APC Applications Best Practices - Use of Aspen Adaptive Modeling

APC Applications Best Practices: Use of Aspen Adaptive Modeling to Maintain and Revamp Aspen DMCplus Models
An Industry White Paper

Stefano Lodolo (AspenTech), Andrea Esposito (ENI R&M)
2012 Aspen Technology, Inc. AspenTech, aspenONE, the Aspen leaf logo, OPTIMIZE, and the 7 Best Practices of Engineering Excellence are trademarks of Aspen Technology, Inc. All rights reserved. All other trademarks are property of their respective owners. 11-1776-0312
About AspenTech
AspenTech is a leading supplier of software that optimizes process manufacturingfor energy, chemicals, pharmaceuticals, engineering and construction, and other industries that manufacture and produce products from a chemical process. With integrated aspenONE solutions, process manufacturers can implement best practices for optimizing their engineering, manufacturing, and supply chain operations. As a result, AspenTech customers are better able to increase capacity, improve margins, reduce costs, and become more energy efficient. To see how the worlds leading process manufacturers rely on AspenTech to achieve their operational excellence goals, visit www.aspentech.com.
About Eni
Eni is one of the most important integrated energy companies in the world, operating in the oil and gas, electricity generation and sale, petrochemicals, oilfield services construction and engineering industries. In these businesses it has a strong edge and leading international market position. Enis commitment to sustainable development is focused on making the most of its people, contributing to the development and well-being of the communities with which the Company works, protecting the environment, investing in technological innovation and energy efficiency, as well as mitigating the risks of climate change. For more information, visit www.eni.com.
Abstract
ENI R&M adopted the new Aspen Adaptive Modeling tool to maintain and revamp its APC applications in the Livorno Refinery. This new tool embodies best practices in model maintenance workflow and compliments a suite of tools that enable a proactive maintenance approach for APC applications. The complete workflow, from problem detection, to diagnosis and repair, is supported and performed on-line through a web interface without the need of off-line activities or moving files across firewalls. Using the new Aspen Adaptive Modeling tools, ENI R&M revamped a hot oil circuit in just a few hours. The performance of the unit showed a significant economic benefit.
Introduction
ENI R&M, like many other operating companies, was struggling to properly maintain its APC applications with a reduced workforce. They were actively looking at new tools and methodologies to improve efficiency. ENI R&M has been working with AspenTech for approximately 15 years to develop new APC applications and to properly maintain the more than 50 existing APC controllers in its refineries. The decision was made to look at AspenTechs Sustained Value tools for APC: performance monitoring, automated testing and adaptive modeling. Frequent oil type changes were being made to capitalize on supply chain opportunities. The limited APC resources were struggling to keep up as these changes required updates to the controller models to keep APC solutions generating the highest value. After ENI R&M tested adaptive modeling in its Livorno Refinery with good results, they decided to deploy in its other refineries.
ENI R&M Livorno Refinery

The Livorno Refinery is a Fuels and Lube Oil refinery with a significant number of APC applications installed. In the following figure, a simplified refinery layout is reported.
Gasoline Components (LCN, ETBE) Intermediate Product (GASOIL, FULL RANGE, LCO) LPG Virgin Naphtha Lamium Gasoline Kerosene Gasoil
Fuel Plant Atomosph. Distill.
CRUDE
ATM. RES.
Vacuum Distill.
Lube Plant
Lube Basis Waxes Petrolatum Aromatic Extracts Bitumen Fuel Oil
5.2 Mton/year Crude Capacity 610 Kton/year Lube Basis Production
The refinery runs 13 Aspen DMCplus MPC controllers and 24 Aspen IQ Inferentials applications for a total of: 210 MVs (Manipulated Variables) 92 Inferential Properties The refinery ranked first in a Solomon 2006 APC/Automation comparative study that involved 18 refining companies and 36 refineries, but was still looking to improve. The APC coverage is reported in the following figure. It can be seen that APC applications cover all major units and others are currently planned on remaining plants. Being a Lube Oil refinery, there are frequent oil type production changes that affect operations and hence the performance of the APC applications. Using manual methods to maintain the models meant that the APC engineers were fairly busy.
Torcia Carb BAL Torcia Lube Stab. C4
MEROX
STAB A GPL Split T2ex
GPL
BAL KERO UNIFINER 1 Split T2 HSW TOPPING HD2 C6 HD3 GAL LVG HVGO DEA MERO AMIUM UNIFINER 2 BAP Split C5
Deiso T104
TIP
Stab. TIP
Isomerata GPL T105
PLAT
Stab. PLAT
Rerun T103
Riformata Fraz A MEA RA Circuito HOT OIL 1 BITUMI MOD. Circuito HOT OIL 2 RVC Other APC DAO SLACK PDA APA CLAUS SCOT VACUUM WAX Fraz C FT2 MEK2 HF2 BASI Fraz B FT1 MEK1 WAX VACUUM HF3
DMC + Ctrl.
Inferential DCS Ctrl.
Sustaining APC Benefits

