Você está na página 1de 16

SPE-173450-MS

Innovative and Automated Workflow for Fast Production Optimization and


Forecast in Gulf of Thailand Gas Fields Using Linear Programing
Optimization
Peerapong Ekkawong, Pannayod Kritsadativud, Pongsathorn Lerlertpakdee, and Anan Amornprabharwat, PTT
Exploration and Production Plc.

Copyright 2015, Society of Petroleum Engineers

This paper was prepared for presentation at the SPE Digital Energy Conference and Exhibition held in The Woodlands, Texas, USA, 3–5 March 2015.

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
Gas fields in the Gulf of Thailand (GOT) share some similar operational complexities and experience
many common challenges. Such challenges include the huge number of wells and platforms, and the large,
complex, interconnected pipeline network. Additionally, each well, of course, exhibits different perfor-
mance, different enhanced recovery as well as different and diverse flow assurance methods. Fluid streams
also vary significantly from well to well; for instance, the differences in condensate to gas ratios (CGR),
water to gas ratios (WGR), and the CO2, and H2S levels. Moreover, production performance in the GOT
remains very dynamic. The decline in production could be seen early, even though proper reservoir
management was achieved because most of the reservoirs were small and compartmentalized. Optimi-
zations aiming to maximize revenue from these fields are very challenging.
State-of-the-art industry solutions to these problems are provided by integrated production modeling,
and reservoir simulation. At first consideration, they appear to be reasonable tools that can physically
describe the flow of fluid, whether in a reservoir, well or surface facility; however, these tools may not
serve well for the complicated compartmentalized characteristics of the gas fields in the Gulf of Thailand.
Currently, determining optimum natural gas production rates in the GOT is performed by manually
fine-tune the production rate using information from the latest well testing data. This method may simple
and convenient but requires large effort and does not guarantee the optimal solution.
This study presents a more efficient production optimization scheme integrating constrained optimi-
zation with decline curve analysis to predict future well production performance. The project net present
value is translated into the objective function, comprising maximizing condensate production and
minimizing waste water production while also honoring daily gas production nomination. Well perfor-
mance, export specification, and the capacity of pipeline networks are formulated as system constraints.
A linear programing optimization algorithm is then used to solve the resulting optimization problem for
a single time step. Next, the optimization is integrated with the production decline trend from the decline
curve analysis to obtain the forecast of future production performance.
Tested against the production data of a large gas field in the Gulf of Thailand, this method showed a
2 SPE-173450-MS

