You are on page 1of 14

SPE-192378-MS

Well-Placement Optimization in a Novel Heavy Oil Recovery Process Using


In-Situ Steam Generated by Thermochemicals
Tamer Moussa, Shirish Patil, Mohamed Mahmoud and Salaheldin Elkatatny, King Fahd University of Petroleum
and Minerals

Copyright 2018, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Kingdom of Saudi Arabia Annual Technical Symposium and Exhibition held in Dammam, Saudi Arabia, 23–26
April 2018.

This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents
of the paper have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Electronic reproduction, distribution, or storage o f any part of this paper without the written
consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may
not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract

Determination of optimal well locations plays an important role in the efficient recovery of hydrocarbon
resources. However, it is a challenging and complex task because it relies on reservoir, and fluid and
economic variables that are often nonlinearly correlated. Traditionally, well placement optimization
(WPO) has been done through experience and use of quality maps. However, reservoir management
teams are beginning to appreciate the use of automatic optimization tools for well placement that will
yield the largest financial returns or highest net present value (NPV). In addition, the performance of a
reservoir is time and process dependent, therefore well placement decisions cannot be based on static
properties alone. On the other hand, well placement optimization requires a large number of simulator
runs in an iterative process, and thus several runs to reach the maximum achievable NPV. Therefore,
there is a real need for automatic well placement approach that uses highly efficient optimization
method, which can improve the result quality, speed of the convergence process to optimal result and
thus decrease the time required for computation.
The objective of this work is to determine the optimal well locations in a heavy oil reservoir under
production using a novel recovery process, in which steam is generated, in-situ, using thermochemical
reactions. Self-adaptive differential evolution (SaDE) and particle swarm optimization (PSO) methods
are used as the global optimizer to find the optimal configuration of wells that will yield the highest
NPV. Comparison analysis between the two proposed optimization techniques is introduced. The CMG
STARS Simulator is utilized in this research to simulate reservoir models with different well
configurations.
Comparison of results is made between the NPV achieved by the well configuration proposed by the
SaDE and PSO methods. The results show that SaDE performed better than PSO in terms of higher NPV
after ten years of production while under in-situ steam injection process using thermochemical reactions.
This is the first known application where SaDE and PSO methods are used to optimize well locations in
a heavy oil reservoir that is recovered by injecting steam generated in-situ using thermo-chemical
reactions. This research shows the importance of well placement optimization in a highly promising and
novel heavy oil recovery process. This also is a step forward in the direction to eliminate the CO2
emissions related to thermal recovery processes.
2 SPE-192378-MS

Keywords
Well placement optimization, Differential Evolution, Particle Swarm Optimization, In-Situ Steam Generation,
Heavy Oil Recovery, In-Situ Chemical Reaction, Net Present Value

