Professional Documents
Culture Documents
Development and Validation Explicitly Coupled Geomechanics Module Compositional Reservoir Simulator
Development and Validation Explicitly Coupled Geomechanics Module Compositional Reservoir Simulator
A R T I C L E I N F O A BS T RAC T
Keywords: Pore pressure-stress analyses in stress-sensitive reservoirs investigate interactions between the in-situ stress
Geomechanics and fluid flow; these interactions help or resist production, or conclude surface subsidence during production.
Compositional reservoir simulator Among the tools for these analyses, an explicitly coupled geomechanics and fluid flow model provides an
Coupled pore pressure-stress analysis essential, reliable, and fast production estimate for field planning and development. In this work, we
Finite element method
implemented this model in an in-house, three dimensional, compositional reservoir simulator, UTCOMP,
Chin's iterative coupling
using Chin's iterative coupling method. This development integrated a stand-alone geomechanics module based
on finite element method with the reservoir simulator, an advantage of our coupling algorithm, and improved
our understanding of the production through various enhanced oil recovery processes such as water and CO2
flooding processes previously coded in UTCOMP. Benefiting from the higher time scales of solution variations
due to the geomechanics module, we lowered the frequency of calling this computationally expensive module.
Also, we reduced the order of the finite element shape functions for displacement from quadratic to linear,
which majorly mitigated the high computational cost of our geomechanics studies while we almost maintained
the solution accuracy. To validate our implementation, we investigated a primary oil production case and
compared the results from UTCOMP with those from two other simulators: (1) CMG software program using
different coupling methods; and (2) another pre-validated in-house reservoir simulator, GPAS. In order to
evaluate our improvements in this work, we compared our results with those from a pre-validated in-house
reservoir simulator, GPAS. We observed a minor discrepancy between the solutions at very early times which
originates from the different structures in these two reservoir simulators, IMPEC in UTCOMP and fully implicit
in GPAS.
1. Introduction Method (FEM). This simulator can solve highly nonlinear problems;
however, this advantage hardly justifies the computational expenses of
In stress-sensitive reservoirs, the interaction between in-situ stress this development. The remedies for the high computational expenses
and fluid flow has some complications, which might be in favor of can be listed as the following: (1) using reduced-order elements for
production or against it (Pan, 2009). Also, due to production from displacement in the geomechanics module; (2) calling geomechanics
some extra loose formations, surface subsidence might occur, which module at lower frequencies; and (3) using IMPEC (IMplicit Pressure
suggests coupling a geomechanics model with the reservoir fluid flow Explicit Concentration) scheme instead of fully implicit scheme for
for the sake of a better simulation of this subsidence. An explicitly reservoir simulation. In this work, we attempted to accomplish these in
coupled geomechanics and fluid-flow model is one of the methods to another reservoir simulator with IMPEC scheme.
provide us with an accurate, reliable, and fast prediction for better well The theme of this work is to develop and validate a geomechanics
planning and reservoir management decisions. module in an in-house reservoir simulator; UTCOMP, which is one of
Pan (2009) integrated a geomechanics module with a fully-implicit the 3D compositional reservoir simulators developed at The University
reservoir simulator, GPAS, using fully coupling and iterative coupling of Texas at Austin (Chang, 1990). This simulator, at its current form,
schemes in conjunction with quadratic elements in Finite Element does not entail any geomechanics module despite the fact that various
Abbreviations: UTCOMP, University of Texas COMPositional oil reservoir simulator; GPAS, General Parallel Adaptive Simulator; IMPEC, IMplicit Pressure Explicit Concentration;
3D, Three Dimensional; EOS, Equation of State; VWP, Virtual Work Principle
⁎
Corresponding author.
E-mail addresses: Mahdi.haddad@utexas.edu (M. Haddad), kamys@mail.utexas.edu (K. Sepehrnoori).
http://dx.doi.org/10.1016/j.petrol.2016.10.044
Received 13 April 2016; Received in revised form 26 August 2016; Accepted 27 October 2016
Available online 29 October 2016
0920-4105/ © 2016 Elsevier B.V. All rights reserved.
