Sunteți pe pagina 1din 12

SPE 124720-PP

Model-Based Optimisation of a Gas Production System - The BP Trinidad


Field Optimiser
A.M. Ramdial, BP Trinidad and Tobago; N.H. Hudson, BP Trinidad and Tobago; R.G. Pike, BP United Kingdom;
J.C. Rodriguez, Aspen Technology; M.S. Stratman, Aspen Technology; S.S. Sama, Aspen Technology

Copyright 2009, Society of Petroleum Engineers


This paper was prepared for presentation at the 2009 SPE Annual Technical Conference and Exhibition held in New Orleans, Louisiana, USA, 47 October 2009.
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
Oil and gas production facilities are being pushed to their limits to achieve maximum throughput in a period of global
economic uncertainty. Cost pressures continue and balancing costs, throughput and maintaining safety remain the unending
challenges. This paper and presentation will address a real example of a gas company that has leveraged real-time data with
engineering models to manage a complex production system.
It will address these business challenges:
1. How to turn real-time data into actionable information (historized process settings, well test data, etc).
2. How to encourage cross discipline communication by integrating sub-sea, offshore, and onshore facilities.
3. Understanding the impact of field and process constraints together in production potentials and capacity
utilization, e.g.:
a. Available separation and transportation capacity.
b. Pipeline back-pressure effects.
c. Wells operational limitations.
d. Onshore treatment constraints (condensate RVP and water contents, capacity of installed stabilization
units, etc).
4. Meeting contractual obligations in the most profitable way (e.g. produce the nominated gas quantities in a way
that maximizes the condensates revenue stream).
5. Reacting to system upsets and planned equipment outages.
6. In depth understanding of the potential of new field developments and of the required changes in the operating
philosophies to maximize the value of the investments.
The Asset system has been tested in a number of real production situations where it has helped to:
1. Pinpoint process behaviors that can be turned into an opportunity for economic improvement,
2. Identify process bottlenecks under various different production scenarios,
3. Co-relate offshore and onshore process constraints to find true overall optimums, and
4. Identify potential oil production increases
Introduction
In Exploration and Production, the combined expertise of process and control engineers, petroleum engineers, and geologists
is required. These disciplines do not necessarily address problems from the same perspective, and they often use different
modeling tools to help them to make day-to-day decisions. Unified decision frameworks which consolidate and encompass
larger scopes and broader data sources help to align otherwise conflicting objectives and allow operations to best tackle asset
issues such as:

Process surveillance/monitoring comparison of actual reservoir performance with simulated expected


performance for fault diagnosis and/or degradation monitoring

SPE 124720-PP

Production optimization maximization of gas production with available onshore and offshore equipment;
changing operating philosophy/conditions (pressures, routing, gas lift, water injection, gas injection, etc.) to best
use available equipment capabilities
Capacity planning assessment of spare plant capacity and available production potential
Maintenance planning impact of equipment outages on production capacity
Contract renegotiation maximum gas nominations and best utilization of spot gas markets
Debottlenecking identification of process bottlenecks and planning of additional investments and/or
redefinition of operating strategies.

The Trinidad Field Optimizer (TFO) is an offline optimization advisory system that links to real time data sources as well
as to modeling and simulation packages (Aspen HYSYS and PROSPER) to provide an overall representation of the bp
Trinidad and Tobago (bpTT) gas production and conditioning facilities.
The tool has been embedded in bpTT operational decision-making processes and is to be employed in the business
Hydrocarbon Value Assurance Team to analyze a number of real operating value realization scenarios. The main advantages
of such an intricate technical system assist the business in:
1) Reacting in a timely manner to changes in system conditions, and
2) Managing gas and associated liquids production more efficiently and
3) Maximizing revenue.
The bpTT Business
BP Trinidad and Tobago (bpTT) holds exploration and production licenses to operate in the marine areas of east coast
Trinidad and Tobago and is considered a big gas business unit which accounts for 10% contribution to the BP Global
business. The present offshore facilities consist of eleven offshore production facilities that feed into three offshore
processing hubs. The three offshore processing hubs, named Cassia Bravo, Amherstia and Mahogany Bravo, have a
combined processing capacity of 3.75 bscf of gas and associated liquid production capability. The bpTT offshore platform
network links to the east coast of Trinidad via four gas pipelines and one liquid pipeline and has a combined maximum
export capability of 3.525 bscf of gas production.