Introduction
Reservoir recovery performance is highly dependent on well locations. Therefore, there is a real need for
a robust well-placement optimization tool especially when recovering a heavy oil reservoir with thermal
recovery processes, because geological and engineering parameters controlling the reservoir performance
are highly nonlinearly-correlated. Therefore, many researches have been done in the area of well-
placement optimization (Badru and Kabir 2003, Güyagüler and Horne 2004, Ermolaev and Kuvichko
2013, Awotunde and Naranjo 2014). On the other hand, thermal recovery methods such as steam injection
are viable and commonly used to recover heavy oil and bitumen primarily by thermally reducing oil
viscosity and improving the displacement of the heavy oil (Scott 2002, Yee and Stroich 2004, Putra,
Rachman et al. 2011, Jinzhong, Wenlong et al. 2013), there are significant heat losses before the steam
reaches the heavy oil. In addition to the concerns of high cost and emission of greenhouse gases. Generally,
thermal recovery processes are employed when the reservoir depth is less than 1100 m (3600 feet) with
oil viscosity of more than 50 cP and density less than 20 °API and well spacing of 2 to 5 acres (Vega
Riveros and Barrios 2011). In other words, the application of a suitable thermal recovery process depends
on the depth of the heavy oil reservoir. Based on thermal efficiency, steam injection has been a good
solution for shallow heavy oil reservoirs. In case of deep heavy oil reservoirs, in-situ combustion is a
promising solution towards increasing the flowability of oil through viscosity reduction (Jinzhong,
Wenlong et al. 2013). Oil viscosity and density, reservoir thickness and well spacing are also considered
during the selection and application of any type of thermal recovery process for a heavy oil reservoir
(Vega Riveros and Barrios 2011). In situ combustion involves generation of thermal energy inside the
heavy oil reservoir and alter its rock and fluid characteristics. However, it has many challenges; such as,
low success rate, gas override, operational issues, not proven for carbonate reservoirs and it is applicable
to on-shore fields only (Castrogiovanni, Fitzpatrick et al. 2011, Al-Nakhli, Sukkar et al. 2016). Therefore,
introducing a heavy-oil recovery approach in which steam could be generated downhole is very important
to overcome the challenges attached to conventional steam injection. One of the promising new heavy-oil
recovery techniques is generating in-situ steam using thermochemical reaction to reduce oil viscosity,
improve the mobility ratio and enhance the heavy-oil displacement. In this approach, steam is generated
in-situ, by injecting exothermic reactants downhole with injected water to create heat and enhance
reservoir pressure for mobilizing heavy oil. The exothermic reaction is triggered by either increasing
downhole temperature or by controlling the downhole pH using a buffer.
Well-locations optimization plays a significant role to improve the recovery efficiency of this new
process. Therefore, the objective of this paper is to determine the optimal well locations in a heavy oil
reservoir developed by a novel heavy oil recovery process using in-situ steam generation by
thermochemical. Two global optimization methods are used in this research, namely; self-adaptive
differential evolution (SaDE) and particle swarm optimization (PSO). The recovery performance is
analyzed based on the achieved net present value (NPV), cumulative oil production (COP) and steam oil
ratio (SOR).
The net present value (NPV) will be used as a main performance indicator in this paper. NPV of a
project is defined as the sum of all the present values of individual cash flows of that project; where the
cash flow is positive for revenue and negative for expenditures (Khan 1993). In the proposed recovery
process, the capital costs at the project’s initialization will include the cost of exploration, drilling and
well completion, thermochemical injection facility and water treatment facility. The expenditures will
include the cost of steam injection, produced water treatment, chemicals handling and recompression,
chemicals cost and operating costs involving well remediation and human resources. All costs are
SPE-192378-MS 3

discounted to the beginning of the project. Revenues are the incoming cash flow that is expected to accrue
from heavy oil sales in after tax dollars (after deducting taxes and royalties), which also are discounted to
the present time. All these cash flows are then combined to give the NPV of the project (Onwunalu and
Durlofsky 2010).

1 Self-adaptive Differential Evolution


Differential Evolution (DE), proposed by Storn and Price (Storn and Price 1997), is considered of the
most effective tools to globally optimize engineering problem over continues domains. There are many
applications in which DE has been applied to engineering related functions. However, the control variables
and alteration methods in DE method are significantly reliant on the function to be optimized. So,
generally, many trials of different alteration strategies and control variables are required to obtain the
optimal result.
To overcome these challenges, a differential evolution algorithm with self-adaptation is required.
Therefore, a self-adaptive differential evolution (SaDE) algorithm introduced by (Qin, Huang et al. 2009)
in which the control variables and alternation methods are adapted based on the knowledge gained from
previous iterations. In other words, this proposed algorithm removed the need of many trials that had to
be conducted to achieve the optimal results.

2 Particle Swarm Optimization


The first step in particle swarm algorithm is to generate particles initially, then give different paces to
them. The objective function is evaluated at each iteration using each particle value and estimate the best
value of the function and best location. It assigns new paces based on the current pace, the best location
achieved by this particle and the best locations of its neighbors. Then with every iteration, these values
are modified and updated. This process continues until reaching to a stopping condition. The PSO
algorithm used in the research is based on the algorithm described in (Eberhart 1995), using modifications
suggested in (Mezura-Montes and Coello Coello 2011) and in (Pedersen 2010).