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
reservoir engineering processes require geomechanics calculations for simulator for the next time step calculations. This procedure is based
a better understanding of production and field development. Moreover, on true Biot's consolidation theory and simplifies Chin's iterative
the IMPEC scheme used in UTCOMP allows easier integration of coupling method (Chin et al., 2002).
various physical phenomena in this reservoir simulator in comparison We simplified the Chin's iterative loop in a time step between the
with fully-implicit reservoir simulators such as GPAS (Pan, 2009). geomechanics module and reservoir simulator to a single execution of
Therefore, in this work, we implement this module and couple it with the geomechanics module per time step. Thereby, we call our im-
the existing code of the UTCOMP reservoir simulator; this coupling is plementation method as “explicit coupling” procedure due to the time-
based on the modified Chin's Iterative coupling method (Chin et al., marching nature of running the reservoir simulator and geomechanics
2002). We validate our results comparing them with the results from module. This simplification may not significantly influence the solution
CMG software program at various coupling procedures (GEM User knowing that UTCOMP is an IMPEC reservoir simulator that applies
Guide, 2015), and GPAS which is a pre-validated in-house reservoir small time steps for the stability and convergence of the solution.
simulator developed at The University of Texas at Austin (Pan, 2009; However, calling the geomechanics module in every small time step in
Wang et al., 1997). The CMG's coupling procedures define different UTCOMP significantly increases the computational expenses.
explicit dependence of porosity on deformation, pore pressure, tem- Therefore, we also attempted to reduce the computational expenses
perature, volumetric strain, and total mean stress (GEM User Guide, by lowering the frequency of calling the geomechanics module.
2015). These functionalities provide various degrees of coupling of We have two porosity definitions in this work: (1) reservoir
geomechanical effects with the reservoir simulator, and thereby, the porosity, ϕ (=∀p /∀b where ∀p and ∀b denote the reservoir pore and bulk
range of solutions to be compared with the results from our imple- volumes, respectively), used in the reservoir fluid-flow model; and (2)
mentation. true porosity, ϕ* (=∀*p /∀*b where ∀*p and ∀*b denote true pore and bulk
volumes, respectively), calculated from the stress analysis model. Eqs.
(1) and (2) relate these two porosity definitions together via the
2. Method volumetric strain, εν , and initial porosity, ϕ0 in virtue of grain
incompressibility and true and reservoir pore volume equality (Pan,
UTCOMP is a non-isothermal EOS compositional reservoir simu- 2009; Chin et al., 2002; Pan et al., 2007):
lator coded in FORTRAN and enables researchers to investigate a wide
variety of important enhanced oil recovery processes (Ghasemi Doroh,
Table 1
2012; Chang, 1990). In this simulator, IMPEC scheme is used for
Equivalent compressibility, c∼p (Pan, 2009).
solving the governing system of equations for fluid flow in porous
media; pressure equation is solved implicitly and concentration values Dimension (n) Equivalent Compressibility (c∼p, n,n = 1, 2, 3)
are calculated explicitly. More details about the IMPEC scheme in
UTCOMP can be found elsewhere (Ghasemi Doroh, 2012). In our 1 (1 − 2ν )(1 + ν )
E (1 − ν ) ϕ0
implementation, geomechanics module is developed and called right 2 2(1 − 2ν )(1 + ν )
after solving the system of equations for pressure every single or Eϕ0
3 3(1 − 2ν )
multiple time step(s), and the geomechanics output, which includes Eϕ0
new porosity and permeability values, is passed to the reservoir
282
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
ϕ*=1 − (1 − ϕ0 ) eεν (1) acceptable (Bathe, 2001; Hughes, 1987). Moreover, calling the geo-
mechanics module in every single time step unnecessarily increases the
⎛ 1−ϕ0 ⎞
ϕ=⎜ ⎟ ϕ* computational expenses without much gains in the solution accuracy.
⎝ 1−ϕ* ⎠ (2) Thereby, we attempted to resolve these issues in our implementation
without compromising the solution accuracy and convergence.