It is something of a misnomer to say that APC applications require maintenance. If nothing in the plant ever changes, then no maintenance (i.e. model updates) is required. However, when significant changes are made to the process or the feedstock characteristics change significantly, then the APC models must be made aware of these changes. When model updates are not done, the performance starts to degrade. Illustrations like the one below are well known and all deliver the same message: poor maintenance, sooner or later, inevitably jeopardizes the APC investment.
Maximize Value of Control Application Pre-APC Implementation Max Value and Efficiency
Degradation of Control Performance APC Degradation
Troubleshoot and Solve Control Issues Lost Opportunity APC Recovery
Enhance APC Performance APC Improvement
Degradation Time Annual Profits ($M) Process unit, equipment & operation changes over time reduce the advanced control benefits Sustained & entranced performance (continuous improvement) Process models are re-identified following the process/unit modifications to recapture advanced control benefits
Lost Profit Major unit turnaround is implemented Controller/optimizer switched off
Controller/optimizer switched on
Time
There are many potential reasons for performance degradation. Some of the most likely are:
Staff Mobility Internal staff originally familiar with the application moves to a different position New staff may not be able to immediately support the application New staff may require significant training to be able to understand and support the application Process Changes Processes are often changed, and these changes can affect controller performance Catalyst changes, exchangers fowling, valves and other instrumentation changes Routine maintenance on instrumentation and equipment Economic Changes These affect the steady state solver solutions, and if they are not recognized and accommodated, performance may degrade or the controller may even lose money instead of accumulating profits
Typical signs of performance degradation are present in many applications. These signs are: Sub-controllers in OFF status; MVs (Manipulated Variables) or CVs (Controlled Variables) routinely out of service or in DCS LOCAL status Some CVs never reach SS (Steady State) targets before SS targets change again Flat lines for some CVs Some CVs remain outside limits for extended periods Many MV limits clamped or MVs at setpoint (i.e. with high/low limits collapsed) Some MVs show noise response with frequent change of directions Almost all MVs in a controller are moving on every controller execution MV dynamics are often being limited by max move size limit CV prediction error tends to be positive, then negative for extended periods Cycling CVs or MVs Optimal solution (i.e. steady state target) flips frequently Primary controls not holding setpoints Control is too aggressive with insignificant CV error; controller is aggressive with secondary objectives Without sustained value tools, the following sequence of events is typical: 1. Control Something changes in the process or in the operating mode Controller begins to oscillate or perform badly (maybe just in some areas and only under some circumstances) Operators start clamping MVs or taking out MVs/CVs or entire sub-controllers 2. Detect Control engineer is usually not automatically alerted about the problem Operators will likely call for help only when the problem becomes big enough Control engineer may spot the issue while checking trends or controller limits or passing by the control room 3. Diagnose At some point the control engineer is somehow notified by a keen operator or spots the issue himself The control engineer will diagnose the problem by speaking with operators and analyzing data either online or on his desktop. This may entail extracting data from other systems before analysis can begin. 4. Repair Diagnosis is completed Problem may simply be ignored or manually repaired and very often a sub-optimal solution is taken (e.g. Why invest days in retesting an area if I can simply de-tune the controller or manually adjust a couple of gains?) Small problems tend to build-up until parts of the controller or the entire application are switched off A major revamping step then has to be undertaken
The control engineer usually needs to manually extract process data to isolate the root cause. Once the nature of the problem has been determined, the manual model building method increases the time it takes to correct the problem and return the controller to full service. If maintenance is deferred, the problems build up until a major revamping activity has to be undertaken to fix all the issues that have been slowly accumulating. This approach is very inefficient and causes a loss of benefits that can be as high as 50-60% during the application lifecycle. With some supporting automation, we can significantly streamline this workflow and reduce the time and effort needed to keep controllers at peak efficiency.
APC Maintenance Methodology