3 Single and Chain In-Situ Thermochemical Reactions


The in-situ thermochemical reaction could be triggered as a single reaction or chain of reactions. In single
reaction, the thermochemical reaction is trigged to produce heat as shown in Figure 1. In chain type, the
thermochemical reactions are trigged and repeated several times to regenerate the heat at specific intervals.
In this research, both processes are simulated and applied to a heavy oil reservoir. Table 1 shows the
parameters of the in-situ steam generated by thermochemical reactions. As shown, each reaction cycle
will take 8 days as the temperature will decrease from 582.9 oF to 213.8 oF and steam quality will decrease
from 0.9 to 0.4.
4 SPE-192378-MS

Figure 1 - Heat profile generated by in-situ thermochemical reaction

Table 1 - Generated steam conditions from thermochemical reaction


Time Temperature
o Steam Quality
days F
2 582.9 0.9
4 306.5 0.8
6 267.7 0.6
8 213.8 0.4

4 Methodology
In order to perform well-placement optimizations, the two optimization algorithms; self-adaptive
differential evolution (SaDE) and particle swarm optimization (PSO) are integrated in a MATLAB
framework to numerical simulator. Steam, Thermal and Advanced Processes Reservoir Simulator
(STARS™) from CMG was used in this work to simulate the proposed in-situ thermochemical steam
injection process where three wells are completed in a heavy oil reservoir as shown in the next sections.
The minimum well spacing (r) is considered in this paper as the main constrain to the optimization
problem. r is estimated from the area (A) surrounding a particular well within which no other well shall
be placed. Therefore, the minimum r of that area could be calculated as shown in Eq.(1):

43,560A
r (1)

Similar to all other evolutionary algorithms (EAs), the evolutionary process of self-adaptive differential
evolution (SaDE) and particle swarm optimization (PSO) starts with initialization step to create a
population size NP . In the initialization step, the lower and upper bounds of each parameter must be
assigned before the initial population to be created, where A L and A U are the lower and upper bound
values respectively. Therefore, the initial value at first generation is:
a j ,i ,0  rand j  0,1 .  A j U  A j L   A j L (2)
SPE-192378-MS 5

Where the random number generator, rand  0,1 , returns a uniform-distributed random number within
the range [0,1] . Therefore, the current population ( Px ,g ), as calculated in Eq.(3), consists of ai , g vectors
as shown in Eq.(4).
Px , g  ai , g : i  1, 2,..., NP  g  0,1,..., g max (3)

Ai , g  (a j ,i , g ), j  1, 2,..., D (4)
where g  0,1,..., g max is the generation number, i is the population index from 1 to NP , and j is
parameter index from 0 to the total number of optimization parameters ( D ). In this paper, the total number
of optimization parameters are 6 as there are 3 wells and each well has 2 axis  x j , y j  .
After population generation, the values are constrained as following for each well j  1 with respect to
the previous well j as calculated in Eq.(5):
x 2  x 1    y 2  y 1   r2
2 2
(5)
where r is the minimum well spacing calculated from Eq.(1).
The main performance indicator is the NPV after 10 years of production. The reason for selecting NPV
as a primary indicator is to study the effect of well-placement optimization on the economic return of the
proposed in-situ recovery process compared with the base case in which no well-location optimization is
applied. The cumulative oil production (COP) and steam oil ratio (SOR) are also monitored in this
research.

4.1 Model Description


Steam, Thermal and Advanced Processes Reservoir Simulator (STARS™) from CMG was used in this
work to simulate the in-situ thermochemical steam process to develop a heavy-oil reservoir (CMG 2017).
A homogeneous 3D Cartesian model with three wells, with 31 grid blocks along the x-axis, 6 grid blocks
along the y-axis and 20 grid blocks along z-axis was used. The reservoir and fluid parameters used in both
process are captured from (Curtis, Kopper et al. 2002, Gates and Chakrabarty 2005) and are shown on
Table 2. Both models were simulated for 10 years (3600 days).