significant increase in the condensate production and a decrease in the water production. This solution not
only enhanced production, but also reduced tedious time required for modeling, history matching, or
manually configuring well production. Main assumptions, limitations and the conclusion of the proposed
method are also included in this study.
Introduction
Various production optimization aspects have gained wide interest in recent years as a result of fast
developments in computer technology. Oil field production optimization studies, such as allocating
limited amounts of gas lifts to specific wells in order to maximize oil production have been conducted1–3.
Fang and Lo4 used a linear programing technique for gas lift allocation by honoring multiple flow rate
constraints. A generalized well-management scheme was formulated to maximize oil production by
integrating reservoir performance, wellbore hydraulics, surface facilities constraints and lift-gas alloca-
tion. Dutta-Roy et al5. studied the optimization problem in oil fields with multiple connected wells through
a surface pipeline network. They presented a scenario where the backpressure in a shared flowline was
relatively large, the flow interaction among wells could be significant and affect the gas lift performance
relationship. A new approach to the gaslift optimization problem was proposed by formulating full-
network solution. It was then solved using a Sequential Quadratic Programing (SQP) method. Wang et al.6
proposed a formulation of a problem which could handle flow interactions among wells and could be
applied to a situation where a well could not be produced without gas lift injection. Most of these
techniques provided reasonable results, especially they furnished these results in conjunction with use of
detailed reservoir simulation models employed as integral parts of the prediction.
Apart from studies on oil field production optimization, studies on gas field optimization have also
been undertaken and they have proven to be simpler because there is no artificial lift involved. However,
some gas field operating philosophies are totally different from those philosophies underlying oil fields.
For example, it could be a given that production in an oil field always means continuously maximizing
oil production, whereas production in a gas field normally means control by nomination. For production
optimization, room for revenue improvement could be achieved not by maximizing gas production, but
rather by increasing by-product condensate production and reducing operational costs. O’Dell P.M.7
applied this concept by proposing a method to recomplete a well within an optimum time and to use a
logical procedure to select the rate for each well when the deliverability exceeded demand through the
approach of creating a numerical computer model that honored all the well economics.
In this study, we have developed a ‘fit for purpose’ approach that combines an optimization algorithm
and a fast innovative workflow to evaluate the reservoir response and to forecast the production profile.
Our method integrates linear programming optimization, which is widely used in other industries and
many applied science fields to optimize production at a single time step, with well performance prediction,
using a decline curve analysis. While reservoir simulation is considered one of the best tools to predict
reservoir performance, it requires a very large amount of information about precise geological data and
fluid characterization. Moreover, the reservoir heterogeneities in the Gulf of Thailand gas fields are very
difficult to accurately model and the results (of modeling) could be non-unique due to limited number of
data available to compare to model parameters. Instead of attempting to characterize the reservoir in detail,
the observed production history using a decline curve analysis has been used to construct the the model
to predict future behavior.
This study also renders options to optimize field production targeting specific objectives, such as
maximizing condenste production with given facility capacity, minimizing water production while
maintaining condensate production target, maximizing gas potential with given maximum condensate/
water handling constraints, and multiobjective trade-off between maximizing condensate production and
minimizing water production. The mentioned schemes capture most cases that one would want to perform
in the standard operation of any gas fields.
SPE-173450-MS 3

The case study on Greater Bongkot North (GBN) field, the largest offshore gas field in the Gulf of
Thailand and its benefits to improve the current work process will be shown throughout this paper.

Methodology

Problem formulation is presented in four parts. The first part describes the formulation of a large scale
gas field as a system of linear equations while Part 2 carries on by presenting the inevitable additional
constraints that would arise from addition of a booster compressor, an optimization problem that
consistently occurs in gas fields with a booster compressor. Part 3 covers the treatment of back pressure
behavior from the interconnecting pipeline between platforms by using an empirical relationship of gas
potential reduction and sealine capacity. These three parts help engineers to solve for optimal production
rates from each well at a single time step. Overall, the last part, Part 4, integrates this approach with a
decline curve analysis to enable engineers to not only supply an optimization calculation in a single time
step, but also supply a short-medium term production forecast.

Description of The Optimization Problem


In a typical gas field, the well production rate is normally monitored well testing at specific choke sizes;
however, that these test rates may not be the optimal rates that satisfy our ultimate goal. Therefore, the
unknown variable at play in the optimization problem would be the production rate from each well that
could maximize the objective function value. In this study, we simplify the production network by
focusing only on the production rate and explicitly detach the pressure from the iterative workflow. Gas
production rates from each well are considered as controllable variables, which can be manipulated in the
field by adjusting the choke valve. Condensate and water production are considered to be by-products of
gas production and are functions of gas rate: condensate gas ratio (CGR), and water gas ratio (WGR). The
objective function is formulated to maximize revenue by producing gas at nomination, meeting the
specification, maximizing condensate production, and minimizing water production.
Gas wells in the Gulf of Thailand are routinely tested to monitor their performance and the test results
are used to construct wellhead performance relationships (WPR) that provide the relationships of
production rate and wellhead flowing pressure. Maximum potential can be realized when the well is
produced at minimum allowable flowing pressure while the shut-in well experiences its maximum
wellhead pressure. This information specifies the well operating range. Other equalities and inequalities
constraints include daily contract quantity (DCQ), gas specification, platform & pipeline capacity, BC
operating range and facility limits.