A proper APC maintenance methodology should have the following characteristics: Incorporate APC best practices Minimize effort
4. Training KPIs 1. People
Automate and simplify maintenance tasks Use proper baselines, KPIs and automated reports to continuously track performance Rapidly detect changes in performance Few KPIs covering the performance of both controllers and models Use diagnostic rules to isolate the root cause of performance degradation Quick assessment of problems Use automated step testing to quickly generate high quality data for improved models Relieves engineering from manual testing and from night-time engineering Preprocessing rules prepare data for modeling Automated data cleaning tasks, consistent preprocessing Minimize the need to manually slice data Automatic generation of new models Generate models without requiring engineering effort Rules for rapid model assessment Quickly assess improvements Avoid manual data collection and moving data through different servers or using media to cross firewalls
2. Work Processes
3. Tools & Technology
Technology keeps improving and tools that enable proper proactive maintenance methodology with the characteristics described above are now available on the marketplace. With that kind of automation, the four steps in maintaining an APC application described in the previous section can now be performed differently as depicted below.
Automated APC Maintenance Workflow

Control Detect Diagnose Repair
Operating mode changes
Control engineer is alerted by process & models KPIs
Drilldown tools to provide performance diagnostics Model quality analysis pinpoints the models to be repaired
Automated model creation leads to faster model repair Automated retesting reduces model revamp
Controller begins to oscillate Operators clamp feed and product draws
Sustained Value Tools

The sustained value tools supporting detection, diagnostics and repair are: Performance Monitoring (Aspen Watch): Includes the capability to historize controller and process data, build baselines, controller and process KPIs and automate reporting. Through these performance KPIs, the user can rapidly detect when the process is not operating at peak performance. Model KPIs show the specific MV/CV pairs that are exhibiting poor performance. Automated Step Testing (Aspen Smartstep): Automates process step testing while maintaining the process within specifications at all times. This tool supports single and multi-test methods. It produces richer data quicker than manual step testing as it enforces APC best practices and estimates the largest possible MV steps while still maintaining the process within constraints. Much of the plant testing can be done without engineering supervision. Adaptive Modeling (Aspen Adaptive Modeling): Automates the maintenance lifecycle of a controller by providing the ability to collect historical data, automate calculations for data cleaning, schedule online model quality assessments, run standard and custom KPIs to assess model quality, run model diagnostics and perform online model identification. All of this automated workflow is performed online from a web interface directly on the running controller without the need to start a data collection task, extract data, model or tune off line, move data between systems, or start/stop applications. The process is fully streamlined and it enforces APC best practices at all stages, still giving the APC engineer the capability to control and influence results while eliminating routine manual activities.
Livorno Refinery Proof of Concept