Here, the volumetric strain, εν , is positive in compaction which is in
In order to improve the computational efficiency of the geomecha-
agreement with the well-established geomechanics stress and strain
nics module, we further improved the code with the following features:
sign convention. Notably, this sign convention is different from that in
(1) implementation of first-order elements besides the second-order
the work done by Pan (2009) and Chin et al. (2002). Substituting Eq.
elements; and (2) reducing the frequency of calling the geomechanics
(1) in Eq. (2) results in Eq. (3), an explicit equation for reservoir
module. The formulation of the lower order finite element method is
porosity:
elaborated in the following section which is followed by the results for
ϕ = e−εν −(1 − ϕ0 ) (3) various test cases. Also, lowering the frequency of calling the geome-
chanics module can be implemented by accumulating the pore pressure
Also, reservoir porosity can be expressed in terms of true porosity
change during the time steps between the old and new geomechanics
after substituting the term (1 − ϕ0 ) from Eq. (1) in Eq. (3):
calls and passing the pore pressure change to the geomechanics module
ϕ ≅ ϕ* (1 − εν ) (4) to calculate new porosity and permeability values.
For clarity, the previous implementation by Pan (2009) in GPAS
A detailed derivation of Eqs. (1) through (4) can be found in
can be outlined as below:
Appendix A. These equations include the effect of the volumetric strain
in porosity, and are implemented after the calculation of the volumetric
1. Use true porosity, ϕ* in mass balance equation instead of reservoir
strain in the geomechanics module. Moreover, as expected, Eqs. (1)
porosity.
and (3) show that increasing the volumetric strain (which coincides
2. Use the approximate equivalent compressibility, c∼p as shown in
with the bulk volume contraction according to the signs convention for
Table 1 in order to improve the stability and robustness of the solver
the strains) results in the reduction of true and reservoir porosities or
of the fluid-flow system.
the pore volume. Considering Eqs. (1) through (4), we need to modify
3. (Optional) use Eq. (7) to update the absolute permeability based on
the mass balance equation as Eq. (5) (Pan et al., 2007):
the variation of porosity (Chin et al., 1998):
D [(1 − εν ) ϕ*Ni ] np ⎡ K krj ⎤ D (ϕ*Ni )
∀b − ∀b ∇. ∑ j =1 ⎢ξ j xij μ (∇Pj − γj ∇D ) ⎥+ ∀b εν Dt − qi=0,
Dt ⎣ j ⎦
k / ki=(ϕ*/ ϕi*)n , (7)
i=1, 2, …, nc +1
(5)
where k and ki are the current and initial permeability values, and ϕ*
Since we assume infinitesimal strains, we can neglect the last term, and ϕi* denote the current and initial true porosities, respectively. Eq.
D (ϕ*N )
∀b εν Dt i . Using Eq. (4) we have: (1) provides the current true porosity based on the initial true porosity
np ⎡ K krj ⎤ and the volumetric strain which is obtained from the geomechanics
D (ϕNi ) ⎢ξ x (∇Pj − γj ∇D ) ⎥− qi=0, module. The power-law exponent n is an empirical constant and
∀b − ∀b ∇. ∑ ⎣ j ij μj ⎦
Dt j =1 strongly depends on chemical and mechanical compaction processes
i=1, 2, …, nc +1 (6)
(David et al., 1994). The typical value of n is 3.0 after P & M theory
where ϕ is calculated using Eq. (4) depending on the true porosity, ϕ* (GEM User Guide, 2015) and for chemical compaction processes
and the volumetric strain,εν ; these two parameters are evaluated in the (David et al., 1994). Generally, this power-law relationship between
geomechanics module and transferred to the reservoir simulator for the permeability and porosity is established for stress-sensitive, weak, or
subsequent time step(s) calculations. unconsolidated rocks, and the exponent n is considered as a matching
Eq. (6) uses the same terms as in the mass balance equation in parameter (Chin et al., 1998) and is obtained empirically through
traditional reservoir simulators without the geomechanics effect, which curve-fitting of the permeability reduction as a function of rock
simplifies the integration of a reservoir simulator with a stress analysis porosity with experimental data. This procedure can conclude n in
simulator using minimal modifications. One important modification the wide range of 4.6–25.4 for sandstones (David et al., 1994). For
(for the sake of the porosity initialization phase) is to replace the instance, for a high permeability reservoir at Gulf of Mexico deep-water
traditional constant pore compressibility scalar with the so-called
equivalent compressibility, c∼p , as shown in Table 1 for different
geometric dimensions.
