You are on page 1of 10

Copyright 2012, Brazilian Petroleum, Gas and Biofuels Institute - IBP

This Technical Paper was prepared for presentation at the Rio Oil & Gas Expo and Conference 2012, held between September, 17-
20, 2012, in Rio de Janeiro. This Technical Paper was selected for presentation by the Technical Committee of the event according to
the information contained in the final paper submitted by the author(s). The organizers are not supposed to translate or correct the
submitted papers. The material as it is presented, does not necessarily represent Brazilian Petroleum, Gas and Biofuels Institute
opinion, or that of its Members or Representatives. Authors consent to the publication of this Technical Paper in the Rio Oil & Gas
Expo and Conference 2012 Proceedings.
______________________________
1
M.Sc., Digital Infrastructure Practice Coordinator Halliburton
2
M.Sc., Senior Reservoir Consultant Halliburton
3
Associate Reservoir Consultant Halliburton
4
M.Sc., Senior Petroleum Engineer Petrobras
5
Petroleum Engineer Petrobras
6
M.Sc., Partnership Coordinator Petrobras
IBP1332_12
INTEGRATED METHOD TO OPTIMIZE WELL
CONNECTION AND PLATFORM PLACEMENT ON A
MULTI-RESERVOIR SCENARIO
Sergio H. G. de Sousa
1
, Marcelo G. Madeira
2
, Martha S. Frana
3
,
Rosane O. Mota
4
, Edilon R. da Silva
5
, Vanessa P. S. King
6

Abstract
This paper describes a workflow created to optimize the platform placement and well-platform connections on a multi-
reservoir scenario using an integrated reservoir simulator paired with an optimization engine. The proposed
methodology describes how a new platform, being incorporated into a pre-existing asset, can be better used to develop
newly-discovered fields, while helping increase the production of existing fields by sharing their production load. The
sharing of production facilities is highly important in Brazilian offshore assets because of their high price (a few billion
dollars per facility) and the fact that total production is usually limited to the installed capacity of liquid processing,
which is an important constraint on high water-cut well production rates typical to this region. The case study asset used
to present the workflow consists of two deepwater oilfields, each one developed by its own production platform, and a
newly-discovered field with strong aquifer support that will be entirely developed with a new production platform.
Because this new field should not include injector wells owing to the strong aquifer presence, the idea is to consider
reconnecting existing wells from the two pre-existing fields to better use the production resources. In this scenario, the
platform location is an important optimization issue, as a balance between supporting the production of the planned
wells on the new field and the production of re-routed wells from the existing fields must be reached to achieve
improved overall asset production. If the new platform is too far away from any interconnected production well,
pressure-drop issues along the pipeline might actually decrease production from the existing fields rather than augment
it. The main contribution of this work is giving the reader insights on how to model and optimize these complex
decisions to generate high-quality scenarios.
1. Introduction
Exploration activities on the Brazilian subsalt frontier are heavily concentrated in nearby pre-existing field
developments so that, in case of a successful exploration campaign, the new development projects can share the
installed means of production and delivering of fluids. This setup brings several advantages to oil companies:
development costs reduction, earlier project payback, risk mitigation, etc. Of course, there is always a downside; in this
case it is the added complexity to both asset design and management. Complexity arises from every aspect of the
project, including multiple reservoir modeling and simulation, multiphase flow, pressure drop along pipelines, artificial
lift, and subsea layout generation.
Integrated reservoir modeling has different meanings to different people (Settari et al. 2001; Coats et al. 2004;
Evans 2007; Campozana 2008; Huang et al. 2011), but the trend shows that this term is generally used when traditional
reservoir simulation techniques (fluids flow modeling on a porous medium) is integrated with disciplines outside of
reservoir engineering, such as facilities modeling, geomechanics, geostatistics, and geochemistry. In this paper,
however, the term, integrated reservoir modeling, refers to the modeling (and simulation) of both the reservoir fluid-
Rio Oil & Gas Expo and Conference 2012
2
flow equations with the pipelines and risers that lead to the production/injection facilities and also part of the production
plant.
This work used a fully-coupled, next-generation, reservoir simulation described by Coats et al. (2004) and also
Al-Matar et al. (2007). It is capable of modeling both the traditional reservoir-simulation equations as well as the
surface-flow equations from the wellbore up to the processing plant. This formulation allows for a more holistic view of
the asset, while allowing multiple reservoirs to be simulated together when they share a common surface network. The
optimization side of the work was performed with an uncertainty analysis and optimization tool, described by Evans
(2007), capable of working with the chosen simulator. Paring both tools allows uncertainty and optimization workflows
to be performed on the reservoir model and/or the surface network.
The goal of this work was to develop and test a methodology of integrating a newly-discovered field in a
previously-developed asset from the determination of the number of wells to develop the new field along with the
location of a new production facility and, finally, determining if it is advantageous to reconnect wells from the existing
reservoirs to the new production facility. The proposed methodology, except for the study of reconnecting existing
wells to the new production facility, was implemented in a real Brazilian offshore asset. The case study shown in this
paper is not, however, the original set of fields because of confidentiality issues.
2. Methodology
The proposed methodology has two main optimization steps. The first one narrows the scope of the analysis
down to the newly-found field, where a local optimization effort attempts to determine how many wells should be used
to develop the field and where to locate them. On the next step, the scope widens to the entire asset, and a new
production facility for the new field and its accompanying surface network is modeled so that both the platform location
and which and how many wells of the existing fields will have their production redirected to it. Both optimization
stages are detailed below.
2.1. New Field Development
This optimization stage was performed in the same fashion described by Sousa et al. (2010), but in a simpler
context, as the new field model used in this work has a strong aquifer and will not need, in this stage, to be developed
with the support of water injection (please refer to the Simulation Case Study section for more details). The basic
optimization steps are:

1. Calculation of a quality map to guide the definition of possible well targets.
2. Find highest-quality targets subject to constraints on target overlapping and cuts on the quality value.
3. Positioning of new production wells on each defined well target.
4. Find a subset of producer wells that maximize oil production while minimizing the capital expenditures.

The following calculation was performed in the static and dynamic properties of the grid to generate a 3-D
quality map.
Quality
]k
= I
]k

]k
So
]k

ln(Kx
i]k
+1)
ln(Kx
mcx
+1)
(1)
Equation 1 shows the calculation of the new grid property Quality that takes place for every cell in the grid,
thus the use of the grid block indexes, i, j, and k. V
ijk
, stands for the rock volume,
ijk
, to the porosity, while the So
ijk

represents the cells oil saturation and Kx
ijk
its horizontal permeability. The first three factors on Equation 1 calculate
the total oil volume of the cell. The last factor is actually a permeability penalizer that will multiply the oil volume by a
fractional amount. The only grid blocks that will not be penalized are the ones whose horizontal permeability are equal
to the maximum horizontal permeability found on the grid (Kx
max
).
In the context of this optimization step, well targets can be conceived as a drainage area of the potential
producer well to be allocated in a region. Thus, an appropriate target radius should be specified to guide the target
location-optimization algorithm described next. The optimization process begins with a new target being traversed
through every grid block that has a quality value above a user-defined threshold, and the objective function of that
location is the sum of quality values of all blocks intersected by the target radius. The location with the highest
objective function accommodates the initial target. The process is then repeated with the additional constraint that the
new target radius cannot overlap with existing ones. The process ends when no new cell has a quality value above the
threshold or when no valid grid block can accommodate a new target as a result of overlapping issues with existing
targets.
Rio Oil & Gas Expo and Conference 2012
3
Despite the quality map having a recurrent property (oil saturation) in its formula, the real quality of the target
location depends on dynamic issues like fluids flow, proximity to water sources and/or other wells, reservoir pressures,
etc. For this reason, an optimization process using simulation is applied to identify an optimized subset of targets that
maximize oil production while keeping capital investments under control. The optimization process starts by locating a
producer well on all targets identified in the previous step. A set of simulation runs are then performed by an optimizer
where a subset of the wells are open to flow (selected wells) while the rest of the wells remain closed (eliminated
wells). It should be clear that if the objective function is simply cumulative oil production, optimized solutions would
tend to select all wells to be open because closing even one well that produces, assuming 1 m/day, would lower the
cumulative oil production unless in the presence of production constraints, where wells with high water cuts could
adversely impact oil production.
In the absence of economic data to model the capital expenditures of drilling a well on the field, one can
represent the cost of drilling in terms of oil volume (i.e., the amount of oil that needs to be produced by a well for it to
be viable). Thus, the objective function in this simplified scenario could be:
(x, c) = Np (x