Problem Formulation as System of Linear Equations


In this work, gas field production optimization can be formulated as a system of linear equations and
linear constraints. The formulation is also dependent upon the objective function. Once the problem is
formulated, a simplex algorithm in the MATLAB optimization toolbox is used to search for optimal
conditions honoring all the given constraints. This work presents four alternative objective functions: (I)
maximize condensate production, (II) minimize water production, (III) maximize gas potential, and (IV)
optimize multiobjective function. These equations can even be modified to support other objective
functions, such as minimizing condensate production when the condensate tanker is full, i.e., its capacity
is reached, or minimizing gas production, so that facilities maintenance can be performed. Since these
cases do not commonly occur in the Gulf of Thailand, the focus here remains on the four numbered cases
above.
4 SPE-173450-MS

Objective I: Maximize condensate production with given maximum facility limit


maximize:
(1)

subject to:
(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

The objective of this formulation is to maximize field condensate production, which is the linear
combination of well gas production rate and its condensate to gas ratio (CGR) (Eq.1). The field gas
production rate is required to be equal to the nomination (Eq.2). The field water and liquid handling are
constrained by water injection facility and field facility limit (Eq.3 and Eq.4). The gas specifications are
constrained via Eq.5 and Eq.6 where field CO2 is controlled between the upper and lower limits of the gas
sales agreement (GSA). The gas rate from each pipeline manifold, platform, or gathering station must be
less than the maximum allowable rate (Eq.7). Each well cannot be produced at higher rates than the
erosional flow rate (Eq.8). Finally, the WPR curve is used to define minimum and maximum production
from each well at each specific time (Eq.9). The Eq. 3 to 9 are basic constraints for this problem that will
be applied to other alternative objectives.
Objective II: Minimize water production while maintaining condensate production target
minimize:
(10)

subject to:
(11)

(12)

(13)

(14)
SPE-173450-MS 5

(15)

(16)

(17)

(18)

This formulation aims at minimizing field water production (Eq.10) to simulate a situation where the
field water production limit is a critical issue. The desired condensate production rate is considered to be
an equality constraint (Eq.12). The liquid facility limit, gas specification requirement, production mani-
fold limit, and well production boundaries are treated as constraints in the same way as Eq. 13 to 18.
Objective III: Maximize gas potential with given maximum condensate/water handling constraints
maximize:
(19)

subject to:
(20)

(21)

(22)

(23)

(24)

(25)

(26)

(27)

This case is directed at calculating maximum operational potential under constraints in order to
investigate maximum gas delivery capability. The objective is to maximize field gas production rate
(Eq.19). Unlike the previous formulations, gas production becomes the main objective instead of an
equality constraint. Eq. 20 to 27 denote other general constraints similar to the other cases.
Objective IV: Multiobjective trade-off between maximizing condensate production and minimizing
water production
maximize:
(28)

subject to:
6 SPE-173450-MS

(29)

(30)

(31)

(32)

(33)

(34)

(35)

(36)

(37)

Multiobjective optimization allows engineers to examine the relationship between these two objec-
tives: maximizing field condensate and minimizing field water along the Pareto front8–9. The ␩ represents
weighting factor between the two objectives which has a value between 0 and 1. In regard to the case
where ␩⫽1 it means that the objective focuses only on minimizing water production (Objective-II) the
case where ␩ means the objective focuses only on maximizing condensate production (Objective-I). The
␩ values between 0 and 1 denote the objective function that focuses focuses partially on both of the
objectives. This case would perhaps serve more flexiblility and ‘closer to real-life’ application.

