Vous êtes sur la page 1sur 0

SPE 128137

Integrated Production Model Calibration Applied to a Gulf of Mexico


Sub-sea Field
L. Saputelli, SPE, S. Rudolph, SPE, and J . Embser, SPE, Hess Corporation
Copyright 2010, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Intelligent Energy Conference and Exhibition held in Utrecht, The Netherlands, 2325 March 2010.

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 of 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
Reservoir, Production and Operations Engineers utilize integrated production models (IPM) to understand and optimize
performance in reservoirs, wells and pipeline networks. This often requires an initial calibration followed by subsequent updates.
IPM calibration may offer several challenges since it involves non-unique solutions for reservoirs, wells and pipeline network
parameters. In addition, suspect results can occur from manual data entry, uncertainties in subsurface parameters, subjective
interpretation and poor quality data. These challenges are exacerbated when the user needs to repeat the procedure periodically
(e.g. every month) or when new alternative scenarios require subsequent evaluation.
In this work, a consistent and systematic procedure was developed to assist in the calibration of an IPM under the presence of
uncertainties. The procedure was successfully applied to a subsea Gulf of Mexico (GOM) field which consists of several
reservoirs, wells, sub-sea pipeline network and surface equipment.
The IPM calibration is based on a workflow engine that adjusts IPM parameters to minimize an error function. The workflow
engine is used to couple the IPM to a spreadsheet and an optimizer. The spreadsheet computes an objective function (weighted
error function of measured vs. simulated results for each IPM variable) and the optimizer follows a gradient-free taboo and scatter
search algorithm. Finally, IPM parameters (e.g. reservoir tank pressures, well productivity indexes (PI), pipeline internal diameter,
roughness, equivalent choke diameter and choke correction) are selected from the objective function global minimum.
The calibration procedure was also tested by different users leading to similar results. The workflow engine permitted the
evaluation of thousands of IPM scenarios in a few days, which would be impossible to do manually. The calibrated IPM predicted
field production rates within 4.2% of the actual performance. The procedure developed in this work can be extended to deal with
any combination of well, reservoir and pipeline models.

Introduction
An IPM is a mathematical representation of a petroleum production system which includes the simultaneous interactions of
reservoirs, wells, and facilities. The main purpose is to compute multiphase flow rates, pressure and temperature through pipelines,
and the deliverability of the reservoir(s) throughout the producing life of the asset.
IPMs are used as the basis to optimize asset performance and its value has been shown in literature (Hepguler et al., 1997;
Coats et al., 2003; Ghorayeb et al., 2005). In general, an IPM may be used to (i) increase awareness of the assets performance by
allowing continuous surveillance (ii) increase asset uptime by identifying performance gaps (Arvalo et al., 2006); (iii) ensuring
asset integrity by operating in safe envelopes; and (iv) optimize asset value and performance by evaluating various producing
scenarios while honoring physics and constraints, (Litvak, et al., 2002; Khan et al., 2006; Rodrguez, et al., 2006).
Model parameter uncertainty is a major challenge in the use of IPM for decision making. Production operation success is
affected by many technical uncertainties [Saputelli, et al., 2002]. These uncertainties, which are often difficult to evaluate, affect
the ability to make reliable decisions. Project teams are often misguided with regards to assessing uncertainties including:
(1) Expected project outcome is evaluated in comfortable ranges of uncertainties, especially in those ranges we know best. For
example, assigning a value to an uncertainty variable in which we know only one value or one model. As a result, we often adhere
to a single deterministic value as the absolute truth; and rarely extend the expected range;
(2) Uncertainties that cannot be evaluated in available models are frequently just ignored; and
2 SPE 128137
(3) Uncertainties are often evaluated by separate departments or disciplines, and not all uncertainties are evaluated for both
subsurface and surface variables.
In the opinion of the authors, multi-discipline team members should ideally identify key uncertainties in which each parameter
set should have an equal probability to satisfy historic data. However, this step is often disregarded given time limitations or lack
of understanding of workflow.