The general time-marching procedure can be expressed as the
following: in every time step, the reservoir porosity and permeability,
and the pore pressure are passed from the reservoir simulator to the
geomechanics module and the updated porosity and permeability
values are returned from the geomechanics module to the reservoir
simulator for the next time step(s). The updated porosity and perme-
ability may be used for multiple subsequent steps in order to reduce the
computational cost of calling the geomechanics module. Moreover, a
type of coupling updates only porosity values after the geomechanics
calculation and permeability remains constant through the simulation.
For further details on the iterative coupling procedure, readers are
referred to Chin et al. (2002) and Tran et al. (2002).
We primarily implemented higher order (quadratic) finite elements
for displacement in the geomechanics module, which caused heavy
computational expenses. Using quadratic elements sufficiently satisfies
severe inf-sup condition for mathematical convergence of our problem;
however, a lower order element not fulfilling this condition may still be Fig. 1. 8-node “brick” element (Smith and Griffiths, 2004).
283
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
where [A] is the operator matrix and {U} denotes the displacement
vector. The aforementioned volumetric strain can be expressed as the
summation of the first three entries of the above strain vector,
according to Eq. (9):
εv=εx +εy +εz (9)
284
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
285
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
Table 4 plete description can be found in GEM User Guide (2015). The higher
Summary of input data for the 3D water injection case. production rate associated with GCOUPLING 2 is in good agreement
with the UTCOMP results, which validates our geomechanics imple-
Model Parameter Value
mentation in UTCOMP.
Reservoir model Length (ft)×Width 350 × 350 × 60 Moreover, CMG simulator is capable of updating geomechanics
(ft)×Thickness (ft) quantities with a constant frequency indicating the number of elapsed
Number of grid blocks 7×7×3
time steps per geomechanics calls (GEM User Guide, 2015). We used
Grid block size (ft3) 50 × 50 × 20
Porosity 0.3
this capability in the primary production case for GCOUPLING=2 in
Permeability (mD) 100 CMG to increase the mentioned frequency value from 1 to 2, which
Initial water saturation 0.2 reduced the CPU time by a factor of 0.58. This computational time
Simulation time (day) 2000.0 reduction should be almost identical between different cases consider-
Injection well water 100
ing the following: (1) the majority of the overall computational time is
content (%)
Reservoir fluid composition C10 (100%) spent during geomechanics calculations; and (2) the geomechanics
Water injection rate (STB/ 400 computational expenses are proportional to the number of geomecha-
day) nics module calls. This hypothesis is later investigated by comparing
Production well pressure 600 the computational time saving in the current case, primary production
(psi)
Injection well grid location 1×1×1 to 1×1×3
in CMG, with that in the water injection and CO2 injection cases when
Production well grid 7×7×1 to 7×7×3 the frequency of the geomechanics calls is reduced.
location From the mechanical point of view, in our implementation where
Relative permeability Corey's model we have no normal displacement in the lateral walls, a higher oil rate is
model (Table 5)
reasonable as a result of active geomechanics effects; pushing against
Geomechanics model Number of elements 7×7×3 the reservoir through the overburden pressure. This simple case is
Element type 8-node brick similar to the case of squeezing a sponge fully saturated with water. As
Element size (ft3) 50 × 50 × 20 expected, leaving the sponge on a table does not change the sponge's
Young's modulus (ksi) 13 water saturation whereas holding it in hand and squeezing it drains
Poisson's ratio 0.3
most of the water out of the sponge.
Modified Rock 0.00030312691
compressibility ( psi−1)
In order to check the performance of our implementation for other
Permeability Power-law 15 major reservoir engineering processes, we applied the developed
Exponent, n program for the geomechanics module in UTCOMP on 3D cases of
water and CO2 injection. Notably, the above comparison with CMG
results sufficiently validated our implementation. Thereby, in the
Table 5 following cases, we only compared our results with those from a pre-
Corey's model relative permeability parameters for the 3D water injection case.
validated in-house reservoir simulator, GPAS, in order to evaluate our
Water Oil improvements in this work.
286
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
time steps with the variation of pore pressure within the current time
step or reservoir porosity within the previous geomechanics call. As a
complementary, UTCOMP reservoir simulator contains multiple cri-
teria to increase or decrease the time step size depending on the change
of the solution variables at the current time step (Chang, 1990). These
criteria may indirectly take into account the effect of the porosity
change in both the convergence rate and the next time step size, e.g.
through the mass balance equation, Eq. (6), where the reservoir
porosity is modified using the geomechanics module.
