optimization of the water alternating gas injection...simultaneous water alternating gas (swag)...
TRANSCRIPT
1
Optimization of the Water Alternating Gas Injection
Compositional fluid flow simulation with Water Alternating Gas Injection optimization on the upscaled synthetic reservoir CERENA-I
Fabusuyi, Oluwatosin John; Quintao, Maria Joao; Azevedo, Leonardo; Soares, Amílcar
Email addresses: [email protected]; [email protected];
[email protected]; [email protected]
Centre for Petroleum Reservoir Modelling
Instituto Superior Técnico
Avenida Rovisco Pais, 1
1049-001 Lisboa
Abstract- This work focuses on the optimization
of the production strategy on an up-scaled
synthetic reservoir CERENA-I, which mimics
some characteristics of a Brazilian Pre-Salt field.
This reservoir has a saturated oil leg with a
retrograde condensation gas cap, both with a
high CO2 content. The production strategy
involved the implementation of a simultaneous-
water alternating gas injection scheme (SWAG).
The objectives for this study were to increase the
oil recovery while reducing the gas production
and the parameters selected for optimization in
this study were the bottom-hole pressure, the
well position, the injection rate and WAG ratio.
The effects of these variables were studied in
order to achieve an optimal solution.
Keywords: Reservoir simulation, compositional
simulation, PVT analysis, Synthetic reservoir,
Simultaneous WAG scheme, Particle Swarm
Optimization, objective function.
1. Introduction
This study is a continued interest in the
CERENA-I reservoir created by Pedro Pinto [10].
from the Brazilian Pre-Salt play (figure 1), which
has a very high content of CO2. This Brazilian
Pre-salt reservoir poses great challenges in
every aspect of its production, from reservoir
modelling and management, to surface facilities.
The reservoir covers an area of 567 km2 about
300km offshore of Rio de Janeiro, in the Santos
basin.
Fig 1: The Brazilian Pre-Salt Play (Source: ANP)
It is situated in water depths of around 2000m,
with the top of the reservoir situated at
approximately 5200m. It has a 90m thick heavy
oil leg with 18o API and 55% (molar) of CO2
content. It also has a gas cap of retrograde
condensation gas which contains approximately
60% (molar) of CO2.
The idea to maximize the production of oil in the
reservoir led to the development of a single well
Simultaneous Water Alternating Gas (SWAG)
injection scheme on the reservoir instead of the
initial injection of water and gas produced in 2
different wells. SWAG is an enhanced oil
recovery process in which gas is mixed with
water outside the well and the mixture is then
injected as a two phase mixture in the well or,
alternatively, both gas and water are injected at
the same time into the well to get better oil
recovery. Water and gas injection are the best
2
solution to cope with the problems such as early
breakthrough which occur only when gas is
injected individually due to unfavorable oil-gas
mobility ratio. Hence, simultaneous injection of
gas and water would be of greater importance to
improve the sweep efficiency by improving the
displacement front [6].
Finding an optimal depletion strategy for
hydrocarbon production has always been a key
subject in reservoir management. The underlying
problem to be solved is generally the
maximization of a key quantity such as oil
production, net present value (NPV), etc. In the
past, optimal settings of the optimization
parameters were almost exclusively determined
manually. This is generally quite time-consuming
procedure with a high likelihood of obtaining
suboptimal results. While manual approaches
are still predominant strategies in the reservoir
management practice, due to the maturity of
most existing major oilfields and gradual
decrease in large oil discoveries, research for
more systematic optimization approaches has
been initiated. The optimization technique used
in this study was the particle swarm optimization
which was used in the Raven software provided
by the Epistemy Company
The initial objective of this research work was to
find a production strategy to optimize oil
production and reduce the quantity of CO2 being
produced, and as the researched progressed,
different ideas were introduced. The different
parameters to be optimized were introduced and
discussed, these include parameters related to
the production wells and others related to the
injection wells. During the course of the thesis,
due to computational constraints for the
simulation and optimization procedure, the
reservoir CERENA-I was up-scaled, and the up-
scaled version was used henceforth.
2. The synthetic reservoir: CERENA-I The CERENA-I model was created to replicate
some characteristics of the Brazilian Pre-salt
carbonate fields and it contains high-resolution
data sets of petro-physical and petro-elastic
properties. For the case study presented herein
only the sets of porosity and permeability were
used. The model is composed of two facies: a
reservoir facies, composed by microbiolites; and
a non-reservoir facies composed by mudstones,
on a corner-point grid with 161x161x300 cells,
with 25x25x1m spacing.