Optimization with Booster Compressor


Booster compressors (BCs) play important roles in improving gas recovery since these units lower the
minimum allowable WHFP, yielding higher production rates, which leads to additional recovery.
Generally, the BC unit can be installed individually in each remote platform, or it can be connected to
multiple platforms by means of a compressing station. In this example, the BC units are installed in the
remote platforms. The wells routed to the BC units demonstrate increased gas potential, which can be
calculated from a WPR curve. On the other hand, the BC units introduce more constraints into the
production system; these are detailed in the following formulations:
subject to:
(38)

(39)

(40)

(41)

The gas feed to BC unit should be controlled in such a way that the BC can maintain its operating
conditions (Eq.38 to 39). Liquid production rates and CO2 levels are limited as shown in Eq.40 and Eq.41
respectively. These additional constraints can be formulated similarly and included into the overall system
SPE-173450-MS 7

Figure 1—Synthetic example of network pipeline system

constraints in the previous section. Note that these constraints need to be applied to each booster
compressor unit separately as clearly group labelled: BC(i).
Back Pressure Treatment
Large offshore gas fields can have many remote production platforms located some distance away from
the Central Processing Platforms (CPP). Gas production from these platforms flow to the CPP via pipeline
network. Long distance pipelines with high gas flow rates can suffer significant pressure losses, due to
friction, which introduce back pressure through the wells along the pipeline. These back pressures increase
the minimum allowable pressure; in other words, decreased maximum gas potential is the result. Since this
work in this paper simplifies the production system by solving only the system of the flowrate, pressure
behavior is not directly included in the system. But it is proposed that, explicit treatment to mimic this
effect be introduced by using the relationship of the flowrate in the pipe and gas potential drop. Eq.18,
27 and 37, which represent maximum allowable flowrates, are modified below to incorporate the effect
of flow along pipelines (Eq.42).
(42)

Eq.42 denotes the upper boundary of flowrate, but instead of denoting the lower boundary, i.e., limiting
the below maximum well potential rate, the effect of back pressure from the pipeline is incorporated. It
has been taken into account by subtracting the reduction factor from the maximum flow rate in the
pipeline. The C factor is a constant value showing percentage of maximum pressure drop if the production
rate in that line were to reach a maximum. It also includes all the pressure drops from the wells sharing
the same flow line as per the following illustration in the following Fig. 1.
In this illustration above, we want to focus on the well#7 from WP-3 (qg,7). The gas from this well has
to flow through pipe #3, #2 and #1 rerespectively. The penalize term from Eq.42 for this well can be
written as:
(43)

Each pipeline has its own discount factor (Ci) and a ratio of gas flow in each line. In this case, flow
from line #1 flow receives input gas from all wells; thus the ratio has to be summation of gas rates must
be from well #1 to #12. Line #3, in contrast, receives input gas from only wells #7,#8, and #9; thus the
summation of gas rates must be from only from these 3 wells.
8 SPE-173450-MS

Figure 2—Updated qg,i(max), CGR, and WGR with increasing Gp from time step i to time step iⴙ1

Integration of DCA for Production Forecast


So far, discussed has centered around the snap-
shot analysis in a single time step of the system.
This section will combine the linear optimization at
each time step with a production decline curve
analysis to move forward in time dimension, i.e.
develop an optimal production forecast. The type of
production decline and its coefficient can be deter-
mined from the production history. For instance, the
maximum potential gas rate and cumulative gas
production (Gp) can be formulated as a linear rela-
tionship with an exponential decline (Eq.44).
(44)

The maximum gas rate in the future can be


predicted, with updates in cumulative gas produc- Figure 3—Optimization and forecast workflow
tion and will be used as upper boundaries for the
next snapshot optimization. For condensate and wa-
ter production, the trend of CGR and WGR versus
cumulative gas production can be observed in the production history which can be easily updated similar
to maximum gas potential in Eq.44. The overall production optimization workflow starts from solving for
the optimal variable for the current time step. Then, produce the optimal production in that step and update
the cumulative production from each well. Finally, the gas potential, CGR, and WGR can be updated in
the next time step as shown in Fig. 2
Additional activities that impact well performance or addition of new production wells can be added
to modify well behavior at the beginning of each time step. The workflow of a single time step snapshot
optimization and production forecast is illustrated in Fig. 3 below.
Field Implementation
This optimization has already been implemented in to one of the largest gas fields in the Gulf of
Thailand, namely, Greater Bongkot North (GBN), a field that has been producing since 1993. Featuring
geological complexity and heavy compartmentalization, GBN requires large number of development
wells and remote platforms. Currently, the field has more than 400 wells producing from 29 remote
platforms, with more than half of the wells still active. The GBN schematic and pipeline network are
shown in Fig.4.
Managing production from this complex system to meet the DCQ and maximize profits most certainly
introduces challenging tasks. One approach employed by production engineers involves gathering data
from the latest well test results and constructing a WPR curve to calculate maximum potential from each
active producer. Following up on this, well rates are adjusted manually to within the range of producible
SPE-173450-MS 9

