1 s2.0 S0920410521009128 Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Journal of Petroleum Science and Engineering 208 (2022) 109257

Contents lists available at ScienceDirect

Journal of Petroleum Science and Engineering


journal homepage: www.elsevier.com/locate/petrol

Coupled optimization of carbon dioxide sequestration and CO2 enhanced


oil recovery
Shahrokh Bahrami Kashkooli a, Asghar Gandomkar a, *, Masoud Riazi b, **, M. Sadegh Tavallali a, c
a
Department of Chemical Engineering, Shiraz Branch, Islamic Azad University, Shiraz, Iran
b
Enhanced Oil Recovery (EOR) Research Centre, IOR/EOR Research Institute, Shiraz University, Shiraz, Iran
c
Department of Advanced Calculations, Chemical, Petroleum & Polymer Engineering Research Center, Shiraz Branch, Islamic Azad University, Shiraz, Iran

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

Keywords: CO2 injection into hydrocarbon reservoirs can simultaneously pave the path for two important processes, namely
CO2 sequestration oil production increase, and CO2 storage. This combined optimization problem can be referred to as carbon
Gas storage efficiency capture and storage - enhanced oil recovery (CCS-EOR) problem. CCS-EOR is a challenging task with many
Enhanced oil recovery
opportunities which should be studied in a timely manner. Moreover, the attitude of the management team
Optimization
Compositional simulation
towards prioritizing CCS or EOR should be considered. Therefore, in the current study, CCS-EOR problem was
addressed by utilizing the dynamic well flow settings (bottom-hole pressure and/or flow rate) as the optimization
variables with a weighted objective function. This approach was tested on a case study. Its huge simulation time
due to high dimensionality remained a serious challenge. One effective approach was to use the underlying
applied knowledge of the system and extract that wisdom to reduce the simulation time in a rational manner.
CO2 breakthrough time was such a knowledge. Therefore, the importance of CO2 breakthrough time was
stressed, and the understanding of the response of the simulator to this event was used to considerably reduce the
simulation time in a black box approach. Moreover, a concise sensitivity analysis was performed which helped to
schedule the workflow based on the given priority to each process. The optimized case could improve the
objective value and gas storage efficiency up to 2.86 % and 12.1 % respectively. The obtained results can be used
to analyze a variety of operational matters such as the CO2 distribution map in the reservoir, and the well
settings.

1. Introduction fossil fuels are continuously consumed as energy source, an option for
reducing CO2 is the carbon capture and storage (CCS) technology
According to the International Energy Agency (IEA, 2015) , the (Gibbins and Chalmers, 2008) by capturing, separating and diverting it
world energy demand will increase by 37 % from its current level by into geological formations (Bachu, 2003).
2040. A major portion of the total energy will continue to depend on Geological storage of CO2 is achieved through a combination of
fossil fuels (oil, gas and coal). Some statistics predict that the worldwide physical and chemical trapping mechanisms that are effective over
oil demand will increase by 20 % by 2040 (Nilsson, 2016). In line with different timeframes and scales (He et al., 2016). Options for geologic
these predictions, the forecasts of an international panel on climate storage include deep saline aquifers, depleted oil and gas reservoirs and
change show that the carbon dioxide (CO2) concentration will reach inaccessible coal seams (Gibbins and Chalmers, 2008; Wang et al., 2013;
570 ppm in 2100, which can lead to an average global temperature in­ Kim et al., 2017). Depleted oil and gas reservoirs are prime candidates
crease of 1.9 ◦ C and this greenhouse gas emission will cause many seen for CO2 storage in the existing oil production facilities (He et al., 2016;
and unseen problems for humans through climate change. Mitigating Barrufet et al., 2010). Despite the negative role of CO2 in climate change,
the drawbacks of these events has attracted the attention of the global CO2 is an attractive displacing agent for enhanced oil recovery (EOR)
public opinion (Safarzadeh and Motahhari, 2014; Asghari and Al-Dliwe, due to its relatively low minimum miscibility pressure (MMP) in a wide
2005; Le Gallo et al., 2002; Javaheri et al., 2009; Yang et al., 2012). As range of crude oils. Since a large portion of the injected CO2 remains in

* Corresponding author.
** Corresponding author.
E-mail addresses: agandomkar@shirazu.ac.ir (A. Gandomkar), mriazi@shirazu.ac.ir (M. Riazi).

https://doi.org/10.1016/j.petrol.2021.109257
Received 10 July 2020; Received in revised form 15 July 2021; Accepted 18 July 2021
Available online 21 July 2021
0920-4105/© 2021 Elsevier B.V. All rights reserved.
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

