ELSEVIER

Integration

1. Introduction

Computers & Chemical

Computers and Chemical Engineering 24 (2000) 1645-1649 Engineering

www.elsevier.com/locate/compchemeng

of planning and real-time optimization in olefins production

David Geddes *, Terry Kubera

Aspen Technology, 1293 Eldridge Parkway, Houston, TX 77077, USA _

Many olefin producers use both advanced planning systems and real time optimizers to maximize overall plant profitability. These two systems are typically not integrated. Recent improvements in both computer hardware and software now make the integration of planning and optimization practical. This can result in significant benefits both in terms of improved results from planning, and improved on-line optimization. This paper will describe the use of planning and opti- mization in world scale olefins plants, and practical methods of integration between the two systems in order to achieve maximum overall site profitability.

2. Advanced planning systems

Many ethylene producers have used linear program- ming (LP) planning tools for a number of years, and report significant benefits in terms of higher profitabil- ity. However, a number of olefin producers do not use this technology, but instead rely on other methods for planning. Companies that do not use LP planning tools, typically make planning decisions, such as feed- stock evaluation, using furnace yield simulators and/or spreadsheet programs. Although these furnace yield simulators are important for accurately estimating fur- nace yields, they typically do not include downstream process efficiencies and constraints, and they typically do not have optimization capability. The advantages of LP planning technology is that models can be devel- oped that include both furnace yields and down stream processing for site-wide optimization. Although eco- nomic decisions based on furnace yield simulators are a good first approximation, significantly better decisions

* Corresponding author. Tel.: + l-281-5841000; fax: + 1-281- 5844329.

E-mail address: david.geddes@aspentech.com (D. Geddes)

can be reached with LP technology that optimizes the synergy between furnace yields and downstream con- straints. The benefits of this optimized decision making technology are typically in the range of $10-20 per ton of ethylene produced.

Linear programming (LP) modeling systems are used as economic decision support tools for a number of planning and scheduling applications. LP is used in these applications because of its ability to optimize complex facilities. This optimization typically results in significant profitability improvements. Applications for olefin producers include:

Investment planning

Annual budgeting

Feedstock evaluation

Ethylene producers typically per- form two types of investment plan- ning. The first is strategic investment planning which involves new plant construction. The second form of investment planning relates to incremental expansion of the ex- isting facilities. Many companies use the model to develop a series of monthly plant performance forecasts in order to establish annual budgets. The plan- ning model is well suited to this ap- plication since this model is an accurate representation of the olefins complex and a series of cases representing the twelve months of the year can be quickly developed. One of the most important applica- tions of planning models is feed- stock evaluation. Proper feedstock selection is vital to plant operating economics since feedstock cost ac- count for approximately eighty per- cent of plant variable operating costs. In this business process, the

0098-1354/00/$ - see front matter 0 2000 Elsevier Science Ltd. All rights reserved. PII: SOO98-1354(00)00440-3

1646 D. Geddes, T. Kubera /Computers and Chemical Engineering 24 (2000) 1445-1649

model is used to determine the economic value of feedstocks that are available for purchase.

Man thly Large process facilities typically em- planning ploy the monthly planning busi-

ness process. The purpose of the monthly plan is to develop agreed production targets for the plant. Inputs include product supply commitments, feedstock supply and what plant shutdowns or plant limits will apply during the upcoming month. The plan- ning model is then run to set fea- sible production targets for the plant. The planning group serves an important role in coordinating activities between feedstock sup- ply, plant operations, and product sales.

Production Although the monthly plan is im- scheduling portant for setting overall pro-

duction targets, it is not particularly useful to plant opera- tions on a day to day basis. This is because the monthly plan is normally developed using monthly average feedstock and product rates. Plant operations require guidance in processing, not the monthly average feed- stock, but the feedstock currently in inventory. The planning tool is used to maximize plant profitabil- ity given the production targets, current feedstock in inventory, and current plant constraints.

Operations de- Operations decision support refers cision support to the business process of provid-

. ing economic based guidance to operations on a daily basis. This is typically related to unexpected changes in plant operations, such as an unexpected furnace shut- down, a pump failure, etc. The planning/scheduling model is used in these situations to deter- mine the economic loss associ- ated with the equipment problem and is used to justify repairs, perhaps on an overtime basis.

Planning models are site-wide models designed to capture overall site economics. These models capture the overall site technology and economics and include

both key constraints in the plant as well as logistics constraints. Planning systems are intended as economic decision support tools as opposed to simulation systems that are primarily intended to support engineering decisions.