(2)
In Equation 2, Np is the field cumulative oil production, x
i
is a binary decision value that is set to 1 when well i
is selected and 0 otherwise, and c
i
is the cost of drilling well, i. It might be convenient to specify constraints on the
number of selected wells during the optimization process to avoid invalid/inadequate solutions. For instance,
considering the concession time and desired recovery factor on the field, a constraint could be set on the decision
variables to specify a minimum number of wells that should be drilled (Equation 3). Conversely, owing to platform well
slot constraints, it might be necessary to specify an upper limit on how many wells can be drilled (Equation 4).
(x

minimumnumberofwells (3)
!

maximumnumberofwells (4)
Thus, the optimization problem can be posed by modeling the selection of targets with binary values, defining
an objective function similar to Equation 2 to be maximized and a set of constraints, like Equations 3 and 4, tailored to
the problem at hand.
2.2. Platform Location and Facilities Sharing Optimization
This part of the optimization process depends largely on the level of detail of the surface network that will be
implemented. On this work, the following assumptions were made:

1. The seabed topology is modeled by a set of UTM coordinates covering the target area.
2. The risers catenary trajectory will be represented by a set of 3 straight segments running from the
platform down to the touch down point.
3. The riser and pipeline connections from the platforms to the wellheads are straight trajectories when
the depth (z) axis is not considered.
4. Technical and safety radius around platforms can be obtained from linear relationships on the water
depth of the platforms locations.

In this stage of the optimization process, there is a new field with optimized locations for its wellheads, which
should all be interconnected with the new platform. At the same time the new platform, x and y, UTM coordinates are
optimized, the reallocation of a subset of the existing production wells from other fields to the new platform will be
tested in an attempt to maximize installed production capabilities on the asset. The following sections detail important
parts of this optimization stage.
2.2.1 Seabed Topology Modeling
The seabed topology is typically given by a set of triangulation points that can be quite expensive to traverse
during an optimization run where dozens of z coordinates on the seabed need to be determined. Some kind of proxy
model for the seabed surface model can be used to speed things up.
A natural choice would be to train a neural network to model the seabed topology. In fact, the authors have
used this approach with some success in the past, but a more straightforward approach inspired by geostatistics was
devised that proved much simpler to implement. The idea is to sample the data points of the seabed so that you end up
with a manageable size data set that can be manipulated in milliseconds. Given a new data point with x
p
and y
p
UTM
coordinates and a z
p
coordinate that needs to be calculated, the following expression can be used to do so:
Rio Oil & Gas Expo and Conference 2012
4
z
p
=
z
i
[d
mcx
2
-d
ip
2

i
[d
mcx
2
-d
ip
2

i
(5)
Equation 5 calculates a weighted average for the coordinate, z, of the new data point, p, where the weights are
the difference between the square of the data point with the maximum horizontal distance to the new data point (J
mux
2
)
and the square of the horizontal distance from the point, p, to the seabed surface data point, i. This way, the greater the
distance of data point, i, the less influence it will have on the calculation of z
p
; the closer these two points are, the
greater the influence coordinate, z
i
, will have.
2.2.2 Reference Nodes TDP, CRF, and RDS
According to assumption 3 on section 2.2, the surface connections from the platform down to the wellhead are
modeled in a straight line. Thus, the horizontal distance between all nodes can be measured along a radius around the
platform coordinates in the direction of the wellhead. Upon research of the current surface projects that the authors were
exposed to, there are 3 main platform radiuses that are modeled on facility design:

1. TDP Touch Down Point: where the riser touches down the seabed.
2. CRF Connection between Riser and Flowline: a distance beyond which the vertical movement of
the riser because tide is no longer observed.
3. RDS a security radius around the platform where no wellhead can be to guarantee platform safety in
case of gas leaks, which can affect floatability.

