8/22/2019 Integrated Advanced Control and Online Optmization in Olefins Plant http://slidepdf.com/reader/full/integrated-advanced-control-and-online-optmization-in-olefins-plant 1/7 Compu rers c/rem. E ngng, Vol. 13, No. I l/12, pp . 1291-1297, 1989 0098.13.54/89 $3.00 + 0.00 Pr inted in Gr eat B ritain. All rights reserved Copyr ight 0 1989 Pergamon Pr ess plc INTEGRATED ADVANCED CONTROL AND ONLINE OPTIMIZATION IN OLEFINS PLANT R. J. LOJEK an d B. D. WHITEHEAD LI NDE AG, TVT Division, D-8023 Hoellriegelskreuth, Munich, F.R.G. (Receivedfor publication 19 June 1989) Abstract-The benefits of optimization cover a broad spectrum. In many cases the rewards are in the range of several million dollars per year. The first step prior to the implementation of optimization is a detailed study of the current operating conditions and philosophy and market demands. Only then can an optimizer be effectively designed and implemented. The system which consists of advanced control of key plant sections linked to a global online plant optimizer will be described. Since optimum operating conditions are almost always at one or more constraint boundaries, a major task of the advanced control strategies is constraint riding. The advanced control strategies and techniques used are described in detail. The online optimizer is based on a detailed model of the whole plant. Although the major benefits are in optimal operation of the cracking section, a detailed model of the separation section is essential for accurate prediction of constraints. Online data is used to identify changes in feed properties and a suitable starting point for the optimization as well as to update the model correlations and improve accuracy. The design of a simple and robust operator interface is critical to the success of such a system and will be described in detail. INTRODUCTION Olefins plants are ideal candidates for application of online optimization. These plants are extremely integrated from the heating and cooling requirements and can be operationally adjusted to reflect the market demands. The areas of global optimization and local optimization/advanced control will be dis- cussed in this paper (see Whitehead and Parnis, 1987 for a typical application). The typical characteristics of olefins plants (see Fig. l), such as high throughput, a complex interacting process, varying feed stocks, wide product slate, changing market conditions for feed stocks and products make it difficult to identify the optimal set of operating conditions for day-to-day operation. Optimization is the process of finding the extreme value of a plant objective function (either maximum or minimum depending on the application) under constrained operating conditions. Typically the ob- jective function takes the form of overall utility consumption, profit or production. On-line control/optimization is divided into several levels. These are basic control (normal PID con- troller, simple cascades), local advanced control, local optimization and finally global optrmization. The functional quality of each level is strongly dependent on the functional quality of the lower levels. It is of course impossible to have good optimizer perfor- mance when the simplest PID control loops do not function properly. In order to implement a global optimization system, one must start at the lowest level and work slowly and carefully to the top. Some of the most valuable players in this game are the plant maintenance personnel. It is absolutely essential that all measurements and control points used by the optimizer function reliably and trouble free. A common phrase used in the computer industry is “garbage in-garbage out” and this of course also applies to optimization. When more than one variable is to be optimized, the problem becomes too difficult to be solved by experience. A computer-based optimization system is then required. The global optimizer enables the user to determine the best set of operating conditions fo r the given plant boundary conditions (such as feed- stock availability, product demand, etc.) This opti- mization has to respect plan t constraints, i.e. the optimum is only valid when the operation does not lead to a bottleneck situation in one or more parts of the unit. A plant model calculates for a given set of input conditions (such as purities and feed quantity) or set by the boundary conditions (such as ambient tem- peratur e, specific cost for feed/pro duct/u tilities or product limitations) the plant profit taking into ac- count the constraint situation for all key plant items. The results from the simulation model are fed back to the optimizer which generates, based on present and past results, a new set of values for the optimized operating variables. The procedure is iterative, the approach to the optimum depends on the number and type of variables to be optimized and the number of iterations. The impact of the individual input variables on the plant profit differs widely. Table 1 shows only those CACE 13-11;12-H 1291
7
Embed
Integrated Advanced Control and Online Optmization in Olefins Plant
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
8/22/2019 Integrated Advanced Control and Online Optmization in Olefins Plant
Compu rers c/rem. E ngng, Vol. 13, No. I l/12, pp. 1291-1297, 1989 0098.13.54/89 $3.00 + 0.00
Pr inted in Gr eat B ritain. All rights reserved Copyr ight 0 1989 Pergamon Pr ess plc
INTEGRATED ADVANCED CONTROL AND ONLINE
OPTIMIZATION IN OLEFINS PLANT
R. J. L O J E K and B. D. WHITEHEAD
LI NDE AG, TVT Division, D-8023 Hoellriege lskreuth, Mun ich, F.R.G.
( R e ce i v ed f o r p u b l i c a t i o n 19 June 1989)
Abstract-The benefits of optimization cover a broad spectrum. In many cases the rewards are in the rangeof severa l million dollars per year. T he first step prior to the implementation of optimization is a detailedstudy of the current operating conditions and philosophy and market demands. Only then can anoptimizer be effectively designed an d implemented.