Furthermore, in another study, we attempted to demonstrate the
grid independency of the solution. For that purpose, we refined the grid
Fig. 6. Comparison of oil production rate in GPAS and UTCOMP with and without used for the 3D water injection case from 7 × 7 × 3 to 25 × 25 × 10
geomechanics module for the 3D water injection case. keeping all other model parameters the same as those in Tables 4 and
5. Having compared the average reservoir pressure, oil production rate,
generation can be found in Tables 4 and 5. and cumulative oil production from the investigated cases, we con-
Fig. 4 compares the cumulative oil production in the 3D water cluded that the discrepancy between the results due to grid refinement
injection case from two simulators, GPAS and UTCOMP with and is negligible, which proved the independence of the solution from the
without geomechanics, and with or without permeability modifications. grid.
As mentioned before, in an explicit coupling procedure, updating
permeability values can be performed optionally while the porosity 3.3. CO2 injection case
modification provides always a geomechanics coupling. Fig. 4 shows
the perfect match between different methods, which verifies our The last investigated case examines the geomechanics effect on
implementation and the negligible geomechanics effects on production. production where the main driving mechanism is oil swelling, reduc-
This negligible effect is reasonable as the reservoir is fully saturated tion of oil viscosity, and IFT through achieving miscibility conditions
with a single component liquid with low compressibility values, which depending on thermodynamic properties including pore pressure.
does not allow the contraction of the pore space or the porosity and Notably, pore pressure is one of the transferring parameters from the
permeability alteration under the overburden loads. Moreover, Fig. 5 reservoir simulator to the geomechanics module. Considering the effect
shows the small difference between the cumulative oil production of pore pressure alteration on porosity and permeability, the indirect
profiles by a closer look into the curves shown in Fig. 4. The difference influence of our geomechanics coupling on thermodynamic calcula-
is around 0.3% of the average value after 2000 s of injection or tions can be easily inferred. Moreover, CO2 injection has recently
production. attracted significant attention and field application as it provides a
Fig. 6 compares the oil production rate from different simulators great potential for carbon sequestration besides enhancing the oil
and coupling methods, and is in agreement with the results in Figs. 4 recovery. Tables 6 and 7 provide the detailed model description for this
and 5. Noticeably, the oil rate from “UTCOMP+GEOM (por+perm)”
curve maintains the highest level right before the abrupt reduction Table 6
Summary of input data for the 3D CO2 injection case.
around 750 days while it drops to the lowest level right after this sharp
reduction. The integration of this variation in time, as shown in Figs. 4 Model Parameter Value
and 5, smooths the observed transient behavior in oil rate and thereby,
the geomechanics effects. Therefore, the oil rate comparison provides a Reservoir model Length (ft)×Width 400 × 400 × 45
(ft)×Thickness (ft)
better method to evaluate the effect of the geomechanics coupling on
Number of grid blocks 10 × 10 × 3
production. Grid block size (ft3) 40 × 40 × 15
We attempted to reduce the frequency of geomechanics calls for the Porosity 0.3
3D water injection case. This concluded a significant CPU time Permeability (mD) 10
reduction by a factor of 0.6 when calling geomechanics module every Initial water 0.17
saturation
two time steps instead of every time step. Moreover, this reduction in
Simulation time (day) 4000.0
the frequency of the geomechanics calls did not influence the results as Injection well content CO2 (95%) and C1 (5%)
a comparison for the average reservoir pressure, oil production rate, Reservoir fluid CO2 (1%), C1 (29%), NC16
and cumulative oil production from different calling frequencies shows composition (70%)
a perfect match between the results. Total gas injection 10
rate (MSCF/day)
Interestingly, this CPU time reduction factor, 0.6, is close to the Production well 500
CPU time reduction factor, 0.58, in the primary production case which pressure (psi)
was obtained using GCOUPLING=2 in CMG. It should be noted that Injection well grid 1 × 1 × 1 to 1 × 1 × 3
both mentioned cases executed the geomechanics module every other location
Production well grid 10 × 10 × 1 to 10 × 10 × 3
time steps, which led to these CPU time reduction factors close to 0.5.
location
However, the difference between these CPU time reduction factors and Relative permeability Corey's model (Table 7)
0.5 reflects the contribution of the reservoir simulation calculations in model
the overall CPU time.