In all projects that were researched, there was a pattern on determining these radii from the platform, which
were linearly dependent to the water depth (WD) of the platform location, as shown in Figure 1.

Figure 1. Schematic representation of network nodes and connections between a platform and a wellhead
Given that the UTM coordinates for the platform location and the wellheads are known, it is an easy task to
calculate the coordinates of the TDP, CRF, and RDS nodes as a function of the platform and wellhead coordinates:
(x

, y

) = (x
pIut
, y
pIut
) +

i

wh
(x
wh
x
pIut
, y
wh
y
pIut
) (6)
Equation 6 is a parameterized equation where x
i
and y
i
are the searched UTM coordinates for nodes TDP, CRF,
and RDS, x
plat
and y
plat
are the UTM coordinates of the platform, r
i
is the node radius (in function of the platform water
depth, see Figure 1), r
wh
is the wellhead radius (horizontal Euclidean distance between the platform and the wellhead).
2.2.3 Riser Modeling
To model the riser nodes CAT1, CAT2, and CAT3 (refer to Figure 1), a normalized equation for the catenary
(u z 1 and u r 1) was created (Equation 7).
z = 1
ucosh[
1-r
c
-ucosh(0)
ucosh[
1
c
-ucosh(0)
(7)
Rio Oil & Gas Expo and Conference 2012
5
In Equation 7, a is a non-zero positive constant that increases the catenarys curvature (values closer to zero) or
decreases it (values tending to one and above). The resulting catenaries for different values of a can be seen in Figure 2.



Figure 2. Normalized catenary curves from Equation 7 for different values of a
To use Equation 7 to find normalized values of z from actual radius, you should first normalize the radius
value, dividing the radius by the radius of the TDP node and inserting this calculated value into the expression. The
normalized z value can be denormalized by multiplying it by the platform water depth. The choice of a convenient
constant a for the catenary expression should be one that forces the acute angle formed by the riser connecting the
platform with the CAT1 node to be somewhere between 5 and 7. Finally, any 3 nodes between the platform and the
TDP can serve as approximations for the riser catenary, but the authors recommend setting their corresponding radii to
0.25WD (CAT1), 0.5WD (CAT2), and 0.75WD (CAT3).
2.2.4 Description of the Optimization Model
The previous topics in this section described the main challenges in building the surface model. This topic will
present the actual optimization model and how it relates to the surface network. There are essentially two optimization
efforts going on: (1) the relocation of the new platform within predefined boundaries and (2) the attempt to determine a
subset of wells from other fields that could be advantageously interconnected with the new platform yielding a better
use of available facilities.
To model the relocation of the new platform, two real valued variables with predefined boundaries can be used.
To model the reconnection of other fields wells to the new platform, binary variables could be used to model two
conditions: (1) reroute to new platform and (2) do not reroute to new platform.
It is important to state at this point that most optimization engines, including the one used in this work, do not
model nonlinear constraints. It so happens that the main constraint on platform relocation problems is that the platform
anchorage radius does not intersect with any wellhead or subsea equipment. These constraints are basically checked by
calculating the Euclidean distance, a nonlinear expression, of all these obstacles against the new platform location.
Given the fact that the objective function has no such limitations, the recommendation is to model nonlinear
constraints as penalizers in the objective function to guide the optimizer away from unfeasible solutions. This was
accomplished by calculating, for every new platform position, the required anchorage radius given the new water depth.
Then the distance of the new platform location to all known obstacle locations is calculated and subtracted by the
anchorage radius. Negative values indicate unfeasibility of the attempted solution. The most negative calculation is then
multiplied by 1 million, and that amount of oil is removed from the simulated cumulative oil production. If there are no
negative values, then the cumulative oil production value is returned.
3. Simulation Case Study
The porosity map of the multiple reservoir case used in this paper is shown in Figure 3. As shown, reservoirs 2
and 3 already have wells, while reservoir 1 does not.
Rio Oil & Gas Expo and Conference 2012
6

Figure 3. Multireservoir simulation case study: reservoirs 1, 2, and 3
Table 1 resumes the main information data regarding the 3 fields. Preexisting fields (2 and 3) have the highest
volume of oil in place, while the new field (1) has approximately 2/3 of the oil in place of the other fields. Reservoir 3
has the highest well count, with 34 (20 producers and 14 injectors, maximum platform capacity), while reservoir 2 has a
total of 21 wells (13 producers and 8 injectors).