place in depleted reservoirs after CO2-EOR, it is a low cost option to optimizer. Similarly, Ampomah et al. (2017a,2017b) used a composi­
permanently sequester CO2 (Dai et al., 2017; Budinis et al., 2018; tional reservoir model and ANN optimization process and considered the
Allinson et al., 2017). In fact, incremental profitability of such projects geological uncertainty in the Farnsworth Unit oil field. They investi­
due to increased oil production helps to offset the cost of CO2 seques­ gated the effects of bottom hole injection pressure, injection and pro­
tration (Bachu, 2003). Governments also can use financial incentives duction rates, and gas-oil ratio on the objective function. Le Van and
(for instance tax rating or waiver) to motivate such projects (Jiang et al., Chon (2017) used a 3D reservoir model and ANN optimization process to
2019). Optimization can be a key tool for simultaneously utilizing the optimize the injection designs. In addition, Li et al. (2016) applied
reservoir storage capacity for CO2 sequestration and CO2 driving force D-optimal algorithm of design-expert to investigate the role of injection
for efficient oil production. Similar to other optimal petroleum exploi­ models. Moreover, Dai et al. (2016) used a response-surface based
tation problems (Tavallali et al., 2016) such as optimal well placement, economic model in another study. The same approach of using response
drilling planning (Tavallali et al., 2013, 2018), and production and/or surfaces (in the intermediate steps) can be seen in the work of Wang
injection throughput optimization (Tavallali et al., 2014), the combined et al. (2018) where CO2 injection path was the target of optimization.
CCS-EOR optimization study is indeed a challenging task; to be fair, Additionally, Guo et al. (2020) used an economic model to optimize the
maybe one of the most complex ones. This problem is inherently CO2 capture/injection path. Using multi-period approach (instead of
extremely nonlinear (due to the conservation and constitutive equations continuous time domain) as well as parallel computing are the other
involved), with many known and unknown intertwined interactions strategies to reduce the simulation burden and are reported in the works
between the many players in the problem (including the injected gas, of Kwak and Kim (2017) and Min et al. (2018).
reservoir formation, wells and their patterns). Therefore, this problem is As can be seen, mathematical programming based studies provide
mathematically nonconvex and finding the best solution is not a simple handy tools for general decisions, however they lack the required and
task at all. Finally, the huge simulation expenses due to the curse of detailed scheduling information. Moreover, the black box approaches
dimensionality in such problems should not be forgotten (Tavallali et al., have mainly focused on proxies to overcome the problem of high
2018). Given this situation, even attaining a sub-optimal solution that dimensionality; this is a practical approach; however, this approach also
can outperform the expert solution seems to be useful for the related eliminates some of the information and stays on the general level. To the
industries. best knowledge of the authors and based on the previous discussions,
Apart from the experimental and simulation analysis studies on none of the above optimization studies have used the invaluable infor­
combined CCS-EOR (Farajzadeh et al., 2020; Rezk et al., 2019; Alfi and mation of CO2 breakthrough event to address the challenge of high
Hosseini, 2016; Han et al., 2016; Wang et al., 2019; Momeni et al., 2012; dimensionality. This can be a promising tool in achieving a practical
Jia et al., 2018), optimization studies have also emerged. One should solution to support the management team with the required operational
notice that the target of optimization is basically different from exper­ settings; and that is done in the current research. Of course, due to the
imental and simulation studies, and the concept of degree of freedom nature of this complicated problem, even a near optimal solution seems
clearly supports that. Based on the optimization techniques used (and to be an invaluable solution for industry.
probably the level of decision making), these optimization studies can be Therefore, in the current study, CO2 breakthrough (Yuncong et al.,
roughly categorized into two groups: those within mathematical pro­ 2014; Khan et al., 2013) event is used as the prime tool to reduce the
gramming domain and the ones with simulation-based black box total optimization time for the combined CCS-EOR process in one of the
approach. There are fewer contributions with the first approach. For Iranian carbonate oil reservoirs during miscible CO2 injection. This is
instance, Tapia et al. (2015) developed a mixed integer linear program addressed with a compositional simulator linked with a black box
(MILP) using the analogy of the strip packing problem and studied the approach and/with gradient free/heuristic algorithms. Finally, one of
optimal CO2 allocation and scheduling in EOR operations with another the prime factors on the CCS-EOR projects is the attitude of the man­
MILP model. Also, Tapia et al. (2016) developed a discrete-time opti­ agement team towards giving priority to CCS or EOR. Therefore, to study
mization approach to consider both economic discounting and reservoir the effect of that, a sensitivity analysis is performed in the current study
depletion. Such linear or linearization approaches are effective for top and its results can be used to develop the corresponding work flow.
level decision making and are usually not that suitable for scheduling
day-to-day operations. In contrast, there are more contributions with the 2. Methodology
second (black box) doctrine; mainly due to the versatility of using sim­
ulators (Safarzadeh and Motahhari, 2014). In such studies, many oper­ 2.1. Problem definition
ational variables (usually linked with economic analysis) such as
injection patterns, rates, paths, supply limits, bottom hole pressure, gas Using the smart well technology, the oil field can be simultaneously
oil ratio (GOR), and CO2 leakage are of concern. Hence, they can be used defined under EOR and CO2 storage operations. Hence, this process for a
for scheduling purposes (at least theoretically in the long term problems), fixed long time frame (0 ≤ t ≤ T) in petroleum field should be opti­
of course with significant computational expenses. The majority of these mized. Here, the problem can be summarized as follows:
studies used proxy models (for instance response-surface method, design Given:
expert, regression and artificial neural network (ANN) which are useful
for reservoir characterization, history matching and uncertainty quan­ • A complete list of petro-physical data for both the fluids and reser­
tification to select the most determining variables for the oil production voir as well as the initial state of the reservoir.
and CO2 storage, and also reduce the number of control variables. To be • The structure, position, geometry and role (it means injection or
specific, Chen and Pawar (2019) employed a machine learning tech­ production) of each well.
nique to prepare empirical models (which can bypass the full scale • The geometry and topology of the reservoir.
simulation requirement) and used it with Stochastic Simplex Approxi­
mate Gradient algorithm to investigate the effects of injection patterns, Assumptions:
rates, and well spacing on the CO2 sequestration and oil production. In
another study, Dudek et al. (2020) combined the compositional simu­ • The pressure drop caused by the skin around the well is neglected.
lator with genetic algorithm to optimize well controls and inflow control • The injection gas is pure CO2.
variables (ICV). In addition, Goodarzi et al. (2015) used a numerical • Dissolved CO2 may exist in the produced oil.
method and an optimization algorithm to maximize the storage capacity • Sufficient amount of CO2 is (gathered from the nearby industry and)
by considering the thermal effect. You et al. (2020) also combined a supplied to the injection point. Maximum oil production and CO2
compositional numerical simulator with ANN and multi-objective

2
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

injection rates are limited by the capacity of the surface facilities.


Rfu = fl − fv (2)
Both limits are given.
The reservoir simulators ensure that Rfu and Rfl are as close as
Policies: possible to zero (Systems, 2010). This is conducted after spatiotempo­
rally discretizing Equations (1) and (2).
• Both miscible and immiscible injections are possible. In addition to Equations (1) and (2), the following equations are also
• The producer well is shut-in in case of violating the upper limits on defined. In these equations, the bottom-hole pressure (BHPtn ) of the wells
the specific water-cut, gas-oil ratio or the lower limits on the oil are defined as control variables, therefore, the following bounds are
production rate. defined for each well:
• The comparative tendency of the management team towards EOR or
CO2 sequestration is clear and quantifiable. BHPtn ≤ BHPU,t
n n ∈ injectors (3)

Determine: BHPtn ≥ BHPL,t


n n ∈ producers (4)

• Bottom-hole pressure (BHP) and/or flow rate vs. time profiles for where n shows perforated cell (well), t is the index used for time, and U
each well. and L superscripts show upper and lower bounds, respectively. Addi­
tionally, bounds for gas injection and oil production flow rates (Qtn ) are
Targeting to simultaneously maximize the overall oil production and also imposed as:
CO2 storage in the aforementioned project. Qtn ≤ QU,t n ∈ injectors ∪ producers (5)
n