Fig 2: CERENA-I porosity model
Fig 3: Histogram of porosity for both facies
Permeability was modelled recurring to the
porosity model and it exhibits a dependence that
was derived from real analogues [4].
3
3. Dynamic simulation
Due to the lack of real data from analogue fields
the oil composition for this study was obtained
from a generic sample of oil from Petrel's®
library, grouped to reduce computation time and
memory requirements, and with the molar
percentages re-adjusted to the known CO2
content of the analogue field (Table 1).
Table 1: Molar percentages of the oil with grouped components.
Component Molar % Mol. weight
CO2 55.00 44.01
C1 16.56 16.043
C2 4.46 30.037 C3 3.15 44.097
C4-6 5.69 70.237
C7+ 15.11 218
For this case study, the three parameter Peng-
Robinson equation of state was chosen, and
tuned to match the estimated PVT observations
(Table 2).
Table 2: Estimated saturation pressures
Bubble point (bar) Dew point (bar)
493 400 Due to the huge number of cells in the reservoir,
the choice was made to run the simulation on a
fine grid sectoral model (figure 4) which, despite
being considerably smaller, when compared to
the original model, reproduces the total variability
of the full field.
Fig 4: Sectorial model area
Despite doing this, the computational time and
memory needed for the number of iterations
needed during the optimization process was
really enormous, hence the sectorial reservoir
was up-scaled. The porosity and permeability
distribution in the up-scaled reservoir were made
to replicate the distribution in the original
sectorial model. The new up-scaled sectorial
model contains a combination of grid cells of
about 22 x 22 x 154 cells compared to the 45 x
42 x 300 cells in the original sectorial model.
Fig 5: Up-scaled perm x and y (left), original perm x and y (right)
Fig 6: Up-scaled perm z (left), original perm z (right)
Fig 7: Up-scaled porosity (left), original porosity (right)
Figures 5 to 7 show the visual differences
between the up-scaled and original sectorial
models of the permeability and porosity models.
The trends, facies and distributions obtained in
the original sectorial model can also be observed
in the up-scaled model with variations. From this
point on, the link to the original full field model
4
and the sectorial model is severed and the study
object is now the up-scaled sectorial model. For
this reason no boundary effects will be added to
the dynamic model, to account for the influence
of the remaining area.
Fig 8: Well Locations
The well pattern chosen for this study was a
traditional five-spot configuration with four vertical
producer wells in the corners and one vertical
injector well in the center (Figure 8).
Table 3: Fluids originally in place
The model was initialized and the fluids in place
(Table 3) were calculated for the equilibrium
conditions.
We first chose to produce the gas cap, to
access its liquid condensate fraction. The fluid
was condensed in surface separators and the
resulting dry gas was re-injected back into the
gas cap, to help keep reservoir pressure. The
gas cap was produced for one year, after which
the completions of the producer wells were
closed in this zone and opened in the oil leg.
4. Optimization results
The optimization technique selected for this study
was the particle swarm optimization technique.
This was implemented in a software called Raven
from the Epistemy Company. Four different
parameters were optimized during this study, the
bottom-hole pressure of the four production wells,
the injection rate, the WAG ratio and the well
position. The results obtained are shown below.
4.1 Production well Bottom-hole
pressure (bhp) Two strategies were considered for this
optimization, we considered having the same bhp
values for the 4 wells and we also considered
varying the bhp values for each well. This two
ideas were tested and optimized and the results
shown below.
Same Bottom-hole Pressure
These results were obtained after 255 iterations
of PSO-based algorithm optimization. It can be
observed that the suitable BHP for these wells
was obtained between 200 to 500 bars.
Fig 9: Same BHP Gas production opt
Reservoir volume of oil Reservoir volume of gas
1.5 x 107 sm3 1.46 x 1010sm3
5
Fig 10: Same BHP Oil production opt
A closer look at the plot indicates that as the BHP
increases from 200 until it peaks at about 454
bars, the Field Oil Production Total increases
gradually while the reverse happens for the Field
Gas Production Total as evidenced by figures 9
and 10.
Fig 11: FGPT vs FOPT for same BHP
The optimal bhp for these 4 wells to operate at
optimal condition would be at the peak pressure
of 454 bars. The plot helps us to understand how
the FOPT and FGPT are inter-related in terms of
the BHP, thus we can infer from this study the
importance of the BHP on the productions of oil
and gas.