Figure 4 —Greater Bongkot North (GBN): platform location and pipeline network

potential that will meet the field DCQ in simulation worksheets. A manual fine-tuning process is then
performed to optimize field production, i.e. to increase condensate and reduce water production through
intensive consideration of all imposed constraints. Results so determined are then applied to the ‘real’
production rates in the field by adjusting the choke size at each wellhead in line with the simulated rates.
It will be demonstrated in this study that the benefits achieved from a new and different methodology
can serve as a viable replacement for the convention workflow approach discussed in the section above.
The same set of data is prepared for both manual fine-tuning and optimization by means of our new

Field constraints
– Maximum Daily Quantity (MDQ) 630 MMscf/d
– Daily Contract Quantity (DCQ) 550 MMscf/d
– Field liquid handling capacity 47000 STB/d
– Field water handling capacity 22000 STB/d
– Field condensate handling capacity 25000 STB/d
– Gas sales CO2 specification 22-23 %
Well and pipeline constraints
– Well erosional flow rate 15-25 MMscf/d (varied on each well)
– Pipeline capacity 60 – 375 MMscf/d (varied on each pipeline)
Booster Compressor constraints
– Minimum gas feed 3 MMSCF/D
– Maximum gas feed 8 MMSCF/D
– Liquid handling capacity 3000 STB/D
– Maximum CO2 limit 40 %

algorithm. Further, all production constraints are listed below:


The objective of this exercise is to maximize condensate production, honoring all imposed constraints.
The result is presented as a look-back comparison, at a snapshot of time between the manual fine-tuning
and the improvement proposed by this newly-proposed workflow. Two exercises are to be presented: (I)
maximum field gas requirement at 630 MMscf/d and (II) average DCQ at 550 MMscf/d.
10 SPE-173450-MS

Table 1—Field production results for maximum gas nomination scenario for both methods: manual fine-tuning and linear optimization

Figure 5—Comparison of field condensate production (left) and field water production (right) for scenario I from both methods: manual fine-tuning
and linear optimization

Scenario I: Maximum Field Gas Production at 630 MMscf/d


Table 1 summarizes the total field production from the two methods. Both of the methods yield
maximum gas delivery at 630 MMscf/d and CO2 specification. However, the manual fine-tuning method
could not reach themaximum water handing constraint (⬍ 22,000 STB/d). Conventionally, engineers
would try to manually adjust the well rate to meet all given constraints. However, with more than 200
active wells and hierarchical constraint levels, all constraints might not be accounted for human having
to conduct such a large number of tasks.
Implementation of computer programming in linear optimization makes it possible to complete the job.
Thus, in this case, field water production reached maximum capacity at 22,000 STB/d, which is -529
STB/d less than the fine-tuning method. Ultimately, the optimization also furnished suggested well rates
that could possibly increase field condensate production to 24,974 STB/d, which is ⫹2016 STB/d more
than the conventional method, the numbers for which are shown shown in Fig. 5. There are no extra
operations or additional cost required and this method only utilized the existing information with
re-engineered workflow to bring sizably inceased more profit to the asset.
Scenario II: Average DCQ at 550 MMscf/d
The second example demonstrates the optimization with the same system, but with a change in gas
delivery to 550 MMscf/d. Similar to scenario I, the manual fine-tuning met the constraint on the gas
requirement, but could not satisfy the sales gas CO2 specification (Table 2). In this case, the field water
production fell far from the maximum capacity level; the optimization was run with two different
scenarios: (I) Optimize with equal water as the fine-tuning method and (II) Optimize to reach maximum
feasible condensate production.
The first scenario keeps the same total field water production at 16964 STB/d as does the conventional
method. With the same field water production, the algorithm adjusted the well configuration to maximize
condensate, which resulted in ⫹841 STB/d increase in field condensate production, as illustrated in the
middle bar of Fig. 6. The second scenario attempts to maximize condensate production with the facility
capacity limit. This resulted in ⫹1263 STB/d condensate production and ⫹2453 water production as
SPE-173450-MS 11