In this study, time invariant upper and lower bounds on the above
2.2. Model description
flow equation are used; however, time dependent values could be
replaced at the cost of increased solution time. Constraints (3–5) are
The reservoir dynamics is represented by a set of partial differential
defined by considering the pros and cons of economic and technical
equations. They are described in many standard references (Ertekin
limitations. For instance, they impose bounds based on the limitations of
et al., 2001; Rios et al., 2018; Aziz, 1979). In a concise form, that can be
the wellhead facilities and/or prevent the damage to both underground
represented as follows (Systems, 2010):
formation and surface facilities. The cumulative oil production (Np ) and
dM cumulative volume of stored CO2 (vCCS ) are important indicators for
Rfl = +F+Q (1)
dt evaluation of the project performance:
∑∑
where, Rfl is a nonlinear residual defined for each fluid, M is the mass per Np = Qtn n ∈ producers (6)
unit surface density at time t, F is the net flow rate into the adjacent cells, n t

and finally Q is the net flow rate into the wells. Since compositional case ( )
∑ ∑ ∑
is at the focus, minimum Gibbs energy of the system should also be vCCS = QtCO2 ,n − QtCO2 ,n′ (7)
imposed, hence the equality of fugacity conditions (fl = fv ) becomes a t n∈injectors ′
n ∈producers
complementary equation for two hydrocarbon phases as below:
The last summation includes the CO2 amount in both gas and oil
phases.
Finally, the following objective function is defined to couple both
purposes of EOR and CO2 storage:
( )
Np vCCS
maxf = w1 + w2 × 100 (8)
OOIP MCCS

where OOIP is the original oil in place, MCCS is the estimated total CO2
injection into the reservoir, and finally w1 and w2 (0 ≤ w1 , w2 ≤ 1) are
weights determined based on the tendency of the management team
towards EOR or CO2 storage. For easier discussion, let us refer to the first
and second fractions in Equation (8) as FOE (field oil production effi­
ciency) and GSE (gas storage efficiency) respectively. Equations (1)–(8)
form our model. Hereafter, we refer it as model EC representing the
EOR-CCS problem.
Solution Strategy: The aforementioned optimization model EC is a
typical nonlinear, nonconvex, non-smooth, dynamic partial differential
equation (PDE) constrained problem. Therefore, in order to solve it
efficiently, its solution strategy can be implemented into two sets of
successive stages: forecasting the production and accordingly updating
the BHPs. A compositional commercial numerical flow simulator
(ECLIPSE 300, version 2018) is used for the former and a serial opti­
mization algorithm employed for the latter. Heuristics and gradient free
search algorithms are arguably the best candidate for such nonlinear and
non-smooth problems. The nonlinear nature of this problem renders that
as a non-convex problem. Therefore, a serial combination of genetic
algorithm (GA) and pattern search (PS) algorithm were used to avoid
trapping in local minima. The former method is implemented using ga
keyword in MATLAB and the latter via patternsearch with default
settings.
Fig. 1. Schematics of solution approach.

3
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

Table 2
Composition of the reservoir fluid.
Components Composition %

N2 0.13
CO2 0
H2 S 1.65
CH4 41.15
C2 H6 9.06
C3 H8 6.53
C4 H10 4.75
C5 H12 3.55

Fig. 2. Top view of the OOIP, the reservoir geometry and the location of C6 H14 3.02
the wells. C7 H16 3.20
C8 8.86
C11+ 18.10
Table 1
Initial reservoir conditions and rock and fluid properties.
Property Value Property Value

Initial reservoir pressure 4249 Psi Formation 250 ft


thickness
Bubble point pressure 2402 Psi Water-oil contact 8005 ft
Reservoir temperature 210 ℉ Formation water 237,000 ppm
salinity
Oil viscosity (reservoir 1.1 cP w 1 , w2 0.5
conditions)
Formation water viscosity 0.66 cP BHPU,t
n∈injectors
9000 psi
Average porosity 13 % BHPL,t 800 psi
n∈producers
Average permeability 4.1 mD QU,t
n∈injectors
4000 MSCF/
Day
Max water cut 93 % QU,t 5000 STB/
n∈producers
Day
Max GOR 3.5 MSCF/ QL,t
n∈producers
200 STB/Day
bbl

GA which is a population based method (Tavallali et al., 2018), ex­


plores a wide range of feasible region of solutions without strong Fig. 3. The water-oil relative permeability and the capillary pressure.
dependence on the initial guess. Once GA is terminated, PS (Edgar et al.,
2001) is initiated with the best GA final solution and it can effectively
search local and fine zones. PS uses a template around each solution
point to evaluate the objective function in its neighborhood. Then the
template is transferred to the neighboring solution with the best current
objective function. The same approach is iterated until the size of the
neighborhood region becomes less than a default tolerance value. The
reader can refer to Edgar et al. (2001) for more information about the
details of these widely used optimization algorithms.
Implementation: To solve this problem the entire time horizon is
discretized into τ periods and for each period the appropriate bounds on
the BHP are determined by the optimizer. Flow rate bounds could also
be determined; yet due to the curse of dimensionality only BHP is used in
this study. These bounds are then transferred to the reservoir simulator
to set the BHP and solve Equations (1) and (2). This is notable that due to
the programming nature of this simulator, the software attempts to push
Fig. 4. The gas-oil relative permeability and the capillary pressure.
the BHPs to their given bounds. At any point in time that the software
fails on that, it changes the control from BHP to flow rate (or vice versa
readers can refer to the study of (Tavallali et al., 2018), for more in­
depending on the initial user’s choice). Therefore, the simulator can
formation on the programming details of such simulator-optimizer
provide a time dependent BHP profile based on the bounds provided by
linking approach.
the optimizer.
The details of this successive approach are revealed in Fig. 1. Here,
3. Case study
the script-based nature of ECLIPSE is utilized to facilitate the data
transfer to and from the optimizer in the MATLAB software. Basically,
To test EC model and the aforementioned solution strategy, both are
MATLAB is used to write the BHP values set by GA or PS into ECLIPSE
implemented on a real 3- dimensional and 3-phase heterogeneous car­
keywords for controlling the well operations in script files and feeding to
bonate oil reservoir with 32 × 28 × 12 anisotropic grid blocks with in
ECLIPSE. Then, system keyword in MATLAB, executes these ECLIPSE
homogeneous sizes. This reservoir is located in the south of Iran, and it is
files and their rsm output files are interpreted into objective values.
under CO2 injection (the required data has been obtained from the South
These values are then automatically judged and used by PS and/or GA to
Zagros Oil and Gas Production Company (one of the subsidiaries of
either prepare the next set of BHPs or terminate the iterations. Interested

4
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

Fig. 5. (a) The kx histogram of the entire reservoir and (b) the top view of kx in the first layer.