Different Bottom-Hole Pressure
With the same objective functions in mind, the
proposed task was to conduct the optimization
simulations by observing the production runs
when the bhp of the 4 production wells varied as
opposed to having the same value as shown
above. The bhp was assigned the letters j, k, l, m
respectively and the optimization simulation was
conducted within the same range. After several
days and over 300 iterations, the results obtained
from this optimization are presented below.
Fig 12: Different BHP Gas production opt
Fig 13: Different BHP oil production opt
The results obtained from this simulation did not
produce any useable results and took several
days and hours to obtain any recognizable
results. This can also be buttressed by the plot of
FOPT and FGPT shown below in figure 14.
4.2 Injection rate and WAG ratio When conducting a WAG scheme, one important
factor inter-twined with the injection rate is the
Water Alternating Gas ratio. This variable is
defined as the ratio of the volume of water
6
injected to the volume of gas injected. In this
scheme, a single injection well is used and both
water and gas are injected together without
mixing at the top.
Fig 14: FGPT vs FOPT for same BHP
In this study, the WAG ratio can be explained by
the expression below,
WAG ratio = Volume of water injected: Volume of Gas
≡ Water Injection rate: Gas injection rate
≡ Water injection rate: (Water injection rate x 𝑘
𝑗)
≡ j: k
To start, a random WAG ratio was selected and
different injection rates were tested with this
WAG ratio, a certain trend was observed in all of
them with the best oil production being observed
when the WAG ratio favours a high injection of
gas over water but with a high volume of gas
produced along with it. The reverse is the case
when a high volume of water injection is favoured
over the gas injection: in this case, the oil
production decreases and the gas production
also decreases. The visible trend in the figure 15
shows the region that satisfies our objective
functions, where we are able to produce oil
maximally and gas minimally. A result from one
of the injection rates used is shown in figure 15,
where the injection rate selected was 7570
sm3/day. Some selected results from this
simulation are shown in figure 16.
Fig 15: WAG ratios at 7570
Fig 16: 7570 FOPT results
The different WAG ratios along the trend lines
were able to fully satisfy our objective functions
and one of them was selected and maintained for
the rest of this study. The WAG ratio that was
selected was the ratio 2:3. With this WAG ratio,
an optimization simulation was conducted to
determine the optimal injection rate for the
injection well in this production, and the results
obtained are shown in figure 17 and 18. We can
observe 3 different sections. The first section
corresponds to the region where the oil
production increases as the injection rates also
increase, until an injection rate of about
32,000sm3/day. Afterwards, an increase in the
injection rate causes a significant jump of
production in about 100,000sm3. This total oil
production is the same despite the continual
increase of the injection rate until a new oil
7
production total is obtained observed at an
injection rate of 40,214 sm3/day. From this point
forward, no further difference is observed. If the
corresponding gas production total is observed
and compared to the oil production total. All
through this first 2 stages observed in the oil
production plot there was still an increase in the
gas being produced up until an injection rate of
about 46,000sm3/day. Since our initial aim was to
increase oil recovery and reduce gas production,
it was necessary to pick the injection rate that
best satisfies this aim. The injection rate with the
most oil recovered and less gas produced is the
40,214sm3/day, which is the optimal injection rate
for the parameters and conditions that were used
for this final simulation.
Fig 17: Optimal FOPT results
Fig 18: Optimal FGPT results
4.3 Well’s positions The aim of this optimization simulation is to help
us make a better decision in the placement of our
wells in the reservoir to enhance better oil
recovery.
Fig 19: Reservoir Quadrants
The optimization was carried out by dividing the
reservoir into 4 quadrants with a well sited in
each quadrant and the assumption is that each
well will be sited in optimal locations in its
respective quadrant in order to maximize the total
oil recovery. The simulation was conducted with
the optimal parameters observed in the previous
simulations, 454 bars for the production bhp and
40,214sm3/day for the injection rate, and
maintained for the whole optimization process. In
the optimization, each well is only allowed to
move and be optimized within its own quadrant,
and the field oil production total for each
simulation is used in understanding the
optimization results. The results obtained from
the optimization results are presented in a layer
map of a reservoir. The image above shows the 4
different quadrants that the reservoir was divided
to, A, B, C, and D. It also shows the different
locations that were tested during the optimization
process. The aim of the optimization is to indicate
the regions with the most likely better oil recovery
in the reservoir. As it can be seen in figure 20, in
the first quadrant, it would be advisable to place
the first well around the edge of the reservoir, the
north-west region, as shown in the figure 40a
below. In the second quadrant, B, the best region
for the well placement for better oil recovery
would be the upper North east region of the
reservoir as seen in figure 40b above. In the third
8
quadrant, C, the best region for the placement of
the well is the south-south region of the reservoir
as shown in figure 40c. The best region for well
placement in the fourth quadrant is the south-
east region of the reservoir. The regions around
the edges with the best porosity and permeability
values also coincide with the regions good for
well placements as shown in figure 21 and 22.