The system which consists of advanced control of key plant sections linked to a global online plantoptimizer will be described. Since optimum operating conditions are almost always at one or moreconstraint boundaries, a major task of the advanced control strategies is constraint riding. The advancedcontrol strategies and techniques used are described in detail.
The online optimizer is based on a detailed mod el of the whole plant. Although the major benefits arein optimal operation of the cracking section, a detailed model of the separation section is essential foraccurate prediction of constraints. Online data is used to identify changes in feed properties and a suitablestarting point for the optimization as well as to update the model correlations and improve accuracy.
The design of a simple and robust operator interface is critical to the success of such a system and willbe described in detail.
INTRODUCTION
Olefins plants are ideal candidates for application
of online optimization. These plants are extremely
integrated from the heating an d cooling requirements
and can be operationally adjusted to reflect the
market demands. The areas of global optimization
and local optimization/advanced control will be dis-
cussed in this paper (see Whitehead and Parnis, 1987
for a typical application). The typical characteristics
of olefins plants (see Fig. l), such as high throughpu t,
a complex interacting process, varying feed stocks,
wide product slate, changing market conditions for
feed stocks and products make it difficult to identify
the optimal set of operating conditions for day-to-day
operation.
Optimization is the process of finding the extreme
value of a plant objective function (either m aximum
or minimum depending on the application) under
constrained operating conditions. Typically the ob-
jective function takes the form of overall utility
consumption, profit or production.
On-line control/optimization is divided into severallevels. These are basic control (normal PID con -
troller, simple cascades), local advanced control, local
optimization and finally global optrmization. The
functional quality of each level is strongly dependent
on the functional quality of the lower levels. It is of
course impossible to have good optimizer perfor-
mance when the simplest PID control loops do not
function properly. In order to implement a global
optimization system, one must start at the lowest
level and work slowly and carefully to the top.
Some of the most valuable players in this game are
the plant maintenance personnel. It is absolutely
essential that all measurements and control points
used by the optimizer function reliably and trouble
free. A common phrase u sed in the computer industry
is “garbage in-garbage out” a nd this of course also
applies to optimization.
When more than one variable is to be optimized,
the problem becomes too difficult to be solved by
experience. A computer-based optimization system is
then required. The global optimizer enables the user
to determine the best set of operating conditions for
the given plant boundary conditions (such as feed-
stock availability, product demand, etc.) This opti-
mization has to respect plan t constraints, i.e. the
optimum is only valid when the operation does not
lead to a bottleneck situation in one or more parts of
the unit.
A plant m odel calculates for a given set of input
conditions (such as purities an d feed quantity) or set
by the boundary conditions (such as ambient tem-
peratur e, specific cost for feed/pro duct/u tilities or
product limitations) the plant profit taking into ac-count the constraint situation for all key plant items.
The results from the simulation model are fed back
to the optimizer which generates, based on present
and past results, a new set of values for the optimized
operating variables. The procedure is iterative, the
approach to the optimum depends on the number
and type of variables to be optimized and the number
of iterations.
The impact of the individual input variables on the
plant profit differs widely. Table 1 shows only those
CACE 13-11;12-H 1291
8/22/2019 Integrated Advanced Control and Online Optmization in Olefins Plant
tOCAl_ ADVANCED CONTROL calculates and implements a new coil uullet tern-
P~,fkiti0?2perature set point. The periodic f-eedhack of
tncasured srvrrity derived frutn the cracked gas
As implied by the title, local advanced control is analysis is used to calculate an outlet temperature
the control of a specific area within the plant using at. the lime of sampling. This temperature is
higher level control strategies. An area is typically compared with the actual outlet tompuraturc at
defined as a column with associated equipment. a that time. The diff‘erencc in the temperature is
reactor system, a compressor or a furnace. then used to calculate :I correction of the outlet
Optimization is typically divided into two sub- temperature set point to achic1.c II-K rcyuiredcategories. These are commonly known as local severity. This prcdicror;corwctor approach en-
optimization and global optimization. Local optimiz- sures that the long period of time between
ation is area-specific and gIoba1 optimization entails cracked gas analyst has tittle et%ct on the <tualtivthe entire unit. The following anaingy illustrates the of controi.relationship between these two types of optimization. . Computer-aided start-up and shut-down control
.4 !arge capacity vcsscl having one non-controlled assists the operator during the prace%s steam
Iiyuid inlet stream and a Aow-controlled outlet phase of start-up or shut-down by adjusting c&I
strram is equipped with a level controller. The flow slcant flow miss and outlet letnpwatur2 set poitats
controller responds very quickly- and can hold its set in accordance with the standard start-up and
point with relative ease whereas the level controller shut-down procedure. Tt also assists the operatorrecIoir<s more time in the event of a process dis- during the feed flow phase of start-up or shut-
turbance or a set point change to hold its set point. down by ad,jusiing coil feed Row rates as wc!l. It
The twu controllers work together to obtain good informs the operator when burners have to bc lit
level control. The set point of the level controller may or extinguished and continues control when rhijhe compared to an objective function. Globai opti- has been done. Following this pha~ the normal
mization and local optimization work together in a tasks take control.similar fashion, where the “set point” to the globail *The total run :imc of the furnaces and transferoptimizer is the tank level set point and the signals lint exchange?- (TLX) based on present operaringwhich are passed down to the focal optimizers arc conditions MC calcule~cd. These arc compared
comparable to the setpoint of the flow controller. with a minimum run lime SCL point. Xi‘ thcrc 1s a
problem. the strategy idcntifics the limiting factor
FWfZU(‘C coizt,m/ (e.g. pressure or temperature dil‘rerence UF iem-
The ndsanced control strategies perform the follow-perature in furnace or TLX). ff ths pressure
ing:diEerence is the Iimiting factor, the tbrnace
throughput set point is r-educed to a valur at
which the minimum run time c‘an he mci at
*The total throughput needed to achieve a spe- otherwise constant operating conditions. If thecified production rate is controlled and the hydro- temperature is the limiting factor either the
carbon feed is distributed to the individual coils throughput or the cracking severity are reduced.