Iranian central oil fields company)). Three producing (P1–P3) and two 4. Results and discussion
injection (I1–I2) wells exist in this reservoir (Fig. 2) and their perforation
spans the entire reservoir depth. Fig. 2 shows the 5-spot pattern which Initial screening: Fig. 7 reveals the field gas oil ratio (FGOR) as well
are considered for optimizing both CO2 sequestration and EOR pro­ as the CO2 fraction in the produced gas and liquid of the base scenario.
cesses. The production is modeled via a compositional simulator. The As can be seen, the CO2 breakthrough occurs on the 2192nd production
initial conditions and reservoir model parameters are provided in day (tbCO2 ). Obviously, the produced gas is only hydrocarbon before this
Table 1 and the composition of the reservoir fluid is given in Table 2. The date. Since initially it is assumed that the management team is neutral to
water-oil and gas-oil relative permeabilities, as well as the relevant prioritizing the EOR or CO2 storage (w1 = w2 = 0.5), this breakthrough
capillary pressures of the reservoir rock, are reported by Fig. 3 and is a pivotal event. Before the gas breakthrough ( t < tbCO2 ), the back­
Fig. 4, respectively. As can be seen from Fig. 3, the reservoir rock has an
ground settings of the reservoir simulator enforces the upper limit on the
intermediate wettability. That is, the location of the junction of the
oil production; and to support this, the upper limit on the gas injection
relative permeability diagrams for water and oil is about 0.5. Moreover,
also remains active. During this period, CO2 storage efficiency is 100 %
the histogram of the x-permeability and porosity field are reported in
since breakthrough has not yet happened, and simultaneously oil pro­
Fig. 5a and Fig. 6a, respectively, and Figs. 5b and 6b show their top view
duction is (locally) maximized. This means that even without using
in the first layer. As can be seen from Figs. 5 and 6, the maximum fre­
external optimizer, the internal reservoir simulator has acted ideally.
quency of permeability and porosity are almost 3.75 md and 0.15,
This fact helped to narrow down the optimization time window from 0 ≤
respectively. Moreover, the region between P1, P2 and I1 is a highly
t ≤ T to tbCO2 ≤ t ≤ T. To do so, the results of the initial simulation for
permeable and porous zone. The EOR and CO2 storage in the above­
mentioned reservoir is to be planned for a 20 years’ time frame by 0 ≤ t ≤ tbCO2 is stored and used as the initial settings of the rest of sim­
providing the BHP (and subsequently) flow rate profiles. For a fair ulations. Since this restart feature of the reservoir simulator is employed,
comparison, a base scenario is defined and its objective value is the objective function could be calculated for the entire time (0 ≤ t ≤ T)
compared with those of the optimized scenarios. The objective function without repeatedly simulating 0 ≤ t ≤ tbCO2 for each simulator call back.
is consisted of two separate EOR and CO2 storage terms, and in some More description is useful to provide more accurate insight to the
cases, they require contradicting operational decisions; therefore, effect of the above action. In the initial study, a simpler case with 3
defining a robust base scenario is not a simple task. The general sense is producers and 1 injector (instead of 2 injectors) was introduced to the
that, the lower/higher BHP for producers/injectors can lead to higher optimization workflow. It took 5.1 × 105 (s), 4.3 × 104 (s) and 4.9 × 104
production. Therefore, this idea is used for defining the base scenario. (s) in order for the GA, PS and GA + PS to terminate the search. Despite
Additionally, since the attitude of the management team affects the this long simulation time for a simpler case the obtained results are not
final solution through the numerical weights of w1 and w2 , a concise satisfactory. Obviously, simulation time for a more difficult case would
sensitivity analysis on these parameters are performed. In what follows, be even higher. However, after shifting the optimization time to tbCO2 ≤
the results for w1 = w2 = 0.5 are presented, unless otherwise stated. t ≤ T and implementing the restart feature for a more difficult case with

5
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

Fig. 6. (a) The porosity histogram of the entire reservoir and (b) the top view of porosity in the first layer.

3 producers and 2 injectors (instead of 1 injector), the final solutions Comparison: The objective values for the various scenarios with
seems to be satisfactory. As can be seen in Table 3, the simulation time different combinations of the optimizer and the sources of initial guesses
has decreased by 39 % for the GA, 69.7 % for the PS, and 67.3 % for the are reported in Table 3. As can be seen, the GA has successfully increased
GA + PS. The GA is a heuristic method, and usually 50 individuals (each the objective value for 2.86 %. Moreover, by shifting the simulation time
set of values for input variables is an individual) are generated in each to the breakthrough event, the simulation time could be reduced by
population by MATLAB and that means 50 simulations should happen to 39.2% (the GA) and 67.3 % (the PS). In this optimized case the BHPs of
find the best individuals before starting the next round. Creating new P1 to P3 and I2 are in order of 800, 800, 3052 and 8149 psi, and I1 is
populations requires even more simulation. However, the grid search closed, whereas the BHPs of P1, P2, I1 and I2 in the base case were 800,
algorithm can rely on almost 4 new simulations in each iteration. Hence, 800, 9000 and 9000 psi, respectively and P3 was closed. Analyzing the
it seems natural to see the simulation time as GA > GA + PS > PS, and data in Table 3 is very informative. The GA has compromised on FOE
the same was observed here. ratio (reduced from 0.310 to 0.238); however, it could increase the GSE
Finally, one should notice that curse of dimensionality limits the from 0.870 (of the base) to 0.975. This can be further explained once the
definition of BHP for each small time steps. Therefore, in this study, we dynamics of the reservoir is analyzed using Fig. 8 to Fig. 14. Note that
decided to take a more realistic approach from computational point of the dynamics of the reservoir fluids depends on both the well through­
view and define the bounds on BHPs from the very beginning of pro­ puts and the topological and geological structure of the reservoir. All
duction life. It is clear that the shorter the time frame, the more effective these factors should be checked before comprehending an optimal or
the bound settings. near optimal solution. Higher FOE is due to higher production, which is
To validate the results, firstly, history matching is performed on the in turn achieved once the reservoir pressure is increased by increased
dynamic model and after that all scenarios are considered. This action injection. This consequently yields a highly distributed CO2 front in the
ensures us that the model parameters are correct. Secondly, as previ­ reservoir which has moved towards the producers. If not controlled
ously mentioned, no external optimizer is used before breakthrough. properly, it can cause high CO2 production, and waste all the measures
However, due to internal settings of ECLIPSE, obtaining the maximum taken for storing CO2 gas. This happened in the base case and the same is
oil production is the first target. In order to increase the reservoir prevented in the optimized scenario. In contrast to the base scenario
pressure, maximum gas is injected while no CO2 is produced, hence both (where P3 was shut-in about the 5900th day due to violating the GOR
FOE and GSE are at their maximum values. If another objective function limit, 3.5 MSCF/bbl), I1 is closed from the very beginning in the opti­
is used, there could be some doubt in the (local) optimality. With the mized scenario. As shown in Figs. 5b and 6b the area enclosed between
current simulator settings and the current objective function, this seems P1, P2 and I1 is a highly permeable and porous region. Therefore, it is
logical to state that (local or near local) optimum is reached. obvious that fully active injection by I1 in this region could facilitate the