Table 2—Field production result for average gas DCQ scenario from both methods: manual fine-tuning and linear optimization

Figure 6 —Comparison of field condensate production (left) and field water production (right) for scenario II from both methods: manual fine-tuning
and linear optimization

shown in the right bar of Fig. 6. Total field condensate production became 24001 STB/d with this scenario
which is still lower than the maximum handling capacity of 25000 STB/d. In spite of this, all the wells
with high CGR ratios were already producing at maximum potential to reach this field production which
implies this rate is the maximum feasible condensate production: 550 MMscf/d.
Another observation made from the two optimized scenarios was the trade-off between increased
condensate production and increased water production. To reach maximum condensate production in
scenario (II), water production increased ⫹2453 STB/d while condensate production increased only
around ⫹422 STB/d compared to scenario (I) seen in middle and right bars of Fig. 6. Thus, it is clear that
the algorithm can assist engineers in finding optimal solutions, and engineers are the ones who must
decide which level of field production operations will be selected based on their experience and
day-to-day operating conditions.
Application to Production Forecast
This section extends snapshot optimization to consider its viability as a practical tool to perform
optimal production forecasting, through integrating a historical production decline, as illustrated in the
workflow shown in Fig. 3. To incorporate the production history, production data from each well is
analyzed by plotting well maximum flow potential and cumulative gas production. Fig. 7 shows potential
the decline history from some active producers. Trend changes can be caused by additional perforations
in the new reservoir. For most of the wells, the decline trend can be defined, or assumed to be an
exponential decline. This is a linear relationship between flow potential and cumulative gas production
(Eq. 44). The latest exponential decline parameters were extracted from each well and input into the
optimization program.
For this algorithm, information regarding decline history, changes in production, and other related
operational data have been gathered to construct a production profile. Fig. 8 shows production profile
results for this example. Field gas production and maximum gas potential were generated, using the
12 SPE-173450-MS

Figure 7—Production decline history from several wells in the Bongkot field with exponential decline trends

Figure 8 —Production forecast results (a-Top left): Gas production rate and gas maximum potential (b-Top right) Field water production (c-Bottom
left) Field condensate production (d-Bottom right) Field CO2 percentage

history of production decline (Fig. 8a). These events were also included in this simulation: DCQ change
at time step 100, start-up of new platform at time step 112 and gas production plateau ends at time step
170. Field water production was kept below limit at 22000 STB/d (Fig. 8b). The program attempted to
maximize condensate production overtime, with initial rate of approximately 25000 STB/d; but this
dropped to 15000 STB/d at the end of the simulation (Fig. 8c). The field CO2 percentage was kept within
the range of 22-23% as per the sales gas requirement (Fig. 8d).
Generally, the prevalent industry approach to generating production profiles for gas fields uses a
material balance network system built around an integrated production model10. This material balance
system provides engineers with full and complete physical properties of well and reservoir interactions if
sufficient data are available. Critically important, however, is the crucial requirement of tedious history
matching before the model becomes economically reliable. Also, computation time and effort could be
SPE-173450-MS 13

Table 3—Comparison of advantages from three production forecast method

huge depending on the size of the model, not to mention possible mathematical instability. These
problems loom prominent and probably in large production systems, as was particularly shown in our
example.
The approach that we have proposed in this paper optimizes production at each time step, and then uses
a DCA to update performance of the wells. The process is iterated through time-steps until the prediction
period is covered. Linear programming optimization solves the optimal production rate for each well to
meet the required objectives. This method yields sensible and reasonable results that are user friendly, and
considerably less computationally demanding. The initial rate from such a prediction hit close to the actual
rate because it was based on the latest well performance without having to slowly trudge through the
tedious and time-consuming tasks of history matching. Table 3 compares advantages among the three
production forecasts: modeling, DCA, and this integrated approach.