Figure 1 The bpTT platform network


Currently, the associated crude and condensate production are not constrained by market demand and generate an
unconstrained revenue value. BpTT has contracts in place to supply natural gas at specific pressures, to the domestic market
through the National Gas Company of Trinidad and Tobago (NGC) and to the Liquefied Natural Gas Company (Atlantic
LNG). The combined demand required from our midstream markets is approximately 2.7 bscf.

SPE 124720-PP

Need for Change


In the case of oil and gas producing assets this is even more important since the industry is under tremendous pressure to
keep the assets producing at full capacity (or agreed-upon rates) which leads to most operators being risk-averse.
The importance of addressing the issues of People, Plant and Process through the successful adoption of new
technologies has been widely documented (Mochizuki et al. 2006; Truschinger 2004). Taking a simplistic approach and
focusing on only one or two of the three areas has been identified as the single most important reason for failed technology
adoption initiatives. These three factors are briefly described before entering in detail into TFO specific matters.
People
A team of gas dispatchers whose job function is to ensure that the market demand is met 24hrs a day all year manages
bpTTs gas. Their operating process includes calling the respective markets to understand their demand profile throughout
the day and to ensure that the gas supply meets their required operating pressures. This process normally is managed hourly
and can often be time consuming and redundant.
The adoption of this new technology will improve productivity by allowing the gas dispatchers
1. To eliminate the guess work with respect to daily optimization
2. Add additional revenue to the saturated gas business by operating more efficiently in maximizing liquids
hydrocarbon based production whilst meeting gas nominations.
3. Reduce the time to react to changes that affect normal operating conditions.
Process
In bpTT, a Production Optimization Team (POT) was formed three years ago primarily to manage daily optimization. The
team requested inputs from petroleum engineers from each processing hub, gas dispatch team and market representatives in a
daily forum. The inputs included Installed Platform Capacity (IPC) data, schedules for planned downtimes for both offshore,
market activities, and production adding opportunities. By carefully analyzing the data inputs, the POT would reschedule
production-impacting events to ensure that bpTT meets market demand for any specific day.
There is an emphasis on maximizing revenues and with the bpTT gas business; by producing wells with greater
condensate to gas ratios (CGR) the business is able to achieve this. From this, a Well Priority Shut in List was developed to
manage all wells. This list advises on which wells should be opened/closed to maintain integrity and maximize crude and
condensate production. The petroleum engineers update the list monthly. Once the new work schedule and supply profiles are
agreed, the gas dispatch team works with the Well Priority Shut in List as a guide to supply demand.
The current workflow between the POT, asset engineers and gas dispatchers is tedious and requires constant
communication to develop a solution to different scenarios encountered. It is the aspiration of all that the implementation of
the TFO tool will simplify the process for small issues. It was equally important to integrate the needs of the shareholders to
ensure that the tool will be robust which allows it to satisfy scenario requests with respect to system optimization.
Technology
In the area of field optimization in bpTT, successful implementation requires the collaboration of the following
technologies (see Stenhouse, B. and Goodwin, S. 2004; Stenhouse, B. 2006):

Fast running modeling and simulation tools

Fast running optimization tools

Digitally enabled metering systems

Communication systems

Real-time databases/process historians

Modeling and simulation tools

User interface infrastructure

Workflow-handling tools
Trinidad Field Optimiser (TFO)
The Trinidad Field Optimizer (TFO) is a model based offline advisory system that uses historical process data and an ad-hoc
HYSYS steady state model to find out the values of key process variables that render an optimum production situation
(typically leading to meeting a certain gas nomination while maximizing the condensates recovery).