6
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

Fig. 7. The analysis of the base scenario: (a) Field gas oil ratio, and CO2 gas fraction in (b) produced gas and (c) produced liquid.

Table 3
The objective values (w1 = w2 = 0.5).
Optimizer Initial guess source Elapsed time (s) FOE GSE GSE improvements Objective value Objective value improvements

None (Base) – 49.4 0.310 0.870 – 58.995 –


PS Base 1.3 × 104 0.276 0.915 5.2 % 59.540 0.93 %
GA – 3.1 × 105 0.238 0.975 12.1 % 60.680 2.86 %
PS GA 1.6 × 104 0.238 0.975 12.1 % 60.680 2.86 %

advance of CO2 front towards P1 and P2. Since breakthrough had forget the positive side of high CO2 injection on increasing the reservoir
already happened, this phenomenon could lead to even more rapid in­ pressure and enhancing the oil production (FOE of 0.310). However,
crease of CO2 production, without fully utilizing the sequestration ca­ even this had led to water production increase. In contrast, in the opti­
pacity (obviously reducing the GSE). This is exactly what has happened mized case, another approach has prevailed. I1 is closed from the very
in the base case with GSE of 0.87 (with 58.4 BSCF injection, 50.81 BSCF beginning and the injection load is transferred to I2 in the region with
stored and 7.59 BSCF produced, Fig. 10). Of course, one should not much lower permeability. This obviously delays the increase of CO2

7
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

Fig. 8. CO2 distribution in the reservoir at the end of production window for the (a) base and (b) optimized scenarios.

Fig. 11. The average field reservoir pressure.

also prevents the CO2 rich region penetrating the producer zones
(Fig. 8). This approach appears effective in increasing GSE and achieved
Fig. 9. CO2 throughput for both injection and production. a final value of 0.975 (with 38.00 BSCF injection, 37.06 BSCF stored and
0.93 BSCF produced, Fig. 10). Of course, the price for this, is the lower
FOE 0.238.
In the base scenario, P3 was closed around 5971th production day,
then, I1 and I2 continuously injected with constant flow rate. Therefore,
the reservoir pressure started to increase. However, in the optimal sce­
nario, I1 is closed and then the reservoir pressure accordingly decreased
compared to the base scenario. Examining Figs. 8 and 9 proves that
although less CO2 volume is stored in the optimized scenario, less CO2
volume is also injected and produced (Fig. 10); and it yields higher GSE.
In fact, the data revealed in Fig. 12 confirm that this optimization is
reduced the CO2 fraction in both produced gas and liquid. The CO2
capillary trapping could prevent the formation of CO2 rich zones in the
vicinity of the producer wells and creates a more uniform distribution of
CO2 in the reservoir. In addition, the gas phase continuity decreases and
also the availability of CO2 reduces in the vicinity of the producer wells.
However, it affects the CO2 mobility (reduces the CO2 fraction in pro­
Fig. 10. Cumulative CO2 injection and production. ducing free gas) and the oil swelling (reduce the CO2 mass transfer to
oil). Therefore, the suggested optimization approach helps to store more
production and provides more contact time in the reservoir. Moreover, CO2 in the reservoir, and the GSE is increased. Additionally, as can be
the BHP of P3 is increased in this case. The former action reduces the seen from Fig. 13 the water production rate has also declined in the
reservoir pressure, whereas the latter one saves the same (Fig. 11), and optimized case. These results strongly show that the idea of “the more

8
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

Fig. 15. The effect of different values of w1 = 0.3, 0.5, ​ and ​ 0.7 on FOE, GSE,
and objective function, Sensitivity analysis.

injection, the better” should be modified by the management team.