Computation time
This section summarizes the runtime of the optimization workflow and compares it with the conven-
tional method. Benchmark CPU in this work is Intel® CORE™ i5-3320M CPU @ 2.6 GHz with 4 GB
RAM. Two types of optimization workflows are analyzed: (I) Snapshot optimization and (II) Production
forecast

(I) Snapshot optimization


Calculation for single time step can be done efficiently, taking only 3.1 sec. Communication between
MS Excel and MATLAB is requires the most time in this process. Linear optimization requires only 0.2
sec for optimal solutions. Speed of this method far outstrips manual fine-tuning, which takes days of man
hours, and could possibly yields poorer results.

(II) Production forecast


This test were completed for 180 time-steps. Each time-stepis one month, which is equivalent to 15
years of prediction time. Total running time is 26.1 sec. Linear programing solver process (11.5 sec)
required the longest time, and were called at every time step. This method yielded much faster results than
constructing the material balance network model, which necessitated hours to perform the same tasks.

Software Prototype (LINOPTT)


The primary deliverable of this work is a fit-for-purpose in-house optimization software implementing
the workflow and algorithm in this study called: “Linear Production Optimization Toolbox (LINOPTT)”.
The workflow and linear programming algorithm were developed using MATLAB optimization toolbox.
The software is designed to match with various objectives and is flexible for modification for any
alternative objectives or varying constraints.
14 SPE-173450-MS

Figure 9 —LINOPTT Software Structure Diagram

To ensure greater specific user friendliness, engineers need to prepare input only the data in the
provided MS Excel template. The software provides results in MS Excel, with the program structure
illustrated in Fig. 9.

Future Development and Implementation


The program has been successfully implemented for production optimization in GBN. Similar work-
flow soluations are commonly used in several fields in SE Asia region. Already planned are software
improvements that will handle additional field constraints, such as gas fields containing high CO2 levels,
fields operated with CO2 removal units, and gas fields with sand production problems. Additionally, the
concept of this workflow can be further developed for an oil field with gas-lift allocation. This is more
complex as it will require involve nonlinear constraint optimization.

Conclusion
● The proposed workflow of the linear programming optimization provides a robust
method to optimize the production from gas fields under complicated production
constraints. Deliverables are the recommended optimal wells configuration and well
production rates that should be produced.
● The example from GBN production optimization demonstrates the improvement of
increasing condensate production and reducing water production while honoring
other constraints, which include gas potential requirements, booster compressors,
and field maximum capacity limits. This new approach guarantees optimal solu-
tions, unlike manual fine-tuning, which cannot honor all constraints, resulting in
sub-optimal solutions. The field gains revenue by producing gas at nomination,
meeting the specification, maximizing condensate production and minimizing water
production.
● The program can extend capability from a single time step snapshot analysis to
short-medium term production forecast by integrating a decline curve analysis.
Historical rate trends versus cumulative production is used as a performance
prediction tool to update well performance at each time step. Using this workflow
can perform similar tasks as using the material balance package, but with much less
complexity, both in terms of required input and computation effort.
SPE-173450-MS 15