There are important differences between planning and scheduling for olefins producers. Planning activi- ties, such as strategic investment planning, annual bud- geting, feedstock evaluation and monthly planning are typically looking ahead for a one-month period or longer. Planning activities are often performed at a corporate headquarters location, often removed from the plant. Scheduling on the other hand is concerned with day to day operations and is typically looking ahead for the next 7-14 days. Scheduling activities are performed at the plant. Because of the different appli- cations for planning and scheduling and also because of the difference in time horizon, there are differences in certain aspects of the models required for both applica- tions. Planning models should be capable of evaluating a broad range of feedstocks and operating conditions. The plant constraints in these models need to cover a broad range of operations. Scheduling models should be used to more accurately predict performance with specific feedstocks and current plant conditions. Plant constraints in these models should accurately reflect current plant conditions. Therefore, even though the planning and scheduling models are very similar there are important differences.

3. Real time optimizers

For the hour to hour operations, real time optimizers determine the proper economic point to operate given the product and feedstock pricing, unit constraints, and energy usage. Complicating the problem is frequent feed changes, downstream disturbances and stream price swings. A robust real-time optimization system must properly handle the non-linearities in an olefin plant by using first principle rigorous models. Rigorous models include distillation, compressors, flash, heat ex- changers and the furnaces. The scope of the optimizer must encompass the entire olefin plant from furnaces through the C2 separation, and C3 separation areas. Typical optimization setpoints include: l Furnaces (feed rate, steam ratio, severity). l Charge gas suction pressure. l Ethylene refrigeration suction pressure. 0 Propylene refrigeration suction pressure. l Demethanizer (overhead ethylene loss, overhead

pressure, bottoms methane composition). l Deethanizer (overhead propylene composition, bot-

toms ethane composition, column pressure). l C2 splitter (bottoms ethylene composition, column

pressure).

D. Geddes, T. Kubera /Computers and Chemical Engineering 24 (2000) 1645-1649 1641

0 Depropanizer (bottoms C3 loss, overhead C4 loss, column pressure). Underneath the real time rigorous optimizer is an

advanced process control (APC) system, which has local optimization capability. Multivariable control is ideal for this application using empirical models to predict steady state, optimize the independent variables to satisfy all the constraints and economic objectives, and then dynamically optimize the movement of the independent variables to maintain operational stability. The APC runs on a frequency of minutes and is capable of implementing rigorous optimizer solutions timely and safely.

Implementation of real-time rigorous optimizers and advanced control systems on olefin plants pro- duces significant economic benefits. Annual bene- fits realized range from $5 to 15 per ton of ethylene produced.

Unlike planning systems, rigorous optimizers and controllers contain much more detailed information about plant constraints and the affect of rates, stream composition, pressure, etc. on the constraint. It is this detail that enables optimizers to adjust operations and move closer to multiple constraints and thereby gain the benefit.

4. Production accounting

Production accounting, or yield accounting, closes the loop between the planned operation and the actual operation. It actually serves a number of functions for this purpose. Some of these include: l Accounting. l Transfer amounts. l Production loss resolution. l Instrument troubleshooting. l Process troubleshooting. l Operations feedback.

For production accounting to be meaningful in oper- ations, balances must be done frequently and consis- tently to get immediate feedback on how the unit is doing according to the latest plan. This becomes espe- cially crucial when the operational plan changes fre- quently or significantly. Currently, many facilities use custom spreadsheets to perform their production ac- counting. This is time consuming and does little to perform data reconciliation and opens the business process up to a number of errors. Therefore, a produc- tion accounting system must be robust and be able to perform gross error detection as well as data reconcilia- tion with a user friendly front-end. A fully robust production accounting system can bring benefits of $1 per ton of ethylene produced.

5. Integration of systems

Because the current market situation demands pro- ducers to be agile, feedstock selection, planning and scheduling are done frequently putting demands on operations to change quickly. It is important that the plant operations be able to keep up with these changes. Specifically, the real time optimization system must be updated with new plans and the production accounting system must be able to accurately reflect the current production in a timely fashion. Conversely, the plan- ning system must adequately reflect the plant opera- tions to avoid calculating unobtainable targets.

Olefin producers find there is significant value inte- gration of the applications. Specifically: l Integration of the planning system and production

accounting. l Integration of the planning system and real-time

optimization and APC. Benefits of integration come from improved planning

and optimization. With consistent models, the planning system will reflect the actual operations more closely and the real time optimization system will reflect what the planning system objectives. Benefits are estimated between $1 and 3 per ton of ethylene produced.

5.1. Integration of planning and yield accounting

The business process of performance monitoring to evaluate plant operations on a regular basis requires the integration of planning and yield accounting. The abil- ity to compare actual performance with planned perfor- mance on demand is a valuable capability. This comparison becomes especially difficult as unexpected disturbances such as a downstream customer having plant problems or a feedstock delay.

To accomplish this, reconciled plant operating data from the yield accounting system is coupled with the information from the planning system. This informa- tion is used to automatically prepare a ‘plan’ versus ‘actual’ report on a daily basis. For example, if this report identifies 2 weeks into the month that the desired propylene production for the month is 5% below plan, then appropriate action can be taken before the end of the month arrives. This integration is best accomplished through an integrated production management system.