to achieve equalized coil radiant zone outlet l Decoking controI assists the operator during the
retnpcratures. decoking procedure by adjusting the decoking
@ Steam-to-feed ratio control with safety features air. process steam Row rates and outlet tempera-
to avoid zero flow. ture in accordance with the standard decoking
l Combustion control where the flue gas damper is procedure. It informs the operator when burners
adjusted to achieve a given oxygen content sub- have to be lit or extinguished and continues
ject to a high limit on the fire box pressure. control when this has been done. It periodically
Feedforward compensations for variations in asks the operator if the t~lbe vcali terrqwatcrre is
both fuel gas calorific value and fuel gas pressure acceptable and the strategy continues onI> atier
are applied. confirmation.
l Outlet temperature control. The operator can t Local optimiza?~on is essentially the globaf opti-
choose to use the outlet temperature or the mization described in the previous section. sinceaverage coil radiant zone temperature in each the furnaces are the heart of the plant.
furnace. Feedforward compensations for changes
in throughput and fuel gas calorific values are CdWF&3 cmt ro
applied,
*Cracking severity control where a yield predic-The advanced control strategy performs the follow-
tion model is used to compensate for changes ining:
the set points for cracking severity. hydrocarbon l Purity control adjusts the reflux rate for the
throughput, steam-to-feed ratio and where ovcrhcad and the boil-up rate for the bottoms to
changes in the feedstock quality and cracked gas achieve the required product purities. Tray tcm-
pressure are automatically compensated for. It peraturc differences together with an analyflcal
1294 R. J. LOJEK and B. D. WHITEHEAD
8/22/2019 Integrated Advanced Control and Online Optmization in Olefins Plant
portant part of any online process control system.
This is more of an art than a science. At this point
one should delineate between the operator interface,
the maintenance interface and the engineering inter-
face.
The op erator interface must be clean and simple.
Only important information should be included. The
operator should not be burdened with inputs which
he is not able to change (e.g. tuning factors, decou-
pling effects, etc.). An examp le of this would be the
advanced con trol of a fractionator. The operator may
set the following inputs:
e Overhead purity.
l Bottoms purity.
o maintenance status of all analyzers used in a
strategy.
mStarting an d stopping of the strategies.
Maintenance personnel requ ire access to controller
tuning factors, data traffic control as well as infor-
mation regarding the overall structure of the system.
Several diagnostic aids are also made available to the
maintenance personnel. One of these is the plausibil-
ity checking of the process variables used in the
advanced control system. For example, using vapor
pressure data it is possibte to comp are a temperature
and pressure in the same service.
The engineer must have easy access to all systems
information. This includes decoupling matrices, opti-
mizer tuning factors and diagnostic systems within
the optimizer.
CONCLUSIONS
The system described above combines the tw o
levels of local advanced control and global on-line
optimization in an oiefins plant. For full realization
of the potential benefits these levels must be well
integrated, for example since the optimum usually lies
on a constraint, the advanced control must be de-
signed to control at the constraints.
In addition the system must be robust and easy-to-
use to ensure operator acceptance.
Finally, the success depends on continued main-tenance of both hardware and software-
REFERENCES
Fuge C., P. Eisele and B. D. Whitehead, Optimization of theoperation of a gas terminal. Presentation to the Inf.Cr>vgenic rMateriuf CoqGrence, Cryogen i c Eng inee r i ng
Con fhv tce , Chicago (1987).Whitehead B. D. and M. Parnis, Computer control im-