Hence, depending on the objective of the project (for instance increasing
FOE, GSE, profit or all) the appropriate injection plan may differ, and
therefore the idea of the higher injection, the better should be revisited
based on the objective defined. The variations in the profile of produced
CO2 in the base case (Fig. 9) are due to the shut-in of P3 based on the
GOR level violation. The effect of this shut-in can be seen in other figures
too (Fig. 11).
Finally, it is useful to mention that since the average reservoir
pressure has remained higher than the bubble point pressure (Table 1
Fig. 12. CO2 fraction in the (a) produced gas and (b) produced liquid. and Fig. 11); the reservoir has remained at the under-saturated condi­
tion. Also, the equilibrium interfacial tensions (IFT) between the oil and
pure CO2 are measured at 210 ◦ F and pressures varied from 1500 to
4000 psia. As it is reported in Fig. 14, the IFT between pure CO2 and oil is
37 and 11 dyne/cm for 2500 and 3500 psia respectively at reservoir
temperature. However, from Fig. 14, the MMP of the oil and pure CO2 is
calculated by vanishing interfacial tension (VIT) method based on the
IFT data (Azizkhani and Gandomkar, 2019). The MMP is 3985 psia at
reservoir conditions. Consequently, the increase of GOR is due to the
injected CO2 and not the free gas production.
Sensitivity analysis for w1 &w2 : The comparative values of FOE,
GSE and the objective value (f/100) are shown in Fig. 15. These results
are obtained by initiating the pattern search algorithm by the previous
base scenario. As can be seen from this figure, w1 value directly/
inversely affects the FOE/GSE; however, the same figure reveals that
GSE has more dominant effect on the final objective value. In another
words, in such under-saturated reservoir, with miscible CO2 flooding
Fig. 13. Field water production rate.
where the gas breakthrough will happen sometime later in the produc­
tion cycle of the reservoir, the integral management of gas injection, gas
production and consequently gas storage have a prime role in the effi­
cient lifetime management of the system; and this is not separate from
the oil production planning. To clarify that, details of Equation (8)
should be reviewed. Firstly, since w1 + w2 = 1, hence once w1 is
increased its effect is to favor more oil production and that yields
increased FOE. Simultaneously w2 decreases and that inhibits the GSE
increase.
Secondly, the dominancy of GSE can be due the numerical order of
Np
OOIP compared to MvCCS
CCS
. An interpretation in this case study can be that the
amount of unexploited petroleum is indeed substantial. One should
notice that, the optimization solution is a strong function of the selected
objective function, and the solution for one objective function cannot be
interpreted for another. This intertwined feature highlights the impor­
tant and irreversible role of process optimization in this problem and
calls for more in-depth interdisciplinary studies. Surely active collabo­
Fig. 14. Minimum miscibility pressure measurements based on the vanishing ration of researchers from process system engineering background with
interfacial tension technique, MMP = 3985 psia. petroleum engineers can significantly provide new opportunities. This
trend has already started and is gaining more and more attractions (Khor

9
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

et al., 2017; Tavallali et al., 2016). 8) w1 value directly/inversely affects the FOE/GSE.
9) The GSE has more dominant effect on the final objective function
5. Conclusions value. In another words, in such under-saturated reservoir, with
miscible CO2 flooding, where the gas breakthrough will happen
The enhanced oil recovery and CO2 storage are both complex pro­ sometime later in the production life of the reservoir, the integral
cesses which are intrinsically nonlinear, non-convex and come with management of gas injection, gas production and consequently
curse of dimensionality. This situation creates a dilemma for using gas storage have a prime role in the efficient lifetime manage­
gradient free and heuristic methods. In this study, a weighed combined ment of the system.
objective function was defined to simultaneously consider both pro­ 10) When the management team is more interested in EOR and in­
cesses. It was shown that once the optimization is used with an under­ creases its weight factor, it enhances the FOE; however, it de­
standing of the process, this problem can be alleviated to some good teriorates the GSE and the objective function.
extend. To be specific, since a commercial reservoir simulator was used
here, it was shown that in an under-saturated oil reservoir with CO2 Credit author statement
flooding, the main benefit of optimization appears after the CO2
breakthrough. Before this event, all injected CO2 is stored (100 % stor­ Shahrokh Bahrami Kashkooli: Conceptualization, Methodology,
age efficiency) and also the heuristics defined for the simulator enforce Validation, Investigation, Resources, Writing – original draft, Writing –
maximum oil production. Hence this part can be left to the simulator review & editing, Visualization, Supervision, Project administration.
itself, which reduces the overall run time. The complementary optimi­ Asghar Gandomkar: Conceptualization, Methodology, Validation,
zation can start just after CO2 breakthrough. We also conclude the Investigation, Resources, Writing – original draft, Writing – review &
following results from this study: editing, Visualization, Supervision, Project administration. Masoud
Riazi: Methodology, Investigation, Writing – original draft, Writing –
1) After shifting the optimization time to tbCO2 ≤ t ≤ T and imple­ review & editing. M. Sadegh Tavallali: Conceptualization, Methodology,
menting the restart feature for a more difficult case, the final Validation, Investigation, Resources, Writing – original draft, Writing –
solutions were satisfactory and the simulation time has decreased review & editing, Visualization, Supervision, Project administration.
by 39 % for GA, 69.7 % for PS, and 67.3 % for GA + PS.
2) The GA has successfully increased the objective value to 2.86 %.
3) The optimization increased the GSE and also, controlled the CO2 Declaration of competing interest
injection level.
4) The optimization reduced the CO2 fraction in both produced gas The authors declare that they have no known competing financial
and liquid. interests or personal relationships that could have appeared to influence
5) The water production rate has also declined in the optimized the work reported in this paper.
case.
6) This approach could prevent the formation of CO2 rich zones in Acknowledgments
the vicinity of the producer wells and create a more uniform
distribution of CO2 in the reservoir. The authors would like to express their appreciation for the support
7) The results from this study indicate that the idea of “the more received from the South Zagros Oil and Gas Production Company (one of
injection, the better” should be redefined by the management the subsidiaries of Iranian central oil fields company) for kindly
team. providing the reservoir data.

Nomenclature

EOR Enhanced oil recovery


MMP Minimum miscibility pressure
ICV Inflow control variable
IFT Interfacial tension
VIT Vanishing interfacial tension
CCS Carbon capture and storage
CCS-EOR Carbon capture and storage-enhanced oil recovery
CO2-EOR CO2 enhanced oil recovery
GOR Gas oil ratio
FGOR Field gas oil ratio
FWPR Field water production rate
FPR Field average pressure
BSCF Billion standard cubic feet
OOIP Original oil in place
EYMF Field CO2 mole fraction in the produced gas
FXMF Field CO2 mole fraction in the produced liquid
FOE Field oil production efficiency
GSE Gas storage efficiency
GA Genetic algorithm
PS Pattern search algorithm
ANN Artificial neural network
MILP Mixed integer linear program
PDE Partial differential equation

10
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

vCCS Cumulative volume of stored CO2


Np Cumulative oil production
MCCS Estimated total CO2 injection into the reservoir
T Production time
t time
P Producer wells
I Injector wells
kro Oil relative permeability
krg Gas relative permeability
krw Water relative permeability
pc Capillary pressure
kx Absolute permeability in x direction
ϕ Porosity
Rfl The nonlinear residual defined for each fluid
M The mass per unit surface density
F The net flow rate into the adjacent cell
Q Flow rate
w1 &w2 Weights of FOE and GSE respectively
f Objective function
fl Liquid fugacity
fv Gas fugacity
QU,t
n∈injectors Injection flow rate (upper bound)

QU,t
n∈producers Production flow rate (upper bound)

QL,t
n∈producers Production flow rate (lower bound)
tbCO2 CO2 breakthrough time
BHPtn Bottom hole pressure
BHPU,tn Upper bound of bottom hole pressure of injector wells
BHPL,tn Lower bound of bottom hole pressure of producer wells

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.petrol.2021.109257.