SPE 124720-PP

Figure 2 TFO network architecture

Figure 3 TFOI user interface snapshot


TFO consists of the following software components:

Library of well models (PROSPER from Petroleum Experts)

Steady state field model and optimizer (Aspen HYSYS from Aspen Technology) (TFO model TFOM)

Excel executive workflow managing application (TFO interface TFOI)

Data historian (PI from OSI Software)

MIRS production database (bpTT proprietary implementation)


TFOI acts as the human interface between the optimizer end users and the linked simulation model and databases. It
handles the data traffic between the various system components and, among others performs the following actions:

Sensitivity analysis of the individual well models to generate a parametric characterization of the well head
pressure-flow behavior.

Retrieval and validation of field data

Field model validation against field data (comparison of snapshot model predictions with historized process
data)

Field model calibration to best match actual process conditions

Configuration of specific optimization scenarios

Presentation of results in terms of potential extra revenues and required actions to achieve them

SPE 124720-PP

Figure 4 TFO application workflow


TFO Design Principles
The essential principles that drove the implementation of the TFO can be summarized as follows:

Connectivity Built-in field data retrieval, validation and visualization mechanisms shall be in place.

Flexibility The optimizer has to be dynamically reconfigurable to represent the actual state of asset equipment
at all times.

Fidelity The optimizer shall allow validation against key plant data and provide the means to be easily tuned
when necessary.

Robustness The optimizer shall be able to run without failure for all possible asset production situations (i.e.
wells, platforms, pipelines, separations trains on/off).

Usability The user interface shall be friendly and intuitive providing simple mechanisms to display/visualize
model results in a way that can be effectively communicated among team members.
TFOM
The engine of TFO is a HYSYS model (TFOM) that combines first engineering principles with ad-hoc empirical
characterizations of critical pieces of equipment, developed from long term historized process data. This dual modeling
approach increases the fidelity of the model without loosing any of its predictive capabilities, which are required for
meaningful optimization runs.
The TFOM is equipped with auto-calibration mechanisms which allow operating the same field model that will be
afterwards used for optimization in parameter estimation mode. This one model approach has clear benefits with respect
to other approaches, as it minimizes optimizer maintenance costs and simplifies the application workflow. An additional
advantage of this approach is that the same model which is used for optimization purposes can also be used for what-if
analysis and other offline studies, since it is already calibrated against field data (Fidelity).

SPE 124720-PP

Gas Pipelines
Liquid Pipelines
Multi Phase
Pipelines
Non-BP Pipelines

Custody Transfer Metering


Gas Metering Liquid
Custody Transfer

Figure 5 - Trinidad Pipeline & Custody Transfer Network

Figure 6 TFO model organization


A key element of the design of field optimizers such as TFO is the ability to automatically activate and deactivate blocks
and sub-systems. A gas asset such as BP Trinidad and Tobago is a continuous movement target. Wells and platforms are
shutdown and brought on line on a regular basis, to control and adjust gas productions to demands, to perform well tests, to
fix flow assurance problems etc. An optimizer which is not able to automatically detect and react to this dynamic nature of
the asset will deliver little value.

SPE 124720-PP

The TFOM has been created in a way that 1) allows switching from validation/calibration mode to optimization mode
with a simple flag change and 2) allows activating/deactivating wells, platforms, pipelines and separation trains through
on/off status indicators (some of them directly retrieved from data historians). A fair bit of spreadsheet logic needs to be
created to cope with relatively high number of possible asset configurations, but that undoubtedly increases the value of the
tool (Flexibility).
The level of Robustness of an optimization model is considerably higher than to what it is typically required for standard
process simulation. Individual pieces of equipment need to be protected against infeasible operating regions and/or be
provided with extrapolation mechanisms to avoid model failures during the optimization process. A good provision of effort
needs to be made for model testing before attempting to run any optimization. The small inefficiencies and short cuts that
will never manifest in a standard simulation model tend to pop-up rather quickly when the model controlled by the
optimization engine.
The common areas that shall be paid special attention when constructing the model are: 1) recycle loops which are solved
in an iterative manner in sequential modular flowsheeters shall be avoided wherever possible, as they are a sink of simulation
time and a source of noise, 2) certain unit operation blocks can be configured in different ways; direct calculations are
preferred to iterative loops inside individual units; these loops can be converged as part of the optimizer problem
configuration.
TFOI
The Excel-based executive TFO application (TFOI) is responsible to handle the Connectivity of the model with the
required data sources and with the library of validated PROSPER well models (typically updated by production and
allocation engineers). TFO uses a centralized OSI PI server that receives real time data from the various Honeywell
Uniformance PHD servers installed in each platform and/or terminal. The PI server also receives cumulative production data
from the bpTT proprietary production database named MIRS.
TFOI performs the following functions:

