Introduction
The Asian-Pacific refinery mentioned in this blog has a crude processing capacity of 100 kb/day with FCC, Hydrocracker, reformer, visbreaker complexes, and three crude units. The refinery has 3 diesel and 2 kero product blend headers and each is blended into 8 run-down product tanks. Gasoline blending configuration uses 6 stock tanks to blend into 7-8 product tanks. Gasoline blending has only trim control on TDC-3000. The tank qualities of final product tanks are measured only by lab analysis of manual samples taken either at a mid-blend point or at the completion of the final blend to determine its suitability for shipment or the need for reblends.
Problem
The refinery was facing the problem of online monitoring of product qualities of process run-down end product tanks for Kero and Diesel tanks. This was necessary to enable unit operators and shift managers to adjust the upstream flows manually to minimize product quality giveaways. Since there was no feedback control system installed to control the product tank qualities, the refinery felt the need for an automated online computer software system that was able to track and calculate the qualities of run-down tanks in near-real-time. This system would give operators and shift managers directional guidance to adjust crude and other process unit parameters to reach the quality targets of Diesel and Kero product tanks.
The shift managers did not know the qualities of final product tanks until the lab analysis results were available 8-10 hours later. At that point, either the tank qualities were off-spec or had unacceptable quality giveaways. This caused huge economic losses to the refinery either due to shipping the product with unacceptable giveaway qualities or time lost in reblending the tanks to minimize the unacceptable give-away or blends them to specs. Frequently, the blending was also suspended to analyze the tank for lab results, diverting the run-down to another tank in the meanwhile.
Solution
Offsite Management Systems LLC (OMS) developed and installed a demo version of an online Tanks Quality Tracking System (gomsTQTS™) in April 2002. OMS worked together with the refinery as a team to further develop, enhance, test, and install additional functionalities of gomsTQTS™. OMS commissioned a full licensed version of gomsTQTS™ to the refinery in June 2003. The Tanks Quality Tracking System (gomsTQTS™) owned by OMS, is based on client-server architecture on the Windows NT/2000 platform, using Visual Basic as the user interface with SQL Server 2000 as the database engine.
The system has a modular design that enables it to integrate seamlessly with other OMS products, third-party software, and plant control DCS and process databases (Figure-1). The system is installed on one server with the following specifications: Pentium-IV 2.3 GHz, 512 MB RAM, 80 GB HDD and is currently accessed by 25+ client machines via the plant network. The system is accessed by the refinery planner, shift managers, process operators, control engineers, operation and business managers in order to use the online product qualities in their respective job roles.
The total time for the customization and implementation for gomsTQTS™ was 3 months including time from the initial process flow diagrams, tags building, system customization, and commissioning in product mode.
Configuration
OMS customized the gomsTQTS™ system by building and integrating process displays (18 in total) for the Kero and Diesel system specific to the refinery. OMS system’s interfaced with Honeywell Ph.D. seamlessly to access plant data via gomsProc™ interface system and also provided many special processing and calculations exit routes for the real-time plant data. gomsTQTS™ is a completely integrated system and has a number of Plug-N-Play sub-modules and distributed SQL server databases (Figure-2). The sub-modules shown in Figure-2 are core system independent and are installed only once for all OMS products.
The qualities tracked were freeze point, the flashpoint for Kero product tanks, 90% distillation, and Sulphur for Diesel tanks. gomsTQTS™ has approximately 80+ qualities already built-in for all fuel, diesel, and gasoline products. The system accesses about 600 plant data tags and online tracking calculations are done once every five minutes with a total cycle time including data access and calculations of approximately 8 minutes.
OMS customized the gomsTQTS™ system by building and integrating process displays (18 in total) for the Kero and Diesel system specific to the refinery. OMS system’s interfaced with Honeywell Ph.D. seamlessly to access plant data via gomsProc™ interface system and also provided many special processing and calculations exit routes for the real-time plant data. gomsTQTS™ is a completely integrated system and has a number of Plug-N-Play sub-modules and distributed SQL server databases (Figure-2). The sub-modules shown in Figure-2 are core system independent and are installed only once for all OMS products.