Among the Livorno Refinery APC applications, there are also two hot oil circuits (HOTOIL1, HOTOIL2). The first circuit delivers around 65 MM Kcal/h and the second around 25 MM Kcal/h to many reboilers and other exchangers in many plants all around the refinery. Below is a simplified screenshot of the circuits.
The evaluation of adaptive modeling focused on the HOTOIL1 circuit controller and mainly on the F1 furnace. HOTOIL1 Controller 10 MVs; 51 CVs; nearly 100% service factor Most MVs are related to F1 furnace Most CVs are valve outputs of hot oil user control loops Controller originally deployed in 2005 Controller objectives and benefits Operations flexibility and maximization of delivered duty whenever required Disturbances rejection Stability of temperature and pressure of the loop Optimization of furnace combustion Controller main constraints Loop pressure and return temperature Feed pump capacity Furnace skin temperature, draft and excess O2 F1 Furnace: 4 cells, 8 passes; mixed fuel gas / fuel oil burners 4 dampers, 1 blower with backup
The evaluation was conducted in a meeting room close to the control room with around 15 APC engineers from all ENI R&M refineries. The unit has been selected because: Efficiency control of F1 furnace in Aspen DMCplus has been running with limited capabilities for some months due to model degradation after field equipment maintenance A model revamp for that section was required (old models couldnt run closed loop anymore) Only a small portion of model matrix involved Relevant impact on controller benefits Ideal candidate for an Aspen Adaptive Modeling pilot The goal of the revamp project was also to evaluate new features provided through the new tool 6 MVs were involved in the maintenance activity. We started from the following situation: The evaluation itself took around 2 full days and workflow went through the following steps:
MV 00DC2AOP 00DC2BOP 00DC2COP 00DC2DOP 90FC23ASP 90FC23BSP Description Chamber A damper position Chamber B damper position Chamber C damper position Chamber D damper position Blower flow rate SP Backup blower flow rate SP Strategy COST Maximize COST Maximize COST Maximize COST Maximize COST Minimize COST Minimize Constraints Chamber draft and balancing Chamber draft and balancing Chamber draft and balancing Chamber draft and balancing Excess O2, air to fuel ratio Excess O2, air to fuel ratio Status Out of service Out of service Out of service Out of service Over-constrained Over-constrained
Controller performance assessment through baselines and KPIs (Aspen Watch, Aspen Adaptive Modeling) Automated Step Testing tool (Aspen SmartStep) configured and run throughout the whole process As-is model quality assessment performed (Aspen Adaptive Modeling) Automated data cleaning and case setup on the Performance Monitoring System (Aspen Watch, Aspen Adaptive Modeling) Model Identification iterations (Aspen Adaptive Modeling) Online model update and deployment (Aspen Adaptive Modeling) Post-revamping model quality assessment (Aspen Watch, Aspen Adaptive Modeling) All this was done smoothly through a virtual machine connected on the ENI R&M control network and all done online from the Production Control Web Server operator interface. During the Aspen SmartStep activity, the group had plenty of time to discuss maintenance methodology and revise baselines and KPIs.
The most interesting KPI that was discussed and enabled is a modified version of the Utilization Factor (UTL) that is available as a part of the collection of built-in KPIs in Aspen Watch. The idea of a Utilization Factor was first proposed by Mr. Allan Kern in Hydrocarbon Processing, October 2005. This KPI, modified by ENI R&M engineers, is defined as follows:
ENI_UTL= (CCS+MFU+MVM) / IPMIND*100
Where: CCS= Number of CVs at high/low limit, setpoint, ramp or external target MFU= Number of MVs at external target or engineering limits MVM= Number of MVs at min move or out of service by engineer IPMIND= Actual number of manipulated variables (excluding feed forward variables) in the controller A good performance for this KPI guarantees that the controller is not just simply ON but its actually moving and using all available MVs to push constraints (i.e. to accumulate benefits). Aspen SmartStep was already enabled in the Aspen DMCplus application and little reconfiguration was required. Multitest mode was used from the beginning to minimize step testing time while minimizing MV correlation and maximizing signal to noise ratio and hence model quality. As it can be noticed in the figure below, all MVs move concurrently, permitting models to converge quickly.
10
While step testing the unit, the group concentrated mainly on the new Aspen Adaptive Modeling usage and results: View/clean MQA (Model Quality Analysis) data User can view data used in Model Quality Analysis Test Some data cleaning is automatically performed Engineer can also manually clean data using the web viewer Calculations for automated data cleaning can be configured (e.g. when an MV is moved to DCS control or a CV control error is too high) Run an MQA test Run test from the web viewer Schedule a recurring test at a designated time and interval Model KPI carpet plots are automatically updated Configure and Run ID (Identification) case Browse Aspen Watch database for tags to include in ID case ID case can be run on demand or scheduled to run automatically at a particular time and interval Review Model and Deploy Multiple model ID cases can be compared with the current model directly in the web viewer Bode plot analysis available in the web viewer to assess model uncertainty Once satisfied, model can be assembled and deployed online It must be stressed again that all these activities have been performed online through a web viewer interface and using data available in the Performance Monitor database. MQA data appear as a KPI plot where each model is flagged with different colors depending on how good models used by the controller are compared to those assessed with just a few moves. In the figure below, the complete model matrix is reported and the models where the group concentrated are highlighted in red within an oval.
11
When an MQA case is executed, an estimated Gain Multiplier (Gmult) value is calculated in such a way that the prediction errors of the corresponding dependent variable are minimized (its not simply a Gain Multiplier to bring steady state error to zero). The estimated Gmult will then include contributions from the model uncertainty; not only in the steady-state gain, but also in the dynamics. An MQA case uses the existing controller model as a reference to calculate a model quality index, which is a combination of the estimated Gmult value and the calculated model uncertainty error bound. This index represents the accuracy of the model pair in predicting the process response: Good (green) means that the model pair has a high degree of accuracy Fair (light blue) means the accuracy is somewhere between Good and Bad Bad (red) means the model accuracy is low Unknown (yellow) means that a clear answer could not be derived from the data provided (i.e. likely not enough significant data) Models can be checked as Step Responses and as Bode Plots as shown below for the Hotoil1 controller.
Bode Plots have been very useful to monitor modeling progress during step testing. In the two figures below, three hours of step test data are compared against nearly twenty hours of step test data and it can be seen that the uncertainty bands get narrow while the signal-to-noise ratio improves as the step test proceeds.
12
The evaluation stopped after around 24 hours of unattended step testing and updated models were replaced online from the web viewer interface without the need to restart the controller. Models and tuning changes can be directly checked online through the Production Control Web Server interface using a what-if simulation that permits a comparison between old and new settings before deployment. Model quality was reassessed after deployment to confirm the improvement as shown below.
Results
The HOTOIL1 controller was brought back in full operation at the end of the evaluation with the following major results: Restoring correct operation for HOTOIL1 Aspen DMCplus allowed tighter control of excess O2 and draft in F1 furnace cells The operating target was increased for dampers and decreased for blowers given the good performance of the new updated models Excess O2 was significantly reduced F1 efficiency increased by 1.2% on average after the revamp, significant for a 65 MM Kcal/h furnace in terms of fuels consumption reduction and worth well above 100 K/year at current fuel oil cost
13
The trends below show furnace efficiency and the O2 excess for 80 days. The evaluation activity is marked with the red vertical bar.
Advantages of such a solution