Obtaining a solution to an IPM requires assimilation of data and adjustment of various parameters. In the literature, this process
is often referred to as history matching, calibration and/or tuning.
Reservoir history matching is a commonly adopted process in which parameters such as structure, initial fluid contacts, water
saturations, vertical transmissibility (Kv/Kh), effective pay thickness, fault transmissibility multipliers, porosity and permeability
area estimated from the assimilation of field data (Schulze-Riegert et al., 2001). The use of streamlines in finite difference models
has been used to improve cycle time. This process has also been referred to as assisted history matching (Emanuel and Milliken, et
al., 1998).
Assisted reservoir history matching has been practiced in particular in the Petroleum Industry (Brun et al., 2001; Feraille et al.,
2003; Kretz et al., 2002; Landa and Gyagler, 2003; Evensen, 2006) for speeding up the finding of well and reservoir parameters
that match historic behavior. Some assisted history matching tools (Roxars Enable, 2008), defines the quality of a history match
parameter by the calculating the difference between the simulated and the historic value divided by a user specified tolerance. This
study is primarily based on calibration of an IPM using one time step of historical data. However, the overall workflow can be
applied to multiple time steps or assist the user in obtaining a solution to match historical performance.
On a similar approach, field data assimilation is also used in many other petroleum engineering applications. Some examples
include: (1) Production Engineering: inflow performance model and multiphase flow correlations calibrated to welltests; (2)
Petrophysics: Lithology model calibrated to core sample analysis; (3) Completions engineering: fracture growth model calibrated
to real-time stimulation job performance and (4) Reservoir Engineering: fluid properties calibrated to laboratory experiments.
Successful calibration has to be consistent and systematic. At the same time, model complexity requires that only key
parameters be identified and calibrated. Under these conditions, a degree of randomness is needed to find unexpected model
deficiencies. Even the most routine calibration requires a willingness to investigate unexpected outcomes. Ideally, the calibration
of the model should be performed by a person knowledgeable of both the calibration process and the asset in question; however
calibration should provide similar results when performed by different users. The challenging aspect of the calibration process
includes identification of parameters which impact model results and understanding its interrelationship with other parts of the
system. Overall, IPM calibration can challenge users given the potential of non-unique solutions for reservoirs, wells, and pipeline
network parameters. As a result, project success broadly depends on our ability to capture technical uncertainties in both the sub-
surface and surface production models; in this sense, our work will focus in developing a procedure for consistent and systematic
capture of IPM uncertainties.
Another misconceived practice in our industry is the siloed decision making process among disciplines. Because of the
complexity and magnitude of an all-encompassing optimization problem for an asset, decisions are usually made in a fragmented
way for various pieces of the asset; however, optimum petroleum production management should be a result of a sound interaction
of disciplines (geosciences and engineering), physical elements (reservoirs, wells and pipelines) and supporting models.
A classical example includes an Operators reservoir management strategy to drill more wells in an effort to increase field
production given favorable market conditions. However, as more wells are added to the exiting facility production may not
increase as expected. With the use of IPM, an Operator could increase production by simply adding additional pipeline and
compressor capacity thus avoiding the drilling of additional wells or capital expenditures. This conveys the importance of
incorporating multi-disciplinary decision making.

This paper presents a structured methodology for calibrating IPMs while accounting for multidisciplinary uncertainties. It also
presents a case study to exemplify the methodology. The methodology considers the calibration of reservoir, wells, and pipeline
parameters. Although, this study is limited to a specific time or time step within the assets producing life, it can be easily extended
to deal with extended time periods.
The calibration problem presented herein involves the assessment of adding a new well to an existing facility in the GOM, with
at times minimal historic or actual measured data, using IPM. Too further complicate the exercise, several non operated wells
exhibited high uncertainties in performance given the lack of pertinent information (e.g., wellbore diagrams, sub-sea
measurements). The objective was to determine the total field incremental production impact of adding this new well to the
system. To achieve this objective, a workflow was developed to assist the calibration process to reduce the error between actual
field measurements and IPM simulation. Again, the field or model included several reservoir tanks and wells coupled to a subsea
pipeline network.
SPE 128137 3
IPM Calibration
IPM calibration is the process by which IPM parameters are selected in such a way that the mismatch or variance between
observed measurements and model response is minimized. Figure 1 depicts a block diagram of a generic calibration process
applicable to any simulation model. First, observed or measured rates and pressures are compared to those calculated by
simulation; then, by adjusting model parameters by hand or through assisted calibration tools, model parameters are selected to
improve the match or variance of the observed and simulated values. This process is repeated as needed until acceptable variance
on key parameters is obtained.
Again, the quality of a calibration is measured by the variance between the measurements and the results of simulation. To
assess the variance or gap between measured (history) and simulated data (history), most works in the literature use the least
squares method (Schulze-Riegert, 2001).


Model Simulation
Observed
rates and
pressures
Simulated
rates and
pressures
Compare
simulated vs.
observed values
Manual
parameter
selection

Figure 1 Generic calibration process block diagram


In this work, the IPM utilizes a commercial workflow engine (Landmark, 2009) to automate the communication between an IPM
simulator (Petex, 2009), an optimizer (Optek, 2001) and a spreadsheet. The optimizer offers a direct minimization technique which
includes a global, stochastic Tabu search algorithm (April 2003, Kelly, 2002) for minimizing error between measured and
simulated data. The simulation optimization process seeks higher-quality solutions (solution improvement rather than an optimal
answer) in recognition that computations may not necessarily provide proof or achievement of true optimization.
Figure 2 outlines the calibration process: (1) Observed field values are stored in a spreadsheet. (2) Workflow engine changes
the IPM parameters (PI, reservoir pressure, equivalent choke ID, pipeline ID and roughness, and choke correction factor) in
accordance to minimize the objective function response. (3) Workflow engine transfers the manipulated parameters to IPM Suite,
(4) executes or runs the IPM software and (5) transfers the IPM iteration results to the database. (6) Within the spreadsheet, a
global error function of measured versus simulated is calculated; (7) constraints and requirements are stored in the workflow
engine job definition; and (8) simulated results are stored in a relational database. The details of the IPM calibration procedure are
presented in Appendix A.