Data retrieval from the PI server:


 Well data well status, well head pressure and choke opening
 Plant data platform separator pressures, gas and condensate pipeline pressures and flow rates, platform
daily productions, redelivery points flow rates (at Altantic, NGC, Atlas, Picton, etc), gas and crude oil
plants process variables, such as separator pressures and temperatures, stabilization columns flow rates, etc

Gross error detection of bad PI readings and option for the users to validate and correct individual data pieces

Sensitivity analysis of PROSPER well models:


 Generation of well head parametric models that will be supplied to the HYSYS field model as a way of
characterizing the well behavior (i.e. flow rate as a function of well head pressure)
 Comparison of latest well test data with PROSPER model predictions and tuning of PROSPER models to
best match well test data

Model validation and calibration:


 Comparison of field data with model predictions and reporting of the deviations as a means of deciding the
need for model calibration
 Execution of TFOM in parameter estimation mode to minimize the differences between model predictions
and the corresponding field data. The scope of the parameter estimation mode includes: gas and liquid
pipelines pressure drops, platform productions gas calorific values in gas delivery points, condensate
quality indicators (e.g. RVP, and water content), separator temperature differences and overall asset gas
balance (i.e. gas produced versus gas sold)

Configuration of the optimization problem


 Selecting decision variables and their minimum and maximum bounds (including swing and fixed wells)
 Editing maximum and minimum bounds of system constraints (maximum separator gas rates, maximum
allowed platform productions, etc)
 Specifying the commercial targets (typically a certain gas nomination value for the day as well as the gas
and condensate prices for the various revenue streams)
 Specifying the optimization problem to solve, by selecting among a list of possible business situations such
as: meeting gas nominations and maximizing condensates production, maximizing the operation revenue,
maximizing capacity utilization, etc

Reporting optimization results, which are presented in the form of:


 Detailed optimizer output sheets, including decision variable values, constraints status and objective
function value
 Optimizer executive summary sheet, that highlights: well ranking (before and after optimization), gas and
condensate productions (before and after optimization).

SPE 124720-PP

Well models tuning


PROSPER well models are updated and calibrated regularly by production and allocation engineers to reflect actual well
performance. The well models are critical to the predictive capability of the asset model, hence the optimization system has
been equipped with specific functions to check well model predictions against latest well test data, and if necessary refine
model predictions to best match test data (see Figure 7).

Figure 7 Well models validation and tuning


For computational efficiency reasons, the PROSPER models are transformed into polynomial expressions that relate gas
flow rate with wellhead pressure; this is performed by running case studies with each PROSPER model and fitting the
obtained results to the desired expression. TFO embodies that functions and libraries that carry out this activity (see Figure 8)

Figure 8 Well models fitting


Model validation and calibration
Simulation models are a mathematical representation of the physical principles that govern the behavior of a certain
process. As such, they cannot be claimed to be powerful enough to picture all events that may happen in a real process,
especially if this is due to fortuity events, such as a pipeline blockage. Nevertheless, these are perfectly valid operating
scenarios, and in fact, when a field optimizer shall be able to demonstrate its value by providing fast and reliable answers on
how to operate under those circumstances. Equipment mechanical degradation is another instance of factors that will affect
the operations but that can not be modeled with enough rigor using first principles engineering models.
For these reasons, equipment models are typically provided with ad-hoc adjustable parameters that can be tuned to match
the actual process performance. These parameters are assumed to be fixed over a wide range of operating conditions. First
principle model equations continue being valid, but their results are shifted with an offset calculated to mirror reality.
The values of these calibration factors have themselves an intrinsic value, as they are an indication of how far the operation
is from the physical principles that govern its behavior. An oscillatory pressure drop offset in a gas pipeline, may indicate the
presence of significant pipeline dynamic effects.