● The proposed framework is much more efficient than the conventional manual
fine-tuning workflow, which takes numerous man-days with no guarantee of opti-
mal results.
● The LINOPTT was developed to serve fit-for-purpose production optimization for
the gas fields in the Gulf of Thailand. The program is developed on MS Excel
spreadsheet and MATLAB where engineers can prepare input, run optimization,
and manage output that is all yielded in the familiar MS Excel.
Nomenclature
qp,n ⫽ Flow rate of phase p (liquid, water, oil, produced gas, injected gas) in well n
qg,DCQ ⫽ Daily contract quantity of gas from total well
q–line(i),n ⫽ Flow rate of phase p (liquid, water, oil, produced gas, injected gas) in flowline i from well
n
qp,i(min) ⫽ Minimum flow rate of phase p (liquid, water, oil, produced gas, injected gas) in flowline
i
qg,i(max) ⫽ Maximum flow rate of phase p (liquid, water, oil, produced gas, injected gas) in flowline
i
qerosional,i ⫽ Erosional flow rate of flowline i
qp,(max) ⫽ Maximum rate of phase p from total well.
qp ⫽ Rate of phase p from total well
WGRn ⫽ Water gas ratio of well n
CGRn ⫽ Condensate gas ratio of well n
LGRn ⫽ Liquid gas ratio of well n
CO2,n ⫽ CO2 production of well n
CO2,max ⫽ Maximum CO2 production from total well
BC ⫽ Booster Compressor
␩ ⫽ normalized weight factor (for weighted sum approach multiobjective optimization)

Acknowledgement
The authors thank the GBN reservoir engineering team for the data in support of this study, Mr.
Thanapong Boontaeng for discussion during the early stage of this paper’s development, Ms.Rapheephan
Laochamroonvoraponse for sharing ideas on Decline Curve Analysis, Mr. Neal R. Davis for proof read
this paper and correct our writing. PTTEP management for permitting this study to go forward and
ultimately its publication.

References
1. Kanu E.P., Mar J. and Brown K.E. (1981) Economic Approach to Oil Production and Gas
Allocation in Continuous Gas Lift. Journal of Petroleum Technology, October 1981, 1887–1892
2. Buitrago S., Rodriguez E. and Espin D. (1996) Global Optimization Techniques in Gas Allocation
for Continuous Flow Gas Lift Systems Paper SPE 35616 presented at the Gas Technology
Conference held in Calgary, Alberta, Canada, April 28-May 1, 1996
3. Giorgio V., Danillo A., Marco D., Almatasem S. (2012) Integrated Production Optimization and
Surface Facilities Management through Advanced Optimization Techniques Paper SPE 156798
presented at the SPE International Production and Operations Conference and Exhibition held in
Doha, Qatar, 14-16 March 2012
16 SPE-173450-MS

4. Fang W. Y., and Lo K.K. A Generalized Well-Management Scheme for Reservoir Simulation,
SPE Reservoir Engineering, May 1996, 116 –120
5. Dutta-Roy K., Kattapuram J. (1997) A New Approach to Gas-Lift Allocation Optimization Paper
SPE38333 presented at 1997 SPE Western Regional Meeting held in Long Beach California, June
25-27
6. Wang P., Litvak M., Aziz K. (2002) Optimization of Production Operations in Petroleum Fields
Paper SPE 77658 presented at SPE Annual Technical Conference and Exhibition, Texas, 29
September – 2 October 2002
7. O’Dell P.M., Steubing N.W., Gray J.W. (1970) Optimization of Gas Field Operations, Paper SPE
2992 presented at SPE-AIME 45th Annual Fall Meeting held in Houston, Oct 4-7
8. Hajizadeh Y., Christie M., Demyanov V. (2011) Towards Multiobjective History Matching:
Faster Convergence and Uncertainty Quantification Paper SPE 141111 presented at SPE Reser-
voir Simulation Symposium, Texas, USA 21-23 February 2011
9. Christie M., Eydinov D., Demyanov V., Talbot J., Arnold D. (2013) Use of Multi-Objective
Algorithms in History Matching of a Real Field Paper SPE 163580 presented at SPE Reservoir
Simulation Symposium, Texas, USA, 18-20 February 2013
10. Ozdogan, U., Keating, J. F., Knobles, M., Chawathe, A., & Seren, D. (2008). Recent Advances
and Practical Applications of Integrated Production Modeling at Jack Asset in Deepwater Gulf of
Mexico Paper SPE 113904 presented at 2008 SPE Europec/EAGE Annual Conference and
Exhibition held in Rome, Italy, June 9-12

Você também pode gostar