Manipulated IPM parameters:
- Well PI
- Reservoir Pressure
- Pipeline ID
- Pipeline Roughness
- Choke restrictions
- Choke correction
Calculated IPM Variables:
- Flow rate
- Node Pressures
- Choke Delta P
Observed
Values
Simulated
Results
Database
Constraints &
Requirements
Definition
Petexs IPM Suite
Excels spreadsheet
Workflow Engine
Landmarks DMS
Field Measurements:
- Flow rate
- Node Pressures
- Choke Delta P
IPM
( )
{ } s constraint process . .
min
*
t s
y y w
i
i i i

( )


i
i i i
y y w
*
1
2
5
6
Objective Function
3
4
7
8


Figure 2 This works IPM Calibration Workflow using Landmarks DMS and Petexs IPM Suite


4 SPE 128137
Case History
The IPM example presented herein includes an oil and gas field located in the GOM. The field consists of multiple wells. The
majority of the wells penetrate multiple stacked reservoirs and are completed as single selectives or within one reservoir. The
majority of wells are completed in single fault blocks or tanks. The field was developed as subsea tie-backs to a producing or host
facility located miles from the wells.
The surface process facilities are capable of processing up to 40,000 BOPD, 200 MMSCFD and 40,000 BWPD. The facility is
installed in approximately 3,500 ft of water. Wells A, B, and C produce through a commingled line to a high pressure (HP)
separator referred as Route 1 (Figure 4). Similarly, existing well D produces stand alone to a low pressure (LP) separator referred
herein as Route 2. Each flow line has a capacity constraint of ~20,000 BLPD.
The objective of the study was to determine the incremental production impact to the system by adding a new well, well E, to
either the LP or HP systems. Figure 2 outlines this works IPM using commercial software (Petex, 2009). The IPM was composed
of five wells, two seabed flowlines and one surface pipeline network (Figure 3). A summary of the problem objectives and
challenges are as follows:

Objectives:
i. Determine the short-term impact on production from adding a new well (Well E) to the existing flow line.
ii. Review current and alternative pipeline configurations (Route 1 and Route 2)

Challenges:
1. Well tests and production allocation exhibit questionable accuracy
2. Minimal historical well production data
3. Incomplete data set for two non-operated wells
4. Geologic uncertainties with regards to new Well E

The IPM for the field in question exhibits extensive challenges and uncertainties in achieving the business objectives. Given
multiple wells flowing within one pipeline segment in a sub-sea network architecture, the confidence in both historical well tests
and production allocation was compromised. Given the uncertainties in well test data, accurately calibrating the model also
became an issue. In addition, availability of pipeline, wellbore and well tests information was limited on two non-operated wells.
Given the lack of historical well test information and pipeline performance measurements, the tuning process was limited to data
obtained over a period of one to two months. Hence, the tuning process was essentially carried out over a single time step.
Complicating matters further, assumptions were made with regards to well geometry, reservoir pressure, and bottom hole flowing
pressures for the non operated and new drill well(s).
Given these multiple uncertainties, the model was manually validated using the traditional work flow of matching well test data
and pressure drops at then current separator conditions. For each pipeline segment, only one flow correlation (Beggs and Brill) was
utilized during the tuning process. Furthermore, a new well, well E also had inherent uncertainties with regards to bottom hole
pressure and formation thickness among others. As a result, the initial production rates for Well E could range significantly
depending on the post drill results. Given the uncertainties with regards to the new wells performance, a range of predicted rates
were evaluated changing well performance variables.
Models involved with this fields IPM went through a maturing process before the implementation of this work. First,
individual well models were built for all the wells in the system. Given the limited data on two existing wells, high confidence
offset analog well data was used as a base. Wherever data was available, pressure build-up analysis was conducted to determine
reservoir parameters for input into the well models. Individual well models were integrated into the pipeline network model and
manually tested for functionality. Once this base model became functional, the model calibration was attempted based on the
historical pressures (i.e. for the given time step). Pressure was selected as a major controlling factor for obtaining a calibrated
model given the availability of bottom hole, wellhead, downstream, and separator pressures. Once the initial tuning process was
complete, the model was used to calculate the incremental production impact for adding the new well to the system. However,
given the significant number of uncertainties the model lacked a high level of confidence and offered many degrees of freedom. As
a result, uncertainties were identified (Table 1) and defined as decision variables within the workflow engine and optimization
software (Landmark, 2009).