The system configuration parameters are:
System Parameters | Diesel | Kero |
Product tanks | 8 | 8 |
Run-down headers | 3 | 2 |
Process streams | 7 | 7 |
Process tags | 220 | 192 |
Online analyzers | 7 | 8 |
Predicted qualities | 3 (90% pt. Sulphur Density) | 2 (flash pt., freeze pt.) |
Types of quality models | Non-linear | Non-linear |
Custom process displays | 9 | 9 |
Features
The system has many advanced features built in to ensure maximum accuracy in the prediction of tank qualities. Some of these features are described below:
- Current and Target Tank Qualities – Process displays (Figures 3 and 4) show the configured current and target volume tank qualities for all sub-systems online. Target tank qualities are calculated based on target volume which can be changed by authorized users.
- Inter-Tanks Transfer Quality Tracking – One of the key features of the gomsTQTS™ system is the ability to track all inter-tank transfer activities (Figure – 5) for source and destination tanks. These transfers can include emptying, transfer to/from another tank to a run-down tank and can work reasonably well from deviations in tank levels with or without flow meters at tank exits.
- Process Stream/Tank Qualities – The system can be configured to input the stream qualities using single or multi qualities analyzers, lab analysis, and manual data for constant qualities. Similarly, the tank qualities can be input via lab analysis or manual data entry. Once configured, the system can access the qualities via sub-system modules to interface with the plant database, Ph.D. in Honeywell DCS-3000 system.
- Flow Offset Corrections – The system has options to either correct the input-based flows on either manual corrections or auto-corrections. This is done by sensing the valve output and corresponding flows and calibrating the meter offsets due to leakage. The configuration includes the flow and output tags to make offset corrections.
- Dynamic Updates of Analyzer and Model Bias – The system has a built-in analyzer checking algorithm that updates analyzer bias using the lab analysis of process streams and the lab analysis of final tank qualities. It also uses predictions to update the qualities model bias and updates and uses online to increase prediction accuracy.
- Online Regression of Model Parameters – The system regresses the model parameters once the run-down blending is completed and lab analysis results are available to fine-tune the model parameters. This is done dynamically and automatically once a day usually at midnight.
- Quality Calculator – This feature helps the process operator and shift manager to perform “What If” analyses of target product qualities by varying the volume % and quality of blend compositions in an offline simulation mode to help him minimize the quality giveaways.
- Inventory Calculations – gomsTQTS™ has a comprehensive tanks inventory information system (gomsTIMS™) built-in using the ASTM table-54 methodology and tanks movements status. The system also shows the overall material balance of all tanks configured.
- Sensitivity Analysis – This feature estimates quality prediction error bands based on estimated errors in flow measurements and tank volume deviations.
- Advanced 4-R Options – The system also has advanced features which are used to start the system from scratch (Reinitialization), reset the system to the last plant conditions (Reset), reconcile data from hot and cold restarts (Restart), and Reconciliation of mid-blend tank lab qualities.
In addition to the above features, the system has numerous functionality of trends, alarms, historical data displays, lab analysis and biases, regression results, etc. making gomsTQTS™ a very useful and comprehensive system.
Reports
The system generates shift and daily reports of all tanks to show detailed quality prediction performance for the shift or a 24 hr/day and can be generated at configurable times and sent via email, fax, or any local or network printer. The system also generates reports in PDF format and archives them for reviewing later. Monthly reports summarize the prediction performance of all tanks and all qualities and documents the deviations from final lab analyses. Additionally, there are about 30+ online reports available for tabular and graphical representation of data.
Performance and benefits
- Performance Criterion – The licensed version of the gomsTQTS™ system has been in operation at the Singapore refinery for the last eighteen months with 90%+ online availability of tank qualities. It had tracked more than 590 and 800 blended batches for Kero and diesel products, respectively in the year 2003. The system also has built-in regression of quality model parameters that feedback to quality tracking models to improve the performance for subsequent run-down batches.
The refinery specified the following Key Performance Indices (KPI) criterion for the errors in the predictions of qualities by gomsTQTS™ as compared to lab analyses.
-
- Freeze point ± 1.0 Deg C
- Flash Point ± 1.0 Deg C
- 90% point ± 3.0 Deg C
The above performance is achieved only for open-loop monitoring of run-down tank qualities without (optional) feedback controls. Our experience is that gomsTQTS™ has performed better than the above error margin when input qualities from online analyzers are valid. In a recent enhancement, gomsTQTS™ now has built-in analyzer checking algorithms to dynamically update analyzers via lab analysis at times and also using the most current good values in case of bad signals from analyzers. This has improved the accuracy of predictions considerably and relies less on analyzer outages.