However, we could not obtain numerical convergence for geome- Geomechanics Number of elements 10 × 10 × 3
model Element type 8-node brick
chanics calls falling three or more time steps apart keeping the
Element size (ft3) 40 × 40 × 15
minimum and maximum time step size as before. This constant user- Young's modulus (ksi) 13
defined frequency is case-dependent, is obtained by trial and error Poisson's ratio 0.3
especially for complicated problems, and depends on the user-defined Modified Rock 0.00030312691
compressibility
allowable minimum and maximum time step size. This constant
(psi−1)
frequency should be replaced by a more rigorous criterion which would Permeability Power- 10
correlate the frequency of calling the geomechanics module for the next law Exponent, n
287
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
Table 7 the ideal CPU time reduction factor which is the inverse of the number
Corey's model relative permeability parameters for the 3D CO2 injection case. of time steps between the geomechanics module calls. For instance, this
normalized value in this CO2 injection case is 1.1 which is close to 1.16
Water Oil Gas
in primary production case or 1.2 in water injection case. Table 8
Residual saturation 0.2 0.1 0 summarizes our analysis on CPU time reduction due to lower calling
End point relative permeability 0.4 0.7 0.9 frequencies of the geomechanics module in all investigated cases.
Exponent of relative permeability 3.0 2.0 2.0
Fig. 10 compares the time step size through time for the 3D CO2
injection case in GPAS and UTCOMP with and without geomechanics
case where all boundaries except the top boundary are constrained for module. This comparison shows that UTCOMP increases time steps
normal displacement and a unit load is exerted on the top boundary. more conservatively mainly due to the IMPEC scheme where the
Figs. 7 through 9 compare respectively the cumulative oil produc- saturation is calculated explicitly through smaller time steps compared
tion, oil production rate, and average reservoir pressure for the 3D CO2 to those in a fully implicit and unconditionally stable reservoir
injection case in GPAS and UTCOMP with and without geomechanics simulator such as GPAS.
module where both porosity and permeability are modified. These
figures show that activating geomechanics slightly increases the 4. Conclusions
cumulative oil production and oil production rate, and slightly
decreases the average reservoir pressure. These slight differences can We developed a geomechanics module and explicitly coupled that
be justified considering the following observations: (1) the preferential with the UTCOMP simulator. We validated this implementation by
compression of the gas phase via the overburden loading, concluding a comparing the results from several cases against GPAS and CMG. The
small pore pressure increase and consequently, a slight porosity and main findings of this work can be summarized as follows:
permeability modification; (2) the positive effect of the overburden
pressure on production enhancement via porosity reduction; and (3) • Due to different coupling mechanisms, the approaches used in the
the transient porosity, and permeability distribution in the reservoir CMG simulator give higher or lower oil production rates with respect
specially in the vicinity of the injector and producer, which builds up to the case where the geomechanics module is inactive. The higher
the average reservoir pressure before the CO2 breakthrough and production rate is in good agreement with the UTCOMP results,
depletes that after the breakthrough. which validates our geomechanics implementation in UTCOMP.
Moreover, the discrepancy between the results from GPAS and • For most of the investigated case studies, UTCOMP results are in a
UTCOMP in the cumulative oil production and the oil production rate, good agreement with those from GPAS and Coupling 2 of CMG. The
as shown in Figs. 7 and 8, may originate from the different numerical slight difference between UTCOMP and GPAS is due to the different
schemes in these two simulators as explained comprehensively regard- structures of these two reservoir simulators; the former is IMPEC
ing the results in Fig. 3. Nevertheless, this discrepancy is transient and and the latter is fully implicit.
at later times, e.g. after 4000 days, the results converge to each other. • Running the geomechanics module every N time steps (where N is
Furthermore, we tried to reduce the number of geomechanics calls greater than one) can significantly reduce the CPU time while
in order to reduce the total CPU time while maintaining the accuracy of maintaining the solution accuracy. The number of time steps per
the solution. For this purpose, we built a new model similar to the CO2 geomechanics calls, however, requires a rigorous criterion other
injection case except that the frequency of geomechanics calls was than a constant user-defined number.
decreased from every time step to every 10 time steps. This calling • Water flooding is more sensitive to the number of time steps per
frequency reduction significantly reduced the computational cost by a geomechanics calls than the gas flooding cases such as CO2 injection.