IPM Calibration Application
Using the IPM workflow presented in Figure 3, IPM parameters (e.g., PI, Reservoir pressure, Equivalent choke ID, Choke
correction) were changed to match field data and to provide best fit solutions to missing data. The workflow engine software
supplied the spreadsheet with the IPM parameter data per iteration. Care was taken to ensure that the IPM naming convention is
descriptive and consistent with the naming convention used in the spreadsheet. The spreadsheet calculated the error function
SPE 128137 5
between the measured value and the simulated IPM value. At this stage, IPM parameters were modified by the workflow engine in
accordance with the objective function. As mentioned previously, the error function includes weighting factors for each IPM
parameter. The equations describing the IPM calibration objective function are presented in Appendix B.
Figure 3 shows that the Route 2 pipeline component is masked or "greyed out thus reflecting no fluid flow through this
section of pipeline. Depending on the mask status of the flowlines Route 1 and Route 2, there are four possible routes for the
wells to flow to surface. There were many other possible routing alternatives in the field. However, for the purpose of this example
only Route 1 and Route 2 are depicted for simplicity. It should be noted that the example presented herein utilizes only Route 1.



Figure 3 - IPM Case history using Petex Suite (Petex, 2009)


As previously mentioned, the IPM calibration function represents the error between the measured and simulated values for oil,
water, and gas rates (four wells and two separators), well tubing head pressures, and subsea pipeline pressures. Matching
parameters (manipulated IPM parameters) included well productivity indexes (PI), reservoir pressure, choke internal diameters
(ID), choke correction factor, pipeline ID, and pipeline roughness factors. These parameters were selected due to insufficient data
and for validation purposes. A total of 36 parameters or variables were selected for the calibration process (Table 1). Ranges were
applied to these parameters to encompass the expected solutions. These ranges were defined based on field history, known data,
and offset well data. Pipe equivalent internal diameter (ID) was allowed to be altered due to the relatively poor subsea pipeline
pressure matches obtained on the flowline between manifold D and the process facilities. This pressure discrepancy along the
pipeline segment was attributed to possible flow regime changes and/or scale or paraffin deposits. Again, reservoir pressure, PI,
and choke diameters were selected due to limited historical information.

Table 1 Calibration parameters ranges
Parameter Range
Productivity Indexes (PI) for four wells [1-5] for mature wells and [15-40] for younger wells
Reservoir Pressure for four reservoir tanks [5000 to 6500] psi
Equivalent choke internal diameters for four wells [0.016 2] e.g. closed to fully open, increments of 1/64
Choke correction factor for four wells [1% 200%]
Equivalent Pipeline ID for 10 pipeline segments [4 to 6] for pipeline and [5 to 7] for risers
Pipeline roughness for 10 pipeline segments [0.0006 to 0.06]


6 SPE 128137
IPM Calibration Results
Figure 4 shows the IPM calibration objective function (y-axis) versus individual well errors after 2,200 iterations (x-axis or
sequence number). Errors in wells B, C and D found a 90% error reduction (from ~4000 to ~400) after approximately 600
iterations. However, Well A error decreased from ~5000 to ~1200 (or 75%) in the same number of iterations. Additional iterations
permitted error improvement in Well D, yet it did not provided any significant improvement for the rest of the wells. This was
expected given that the historical data for well D exhibited a high degree of confidence especially with regards to recent well test
data, real time pressures and recent pressure build-up data.

Obj fun
Well A
Well C
Well B
Well D
Run Sequence
O
b
j
e
c
t
i
v
e

F
u
n
c
t
i
o
n
,

W
e
l
l

E
r
r
o
r
20000
200
2000
200 2000 1000

Figure 4 - Case history IPM calibration objecti ve function evolution after 2,200 iterations