Table 2 - Simulator input parameters (Curtis et al. 2002; Gates and Chakrabarty 2005)

Input Parameter Value


Reference Depth, ft 900
Reservoir Initial Pressure @ 900 ft, psi 145
oF
Reservoir Temperature, 68
Porosity 38%
Average Horizontal Permeability, mD 7000
Average Vertical Permeability, mD 3000
Formation Compressibility, 1/psi 1.8  10 5
Rock Heat Capacity, Btu/ft3 F 35.04
Thermal Conductivity of Rock, Btu/(ft day F) 44
Thermal Conductivity of Water, Btu/(ft day F) 8.6
Thermal Conductivity of Oil, Btu/(ft day F) 1.8
Thermal Conductivity of Gas, Btu/(ft day F) 0.3
Thermal Conductivity of Overburden and Underburned, Btu/(ft day F) 24
Oil Gravity @ Standard Condition, oAPI 8
oF,
Viscosity @ 68 cP 21,109
Viscosity @ 104 oF, cP 2,918
6 SPE-192378-MS

5 Results and Discussion

5.1 Simulation of Single In-Situ Thermochemical Reaction


Figure 2 shows the steam quality and temperature generated from a single in-situ thermochemical reaction
and the cumulative heavy oil produced as a function of time, under these injection conditions. It could be
shown that the reaction cycle is 8 days, then 4 days for soaking and the production period is 8 days. The
figure shows that cumulative oil production is 103 bbl after 8 days of production.

Injection Soaking Production

Figure 2 – Simulation of single in-situ thermochemical reaction

5.2 Simulation of Chain In-Situ Thermochemical Reactions


5.2.1 In Series Two Thermochemical Reactions
In this case, two in-situ thermochemical reactions are trigged in series (chain reaction) as shown in Figure
3. The second thermochemical reaction starts after completing the first reaction. In other words, the second
reaction is trigged once the steam quality and temperature reach 0.4 and 213 oF respectively. In this case,
the injection phase lasts for 16 days as it consists of 2 cycles of thermochemical reactions. The soaking
period is 4 days, similar to the single reaction case. The figure shows that the cumulative oil production
increased in this case to 197 bbl during the 8 days of production phase.

5.2.2 In Series Three Thermochemical Reactions


Three reactions are trigged in this case as shown in Figure 4. The reaction is repeated once the steam
quality and temperature reach to 0.4 and 213 oF respectively. The soaking period is kept 4 days as in the
previous cases. The cumulative oil produced after 8 days of production is 184 bbl which is less than the
oil produced at the previous two-reactions case by 5%.
SPE-192378-MS 7

Injection (2 Reaction Cycles) Soaking Production

Figure 3 - Simulation of two in-situ thermochemical reactions

Injection (3 Reaction Cycles) Soaking Production

Figure 4 - Simulation of three in-situ thermochemical reactions


8 SPE-192378-MS

5.2.3 In Series Four Thermochemical Reactions


In this case, four consequence thermochemical reactions are trigged as shown in Figure 5. In this case,
the thermochemical reaction is repeated four times when the steam quality and temperature reach to 0.4
and 213 oF respectively. The soaking period is 4 days and the cumulative oil produced during the 8 days
of production is 108 bbl which is less than the two-reactions case by 45%.

Injection (4 Reaction Cycles) Soaking Production

Figure 5 - Simulation of four in-situ thermochemical reactions

It could be seen that the maximum cumulative oil production is achieved in the second case when injection
phase has two thermochemical reaction cycles. Figure 6a and b show comparison between single reaction
case, two-reactions case and four-reactions case in terms of cumulative oil and liquid (oil and water)
production respectively. It can be seen that although the second case, achieved the maximum cumulative
oil production (COP), the cumulative liquid production (CLP) is almost the same. Therefore, the decrease
in oil production at the four-reactions case is due to the increase of water cut and not due to total
production. This increase in water cut is due to the increase of injection time (32 days) compared with 16
days in the two-reactions case. On the other hand, as shown in Figure 7a, the in-situ heat generated from
the single reaction case is not sufficient to reduce the reservoir oil viscosity around the wellbore to
continue production for 8 days. Figure 6 also shows that the oil production is significantly decreased just
5 days after starting the production. But the reduction in oil viscosity resulted from the two and four
reactions cases as shown in Figure 7b and c respectively is sufficient to keep the production for the 8
days.
In addition, Figure 6a shows that at the two and four reactions cases, the oil production rate in the first
two days is very low, then start to increase. This is due to the higher water cut at the beginning of the
production. It can also be observed that after soaking period the water cut is very high near to the wellbore.
Therefore, the number of thermochemical reaction cycles needs to be optimized during the injection
stage in order to allow sufficient heat generation, which will be able to reduce the reservoir oil viscosity
around the wellbore and at the same time not to increase the water cut around the wellbore and hence
increasing the water cut at the producing stream with its associate cost of water treatment.
SPE-192378-MS 9