Table 1. Reservoir comparison table
Reservoir 1 2 3
Oil in Place (MM m) 648.4 987.3 988.1
Number of Producer Wells - 13 20
Number of Injector Wells - 8 14
Total Number of Wells - 21 34

Additionally, reservoirs 2 and 3 have dedicated platforms, each with a processing capacity of 150 thousand
barrels of liquid and a maximum number of 34 wells. The new platform will have the same capabilities.
4. Optimization Results and Analysis
In this section, the main results of both optimization efforts, determination of optimized number and locations
for new wells and new platform locations, and selection of wells from other fields to be connected to it are presented.
4.1. New Field Development Results
Figure 4 shows (left and right images) the quality map calculated using Equation 1. As can be seen by
contrasting the color scale with the colors on the top layer, the highest-quality blocks are not in this layer. The objective
function used in this optimization step was the one detailed in Equation 2 where the cost of all wells was fixed at 500
thousand m. Furthermore, the total number of wells was limited between 11 and 17 (Equations 3 and 4). Figure 4 also
shows the result of the optimization run where, from a total of 34 initial wells (left image), 11 wells remained after the
optimization (right image).
Figure 5 shows how the optimization algorithm generated optimized solutions during the 200 simulation run.
There were essentially two optimized plateaus found, the first starting at simulation 11 and the other starting at
simulation 86 (the best solution found). The color code shows how many wells were selected in each solution. Dark
blue indicates solutions with 17 wells, while light blue shows solutions with 11 wells. All colors in between shows
solutions with an intermediary number of wells. Observe that both plateaus show only solutions with 11 wells, the
difference between one plateau and the other are which main wells are always present on the solutions.
Figure 6 shows a more detailed analysis of the optimized number of wells selected by plotting the total number
of wells versus the objective function value of the optimization. On the solution space exploration, the best results were
Rio Oil & Gas Expo and Conference 2012
7
obtained with a group of 11 wells; the worst solutions had 15 wells, and there is a tendency to improve the objective
functions for solutions with more than 17 wells. It is interesting to notice that, considering all solutions with a fixed
number of wells, there is a dispersion observed in the objective function. This means that although all such solutions
have a fixed number of wells, which wells out of the total available comprise the group determine the quality of the
solution.

Figure 4. Reservoir 1: quality map showing potential wells (left) and selected wells (right)

Figure 5. Evolution of the objective function after 200 simulations
Rio Oil & Gas Expo and Conference 2012
8

Figure 6. Objective function value versus number of selected wells
4.2. Platform Location and Facilities Sharing Optimization Results
In this optimization step, the main goals were to find an optimized location for the new platform to be installed
and select which wells of the pre-existing fields could be rerouted to the new platform. Owing to geometry constraints,
only 3 wells from each existing reservoir could be rerouted to the new platform.
Figure 7 shows how the optimization search progressed. Because the non-linear constraints had to be modeled
as penalizers on the objective function, infeasible solutions greatly distorted the visualization of the optimization
process (chart on the left). If these deviant points are filtered out, it is a lot easier to visualize the optimization process.
Differently from the previous optimization effort, where there were well-defined plateaus, here the objective function
has a much more complicated topology, and optimized solutions are harder to find. The best solution was found on
simulation run 83 (out of 147) and represented an increase in cumulative oil production of 148 MM m (930.8 MM bbl).
At the end of the process, 1 well of reservoir 2 and 2 wells from reservoir 3 were rerouted to the new platform (3
reroutes out of 6 possible), and the platform location shifted from UTM coordinates (415000; 95500; 0) to
(420000; 90000; 0).