Well C Calibration Results
Figure 5 shows the Well C calibration performance results. Two particular values in each plot are highlighted: the Best Well
Match (BWM) with a red circle and the Best Overall Match (BOM) with a red diamond. Best Well Match (BWM) refers to
the iteration which resulted in the minimum well error, and Best Overall Match (BOM) refers to the iteration that provided the
global minimum objective function. A plot where the circle and diamond are very close indicates a high quality calibration.
Conversely, a plot where the circle and the diamond diverge indicates a poor history match and the objective function can not
improve despite a large number of iterations.
The upper-left plot shows the Well Error Tornado Plot indicating the relative influence between IPM calibration parameters (y-
axis) and the Well Error (x-axis). This output alone is very beneficial and can save the user a significant amount of time. The most
influential parameters in this wells error function were PI, pipe roughness, and choke differential pressure (DP) which can also be
verified in the correlation behavior in plots of Well Error versus PI (upper-right) and the plot of Well Error versus DP choke
(lower-center).
SPE 128137 7
The upper-center plot shows the behavior between DP choke (y-axis) and choke diameter (x-axis). The measured (~2 psi) and
simulated value (0.05 psi) are in reasonable agreement. However, the BWM (~0.1) differs from the BOM (~1).
The upper-right plot shows the Well Error performance (y-axis) as a function of the wells productivity index PI (x-axis). It is
possible to identify an absolute minimum in this case since the error performance curve shows a correlated quadratic behavior with
respect to PI. The BWM and the BOM are both in agreement at ~0.7 stb/D/psi. In this case, the PI for this well was determined to
be a known value. Once the calibration exercise was run, the model diverged on a different value for PI than originally calculated.
Since the PI was determined to be a known value with low uncertainty, the newly calibrated PI value was questioned. Upon
further review, the value found by this calibration workflow was indeed correct. The initial manually derived PI was incorrect and
this error had not been previously discovered.
The lower-left plot shows the Well Error performance (y-axis) as a function of the wells choke diameter (x-axis). The BWM
(~0.1) differs from the BOM (~1). It is difficult to identify an absolute minimum in this case since the error performance curve is
not correlated to wells choke diameter. Hence, the BWM and BOM are not in agreement.

Error vs Choke Diameter Error vs DP Choke Error vs Reservoir Pressure
Well Error Tornado Plot
DP Choke vs Choke Diameter Error vs PI
PI
Pipe Rough
DP Choke
Choke ID
FL Rough
Best Well Match
Best Overall Match
History

Figure 5 Case history Well C calibration performance results

The lower-center plot shows the Well Error performance (y-axis) as a function of the wells DP choke (x-axis). It is possible to
identify an absolute minimum since the error performance curve shows a correlated quadratic behavior with respect to DP choke.
The BWM (DP Choke ~0.12 psi) and BOM (DP Choke ~0.07 psi) are in agreement; although the historic field-measured data
pointed to be around 2 psi.
Finally, the lower-right plot shows the Well Error performance (y-axis) as a function of the wells reservoir pressure (x-axis).
The BWM (~4,200 psi) differs from the BOM (~4,500 psi) by 6%. It is difficult to identify an absolute minimum in this case since
the error performance curve is not correlated to wells reservoir pressure; however both BWM and BOM provided a reasonable
low well error. As a result, all parameters for Well C could be determined (Table 2). The minimum well error was 492, or 11%
uncertainty (Table 3).

8 SPE 128137
Well D Calibration Results
Figure 6 shows the Well D calibration performance results. The upper-left plot shows the Well Error tornado plot indicating the
most influencing parameters in this wells error function were Wellhead pressure and DP choke. This can also be verified in the
correlation behavior in plots of Well Error vs. DP Choke (lower-center plot).
The upper-center plot shows the behavior between DP choke and Choke Diameter. The measure and simulated values were in
agreement, i.e. ~120psi. In addition, the BWM (~0.4) is in agreement with BOM (~0.6) by 10%.
The upper-right plot shows the Well Error performance as a function of the wells productivity index PI. The BWM (4.1
stb/D/psi) and the BOM (~3 stb/D/psi) are in close agreement or within 20%. From Figures 5 and 6, one can notice a poor
correlation between PI and well error. In comparison, the process of calculating the PI in Well D was less effective that the PI
calculated for Well C.

Error vs Choke Diameter Error vs DP Choke Error vs Reservoir Pressure
Well Error Tornado Plot
DP Choke vs Choke Diameter Error vs PI
WHP
DP Choke
PI
Choke ID
FL Rough
Best Well Match
Best Overall Match
History

Figure 6 - Case history Well D calibration performance results

The lower-left plot shows the Well Error performance as a function of wells choke diameter. The BWM (~0.4) differs very
little from the BOM (~0.6). It is possible to identify an absolute minimum in this case since the error performance curve shows
two minima with respect to choke diameter. The BWM (ID ~0.4) is in agreement with the BOM (ID ~0.6) by 10%.
The lower-center plot shows the Well Error performance (y-axis) as a function of wells DP choke (x-axis). It is clear to
identify an absolute minimum in this case since the error performance curve shows a correlated quadratic behavior with respect to
DP choke. The BWM (DP choke ~200 psi) and the BOM (DP choke ~220 psi) are in agreement; although the historic field
measured data indicated a DP choke value of 120 psi.
Finally, the lower-right plot shows the Well Error performance as a function of wells reservoir pressure. The BWM (~5,375
psi) agrees with BOM (~5,325 psi) which is within 0.1%. Finally, both the BWM and BOM provided a reasonable low Well Error.
As a result, all parameters in Well D were determined using the calibration process as shown in Table 2. The minimum well
error was 288, which translates into a 2% uncertainty (Table 3).