(a) (b)
Figure 6 - Comparison of different cases
Heavy Oil Project Heavy Oil Project

1,0401,0301,0201,0101,000 990 980 970 960 950 940 930 920 910 900 890 880 870 860
Oil Viscosity (cp) 2017-01-21 J layer: 1 Oil Viscosity (cp) 2017-01-29 J layer: 1
1,0401,0301,0201,0101,000990 980 970 960 950 940 930 920 910 900 890 880 870 860

File: insitu_field_1css_1exo.irf File: insitu_field_1css_2exo.irf


870 880 890 900 910 920 930 940 950 960 970 980 9901,0001,0101,0201,0301,0401,050

0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300
870 880 890 900 910 920 930 940 950 960 970 980 9901,0001,0101,0201,0301,0401,050

User: engin User: engin


Date: 03/03/39 Date: 03/03/39
Scale: 1:2470 Scale: 1:2393
Z/X: 6.00:1 Z/X: 6.00:1
Axis Units: ft Axis Units: ft

PRODUCER
INJECTOR PRODUCER
INJECTOR
5,387 5,387

4,861 4,848

4,335 4,310

3,809 3,771

3,283 3,232

2,757 2,694

2,231 2,155

1,705 1,616

1,179 1,078
0.00 160.00 320.00 feet 0.00 155.00 310.00 feet
653 539
0.00 50.00 100.00 meters 0.00 50.00 100.00 meters
127 0
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300

(a) (b)
Heavy Oil Project
Oil Viscosity (cp) 2017-02-06 J layer: 1
1,0401,0301,0201,0101,000990 980 970 960 950 940 930 920 910 900 890 880 870 860

File: insitu_field_1css_4exo.irf
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300
870 880 890 900 910 920 930 940 950 960 970 980 9901,0001,0101,0201,0301,0401,050

User: engin
Date: 03/03/39
Scale: 1:2470
Z/X: 6.00:1
Axis Units: ft

PRODUCER
INJECTOR
5,387

4,848

4,310

3,771

3,232

2,694

2,155

1,616

1,077
0.00 160.00 320.00 feet
539
0.00 50.00 100.00 meters
0
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300

(c)
Figure 7 - Reservoir oil viscosity distribution (a) single-reaction case, (b) two-reactions case and (c) four-reactions case
10 SPE-192378-MS

5.3 Optimization Results


Figure 8 shows the performance of the two optimization methods; self-adaptive differential evolution
(SaDE) and particle swarm optimization (PSO), based on the achieved net present value (NPV) of the
proposed in-situ thermochemical recovery process while optimizing the well-locations, in terms of the
achieved NPV against the number of simulation runs. It could be shown that generally, the two
optimization methods assisted to increase the NPV significantly. In other words, SaDE and PSO
optimization algorithms enhanced the achievable NPV from approximately $3 and $4 million to $11.3
and $12.6 million respectively (i.e. more than 200% increase in the achieved NPV). This assure the
importance of well-location optimization approach and how it is really effective when optimizing the well-
location in the proposed in-situ thermochemical recovery process.
On the other hand, the figure shows the outperformance of the SaDE optimization method compared
to PSO. It could be seen that using the SaDE method achieved higher NPV than PSO by more than 12%.
It could also be shown that after 20 simulation runs, the SaDE method achieved higher NPV than what
achieved by the PSO after 150 runs. Therefore, the SaDE not only has better exploration capabilities and
achieves better results by the end of iterations, it is even faster than PSO and requires less computational
costs. This is very important feature in the SaDE algorithm when optimizing problems related to reservoir
simulation especially in thermal and thermal-solvent recovery processes. Because in these cases, each
simulation run is timely expensive (i.e. each run requires long time to be completed in field-scale).
Therefore, the optimization algorithm to be used in these cases have to have good exploration capabilities
to explore the search space efficiently and be fast as well.