The entire process is done online directly from a web viewer and on the running controller. It enforces best practices and moves maintenance from reactive to proactive, maximizing controller uptime and benefits. Controller performance check-up becomes a regular activity that requires limited effort. With the sustained value tools, maintenance activities are triggered by a few properly designed controller KPIs and model KPIs. These KPIs can easily be compared against one or more baselines that can be manually or automatically built in minutes. Automatic reports can be scheduled, designed to include KPIs, calculations and trends, and sent automatically to Operators, Engineers, and Managers filtered by role. KPIs, carpet plots, diagnostics and drill down capabilities enable users to rapidly detect and diagnose the problem. Fixing the problem is then fairly automated but still under engineer control. Proactive maintenance prevents benefits degradation and reduces the need for costly full controller revamping activities.
14
Conclusions
The evaluation performed in ENI R&M Livorno Refinery clearly demonstrated the validity of the methods and tools: HOTOIL1 Aspen DMCplus controller section was successfully revamped Models were updated and all MVs were put back in service This activity delivered immediate and significant benefits The whole process took just 2 days Non-continuous work, as Aspen SmartStep took care of plant testing Aspen Adaptive Modeling features helped to speed up the process Capability to run Model Quality and Identification from web interface Aspen Adaptive Modeling was evaluated with satisfaction Ranked within ENI R&M circuit as a powerful tool to keep Aspen DMCplus controllers efficient over time New proactive approach to Aspen DMCplus applications maintenance The maintenance activity was completed in approximately 24 hours with almost no supervision during step testing and plenty of time to get familiar with the tools and technology. In a refinery like ENI R&M Livorno, with many APC applications, even with very good on-stream factors, there are lots of opportunities to improve performance that are not detected or simply left behind because of the lack of proper tools and methodology and because there is not enough time to address them when working in the old approach.
15
Worldwide Headquarters
Aspen Technology, Inc. 200 Wheeler Road Burlington, MA 01803 United States phone: +17812216400 fax: +17812216410 info@aspentech.com
Regional Headquarters
Houston, TX | USA phone: +12815841000 So Paulo | Brazil phone: +551134436261 Reading | United Kingdom phone: +44(0)1189226400 Singapore | Republic of Singapore phone: +6563953900 Manama | Bahrain phone: +97317503000
For a complete list of offices, please visit www.aspentech.com/locations

APC Applications Best Practices - Use of Aspen Adaptive Modeling

Caricato da

Informazioni sul documento

Descrizione originale:

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

APC Applications Best Practices - Use of Aspen Adaptive Modeling

Caricato da

Copyright:

Formati disponibili

APC Applications Best Practices: Use of Aspen Adaptive Modeling to Maintain and Revamp Aspen DMCplus Models

An Industry White Paper

ENI R&M Livorno Refinery

Fuel Plant Atomosph. Distill.

Lube Basis Waxes Petrolatum Aromatic Extracts Bitumen Fuel Oil

5.2 Mton/year Crude Capacity 610 Kton/year Lube Basis Production

Torcia Carb BAL Torcia Lube Stab. C4

STAB A GPL Split T2ex

Isomerata GPL T105

Inferential DCS Ctrl.

Sustaining APC Benefits

Degradation of Control Performance APC Degradation

Troubleshoot and Solve Control Issues Lost Opportunity APC Recovery

Enhance APC Performance APC Improvement

Lost Profit Major unit turnaround is implemented Controller/optimizer switched off

APC Maintenance Methodology

3. Tools & Technology

Automated APC Maintenance Workflow

Operating mode changes

Control engineer is alerted by process & models KPIs

Controller begins to oscillate Operators clamp feed and product draws

Sustained Value Tools

Livorno Refinery Proof of Concept

ENI_UTL= (CCS+MFU+MVM) / IPMIND*100

Advantages of such a solution

For a complete list of offices, please visit www.aspentech.com/locations

Potrebbero piacerti anche