5.2. Integration of scheduling and closed-loop real-time optimization

An important consideration for olefin producers is the integration of scheduling and real time optimiza- tion. Closed-loop real-time optimizers are widely be- coming a standard in the olefins industry. Real-time optimizers are based on non-linear models and optimize

1648 D. Geddes, T. Kubera /Computers and Chemical Engineering 24 (2000) 1645-1649

plant operations on an hour by hour basis. This opti- mizes the plant based on changes in ambient air temper- ature, feedstock quality, active constraints, etc. on a regular basis.

Real-time optimizers require accurate price informa- tion relating the intermediate products produced by the unit to function properly. The best source of this inter- mediate pricing information is the production schedul- ing LP model. This pricing information is passed to real-time optimizers in the form of marginal values for feeds, products and utilities. In addition to the pricing information, the production scheduling LP also contains information for upstream or downstream constraints that may not be included in the scope of the optimizer. Therefore. this information is a_!~0 reouired for the _ -- - _ _ - - _ _ , ____ - ________ --_ _L ---l----- --- optimizer to accurately calculate the optimized set- points.

In addition to the optimizer’s constraint requirements, the production scheduling LP model also needs informa- tion regarding current plant constraints. This informa- tion is readily available in the plant optimizer and APC system. This information is needed so the production scheduling LP model produces accurate results. An example might be the propylene refrigeration compres- sor is at a maximum speed constraint. As a result, some production schedules would cause this constraint to be further violated. However, if the LP model receives this . -0. _._L:.__:__.. Al__- LL_ .___1.._ inIomiatioii fKxIi the piailt opiinmer, tnen ant: pruuuc;- tion plan produced from the LP will calculate the optimum plan, taking into account that the speed con- straint is active. This transfer of data requires an inte- grated production management system to facilitate data generation, data storage, and transfer between systems.

The intermediate process stream prices are required as input data for the olefins plant real time closed loop optimizer that maximizes plant profitability on an hour by hour basis.

6. Integrated production management

Planning, scheduling, optimization and control all involve optimization and, therefore, one would want to use the same model for each, however, each application is different both regarding the time horizon and the scope. An illustration of the difference in time horizon is shown below.

Technology Time basis Scope

Planning

Scheduling Real-time

optimization

Monthly/ Single OY multi- yearly site Daily/weekly Site-wide Hourly Single plant/unit

Advanced process Minutes Equipment control

Consistency between models is critical to ensure the planning system and real time system drive towards the same objective. Items that affect consistency include constraints, pricing, and operating characteristics (column splits, horsepower usage, etc.). Transfer and mapping of data from one model to the other can bring about this consistency. Integration of these systems has been attempted in the past with limited results. The integration faces many difficulties:

Integration solutions have a difficult workflow man- aopmpnt For .-YamnlP mnltinlP imnnrtc/eunnrtc and .+b”“-V”” * V& V‘.u”‘y’-, *Uu”‘~ I” uuyv’L LY, w‘.y”‘cY Ull..

multiple operations within multiple spreadsheets. Custom interfaces which are difficult to maintain. The user of the system moves on to other responsi- bilities and no one understands the workflow methodology. One of the major systems is upgraded/changed and the interface to other systems must be completely re-done The key to capturing the benefits is easy workflow

management. The system must integrate different appli- cations produced by different vendors, must be up- graded easily, and must provide the tools and environment to improve the overaii business processes. In essence, an integrated production management sys- tem is required.

Integrated Production Management

The integrated production system is PC based and uses Microsoft’s DNA technology for data storage, business logic, and user presentation. The system has the proper security levels for different users and can be accessed at the site or at headquarters.

An example business process is updating the plan- ning model with current operations. Within the inte- grated production management system, the planner can access current constraint information from the plant data historian, the real time optimization system, and the real time control system in one application. The pianner can then iook at the various constraints, seiect

D. Geddes, T. Kubera /Computers and Chemical Engineering 24 (2000) 16451649 1649

changes, and re-run the planning model to see the effect. All this is accomplished through one interface without needing to know the specifics of a plant historian, real-time optimizer, or control system. Interfaces to these various applications are transparent to the user, which frees up the user to do more analysis and less data acquisition and manipulation. It is at this point where integration benefits are obtained.

7. Summary

This historical planning and scheduling business pro- cess is not sufficient to fully optimize olefins plant performance. However, modern planning and schedul-

ing tools that are integrated with other key technologies such as real-time optimization and production account- ing can he used to significantly improve overall olefin plant performance. In summary, these planning and scheduling tools allow producers to: l Support profit oriented decisions. l Identify problems early. l Quickly quantify economics of problem. l Improve communications and efficiency. l Improve site-wide profitability.

Leading companies are actively pursuing the integra- tion of the planning and scheduling processes. They are also expanding the use of their existing planning and scheduling tools to include the business processes dis- cussed here.