The operation of any petrochemical plant requires the transfer of energy to and from the
process, movement of material through the process piping and vessels and suppression of
unwanted reaction or combustion. This is mainly achieved through the use of utilities which
serve as auxiliary variables in plant operation and include electricity, steam, cooling water,
hydrogen, nitrogen, compressed air and fuel gas.
These utilities have costs associated with them and it is therefore common practice to optimise
on the use of these utilities. Other aspects to consider are the generation and transportation/
transmission of these utilities to the point of consumption. These areas have not received
much attention historically mainly due to the large differences that existed between product
prices and utility costs. Recently, there has been a revival in the focus on efficient operation
fuelled by global economic turbulence, higher energy cost, stricter environmental policies,
dwindling fossil fuel supplies and the threat of climate change.
The losses encountered in the generation and transmission of utilities are in many cases
substantial, especially in those utilities where releases to the atmosphere do not present
significant safety or environmental risk such as compressed air, cooling water and steam systems. Also in many cases, the transmission distance is substantial which gives rise to
more losses and inefficiencies.
Furthermore, many electrical utility providers are using time-of-use electricity costs and maximum
demand penalties. This necessitates scrutinising the exact time of consumption and
peak consumption rather just the overall amount consumed.
This study explores the potential benefits of utility control and optimisation from a supply/
generation point of view through the application of modern advanced control and optimisation
techniques. The approach and results for a dual circuit cooling water system are
illustrated.
The cooling water system is an example of a hybrid system where both continuous and discrete
input variables are present. This complicates the formulation of control and optimisation
solutions. A model of the system is developed and verified using real plant data. To improve
model accuracy, a parameter estimation exercise is performed using a genetic algorithm as
the optimisation platform.
Several control and optimisation schemes are then developed with varying degrees of complexity
including advanced regulatory control (ARC), hybrid non-linear model predictive control
(HNMPC) and economic hybrid non-linear model predictive control (EHNMPC). The ARC
scheme uses classical advanced base layer control techniques, together with time and condition
based switching logic, which do not require the use of a system model. The HNMPC and
EHNMPC schemes make use of a system model in the formulation of the control solutions
and use a genetic algorithm for optimisation.
The model is then used to evaluate the performance of the control and optimisation techniques
in several simulation studies by comparing the results to that of a base case study. The results
indicate that substantial energy and cost savings may be achieved without the need to install
additional plant equipment.
Die suksesvolle bedryf van enige petro-chemiese aanleg is afhanklik van die oordrag van energie
van en na die proses, beweging van materiaal deur die aanleg se pyp- en struktuurnetwerk
en die onderdrukking van ongewensde reaksie of ontbranding. Hierdie doelwitte word hoofsaaklik
bereik deur gebruik te maak van utiliteite wat as hulpveranderlikes dien en sluit
onder andere elektrisiteit, stoom, verkoelingswater, waterstof, stikstof, saamgepersde lug en
brandstofgas in.
Daar is koste aan sulke utiliteitstrome verbonde en dit is daarom algemeen om die benutting
daarvan te optimeer. Ander areas om in ag te neem, is die opwekking en vervoer van hierdie
strome tot by die verbruikspunt. Hierdie areas het histories nie soveel aandag geniet nie,
hoofsaaklik as gevolg van die groot verskille wat tussen produkpryse en utiliteitskoste bestaan
het. As gevolg van die onstuimighede in die wêreldekonomie, hoër energiekoste, strenger
omgewingswette, kwynende fossielbrandstofreserwes en drygende klimaatsverandering word
daar egter tans nuwe klem op meer effektiewe aanlegsbedryf geplaas.
Die verliese wat aangetref word in die opwekking en vervoer van utiliteite is in baie gevalle noemenswaardig, veral in gevalle waar verliese na die atmosfeer nie enige ware risiko s vir
veiligheid of die omgewing inhou nie, byvoorbeeld in die geval van saamgepersde lug, verkoelingswater
en stoom. In baie gevalle is die afstand waaroor die oordrag plaasvind groot wat
verder tot verliese en ondoeltreffendheid bydra.
Verder maak heelwat elektrisiteitsverskaffers deesdae gebruik van tyd-van-verbruik tariewe
en maksimum aanvraag boetes, waar die hoeveelheid energie wat gebruik word nie die enigste
faktor is wat in ag geneem moet word nie, maar ook wanneer die energie gebruik word.
Hierdie studie ondersoek die moontlike voordele verbonde aan die beheer en optimering van
utiliteite uit n opwekkings- en oordragsoogpunt, deur gebruik te maak van moderne beheeren
optimeringstegnieke. Die benadering en resultate word geïllustreer deur die toepassing
daarvan op n dubbel-baan verkoelingswaterstelsel.
Die verkoelingswaterstelsel is n voorbeeld van n hibriede stelsel waar beide diskrete en
kontinue insetveranderlikes teenwoordig is, wat die formulering van beheer- en optimeringsoplossings
kompliseer. n Model van die stelsel word ontwerp en geverifieer deur gebruik te
maak van ware aanlegdata. Ten einde die model se akkuraatheid te verbeter word n genetiese
algoritme gebruik in n parameterpassingsoefening.
Verskeie beheer- en optimeringskemas word dan ontwikkel met wisselende grade van kompleksiteit,
insluitend gevorderde regulerende beheer, hibriede nie-lineêre model voorspellende
beheer en ekonomiese hibriede nie-lineêre model voorspellende beheer. Die gevorderde regulerende
beheer maak gebruik van klassieke gevorderde basisvlak beheer tegnieke, tesame met
tyd- en voorwaarde-afhanklike skakelingslogika, wat nie die ontwikkeling van n stelselmodel
benodig nie. Die hibriede nie-lineêre model voorspellende beheer en ekonomiese hibriede
nie-lineêre model voorspellende beheer skemas maak gebruik van n stelselmodel in die formulering
van die beheeroplossing en gebruik n genetiese algoritme vir optimering.
Die model word daarna gebruik om die prestasie van die verskillende beheer- en optimeringstegnieke
te evalueer in n verskeidenheid van simulasiestudies deur die resultate te
vergelyk met die van n basisgeval. Die resultate dui daarop dat n noemenswaardige energieen
kostebesparing bewerkstellig kan word sonder dat addisionele toerusting geïnstalleer hoef
te word.