This would imply that there is ease of flow and
also larger storage of oil in those regions.
The conclusions taken from these results were
also observable in the simulation with the best
result of about 10.2 million sm3.
Fig 20: Well placement vs FOPT optimization
This was obtained with the iteration 95 with
coordinates (1, 18) in quadrant A, (21, 21) in
quadrant B, (9, 1) in quadrant C, and (21, 4) in
quadrant D.
Fig 21: Well optimization regions
Fig 22: Suitable regions and porosity models
5. CO2 Capture One of the tasks of this thesis was also to help
take care of the CO2 produced. In the work
previously done by Pedro Pinto (2014), all the
gas produced was reinjected into the reservoir
thereby avoiding the separation of this CO2 from
the gas being produced because of the
percentage composition of the gas which is about
60 percent molar content of the CO2. There are
several processes available for the removal of
this CO2 irrespective of the high molar content
that is observed in this field. The best method for
this peculiar case in this study is the Fluor solvent
process. The Fluor solvent process is one of the
most attractive processes for gas treating when
the feed gas CO2 partial pressure is high (> 60
psia), or where the sour feed gas is primarily
CO2. The process is based on the physical
solvent propylene carbonate (FLUORTM) for the
removal of CO2. Propylene carbonate (C4H6O3),
is a polar solvent with high affinity for CO2 and αij
values of C1 or C2 to CO2 are high, therefore
hydrocarbon pickup in the rich solvent and
subsequent hydrocarbon losses in the CO2 vent
stream are minimal. Earlier the FLUOR solvent
process was configured to treat very narrow
range of feed gas compositions. Recently new
configurations have been developed for treating
high CO2 content sour gas [11]. The feed gas
pressure in this case varies from 400 – 1200 psig
with the CO2 content varying from 30-80 % and
more. High CO2 content in the feed gas
increases the amount of refrigeration produced
9
by the flash regeneration of the rich solvent. At
very high CO2 partial pressures, the cooling effect
from flash regeneration will exceed the cooling
required for CO2 absorption. Also the viscosity
and surface tension of propylene carbonate
increases dramatically and the absorber mass
transfer rate drops drastically. This negatively
impacts the process, therefore overcooling of the
solvent should be avoided. The excess
refrigeration is harnessed in this application by
lowering the absorption column temperature with
refrigeration generated from flashing the rich
solvent from high to medium pressure. This
allows the absorber to operate at a lower
temperature and increases the solvent loading.
The flashed gasses are compressed and
recycled to reduce hydrocarbon losses in the
CO2 vent. Excess refrigeration generated by
flashing of the rich solvent flowing to the first
stage flash drum is used to cool and condense
the CO2 vent stream from the atmospheric and
vacuum flashes. The condensed CO2 can be
used for EOR or disposed of by injecting the
liquid into an underground formation.
6. Conclusions The aim of this work is to optimize the oil
recovery of the reservoir under study, through a
multidisciplinary approach that includes not only
reservoir modelling, reservoir engineering but
also a glimpse of chemical engineering. An
original reservoir from the Brazilian pre-salt was
modelled to form the CERENA-I static model, to
test the reservoir performance and production
strategies. Reservoir conditions were borrowed
from a real analogue pre/salt field close-by and
the equation of state was tuned to match an
estimated bubble point and dew point.