factor of 0.11, almost proportional to the inverse of the number of time
steps per geomechanics calls. Furthermore, a comparison for the SI Metric Conversion Factors
average reservoir pressure, oil production rate, and the cumulative
oil production from the investigated cases in Figs. 7 through Fig. 9
ft×3.048 E−01=m
shows that calling geomechanics module every 10 time steps does not
ft 3×2.831685 E−02=m3
influence the accuracy of the solution.
mD×9.869233 E−16 = m2
This CPU time reduction factor can be compared with that in the
primary production or water injection cases after normalizing that by day×8.640000 E+04=s
Fig. 7. Comparison of cumulative oil production in GPAS and UTCOMP with and without geomechanics module, and various calling frequencies of geomechanics module for the 3D
CO2 injection case.
288
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
Fig. 8. Comparison of oil production rate in GPAS and UTCOMP with and without geomechanics module, and various calling frequencies of geomechanics module for the 3D CO2
injection case.
Fig. 9. Comparison of average reservoir pressure in GPAS and UTCOMP with and without geomechanics module, and various calling frequencies of geomechanics module for the 3D
CO2 injection case.
Table 8
CPU time reduction using lower calling frequencies of the geomechanics module. The
denominator values in the far right column represent the ideal CPU time reduction factor
and equal to the inverse of the number of time steps between geomechanics calls. .
Acknowledgments
psi×6.894757 E+03=Pa
ksi×6.894757 E+06=Pa The authors acknowledge the participants of Reservoir Simulation
psi−1×1.450377 E−04=Pa−1 Joint Industry Project (RS-JIP) at the Center for Petroleum and
bbl/day ×1.840131 E−06=m3/s Geosystems Engineering at The University of Texas at Austin for their
STB/day ×1.840131 E−06=m3/s financial support of the current work. Drs. Jalil Varavei and Feng Pan
STB×1.589873 E−01=m3 are highly acknowledged for their collaboration during the early stages
of our work on GPAS and UTCOMP. Also, we appreciate Dr. Lee Y.
289
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
Chin for his comments on our work in UTCOMP and the derivation of Engineering are highly appreciated for their valuable and meticulous
the porosity equations. We express our deep gratitude to Dr. Emad W. comments to improve the technical and verbal content of this paper.
Al-Shalabi who critically reviewed the manuscript. Furthermore, the Moreover, Computer Modeling Group, LLC is recognized for providing
anonymous reviewers of the Journal of Petroleum Science and academic CMG license for comparison and verification purposes.
In this appendix, we attempt to demonstrate the detailed derivation of Eqs. (1) through (4) starting from the definition of the volumetric strain
according to the geomechanics sign convention for strains; compaction strains are positive. The differential volumetric strain can be expressed as
Eq. (A-1):
d ∀′b
dεv=− ,
∀′b (A-1)
where ∀′b represents the true bulk volume, and the minus sign on the right hand side is due to the strains sign convention. Integrating Eq. (A-1) from
the initial state (with zero volumetric strain and initial true bulk volume of ∀ 0b ) to an arbitrary state (at volumetric strain equal to εv and true bulk
volume of ∀*b ) leads to Eq. (A-2):
εv ∀*b d ∀′b
∫0 dεv= ∫∀ 0
−
∀′b
.
(A-2)
b
290
M. Haddad, K. Sepehrnoori Journal of Petroleum Science and Engineering 149 (2017) 281–291
References Terzaghi, K., 1936. Theoretical Soil Mechanics. John Wiley and Sons, Inc., New York,
USA, ISBN: 978-0471853053.
Tran, D., Settari, A., Nghiem, L., 2002. New iterative coupling between a reservoir
Bathe, K., 2001. The Inf-sup condition and its evaluation for mixed finite element simulator and a geomechanics module. Paper SPE/ISRM 78192 Presented at the
methods. Comput. Struct. 79 (2), 243–252. http://dx.doi.org/10.1016/S0045- SPE/ISRM Rock Mechanics Conference, Irving, Texas, USA. DOI: http://dx.doi.org/
7949(00)00123-1. 10.2118/78192-MS.
