Innovation For Refinery With Modelling
Innovation For Refinery With Modelling
Innovation For Refinery With Modelling
I
n this paper the authors discuss a range of new straight-run (LSR) naphtha prior to gasoline blending.
modelling technologies available to meet these Typically, 80% or more of the benzene in the gasoline
challenges. Emphasis is given to the benefits of pool comes from reformate, which then becomes the
judicious use of mathematical models that may be best target for benzene reduction. Either of two basic
used to simulate and thus screen alternatives for meet- approaches can be adopted: minimise benzene forma-
ing product specifications. Mathematical models of tion by removing benzene precursors from the reformer
simple and complex refineries are becoming strategic feedstock, or fractionate out light reformate for subse-
assets of small and large companies. Implemented quent conversion or extraction of the benzene. UOPs
properly, they can be used to find solutions to many Penex-Plus process was developed to reduce the ben-
problems faced by refineries in a very cost-effective zene content and increase the octane of gasoline by
manner. UOP has developed state-of-the-art technol- accepting feedstock containing more than 5 vol-%
ogy for implementation of critical models in various benzene. The integrated benzene-saturation section
simulation environments to facilitate the use of such provides good control of the highly exothermic reac-
studies. tion of benzene with hydrogen. Saturating benzene
The implementation is based on modular technology causes a loss of octane, but because of subsequent
that allows users, to link their model to any simulation isomerisation of the C 6 paraffins and naphthenes, the
environment, to perform simulation, data reconciliation Penex-Plus process yields an increase in the octane of
and optimisation studies. Additionally, this modular the product.
approach allows the use of modern statistical estima- The issue of other heavier aromatics in gasoline is
tion theory to be used in order to tune adjustable model more complex. The aromatics in gasoline are of course
parameters from pilot plant and/or commercial data. a major source of octane, both research octane number
This is necessary to establish the ground rules for (RON) and motor octane number (MON). The main
model performance and fidelity. source of toluene, ethylbenzene, xylenes, and C 9+
The UOP suite of models also features detailed two- heavier aromatics is the reforming unit. The secondary
phase fixed and moving bed reactors (e.g., catalytic source of aromatics is the fluid catalytic cracking (FCC)
reforming, fluid catalytic cracking), as well as new tech- unit. To substantially reduce the concentration of aro-
nology to characterise streams from bulk properties matics in gasoline but maintain its octane, large
and to allow connectivity of multiple refinery units within quantities of oxygenates or alkylates must be blended
the same simulation framework. into the gasoline to substitute for the octane of the
missing aromatics.
Introduction It is evident from the above considerations that the
The reduction and elimination of lead in gasoline is issues in the production of gasoline with the current
well advanced worldwide. The majority of the refiners strict specifications are very complex. In this paper we
now produce a high proportion, if not all, of their gaso- concentrate on the role of mathematical models and
line without lead. To meet the octane and benzene simulation in general, in meeting these challenges. More
challenge, many refiners have installed C 5-C6 isomeri- specifically, we will discuss how, by building suitable
sation units to increase the octane of hydrotreated light mathematical models of various refinery process units
Gasoline Pool
Crude Fractionation
Data analysis issues down within the PONA groups, becomes available.
Data analysis for identification of systematic and gross Although this information is usually not available on-
errors line, it can be augmented when lab data become
Data reconciliation for smoothing the random errors available.
Tuning of the models to the commercial test runs to This feed characterisation procedure may be further
increase their fidelity and confidence levels of the enhanced with data reconciliation by incorporating in-
people that will exercise them. formation that is available from the local upstream and
downstream units. The feed composition greatly af-
Optimisation issues fects the performance of the reactors in terms of heat
Identification of the proper object function(s) generation. Consequently, reactor temperature infor-
Gathering of the proper cost data mation can be used to modify the base composition of
Identification of the scope of the optimisation prob- the naphtha feed. This approach requires accurate re-
lem: on-line or off-line optimisation, single unit or actor models and elimination of other disturbances
multi-unit optimisation. such as catalyst deactivation or poisoning as the prob-
able cause of the model-plant mismatch.
An analysis of the modelling, optimisation and data The feed characterisation problem also can take a
issues along with some solutions proposed by UOP are different form. To demonstrate this problem, consider
presented in detail in the sections that follow. a multi-unit simulation that involves a crude tower and
a naphtha complex. If the feed to the naphtha splitter is
Feed characterisation for kinetic and to be connected to an upstream crude distillation unit,
empirical models the connection must be made in a thermodynamically
Detailed feed characterisation is a critical require- consistent manner. In general, the simulation stream
ment for the successful implementation of simulation definitions around the crude unit are vastly different
technology in a refinery complex. This requirement from the ones required by the Platforming and isomeri-
arises from the need to implement rigorous kinetic sation models, which are based on pure components.
models for the reactor models that include pure com- UOP has implemented the concept of stream finger-
ponents in the underlying reaction mechanism. printing technology that connects two different
Complete information about these pure components is manifestations of the same stream: the upstream mani-
not available on-line. As a result, robust feed charac- festation that in general includes light components and
terisation procedures need to be developed. pseudo-components that are derived by an oil charac-
The calculation naphtha like feed composition that terisation, and the downstream manifestation that
was developed at UOP uses P(araffin) O(lefin) consists of a slate of pure components recognised by
N(aphthene) A(romatic) analysis results, ASTM-D86 dis- the kinetic models for reforming and isomerisation.
tillation curve, and specific gravity of the naphtha feed. Similar technology applies to the cases where a
In mathematical terms the feed characterisation prob- pseudo-component based stream will have to be linked
lem can be formulated as follows: to empirical reactor models. The pseudo-component
manifestation is converted to bulk properties such as
minimize: D1160 curves, API gravity, molecular weight and UOP
subject to: K that are usually recognised by the reactor model.
RX Output
Bulk Data
optimisation include specific tech-
Feed nology for the incorporation of such
Calculation Product
Lights Engine Lights
reactor models. Figures 2 and 3
Pseudocomponents I
Properties
Pseudocomponents II show schematically how a kinetic
Properties
based and/or bulk data based reac-
tor model would fit into such an
Figure 2. UOP bulk data based reactor model implementation environment. In a typical implemen-
in flowsheets tation, simulator streams are
mapped to streams recognised by
the reactor model. When the model
executes in simultaneous solution
Component Based Reactors mode, it supplies the effluent on de-
mand from the simulator solver. This
Kinetic Lumps
Kinetic Lumps