Due to lack of computational memory, a sectorial
model was carved out of the original reservoir
and also due to amount of iterations that would
be needed for the optimization process in terms
of the computational time, the sectorial model
was further up-scaled. This was an attempt to
produce a sectorial model with a smaller amount
of cells but retaining the same variability in terms
of its porosity, permeability and also faces
distribution. From this point onwards, the up-
scaled sectorial model was used for the
optimization process. From the results obtained
we can see the effect of the bottom-hole pressure
in determining the inter-relationship between the
oil and gas produced. It can be concluded that for
maximum oil recovery, it would be advisable to
maintain the same bottom-hole pressure for the 4
production well rather than varying them. We
could also observe that the oil production
increases as the bhp increases until it got to a
threshold of about 454 bars and consequently
decreases despite the increase in the bhp. A
trend walk towards the optimal bhp shows a
gradual decrease in the amount of gas being
produced, which shows the importance of
maintaining the bhp for as long as possible to
improve oil recovery before the oil converts to the
gaseous phase. Another part of this work was the
optimization of the injection rate in order to
maintain the reservoir pressure for as long as
possible for better oil recovery. This was carried
out by simultaneous injection of gas and water.
The injection rate and the WAG ratio were
optimized for this case being considered. Initially,
we attempted to optimize the WAG ratio for each
injection rate that was tested, but we discovered
a similar trend as we increased the injection rate
and since this kept leading to a further increase
in oil production, we reversed the idea by actually
selecting a common WAG ratio to all the injection
10
rates we tested and then tried to optimize the
injection rate. The optimal WAG ratio selected
from the available result was the ratio 2:3. An
optimization of the injection rate at this WAG ratio
was conducted. The result, as shown on the plot
obtained (fig 17), indicates that the ideal injection
rate would be the injection of water at about
40,214sm3/day while the gas would be at about
60,000sm3/day. At this rate we were able to
recover about 10,200,000 sm3 of oil from a
possible 15 million sm3 of oil in place in the
reservoir, which is about 68% oil recovery. We
also considered how the placement of the wells
affects the total oil production. We can observe
better oil recovery in the well locations with
corresponding configurations around the north-
east and north-west of the upper region and the
south-south and south-east of the reservoir.
These regions are characterized by much higher
porosity than the other part of the reservoir which
indicates possible oil storage in this region and
also high permeability values which indicates
ease of movement of the oil into the wellhead.
7. References
[1] Archer, J. S., & Wall, C. G. (1986). Petroleum
engineering: principles and practice. London:
Graham and Trotman Ltd.
[2] Christensen, J. R.; Stenby, E. H., Skauge,
(2001) A. Review of WAG field experience. SPE
Reservoir Evaluation & Engineering.
[3] Doghaish, N. M. (2008). Analysis of Enhanced
Oil Recovery-A Literature Review. Dalhousie
University. Halifax: unpublished work.
[4] Horta, A., & Soares, A. (2010). Direct
Sequential Co-Simulation with Joint Probability
Distributions. Mathematical Geosciences.
[5] Kansas Geological Survey (2004).
Sedimentologic and Diagenetic Characteristics of
the Arbuckle Group.
[6] Meshal, A., G. Rida and M. Adel, 2007. A
parametric Investigations of SWAG injection
technique. SPE paper # 105071 prepared to be
presented in 15th SPE Oil and Gas Show,
Bahrain 11-14th March
[7] Nezhad, S., Mojarad, M., Paitakhti, S.,
Moghadas, J., & Farahmand, D. (2006).
Experimental Study on Applicability of
Water.Alternating-CO2 injection in the Secondary
and Tertiary Recovery. First International Oil
Conference and Exhibition in Mexico (pp. 1-4).
Cancun: Society of Petroleum engineers
[8] Nocedal, J. and Wright, S.J.: “Numerical
Optimization”, Second Edition, Springer press,
2006.
[9] Onwunalu, J., Durlofsky, L., 2011. A new well-
pattern-optimization procedure for large-scale
field development. SPE Journal 16 (3), 594–607.
[10] Pedro Pinto (2013). Dynamic simulation on
the synthetic reservoir CERENA I; Compositional
fluid flow simulation with 4D seismic mitoring on a
reservoir with a large content of CO2.
[11] Salako Abiodun Ebenezer, 2005 Removal of
Carbondioxide from Natural gas for LNG
production. Semester Project Work, Institute of
Petroleum Technology, Norwegian University of
Science and Technology, Norway.
[12] Saleem Qadir Tunio, Tariq Ali Chandio and
Muhammad Khan Memon, 2012. Comparative
Study of FAWAG and SWAG as an Effective
EOR Technique for a Malaysian Field. Research
Journal of Applied Sciences, Engineering and
Technology 4(6): 645-648, 2012.