SPE 128137 9
Table 2 IPM Calibration Resulting Well Parameters
598_1 598_2 599_1 599_2
PI 1 6.9 0.7 16.2
Res Pres 6435 5852 4488 5487
Pipe ID 4.99 4.71 5.62 4.09
Pipe Rough 0.0006 0.0018 0.0013 0.0426
Choke ID 0.953 2.000 0.078 1.453
Choke Correction 0.500 0.530 1.430 1.230
Choke DP 28.2 6.6 0.1 15.7
Well A Well B Well C Well D



Table 3 IPM Calibration Resulting Well Errors and Uncertainty Range
Well_A Well B Well C Well D
Minimum Error 1155 477 492 288
Uncertainty Range 20% 5% 11% 2%



Field Results
First, the calibration process resulted in several realistic equipment or parameter modifications including riser roughness factor,
flow line equivalent ID, choke sizes, and reservoir pressure. Without the IPM calibration workflow these equipment related
changes would not have been identified in a satisfactory time frame.
Second, the pre-drill best technical estimate (BTE) of the initial production (IP) rate for well E was estimated at 8,000 BOPD.
This IP rate was obtained using the BTE expected intersection of the vertical lift and IPR curves. Geologic and offset well data
was used in developing the well model for well E. The short-term field production impact of adding a new well, well E, to the
existing commingled HP flow line (Route 1) was estimated at about 5,000 BOPD.
Given the uncertainty in the desired operating drawdown, choke settings were varied to obtain a range of IP rates. For each
alternative flow route, a plot of new drill IP rate(s) versus field incremental production rate(s) was generated for the purpose of
economics. Again, for HP Route 1, the incremental new drill BTE IP rate was estimated at 8,000 BOPD with a respective field
impact of 5,000 BOPD.
Third, once Well E was logged and reservoir pressures obtained, the model was updated with actual parameters; e.g. reservoir
thickness. Given this new information, the model was re-run to determine the estimated IP rate for well E, and the incremental
field impact. The model results included a predicted well IP rate of 7,220 BOPD and incremental field production gain of 4,250
BOPD based on a drawdown equal to the nearest offset well.
Finally, the post drill actual IP rate for well E was 7,533 BOPD which is within 4.2% of the IPM predicted value. The well was
placed into service using HP Route 1, resulting in an actual field incremental production gain within ~7% of the predicted value.

Summary and Conclusions
A workflow was developed to assist the IPM calibration process for a sub-sea offshore GOM field. The integrated production
model composed of several wells, and one seabed surface pipeline network. As a result of this work, several benefits were realized
including the following:

1. Simultaneously calibrated multiple parameters for wells, separators, and pipeline network by preserving global field level
measurements. As a result, reservoir tank pressures, well productivity indexes, pipeline internal diameter, roughness,
equivalent choke diameter, and choke correction parameters were determined.
2. Minimized the time spent to perform IPM calibration in more than 80%: without the automated workflow, key parameters
would not have been identified in a satisfactory time frame;
3. Minimized human data entry. Every iteration was handled automatically providing a traceable database
4. Permitted a repeatable consistent and systematic procedure since this can be applied to future evaluations (e.g. different
time periods) or different fields with minimum work, and
5. Provided a valuable tool to assist the user in obtaining accurate results from an IPM with many uncertainties.

Acknowledgement
The authors would like to thank Hess Corporation for providing support in the review and publication of this paper. In addition,
the authors would like to thank Halliburton-Landmark employees Alejadro Chacon, Stan Cullick and J erome David for providing
support in the building of the DMS workflows.