- Past and Present Performance – The minimum and maximum specs for Kero and Diesel products at the refinery are as follows:
Products | Quality | Min. Degree C | Max. Degree C |
Kero JP5 | Flash pt., freeze pt. | 60 | -47 |
Kero Jet | Flash pt., freeze pt. | 38 | -47 |
Diesel products | 90% | 370 |
OMS and the refinery analyzed the data for the period 1/1/2001 – 12/31/2001 to establish a basis for the estimation of benefits before the pilot gomsTQTS™ project was started in June 2002. Since gomsTQTS™ was going through trial and enhancement phases in 2002, both refinery and OMS considered the representative data period of 1/1/2003-12/31/2003 after a relatively stable version of gomsTQTS™ was commissioned late in 2002.
2001 (Without gomsTQTS™) |
2001 (Without gomsTQTS™) |
2003 (With gomsTQTS™) |
2003 (With gomsTQTS™) |
Delta Improvement |
Delta Improvement |
||
Product | Quality | Average | Std. Dev | Average | Std. Dev | Average | Std. Dev |
Kero products | Flashpoint | 0.935 | 1.09 | 0.69 | 2.75 | 0.245 | -1.66 |
Freeze point | -2.65 | 1.72 | -2.85 | 1.62 | -0.198 | 0.098 | |
Diesel products | 90% | -2.19 | 3.22 | -1.74 | 1.96 | 0.454 | 1.252 |
We can conclude from the above statistics that the 90% point has realized maximum benefits from the reduction in average and standard deviation of predicted qualities versus target specs by as much as 20%. There are also some advantages to flash and freeze point predictions. Since gomsTQTS™ depends strictly on the input values of run-down streams measured by online analyzers, their outages at times ahs also affected the predictions in gomsTQTS™. However, the updated version of gomsTQTS™ has built-in analyzer checks and reconciliation algorithm and we would expect the prediction performance to improve drastically in the future.
- Estimated Tangible Benefits – The benefits from gomsTQTS™ for Kero and Diesel products qualities were estimated using the following formulae:
The above formula accounts for both average and spread of deviation of qualities from the quality spec limit and will penalize or reward for poor and better performance. refinery/OMS has estimated that the potential tangible benefits from gomsTQTS™ for Kero and Diesel products for the year 2003 were in the range of 2.5-4.4 MUS$ depending on the criterion of reduction in average or standard deviation of predicted qualities from the spec limits.
However, we also recognized that it is almost difficult, if not impossible, to arrive at exactly claimable benefits from an open-loop system like gomsTQTS™ as it does not and can not record shift managers’/operators’ indirect action to adjusting the run-down flows to reduce giveaways and avoid off-specs of run-down tank qualities. Nevertheless, the drastic improvements in the quality giveaways can not be ignored and can be safely assumed the role of the gomsTQTS in giving directional help to shift managers or blend operators. Even if we assume conservative claimable benefits of 25% by the gomsTQTS, it is still in the range of 0.65-1.10 MUS$/year, giving an ROI of less than 3-6 months.
These savings are presumably realized from:
- Minimization of quality giveaways
- Elimination of reblending and inter-tanks transfers
- Better planning for shipment and manufacturing as tanks qualities are available online
- Optimization of tank sampling schedule and minimization of lab-analysis
User training
The refinery does not have adequate field equipment automation for an application of this kind to work in a totally automatic mode. For example, the tank’s quality tracking system needs to know the lineup tanks either for run-down, inter-tank transfers, or flow diversions. Normally, in a fully automated refinery, this would have been accomplished by MOV’s, limit switches, and flow meters at tank inlet and outlets of tanks. OMS built custom software switches and created the logic to process them for the refinery to complement the lack of field instrumentations in the refinery.
This required that shift managers and blending operators set/reset these switches in order to change product tanks, transfer tank line-ups, flow diversions, etc., for accurate predictions of qualities. This took quite a bit of effort and is still continuing on part of the refinery management in order to makeshift managers and blending operators owner of the applications. Additionally, the refinery has committed a dedicated control engineer to support the system in the refinery as well as interface with OMS for updates, upgrades, and other support issues.
In order to ensure the success of this application, it required support and commitment from the refinery user community and management to realize the intended tangible benefits. With support from the refinery, OMS has made significant progress in this aspect. OMS has also trained a refinery support engineer to build, customize and configure simple sub-systems in gomsTQTS™ with remote guidance and adequate training documentation from OMS.