Figure 8 - NPV achieved by optimizing well locations by SaDE and PSO methods

Figure 9 and Figure 10 show heat and viscosity distributions, respectively, after 10 years of field
development with and without well-placement optimization. It could be shown in Figure 9 that the heat
distribution when the well-locations are optimized is better than the base case in which the well-locations
are set without optimization. This better heat spreading achieves better viscosity reduction and distribution
in the heavy oil reservoir as shown in Figure 10. Therefore, well-placement is a significant parameter that
has to be optimized to achieve efficient heat distribution and viscosity reduction in oil reservoirs, and more
particularly in heavy oil reservoirs where the main drive mechanism is reducing reservoir oil viscosity by
heat energy. In other words, more efficient heat delivery achieves better oil viscosity reduction and
enhance the oil recovery, and this could be accomplished by efficient well-location optimization approach.
Steam oil ratio (SOR) is defined as the required volume of steam to produce a unit volume of oil. So,
in this work, it is defined as the barrels of steam injected into the reservoir to produce one barrel of oil. It
SPE-192378-MS 11

is common in the literature to use this parameter as a performance indicator of the thermal recovery
processes. Therefore, it is also discussed in this paper as well as the cumulative oil production (COP)
reached after 10 years of production. The recovery performance comparison between the well-location
optimized case and the base case, in terms of cumulative oil production (COP) and steam oil ratio (SOR),
is shown in Figure 11. It could be seen that when the well-locations are optimized, the COP is increased
significantly by more than 40% over what is produced with optimizing well-locations. On the other hand,
the SOR in the optimized case is much lower than its value in the base case during the reservoir
development life time. It is also shown in the base case, it takes very long time (9 years of production) to
achieve similar SOR that attained by the optimized case very earlier (3 years of production). Therefore,
the higher COP and lower SOR achieved by the optimized case are the main reasons that make the
optimized case achieves significantly higher NPV compared to the base case.

Heavy Oil Project Heavy Oil Project


Temperature (F) 2026-11-10 J layer: 1 Temperature (F) 2026-11-10 J layer: 1

1,0401,0301,0201,0101,000 990 980 970 960 950 940 930 920 910 900 890 880 870 860
1,0401,0301,0201,0101,000 990 980 970 960 950 940 930 920 910 900 890 880 870 860

0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 File: exo_sade.irf 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 File: exo_basecase.irf

870 880 890 900 910 920 930 940 950 960 970 980 990 1,0001,0101,0201,0301,0401,0501,060
870 880 890 900 910 920 930 940 950 960 970 980 990 1,0001,0101,0201,0301,0401,0501,060

User: Conference Room User: Conference Room


Date: 3/22/2018 Date: 3/22/2018
Scale: 1:2471 Scale: 1:2471
Z/X: 6.00:1 Z/X: 6.00:1
Axis Units: ft Axis Units: ft

INJECTOR-2
PRODUCER-2 INJECTOR-1 PRODUCER-3
PRODUCER-1 INJECTOR-3 INJECTOR-2
PRODUCER-2 INJECTOR-1
PRODUCER-1 INJECTOR-3
PRODUCER-3
644 645

593 594

542 543

491 492

440 440

389 389

338 338

287 287

236 236

0.00 160.00 320.00 feet 0.00 160.00 320.00 feet


185 185
0.00 50.00 100.00 meters 0.00 50.00 100.00 meters
134 134
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300

(a) (b)
Figure 9 - Heat distribution in the reservoir. (a) SaDE-optimized case, (b) base case

Heavy Oil Project Heavy Oil Project


Oil Viscosity (cp) 2026-11-10 J layer: 1 Oil Viscosity (cp) 2026-11-10 J layer: 1

1,0401,0301,0201,0101,000 990 980 970 960 950 940 930 920 910 900 890 880 870 860
1,0401,0301,0201,0101,000 990 980 970 960 950 940 930 920 910 900 890 880 870 860

0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 File: exo_sade.irf 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 File: exo_basecase.irf
870 880 890 900 910 920 930 940 950 960 970 980 990 1,0001,0101,0201,0301,0401,0501,060
870 880 890 900 910 920 930 940 950 960 970 980 990 1,0001,0101,0201,0301,0401,0501,060