10 SPE 128137
References
April, J ., et al: Advanced Optimization Methodology in the Oil and Gas Industry: The Theory of Scatter Search Techniques with Simple
Examples, SPE 82009, presented at the SPE Hydrocarbon Economics and Evaluation Symposium, Dallas, TX, April 5-8, 2003.
Arvalo, J .A., et al.: Successful Implementation of well management to increase hydrocarbon production: A South Region of PEMEX E&P
Case History, SPE 103758, presented at the First International Oil Conference and Exhibition in Mexico held in Cancun, Mexico, 31
August 2 September 2006.
Brun, B.; Gosselin, G.; Barker, and J . W.,: Use of Prior Information in Gradient-Based History-Matching paper SPE 66353, presented at the
SPE Reservoir Simulation Symposium, Houston, Texas, 11-14, February. 2001
Bruni, T., et al.: A Technically Rigorous and Fully Automated System for Performance Monitoring and Production Test Validation, SPE
84881, presented at the SPE International Improved Oil Recovery Conference in Asia Pacific, Kuala Lumpur, Malaysia, 20-21 October
2003
Coats, B.K., Flemming, G.C., Watts, J.W., Rame, M., and Shiralkar, G.S.: A Generalized Wellbore and Surface Facility Model, Fully Coupled
to a Reservoir Simulator paper SPE 79704 presented at the SPE Reservoir Simulation Symposium held in Houston TX, U.S.A., 3-5
February 2003.
Emanuel, A.S., Milliken, W.J .: History Matching Finite Difference Models with 3D Streamlines paper SPE 49000-MS, 1998
Energy Scitech Ltd. Roxars EnABLE v 2.2 User Manual, 2008
Evensen, G., Data Assimilation: The Ensemble Kalman Filter, Springer, 2006.
Evensen, G., Hove, J ., Meisingset, H. C., Reiso, E., Seim K. S., and Espelid, .,: Using the EnKF for Assisted History Matching of a North Sea
Reservoir Model, paper SPE 106184, presented at the 2007 SPE Reservoir Simulation Symposium, held in Woodlands, Texas, 2628
February 2007.
Feraille, M.., Roggero, F., Manceau, E.; et al.,: Application of Advanced History Matching Techniques to an Integrated Field Case Study SPE
84463 Annual Technical Conference and Exhibition, Denver, Colorado, U.S.A., 5-8, October, 2003
Ghorayeb, K., Holmes, J ., and Torrens, R.,: Field Planning using Integrated Surface/Subsurface Modeling, paper SPE 92381 presented at the
14
th
SPE Middle East Oil & Gas show and Conference held in Bahrain International Exhibition Centre, Bahrain, 12-15 March 2005.
Ghorayeb, K., Holmes, J ., Torrens, R., and Grewal, B.: A General Purpose Controller for Coupling Multiple Reservoir Simulations and Surface
Facility networks, paper SPE 79702 presented at the SPE Reservoir Simulation Symposium held in Houston TX, U.S.A., 3-5 February
2003.
Hepguler, G., Barua, S., and Bard, W.: Integration of a field Surface and Production Network with a Reservoir Simulator, SPE Computer
Applications (J un 1997) 88-93.
Kelly, J .: Simulation Optimization is Evolving, INFORMS J ournal of Computing, 14 (2002), p. 223-225.
Khan, K., et al.: Optimized Field Development Strategy under Uncertainty in El Carito Field North of Monagas, Venezuela, IBP1786_06,
presented at the 2006 Rio Oil & Gas Expo and Conference, Rio de J aneiro, September, 11-14.
Kretz, V., Le Ravalec-Dupin, M., and Roggero, F.: An Integrated Reservoir Characterization Study Matching Production Data and 4D Seismic
paper SPE 77516, presented at the SPE Annual Technical Conference and Exhibition held in San Antonio, Texas, 29 September-2 October,
2002
Landa, J . L., and Gyagler, B.: A Methodology for History Matching and the Assessment of Uncertainties Associated with Flow Prediction
SPE 84465, presented at the SPE Annual Technical Conference and Exhibition held in Denver, Colorado, U.S.A., 5-8 October, 2003
Landmark, DMS - Decision Management System, User Manual, 2009
Nikolaou, M., Cullick, A.S., and Saputelli, L. Production Optimization A Moving-Horizon Approach, SPE 99358, presented at the 2006
Intelligent Energy Conference, Amsterdam.
OptTek Systems, Inc. 2001. OptQuest Callable Library Users Manual. OptTek Systems, Inc.
PETEX (Petroleum Experts), Suite IPM v. 7.1 User Guide, Scotland, 2009.
Rodrguez, R., et al.: Integration of Subsurface, Surface and Economics under Uncertainty in Orocual Field, SPE 107259, presented at the
2007 SPE Latin American and Caribbean Petroleum Engineering Conference held in Buenos Aires, Argentina, 1518 April 2007.
Saputelli, L., et al.: A critical overview of artificial neural network applications in the context of continuous oil field optimization, SPE 77703,
presented at the 2002 SPE Annual Conference and Exhibition, San Antonio.
Saputelli, L., Nikolaou, M., and Economides, M.J .: Real-Time Reservoir Management: A Multi-Scale Adaptive Optimization and Control
Approach, Computational Geosciences, Special Issue on Modeling, 2005
Schulze-Riegert, R. W., Axmann, J .K., Haase, and O., Rian. D.T.: Optimization Methods for History Matching of Complex Reservoirs SPE
66393, presented at the SPE Reservoir Simulation Symposium, Houston, Texas, 11-14 February, 2001


SPE 128137 11
Appendix A IPM Calibration Procedure

The overall procedure to calibrate an IPM consisting of several reservoirs (tanks), wells and a pipeline network is as follows:

1. Build the IPM

a. Collect available data for all wells and pipelines (produced fluids, pressure, temperature, geometries, etc.)
b. Build tank models and establish pressure ranges (20%) based on produced volumes, shut-in tests, well tests, etc.
c. Build well models and establish PI ranges (50%) based on completion data, well tests, analogs and log data.
d. Build pipeline network and establish multiphase flow correlations corrections (10%) based pressure and
temperature measured data;
e. In pipeline network model, establish choke model performance based pressure and temperature measured data;
f. Integrate all elements of the field network, wells and reservoirs
g. Build a table with expected uncertainty ranges for IPM adjustable parameters