Customization
OMS’ Tanks Quality Tracking System™ is a stand-alone system except for the customized process-related displays shown in Figures 3-5. The refinery version of gomsTQTS™ has about 18 customized displays covering overview and process details and advanced features like context-sensitive information displays of analyzers, tank qualities, header, lab analysis, etc. Alternatively, any plant can build some basic displays in DCS or ODBC compliant HMI system to show final tank qualities and gomsTQTS™ can be configured to send data to DCS tags to show on displays. This can be done by plant gomsTQTS™ support engineers with minimum involvement from OMS.
Support
The system has been up and running online 85%+ of the time since June 2002. Downtime has been mainly due to plant networks, system updates, initial troubleshooting, etc. and uptime has increased to 95%+ in 2004. The system monitors and records the system’s online availability information and includes the information in reports sent to OMS every day at midnight. Online availability increased as the system became robust and bug-free. The system completes one cycle of data retrieval and calculations for 16 tanks and 5 qualities in about 10 minutes, thus performing optimally about 145-150 calculations on a 2.4 GHz machine.
The system is designed to generate all relevant data for 24 hours in data folders which are then sent to OMS automatically by email every day at midnight. OMS uses this information to troubleshoot or diagnose problems with gomsTQTS™ systems and processes. The information in these files enables OMS to reproduce snapshots of process conditions at any point in time and simulates the exact conditions in the refinery. These fills also contain daily and shift reports in PDF format for OMS to review in case of any discrepancies in the information presented and pinpointed to the refinery the exact source and reason for it.
Additionally, the system allows the gomsTQTS™ users to review any historical data online in detail to determine when analyzers failed, for example, or the exact time of the beginning of any bad results as nothing is left to guesswork.
Integration with other systems
The tanks quality tracking system by OMS (gomsTQTS™) is a stand-alone and DCS independent system which can seamlessly plugin with many OMS or third-party software products such as online/offline blend control system, DCS based regulatory blend control. Most important of all, gomsTQTS™ can provide feedback to upstream units, process control, and optimization systems as the predicted qualities of run-down tanks from process units will be available online via gomsTQTS™. An online blend control and optimization system depends on the accurate prediction of the final product tank and blending stock tanks. The refinery planner can use the qualities of blending stock tanks available online for recipe planning and optimization quite efficiently without waiting for the lab analysis of stock tanks’ qualities. Figure – 7 shows an architecture diagram for the integration of gomsTQTS™ with feed-forward and feedback control systems.
Refinery/OMS has also integrated gomsTQTS™ with Gasoline Blending Systems using NIR analyzers for stock and product headers to predict the gasoline stock tank qualities online and interface it with online/offline control, optimization, and planning system in the future to realize the maximum benefits of the system. The discussion of this implementation for gasoline blenders will be presented in the future.
Summary
An online tank quality tracking system such as gomsTQTS™ in a refinery offers tangible benefits from better planning of blend recipes, faster shipment to customers, better inventory management, minimization of reblends and giveaways, and can improve the bottom-line. The Singapore refinery (refinery) is a pioneer in implementing such a system developed by Offsite Management Systems LLC and has potential tangible benefits of 2.5-4.4 MUS$/year for the Kero and Diesel blending system. However, the refinery has conservatively estimated a claimable savings of 0.65-1.1MUS$/year with a return-on-investment (ROI) of 3-6 months.
Authors
Dr. Suresh S. Agrawal, President
Offsite Management Systems LLC, US
Mr. Leong Kitt Mun KM, Lead Control Engineer
Singapore Refining Company Pte Ltd., Singapore
Mr. Wee Liang Lum Control Systems Manager
Singapore Refining Company Pte. Ltd, Singapore
Mr. Er Chong Tea James CTJ, Operations Manager
Singapore Refining Company Pte. Ltd, Singapore
#RefineryOffsiteManagementSystems #TanksQualityTrackingSystem #OnlineBlendControl #GasolineBlendingSystems #NirStockAnalyzer #DcsBlendControl #OfflineBlendControl #DcsBlendControl #TanksQualityTrackingSystem #MinimizeGive-AwayQualities
1 2 3 4 5 100%
1 2 3 4 5 0%
1 2 3 4 5 0%
1 2 3 4 5 0%
1 2 3 4 5 0%