Biot, M.A., 1941. General theory of three-dimensional consolidation. J. Appl. Phys. 12 Wang, P., Yotov, I., Wheeler, M.F., Arbogast, T., Dawson, C., Parashar, M., Sepehrnoori,
(2), 155–164. http://dx.doi.org/10.1063/1.1712886. K., 1997. A new generation EOS compositional reservoir simulator: Part I –
Chang, Y., 1990. Development and Application of an Equation of State Compositional formulation and discretization. Paper SPE 37979 Presented at the SPE Reservoir
Simulator (Ph.D. dissertation). The University of Texas at Austin, Austin, Texas. Simulation Symposium, Dallas, Texas, USA. DOI: http://dx.doi.org/10.2118/37979-
Chin, L.Y., Raghavan, R., and Thomas, L.K. 1998. Fully-Coupled Geomechanics and MS.
Fluid-Flow Analysis of Wells with Stress-Dependent Permeability. Paper SPE 48857 Zienkiewicz, O.C., 1972. The Finite Element Method in Engineering Science. McGraw-
Presented at the SPE International Oil and Gas Conference and Exhibition, Beijing, Hill, New York, ISBN: 0070941386.
China. DOI: http://dx.doi.org/10.2118/48857-MS.
Chin, L.Y., Thomas, L.K., Sylte, J.E., Pierson, R.G., 2002. Iterative coupled analysis of Mahdi Haddad is a graduate research and teaching assistant and Ph.D. degree
geomechanics and fluid flow for rock compaction in reservoir simulation. Oil Gas Sci. candidate in petroleum engineering at The University of Texas at Austin. He has in-
Technol. – Rev. IFP 57 (5), 485–497. http://dx.doi.org/10.2516/ogst:2002032. depth knowledge and experience in developing of 3D, multi-stage hydraulic fracturing
David, C., Wong, T., Zhu, W., Zhang, J., 1994. Laboratory measurement of compaction- models in shale resources using fully coupled pore pressure–stress analyses, Cohesive
induced permeability change in porous rocks: implications for the generation and Zone Model (CZM), and eXtended Finite Element Method (XFEM). Moreover, he
maintenance of pore pressure excess in the crust. PAGEOPH 143 (1), 425–456. developed a wax deposition model in pipelines using a mesh-less method, Smoothed
http://dx.doi.org/10.1007/BF00874337. Particle Hydrodynamics (SPH) in his master's research. He holds two bachelor's degrees
GEM User Guide, 2015. Computer Modelling Group Ltd., Calgary, Alberta, Canada. in mechanical and petroleum engineering and a master's degree in mechanical
Ghasemi Doroh, M., 2012. Development and Application of a parallel Compositional engineering, Energy Conversion, all from Sharif University of Technology. Haddad's
Reservoir Simulator (MS thesis). The University of Texas at Austin, Austin, Texas. current research focuses on the simulation of natural fracture reactivation and micro-
Hughes, T.J.R., 1987. The Finite Element Method: Linear Static and Dynamic Finite seismicity during hydraulic fracturing including strong stress interactions of intersecting
Element Analysis. Prentice-Hall, Englewood Cliffs, NJ, ISBN: 0-13-317025-X. fractures using CZM and Material Point Method (MPM).
Pan, F., Sepehrnoori, K., Chin, L.Y., 2007. Development of a coupled geomechanics
model for a parallel compositional reservoir simulator. Paper SPE 109867 Presented
at the SPE Annual Technical Conference and Exhibition, Anaheim, California, USA. Kamy Sepehrnoori is W.A. ‘‘Monty’’ Moncrief Centennial Chair in the Department of
DOI: http://dx.doi.org/10.2118/109867-MS. Petroleum and Geosystems Engineering at The University of Texas at Austin. His
Pan, F., 2009. Development and Application of a Coupled Geomechanics Model for a research areas include computational methods, reservoir simulation development and
parallel Compositional Reservoir Simulator (Ph.D. dissertation). The University of application, enhanced-oil-recovery modeling, naturally fractured reservoirs, and uncon-
Texas at Austin, Austin, Texas. ventional resources. Sepehrnoori holds a Ph.D. degree in petroleum engineering from
Smith, I.M., Griffiths, D.V., 2004. Programming the Finite Element Method. John Wiley The University of Texas at Austin.
& Sons, Ltd, Chichester, England, ISBN: 0-470-84969-X.
291