SPE 124720-PP

In a gas operation such as BP Trinidad and Tobago, the system hydraulics governed by the gas transportation network
from offshore to the coastal line and to the processing terminals onshore has a paramount influence in the operation of the
individual platforms as well as in the overall operating margins.
Special care has been taken in modeling the pipe networks as accurately as possible and in providing calibration
mechanisms that can adjust those models to reflect actual system pressure drops in most scenarios.

Figure 9 Model adjustable parameters for calibration


Pipe models were created for each pipeline segment in the network using two software packages (OLGA and Aspen
HYSYS) and the results were compared and checked against field data extracted for prolonged periods of time-on-stream.
This information was used to develop empirical models that best fitted all data. Each pipeline model is provided with an
offset factor that allows fine tuning predictions to match actual pipeline pressure readings.

Figure 10 Model predictions (black) versus field data for a main bpTT gas pipeline in two different operating situations
All pipe pressure drop offsets, together with other process tuning factors, are automatically calculated by TFOM (during a
calibration step) to match field data with the minimal degree of model adjustment (minimizing the sum of correction factors
values). As part of the testing phase of the system, the resulting pipeline tuning factor values were calculated at relatively
different sets of actual process conditions, not observing important discrepancies. This indicates that the modeling and
calibration approaches employed allow a correct representation of reality within the allowed operating envelopes.

10

SPE 124720-PP

Optimisation
As it has been already mentioned in this paper, the operation of the BP Trinidad and Tobago gas system is typically
constrained and driven by the need of producing the gas quantities contracted with the various gas consumers. Gas
deliverability is therefore the key element for bpTT. This is impacted, either positively or negatively, by a multitude of
engineering and commercial issues. Just to mention a few of them:

Maintenance activities in platforms

Well testing

Pipelines pigging

Valve failures in pipes

Renegotiation of transportation contracts

Availability of new gas transportation routes

And others
The common factor of all these situations is that the asset will be requested to keep gas deliverability at required levels by
utilizing in the best possible way the available system capacity, as well as to react to unexpected events in order to recover
normal productions as quickly as possible. The TFO is conceived to be used as:
1) A day-to-day optimizer to guide the operation towards the most profitable way (i.e. delivering required gas flow
rates while getting the most value out of the associated condensate streams), and
2) As a what-if optimization system that allows finding the optimum operating strategies when major events occur.
To do so, the HYSYS optimizer has been configured with enough flexibility to allow:

Specifying well status (on/off)

Specifying well pressure and flow rate limits

Specifying fixed and swing wells

Specifying pipelines availability

Specifying separation equipments availability

Defining market and commercial information (gas nominations for each consumer, gas values)

Defining the optimization target, among a set of pre-configured scenarios


The set of business scenarios that can be executed by TFO include:

Meeting gas nominations while maximizing condensates recovery

Finding the maximum attainable operation revenue

Finding the maximum attainable gas production

Maximizing the value obtained from gas

Maximizing the condensates recovery

Finding the maximum attainable LNG gas production

Finding the maximum attainable gas production for domestic consumption

Figure 11 TFO Optimization results page comparing field value with optimum set points
This provides a great level of flexibility to the tool and allows to virtually addressing all business concerns and question
marks that the asset typically faces, providing a solid engineering baseline to benchmark individual process engineers
estimates, which often are strongly dependent on their particular experience and preferences.