References Dai, Zhenxue, Viswanathan, Hari, Xiao, Ting, Middleton, Richard, Pan, Feng,
Ampomah, William, Yang, Changbing, Zhou, Youqin, Jia, Wei, Lee, Si-Yong, 2017.
CO2 sequestration and enhanced oil recovery at depleted oil/gas reservoirs. Energy
Alfi, Masoud, Hosseini, Seyyed A., 2016. ’Integration of reservoir simulation, history
Procedia 114, 6957–6967.
matching, and 4D seismic for CO2-EOR and storage at Cranfield, Mississippi, USA.
Dudek, J., Janiga, D., Wojnarowski, P., 2020. ’Optimization of CO2-EOR process
Fuel 175, 116–128.
management in polish mature reservoirs using smart well technology. J. Petrol. Sci.
Allinson, Ken, Burt, Dan, Campbell, Lisa, Constable, Lisa, Crombie, Mark, Lee, Arthur,
Eng. 108060.
Lima, Vinicius, Lloyd, Tim, Lee, Solsbey, 2017. ’Best practice for transitioning from
Edgar, Thomas F., Himmelblau, David M., Lasdon, Leon S., 2001. Optimization of
carbon dioxide (CO2) enhanced oil recovery EOR to CO2 storage. Energy Procedia
Chemical Processes.
114, 6950–6956.
Ertekin, Turgay, Abou-Kassem, Jamal H., King, Gregory R., 2001. Basic Applied
Ampomah, W., RS Balch, M Cather, Will, R., Gunda, D., Dai, Z., Soltanian, M.R., 2017a.
Reservoir Simulation.
Optimum design of CO2 storage and oil recovery under geological uncertainty. Appl.
Farajzadeh, R., Eftekhari, Ali Akbar, Dafnomilis, G., Lake, L.W., Bruining, J., 2020. On
Energy 195, 80–92.
the sustainability of CO2 storage through CO2–Enhanced oil recovery. Appl. Energy
Ampomah, William, Balch, Robert, Will, Robert, Cather, Martha, Gunda, Dhiraj,
261, 114467.
Dai, Zhenxue, 2017b. ’Co-optimization of CO2-EOR and storage processes under
Gibbins, Jon, Chalmers, Hannah, 2008. ’Carbon capture and storage. Energy Pol. 36,
geological uncertainty. Energy Procedia 114, 6928–6941.
4317–4322.
Asghari, Koorosh, Al-Dliwe, Adal, 2005. ’Optimization of carbon dioxide sequestration
Goodarzi, S., Settari, A., Zoback, M.D., Keith, D.W., 2015. ’Optimization of a CO2 storage
and improved oil recovery in oil reservoirs. In: Greenhouse Gas Control
project based on thermal, geomechanical and induced fracturing effects. J. Petrol.
Technologies, vol. 7. Elsevier.
Sci. Eng. 134, 49–59.
Aziz, Khalid, 1979. ’Petroleum Reservoir Simulation. Applied Science Publishers, p. 476.
Guo, Jian-Xin, Huang, Chen, Wang, Jian-Liang, Meng, Xiao-Yan, 2020. ’Integrated
Azizkhani, Ashkan, Gandomkar, Asghar, 2019. ’A novel method for application of
operation for the planning of CO2 capture path in CCS–EOR project. J. Petrol. Sci.
nanoparticles as direct asphaltene inhibitors during miscible CO2 injection. J. Petrol.
Eng. 186, 106720.
Sci. Eng. 106661.
Han, Jinju, Lee, Minkyu, Lee, Wonsuk, Lee, Youngsoo, Sung, Wonmo, 2016. ’Effect of
Bachu, Stefan, 2003. ’Screening and ranking of sedimentary basins for sequestration of
gravity segregation on CO2 sequestration and oil production during CO2 flooding.
CO 2 in geological media in response to climate change. Environ. Geol. 44, 277–289.
Appl. Energy 161, 85–91.
Barrufet, Maria A., Bacquet, Alexandre, Falcone, Gioia, 2010. ’Analysis of the storage
He, Liping, Shen, Pingping, Liao, Xinwei, Li, Fangfang, Gao, Qichao, Wang, Zhilin, 2016.
capacity for CO2 sequestration of a depleted gas condensate reservoir and a saline
’Potential evaluation of CO2 EOR and sequestration in Yanchang oilfield. J. Energy
aquifer. J. Can. Petrol. Technol. 49, 23–31.
Inst. 89, 215–221.
Budinis, Sara, Krevor, Samuel, Mac Dowell, Niall, Brandon, Nigel, Adam, Hawkes, 2018.
IEA, Energy, and Climate Change, 2015. World Energy Outlook Special Report. 2015.
An assessment of CCS costs, barriers and potential. Energy strategy reviews 22,
Organisation for Economic Co-operation and Development and the International
61–81.
Energy Agency, Paris.
Chen, Bailian, Pawar, Rajesh J., 2019. ’Capacity assessment and co-optimization of CO2
Javaheri, M., Abedi, J., Hassanzadeh, H., 2009. ’Onset of convection in co sequestration
storage and enhanced oil recovery in residual oil zones. J. Petrol. Sci. Eng. 182,
in deep inclined saline aquifers. J. Can. Petrol. Technol. 48, 22–27.
106342.
Jia, Wei, McPherson, Brian, Pan, Feng, Dai, Zhenxue, Nathan, Moodie, Xiao, Ting, 2018.
Dai, Zhenxue, Viswanathan, Hari, Middleton, Richard, Pan, Feng, Ampomah, William,
Impact of three-phase relative permeability and hysteresis models on forecasts of
Yang, Changbing, Jia, Wei, Xiao, Ting, Lee, Si-Yong, McPherson, Brian, 2016. ’CO2
storage associated with CO2-EOR. Water Resour. Res. 54, 1109–1126.
accounting and risk analysis for CO2 sequestration at enhanced oil recovery sites.
Jiang, Jieyun, Rui, Zhenhua, Hazlett, Randy, Lu, Jun, 2019. An integrated technical-
Environ. Sci. Technol. 50, 7546–7554.
economic model for evaluating CO2 enhanced oil recovery development. Appl.
Energy 247, 190–211.

