdel Azim, R.G.D.N., Rahman, S.S., Tyson, S. and Regenauer-Lieb, K. (2014)
‘3D poro-thermo-elastic numerical model for analysing pressure transient response to improve
the characterization of naturally fractured geothermal reservoirs’, Geothermal Resources
Council, Vol. 38, No. 1, pp.907–915.
Abdel Azim, R. (2016) ‘An integrated approach for relative permeability estimation of fractured
porous media: laboratory and numerical simulation studies’, Journal of Petroleum Exploration
and Production Technology, pp.1–18.
Abdel Azim, R. and Rahman, S.S. (2016) ‘Estimation of permeability of naturally fractured
reservoirs by pressure transient analysis: an innovative reservoir characterisation and flow
simulation’, Journal of Petroleum Science and Engineering, September, Vol. 145,
pp.404–422.
Barker, J.W. and Thibeau, S. (1997) ‘A critical review of the use of pseudorelative permeabilities
for upscaling’, SPE Reservoir Engineering, Vol. 12, No. 2, pp.138–143.
Carlson, M. (2003) Practical Reservoir Simulation: Using, Assessing, and Developing Results,
PennWell Books, Houston, TX 77027, USA.
Christie, M. and Blunt, M. (2001) ‘Tenth SPE comparative solution project: a comparison of
upscaling techniques’, SPE Reservoir Evaluation & Engineering, Vol. 4, No. 4, pp.308–317.
Corey, A.T. (1954) ‘The interrelation between gas and oil relative permeabilities’, Producers
Monthly, Vol. 19, No. 1, pp.38–41.
Darman, N.B.H. (2000) Upscaling of Two-Phase Flow in Oil-Gas Systems, PhD thesis, Heriot-Watt
University.
Ding, Y., Basquet, R. and Bourbiaux, B. (2006) ‘Upscaling fracture networks for simulation of
horizontal wells using a dual-porosity reservoir simulator’, SPE Reservoir Evaluation &
Engineering, Vol. 9, No. 5, pp.513–520.
Dongxiao, Z. and Tchelepi, H. (1999) ‘Stochastic analysis of immiscible two-phase flow in
heterogeneous media’, SPE Journal, Vol. 4, No. 4, pp.380–388.
Fahad, M. (2013) Simulation of Fluid Flow and Estimation of Production from Naturally Fractured
Reservoirs, PhD thesis.
Fahad, M., Hussain, F., Rahman, S.S. and Cinar, Y. (2017) ‘Experimental investigation of
upscaling relative permeability for two phase flow in fractured porous media’, Journal of
Petroleum Science and Engineering, January, Vol. 149, pp.367–382.
Farag, S.M., Mas, C., Maizeret, P-D., Li, B. and Le, H.V. (2010) ‘An integrated workflow for
granitic basement reservoir evaluation’, SPE Reservoir Evaluation & Engineering, Vol. 13,
No. 6, pp.893–905.
Farmer, C. (2002) ‘Upscaling: a review’, International Journal for Numerical Methods in Fluids,
Vol. 40, Nos. 1–2, pp.63–78.
Jones, C., Al-Quraishi, A.A., Somerville, J.M. and Hamilton, S.A. (2001) Stress Sensitivity of
Saturation and End-Point Relative Permeabilities, Society of core analysts, Edinburgh,
Scotland.
Kumar, A.T. and Jerauld, G. (1996) ‘Impacts of scale-up on fluid flow from plug to gridblock scale
in reservoir rock’, Paper presented at the SPE/DOE Improved Oil Recovery Symposium.
Kyte, J.R. and Berry, D. (1975) ‘New pseudo functions to control numerical dispersion’, Society of
Petroleum Engineers Journal, Vol. 15, No. 4, pp.269–276.
Lough, M.F., Lee, S.H. and Kamath, J. (1996) ‘A new method to calculate the effective
permeability of grid blocks used in the simulation of naturally fractured reservoirs’, Paper
presented at the SPE Annual Technical Conference.
Pickup, G.E. and Stephen, K.D. (2000) ‘An assessment of steady-state scale-up for small-scale
geological models’, Petroleum Geoscience, Vol. 6, No. 3, pp.203–210.
Rossen, R. and Shen, E. (1989) ‘Simulation of gas/oil drainage and water/oil imbibition in naturally
fractured reservoirs’, SPE Reservoir Engineering, Vol. 4, No. 4, pp.464–470.
3D numerical and experimental study 217
Stone, H. (1991) ‘Rigorous black oil pseudo functions’, Paper presented at the SPE Symposium on
Reservoir Simulation.
Talukdar, M., Banu, H., Torsæter, O. and Kleppe, J. (2000) ‘Applicability and rate sensitivity of
several up scaling techniques in fractured reservoir simulation’, Paper presented at the SPE
International Petroleum Conference and Exhibition, Mexico.
Yang, Y., Wang, X., Wu, X-H. and Bi, L. (2013) ‘Multiphase upscaling using approximation
techniques’, Paper presented at the SPE Reservoir Simulation Symposium.