Figure 7. Evolution of the optimization
Figure 8 shows the benefits of this optimization in terms of cumulative oil production (top chart, the real
objective function when infeasible solutions are filtered out), oil production rate (middle chart), and liquid production
rate (bottom chart). It can be seen that the optimization process was efficient towards taking the total asset liquid
production towards its limit (450k bbl/day or 71.5k m/day). It is interesting to point out that there were intermediate
Rio Oil & Gas Expo and Conference 2012
9
solutions that actually took the asset liquid production to the limit, but these solutions did not yield a greater cumulative
oil production at the end of the simulation run.

Figure 8. Asset performance comparison
Finally, Figure 9 compares the base-case configuration for the three platforms when they are only linked to
their own wells (left configuration) and when two wells from reservoir 3 and one well from reservoir 2 are connected to
the new platform. Observe that this improved location does not reduce the length of the flowlines but rather locates the
platform closer to the best performing wells of reservoirs 1 and 3.

Figure 9. Base-case configuration (left) and improved configuration (right)

Rio Oil & Gas Expo and Conference 2012
10
5. Conclusion
This work presented a very useful integrated simulation optimization workflow for newly-discovered fields in
the vicinity of currently-developed assets. The details provided on the modeling of both optimization problems
addressed in this work should provide reservoir engineers with insight on how these principles can be adapted to their
own projects. There is, of course, room for improvement with this methodology, specifically on the matter of the level
of detail of the asset modeling, including more sophisticated routing of the seabed flowlines, such as flowline-length
constraints on the network model, and taking into consideration the differences in capital expenditures related to
relocating the production platform. Nonetheless, the provided framework as presented should be enough to support
decision making in the early stages of field-development planning.
6. Acknowledgement
The authors would like to thank Srgio Sancevero Sacani for his efforts in generating geostatiscal porosity and
permeability maps for the synthetic reservoir model identified in this work as reservoir 1, Halliburton for providing the
structural data for reservoir 1 and all the data for reservoir 2, and BP for making the data for reservoir 3 available.
7. References
AL-MATAR, B.S., PATHAK, A., MANDAL, D., KILLOUGH, J., FLEMING, G., ENGLE, C., BROCK, N.,
VARMA, S. 2007. Next-generation modeling of a Midle Eastern multireservoir complex. Paper SPE 106069
presented at the SPE Reservoir Simulation Symposium, Houston, Texas, U.S.A., 26-28 February.
CAMPOZANA, F.P., ALMEIDA, R. L., MADEIRA, M.G., SOUSA, S.H.G. de. 2008. Optimization of surface network
and platform location using a next generation reservoir simulator coupled with an integrated asset optimizer an
application to an offshore deep water oil filed in Brazil. In: Internation Petroleum Technology Conference, Kuala
Lumpur, Malaysia : IPTC.
COATS, B.K., FLEMING, G.C., WATTS, J.W., RAM, M., SHIRALKAR, G.S. 2004.A generalized wellbore and
surface facility model, fully coupled to a reservoir simulator. SPEREE 7(2). SPE 87913.
EVANS, S. 2007. A modern approach to E&P asset valuation, development, and decision making. JPT.
HUANG, N.S., AHO, G.E., BAKER, B.H., MATTHEWS, T.R., POTTORF, R.J. 2011. Integrated reservoir modeling
of a large sour-gas field with high concentration of inerts. SPEREE 14(4). SPE 146082.
SETTARI, A., WALTERS, D.A., BEHIE, G.A. 2000. Use of a coupled reservoir and geomechanical modelling for
integrated reservoir analysis and management. Paper 2000-078 presented at the Canadian International Petroleum
Conference, Calgary, Alberta, 48.
SOUSA, S.H.G. de, MADEIRA, M.G., CARVALHO, C.P.V., PESSOA, T.C. de S. 2010. An optimized full-field
waterflood development workflow. Paper presented at the 2010 Rio Oil & Gas Expo and Conference, Rio de
Janeiro, Brazil.

You might also like