11
S.B. Kashkooli et al. Journal of Petroleum Science and Engineering 208 (2022) 109257

Khan, Chawarwan, Amin, Robert, Madden, Gary, 2013. ’Carbon dioxide injection for Systems Schlumberger Information, 2010. Eclipse Technical Description.
enhanced gas recovery and storage (reservoir simulation). Egyptian J. Petrol. 22, Tapia, John FD., Lee, J.-Y., Ooi, Raymond EH., Foo, Dominic CY., Tan, Raymond R.,
225–240. 2015. ’Design and scheduling of CO2 enhanced oil recovery with geological
Khor, Cheng Seong, Ali, Elkamel, Shah, Nilay, 2017. ’Optimization methods for sequestration operations as a strip packing problem. Chem. Eng. Trans. 45,
petroleum fields development and production systems: a review. Optim. Eng. 18, 1615–1620.
907–941. Tapia, John Frederick D., Lee, Jui-Yuan, Ooi, Raymond EH., Foo, Dominic CY.,
Kim, Youngmin, Jang, Hochang, Kim, Junggyun, Lee, Jeonghwan, 2017. ’Prediction of Tan, Raymond R., 2016. Optimal CO2 allocation and scheduling in enhanced oil
storage efficiency on CO2 sequestration in deep saline aquifers using artificial neural recovery (EOR) operations. Appl. Energy 184, 337–345.
network. Appl. Energy 185, 916–928. Tavallali, M Sadegh, Karimi, I.A., Baxendale, D., 2016. ’Process systems engineering
Kwak, Dong-Hun, Kim, Jin-Kuk, 2017. ’Techno-economic evaluation of CO2 enhanced oil perspective on the planning and development of oil fields. AIChE J. 62, 2586–2604.
recovery (EOR) with the optimization of CO2 supply. Int. J. Greenhouse Gas Control Tavallali, Mohammad Sadegh, Bakhtazma, Farnoosh, Ali, Meymandpour, Karimi, I.A.,
58, 169–184. 2018. Optimal drilling planning by considering the subsurface dynamics—combing
Le Gallo, Yann, Couillens, Philippe, Manai, Taoufik, 2002. CO2 sequestration in depleted the flexibilities of modeling and a reservoir simulator. Ind. Eng. Chem. Res. 57,
oil or gas reservoirs. In: SPE International Conference on Health, Safety and 16367–16378.
Environment in Oil and Gas Exploration and Production. Society of Petroleum Tavallali, M.S., Karimi, I.A., Halim, A., Baxendale, D., Teo, K.M., 2014. ’Well placement,
Engineers. infrastructure design, facility allocation, and production planning in multireservoir
Le Van, Si, Chon, Bo Hyun, 2017. ’Evaluating the critical performances of a oil fields with surface facility networks. Ind. Eng. Chem. Res. 53, 11033–11049.
CO2–Enhanced oil recovery process using artificial neural network models. J. Petrol. Tavallali, M.S., Karimi, I.A., Teo, K.M., Baxendale, D., Ayatollahi, Sh, 2013. Optimal
Sci. Eng. 157, 207–222. producer well placement and production planning in an oil reservoir. Comput.
Li, Y., Li, J., Ding, S., Zhang, H., 2016. ’Co-optimization of CO2 sequestration and Chem. Eng. 55, 109–125.
enhanced oil recovery in extra-low permeability reservoir in Shanbei’, Energy Wang, Qian, Yang, Shenglai, Han, Haishui, Wang, Lu, Qian, Kun, Pang, Jieqiong, 2019.
Sources, Part A: recovery, Utilization. Environ. Effects. 38, 442–449. Experimental investigation on the effects of CO2 displacement methods on
Min, Baehyun, Sun, Alexander Y., Wheeler, Mary F., Jeong, Hoonyoung, 2018. petrophysical property changes of ultra-low permeability sandstone reservoirs near
’Utilization of multiobjective optimization for pulse testing dataset from a CO2-EOR/ injection wells. Energies 12, 327.
sequestration field. J. Petrol. Sci. Eng. 170, 244–266. Wang, Xiao, van’t Veld, Klaas, Marcy, Peter, Huzurbazar, Snehalata, Alvarado, Vladimir,
Momeni, A., Aghajani, M., Zargar, G., 2012. A Simulation study of carbon dioxide 2018. ’Economic co-optimization of oil recovery and CO2 sequestration. Appl.
sequestration in a depleted oil reservoir. Petrol. Sci. Technol. 30, 751–765. Energy 222, 132–147.
Nilsson, Martin, 2016. ’Organization of the petroleum exporting countries (OPEC). In: Wang, Yang, Zhang, Keni, Wu, Nengyou, 2013. ’Numerical investigation of the storage
International Organizations and the Rise of ISIL. Routledge. efficiency factor for CO2 geological sequestration in saline formations. Energy
Rezk, Mohamed Gamal, Foroozesh, Jalal, Zivar, Davood, Mumtaz, Mudassar, 2019. ’CO2 Procedia 37, 5267–5274.
storage potential during CO2 enhanced oil recovery in sandstone reservoirs. J. Nat. Yang, Guang, Zhang, Ye, Li, Shuiquan, 2012. Uncertainty analysis of carbon
Gas Sci. Eng. 66, 233–243. sequestration in an inclined deep saline aquifer. In: Masters Abstracts International.
Rios, V.S., Santos, L.O.S., Quadros, F.B., Schiozer, D.J., 2018. ’New upscaling technique You, Junyu, Ampomah, William, Sun, Qian, 2020. ’Development and application of a
for compositional reservoir simulations of miscible gas injection. J. Petrol. Sci. Eng. machine learning based multi-objective optimization workflow for CO2-EOR
175, 389–406. projects. Fuel 264, 116758.
Safarzadeh, Mohammad Amin, Motahhari, Seyyed Mahdia, 2014. ’Co-optimization of Yuncong, G.A.O., Zhao, Mifu, Wang, Jianbo, Chang, Z.O.N.G., 2014. ’Performance and
carbon dioxide storage and enhanced oil recovery in oil reservoirs using a multi- gas breakthrough during CO2 immiscible flooding in ultra-low permeability
objective genetic algorithm (NSGA-II). Petrol. Sci. 11, 460–468. reservoirs. Petrol. Explor. Dev. 41, 88–95.

12

You might also like