2. Build objective function spreadsheet

a. Collect historic data (e.g. one or more date sets)
b. Build error function between measured and simulated data (as in Appendix B)
c. Make sure to have consistent equipment naming between Pipeline network and spreadsheet
d. Check IPM calibration error on spreadsheet, plug different parameter range and solve IPM
e. Estimate initial uncertainty ranges for IPM (as in Appendix B)
f. If a reasonable calibration exists, i.e. uncertainty is less than 1% (as in Appendix B), then stop the procedure

3. Design and configure decision model in workflow engine (Landmark, 2009)

a. Select IPM input parameters (parameters to be adjusted) as decision variables
b. Select spreadsheet file, i.e. assign it as the objective function calculator
c. Build a job: Simulator +Spreadsheet Optimization type
i. Select IPM results to be analyzed for output and objective function calculation
ii. Select spreadsheet results (e.g. objective function name) for output
iii. Map IPM results output to spreadsheet input
iv. Select objective function (spreadsheet name) for optimization
d. Submit initially few runs, i.e. submit 1 to 3 runs
e. Check data consistency, i.e. verify connectivity, spreadsheet calculations, and verify validity of results
f. Validate objective function calculated values, i.e. check IPM results offline, plug selected parameter and solve

4. IPM calibration

a. Submit job runs (defined in 3c)
b. Submit 100-200 runs at a time. If run time is large (>8hrs), then reduce the number of runs submitted in one time.
c. Observe the evolution of objective function, i.e. convergence to minimum error vs. oscillating error
d. Analyze the results, e.g. plot objective function vs. adjusted parameters, generate tornado plots
e. Check requirements compliance, e.g. how good are predicted IPM rates vs. production accounting allocated rates
f. Estimate uncertainty ranges for IPM (as in Appendix B)
g. If a reasonable calibration exists, i.e. IPM component uncertainty is less than 1% (as in Appendix B), then stop
the procedure and select IPM parameters from global minimum run.
h. If further improvement is expected from current model set up, then repeat 4e
i. If suspect results are obtained, review sources of error in 1 (e.g. data, model building and parameter range)





12 SPE 128137
Appendix B IPM Calibration Objective Function

A generic objective function was developed to calculate the error between the measured and simulated values. The function
consists of measured absolute error in all key signals such as well oil, water and gas rates; well pressures (wellhead, subsea line
and choke pressure drop). The objective function is computed by adding well multiplied by a certain weight (w
i
), separator
multiplied by a certain weight (w
i
) and total errors (Eq. B-1). The well error (w
i
) is computed so that error quantities are
normalized by their average value. Error units are dimensionless since intermediate quantities are divided by its historic average so
that they get normalized to a common reference and added to similar quantities (i.e. same order of magnitude).

( ) ( ) Error Total Error Sep Error Well Function Objective
,
+ + =

B A
i
i
i i
w w
.................................................................. (B-1)

Well error is computed by adding the absolute errors in wellhead pressure, subsea line pressure, choke pressure drop and
allocated well rates (Eq B-2). Each error is multiplied by respective weight factors (w
hp
, w
lp
, w
dp
and w
q
) so that error quantities are
normalized by their average value. Notice that allocated well rates will account for one quarter of total well error. Under some
circumstances, well allocated rates may be further investigated as possible outliers by computing total uncertainty with and without
the last factor of Eq B-2.

HIS SIM q HIS SIM dp HIS SIM lp HIS SIM hp
Qo Qo w DP DP w SSLP SSLP w WHP WHP w + + + = Error Well
......................... (B-2)

For each separator (A and B), an error is computed by calculating absolute difference between simulated and historic values
(Eq. B-3).

HIS SIM
Qo Qo = Error Separator
................................................................................................................................. (B-3)

Total error is computed by adding the absolute errors in each separators absolute error (Eq B-4).

( )
HIS HIS SIM SIM
QoB QoA QoA QoB + + = ( Error Total
......................................................................................................... (B-4)

Where:
DP = Choke pressure drop
HIS = Historic, measured
Qo = Oil rate
SIM = Simulated
SSLP = Subsea line pressure
WHP = Well head pressure
w
i
= Well error weight
w
hp
= Wellhead pressure error weight
w
lp
= Line pressure error weight
w
dp
= Choke pressure drop error weight
w
q
= Rate error weight

We define the well Uncertainty (Eq. B-5) of the assisted history match process as a measure of the error spread, i.e. the
Absolute Error in each variable divided by its historic average so that they get normalized to a common reference.

( )

+ + +
=
HIS HIS HIS HIS
Qo DP SSLP WHP
Error Well
y Uncertaint
......................................................................................................... (B-5)

Vous aimerez peut-être aussi