User: Conference Room User: Conference Room


Date: 3/22/2018 Date: 3/22/2018
Scale: 1:2471 Scale: 1:2471
Z/X: 6.00:1 Z/X: 6.00:1
Axis Units: ft Axis Units: ft

INJECTOR-2
PRODUCER-2 INJECTOR-1 PRODUCER-3
PRODUCER-1 INJECTOR-3 INJECTOR-2
PRODUCER-2 INJECTOR-1
PRODUCER-1 INJECTOR-3
PRODUCER-3
5,387 5,387

4,848 4,848

4,310 4,310

3,771 3,771

3,232 3,232

2,693 2,693

2,155 2,155

1,616 1,616

1,077 1,077

0.00 160.00 320.00 feet 0.00 160.00 320.00 feet


539 539
0.00 50.00 100.00 meters 0.00 50.00 100.00 meters
0 0
0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300 0 100 200 300 400 500 600 700 800 900 1,000 1,100 1,200 1,300

(a) (b)
Figure 10 - Viscosity distribution in the reservoir. (a) SaDE-optimized case, (b) base case
12 SPE-192378-MS

Figure 11 - Recovery performance comparison between the optimized and base case

Conclusions
Well-placement optimization is a significant factor that affects on reservoir recovery performance because
of the nonlinearity behavior of the geological and engineering parameters controlling the recovery
performance of oil reservoirs. In particular in heavy-oil reservoir that are developed by thermal and
thermal-solvent recovery processes. In such cases there is a real need for a robust well-placement
optimization approach. This paper proposed a new application of the self-adaptive differential evolution
method to optimize the recovery performance of a novel heavy-oil recovery process in which steam is
generated downhole using in-situ thermochemical reactions to reduce heat losses and greenhouse gas
emissions. The performance indicators utilized in this paper are net present value (NPV), cumulative oil
production (COP) and steam oil ratio (SOR). A performance comparison between the proposed SaDE
method and particle swarm optimization (PSO) method was conducted as well. SaDE method
outperformed the PSO and achieved more efficient well-placement optimization that attained more
economic profitability in terms of higher NPV. The recovery performance of the proposed recovery
process when the well-locations are optimized is superior compared to non-optimized well-locations case,
in terms of higher NPV and COP, and lower SOR.

Nomenclature
DE : Differential Evolution
SaDE : Self-adaptive Differential Evolution
PSO : Particle Swarm Optimization
EA : Evolutionary Algorithm
NPV : Net present value, $
COP : Cumulative oil production, STB
CLP : Cumulative liquid production, STB
RF : Oil recovery factor, %
SOR : Steam oil ratio, STB/STB
SPE-192378-MS 13