SPE 124720-PP

11

Results
The rigorous model of the complete asset has been checked against plant data taken over a period of several months. The
results of these runs show that the model is in good agreement with the measured plant data. The values of most of the model
calibration tuning factors are relatively small, which indicates a good prediction of the asset behavior. In those cases where
certain tuning factors were unexpectedly high, they were reflecting a true operating issue (i.e. physical restrictions in certain
system pipelines) or deficient data quality (N. Jalilova et al. 2008). This was identified as a side benefit of the full field
calibrated asset model: i.e. its ability to quickly spot the areas of the production system where it deviates from the
engineering principles that govern it.
The TFO was used to analyze a number of real operational scenarios that included, among others:

Meeting gas demands while maximizing condensates production

Unplanned shut downs of Atlantic LNG trains Optimum strategies to best accommodate the operation to a
lower gas demand situation

Planned shut downs of Kapok and Mango wells Impact in gas deliverability

Bringing on a new high yield well (e.g. in Mango or in Cashima) and analyzing the extra gas production
potential

Analyzing the extra gas production to be attained by using the Cassia-B Immortelle by pass

Simulating a pig blockage of the 40 subsea pipeline Impact on overall gas deliverability

Impact of capacity restriction in the 12 condensates on the ability to meet gas market demand

Maximum gas deliverability as a function of the Atlantic LNG terminal pressure requirements
Typical Optimization Scenario: Meeting market demands with the maximum associated condensates revenue
In the most common optimization scenario, TFO is configured to produce a given set of gas volumes to be delivered to
the multiple gas consumers. The model is initialized with actual field data, which includes well head pressures, platform
pressures, pipeline flows, separator temperatures and pressures, etc. Upon checking the accuracy of model predictions versus
actual field data (i.e. model validation) the TFO is executed to find alternatives of improving the way in which the gas
nominations were produced in that particular day. TFO will come up with an optimal gas production distribution among
available fixed and swing wells, taking into account to quality and capacity constraints affecting the associated condensates
production as well as the back pressure effects on the gas transportation network that make the operation of the various
offshore platforms very interdependent.
Preliminary results indicate a potential of 7% extra condensate production, equivalent to 1900 oil bpd.
Adopting Technologies
The successful implementation of optimization technologies such as TFO is heavily dependent on:
1. Having high-level management support and
2. Willingness of the production engineers and gas dispatchers to adopt it as way of improving existing operating
procedures.
bpTT fully endorses both and have set up the necessary frameworks to embed TFO into the Production Optimization
Team processes. Field Trial experiments are being developed to benchmark TFO results, before it is transferred to the Gas
Dispatch team as a control room application.

Acknowledgements
The authors wish to acknowledge Roland Pike and Bryn Stenhouse (BP EPTG Sunbury, UK) for his invaluable contribution
to the TFO project, and Scott Johnston (Ingen-Ideas, Aberdeen, UK) for his continuous support in the development of the
system graphical interfaces.

12

SPE 124720-PP

References
Mochizuki, S. et al. 2007. Real-Time Optimization: Classification and Assessment. Paper SPE 90213 first presented at the
2004 SPE Annual Technical Conference and Exhibition, Houston, 26-29 September. Revised paper published in Nov. 2006
SPE Production & Operations.
Truschinger, J. 2004. Focus on People. Paper presented at the SPE Digital Energy Conference, Houston 7 April 2004.
Mullick, S. and Strathman, M. 2007. Operating Excellence Modeling for Profits. Upstream Technology Magazine,
December 2007.
Stenhouse, B. and Goodwin, S. 2004. Barriers to Delivering Value from Model Based Gas Field Production Optimization.
Paper presented at the Gas Processing Association Conference, Dublin, 20 May 2004.
Stenhouse, B. 2006. Learnings on Sustainable Model-Based Optimization The Vallhall Optimizer Field Trial. Paper SPE
99828, April 2006.
N. Jalilova, A. Tautiyev, J. Forcadell, J.C. Rodrguez and S. Sama. 2008. Production Optimization in an Oil Producing Asset
The BP Azeri Field Optimizer Case presented at SPE Gulf Coast Section 2008 Digital Energy Conference and Exhibition.
Houston, Texas USA, 20-21 May 2008.

S-ar putea să vă placă și