References
Al-Nakhli, A. R., L. A. Sukkar, J. Arukhe, A. Mulhem, A. Mohannad, M. Ayub and M. Arifin (2016). In-
Situ Steam Generation A New Technology Application for Heavy Oil Production. SPE Heavy Oil
Conference and Exhibition, Kuwait City, Kuwait, 6-8 December 2016, Society of Petroleum
Engineers. doi: 10.2118/184118-MS
Awotunde, A. A. and C. Naranjo (2014). Well Placement Optimization Constrained to Minimum Well
Spacing. SPE Latin America and Caribbean Petroleum Engineering Conference, Maracaibo,
Venezuela, 21-23 May 2014, Society of Petroleum Engineers. doi: 10.2118/169272-MS
Badru, O. and C. S. Kabir (2003). Well Placement Optimization in Field Development. SPE Annual
Technical Conference and Exhibition, Denver, Colorado, 5-8 October 2003, Society of Petroleum
Engineers. doi: 10.2118/84191-MS
Castrogiovanni, A., A. N. Fitzpatrick and C. H. Ware (2011). Benefits and Technical Challenges of
Downhole Steam Generation for Enhanced Oil Recovery. Canadian Unconventional Resources
Conference, Calgary, Alberta, Canada, 15-17 November 2011, Society of Petroleum Engineers. doi:
10.2118/149500-MS
CMG, S. (2017). Advanced Processes & Thermal Reservoir Simulator, User's Guide. C. M. Group.
Calgary, Canada, Computer Modelling Group Ltd.
Curtis, C., R. Kopper, R. Decoster, A. Guzman-Garcia, C. Huggins, L. Knauer, M. Minner, N. Kupch, L.
M. Linares, H. Rough and M. Waite (2002). Heavy-Oil Reservoirs. Oilfield Review, Schlumberger.
14: 30 - 51.
Eberhart, J. K. a. R. C. (1995). Particle swarm optimization, Perth, Australia, IEEE Service Center,
Piscataway, NJ.
Ermolaev, A. and A. Kuvichko (2013). Non-Regular Well Placement Optimization. SPE Arctic and
Extreme Environments Technical Conference and Exhibition, Moscow, Russia, 15-17 October
2013, Society of Petroleum Engineers. doi: 10.2118/166903-MS
Gates, I. D. and N. Chakrabarty (2005). Optimization of Steam-Assisted Gravity Drainage in McMurray
Reservoir. Canadian International Petroleum Conference, Calgary, Alberta, Jun 7 - 9, 2005,
Petroleum Society of Canada. doi: 10.2118/2005-193
Güyagüler, B. and R. N. Horne (2004). "Uncertainty Assessment of Well-Placement Optimization." doi:
10.2118/87663-PA
Jinzhong, L., G. Wenlong, W. Bojun and H. Jihong (2013). Feasibility Study of In-Situ Combustion Huff
and Puff for EOR in Super-Deep Heavy Oil Reservoir. International Petroleum Technology
Conference, Beijing, China, 26-28 March 2013, International Petroleum Technology Conference.
doi: 10.2523/IPTC-16408-MS
Khan, M. Y. (1993). Theory & Problems in Financial Management. Boston, Boston: McGraw Hill Higher
Education.
Mezura-Montes, E. and C. A. Coello Coello (2011). "Constraint-handling in nature-inspired numerical
optimization: Past, present and future." Swarm and Evolutionary Computation 1(4): 173-194. doi:
https://doi.org/10.1016/j.swevo.2011.10.001
Onwunalu, J. E. and L. J. Durlofsky (2010). "Application of a particle swarm optimization algorithm for
determining optimum well location and type." Computational Geosciences 14(1): 183-198. doi:
10.1007/s10596-009-9142-1
Pedersen, M. E. H. (2010). Good Parameters for Particle Swarm Optimization. Hvass Laboratories, Hvass
Laboratories. HL1001: 1 - 12.
Putra, E. A. P., Y. A. Rachman, T. Firmanto, R. Arsyadanie and G. Hafizh (2011). Case Study : Cyclic
Steam Stimulation in Sihapas Formation. SPE Asia Pacific Oil and Gas Conference and Exhibition,
Jakarta, Indonesia, 20-22 September 2011, Society of Petroleum Engineers. doi: 10.2118/147811-
MS
Qin, A. K., V. L. Huang and P. N. Suganthan (2009). "Differential Evolution Algorithm With Strategy
14 SPE-192378-MS

Adaptation for Global Numerical Optimization." IEEE Transactions on Evolutionary Computation


13(2): 398-417. doi: http://dx.doi.org/10.1109/TEVC.2008.927706
Scott, G. R. (2002). Comparison of CSS and SAGD Performance in the Clearwater Formation at Cold
Lake. SPE International Thermal Operations and Heavy Oil Symposium and International
Horizontal Well Technology Conference, Calgary, Alberta, Canada, November, 2002, Society of
Petroleum Engineers. doi: 10.2118/79020-MS
Storn, R. and K. Price (1997). "Differential Evolution – A Simple and Efficient Heuristic for global
Optimization over Continuous Spaces." Journal of Global Optimization 11(4): 341-359. doi:
10.1023/a:1008202821328
Vega Riveros, G. L. and H. Barrios (2011). Steam Injection Experiences in Heavy and Extra-Heavy Oil
Fields, Venezuela. SPE Heavy Oil Conference and Exhibition, Kuwait City, Kuwait, 12-14
December 2011, Society of Petroleum Engineers. doi: 10.2118/150283-MS
Yee, C. T. and A. Stroich (2004). "Flue Gas Injection Into a Mature SAGD Steam Chamber at the Dover
Project (Formerly UTF)." Journal of Canadian Petroleum Technology 43(1): 21 - 29. doi:
10.2118/04-01-06

You might also like