Molecular management for refinery-

petrochemical complexes
Designing a digital twin for process simulation of an integrated complex
enables enhanced molecular management to deliver higher margin

Jitendra Chellani and Sachin Srivastava KBC (A Yokogawa Company)

T
he future comprises a world
of fuels refineries, refinery-in- 50
tegrated petrochemical com- Ethylene Butadiene

plexes and crude oil to chemicals.

Once-through cracking yield, wt%

Propylene BTX
40
Through the energy transition, the
latter two will no doubt be more
30
resilient toward future demand-
and/or supply-side dynamics.
20
The potential for incremen-
tal improvement with increased
conversion capabilities of a fuels 10

refinery with added petrochem-

ical integration is $1.5-2/bbl of 0
Ethane n-butane C5-C6P C7-C8P C9P C10P
crude processed. The value gained Propane i-butane C5-C6N C7-C8N C9N C10N
Cracker feed components
from effective molecular manage-
ment is significant. Key objectives Figure 1 Once-through steam cracking yield for feed components
of molecular management for the
overall complex are: grated complex, as well as ongoing sustainment
• Improvement in gross product worth by max- of benefits through real-time monitoring and
imising high value products re-optimisation.
• Reduced cost of feedstock by replacing expen-
sive imports with low value streams Why manage molecules?
However, with the increase in complexity of To recognise the need for molecular manage-
refinery-integrated petrochemical complexes, a ment and carbon number level simulation, it is
gap has developed in the tools needed to effec- important to understand the difference in nature
tively drill down to the stream carbon number of refining and petrochemical processes. The pur-
level on a continuous basis. Traditional tools pose of refinery naphtha processing units is to
used for refinery optimisation rely on bulk maximise octane barrels which can be blended
properties such as cut point for stream optimi- into gasoline. Isomerisation and dehydrocycli-
sation. Such methods have been found to have sation reactions improve octane, with thermal
limitations in fully achieving the objectives of cracking a side reaction that is minimised to pre-
molecular management. The latest process sim- vent yield loss. However, for steam cracking in
ulation tools address this by enabling detailed petrochemicals, thermal cracking of naphtha is
carbon number breakdown from crude assays optimised to produce olefins.
through to blending and petrochemical units, for Due to differences in the units’ reactions, selec-
the whole integrated complex. These enhanced tion of molecules to be processed is extremely
capabilities enable identification of site-wide important. Replacing molecules from one process
optimisation opportunities across the inte- unit to another can improve yields from both

www.digitalrefining.com/article/1002475 PTQ Q3 2020 1

 Order of preference
naphthenes and benzene in the
reformer feed will contribute to
C3 SC LPG benzene product for an aromatics
C4 SC MS LPG SC – Steam cracker
reformer and can produce off-spec
C5 SC ISOM MS MS – Gasoline
product for a gasoline reformer.
• C7s tend to crack more in the
C6 P SC ISOM MS CCR PX CCR MS ISOM – Isomerisation unit

C6 N and A SC CCR PX MS CCR MS ISOM CCR PX – Naphtha reformer

for p-xylene ISOM unit compared to C5/C6 and
C7 SC CCR PX CCR MS MS ATF/HSD
C8 CCR PX SC CCR MS ATF/HSD
CCR MS – Naphtha reformer
for gasoline
therefore should be minimised.
C9 CCR PX SC CCR MS ATF/HSD ATF / HSD – Jet or diesel
Routing of C7 molecules to the
C10+ SC ATF/HSD aromatics reformer is preferred
High preference Low preference
over the gasoline reformer. This
routing leads to a reduction in the
Figure 2 Feed selection criteria A7/A9 ratio in a transalkylation
unit and improves octane barrels
the specific units as well as reducing operating of gasoline pool.
expenses such as energy consumption, other util- • C7+ components in heavy and medium naph-
ities, catalyst, and chemicals. tha can also be used for middle distillate blend-
The yields from a steam cracker can vary ing. If C7+ components cannot be processed in
extensively depending on the feedstock carbon any of the process units, they can be blended in
number and different isomers (see Figure 1). middle distillates diesel rather than selling them
Yield of ethylene and propylene is expected as open grade naphtha.
to reduce while yield of heavier products is • C8s are preferred feedstock for an aromatics
expected to increase with an increase in the feed reformer as they reduce the throughput of most
carbon number. Even for feeds with the same of the process units in the aromatics and recy-
carbon number, cracking yields change with cles as well. This reduces cracking to light ends
changes in C/H ratio, for instance paraffinic and the opex of the units, and improves p-xylene
feeds have higher ethylene yield. Different feeds yield.
can be cracked separately or co-cracked, all hav- • C9s in aromatics reformer feed reduce the A7/
ing an impact on the cracking yields and furnace A9 ratio of a transalkylation unit which improves
run length. Aromatics components in the feed the yield of xylenes and reduces benzene yield.
rarely crack to produce ethylene and propylene, • C10+ components are generally not preferred in
but these may be too costly to separate prior to aromatics, but these can be processed in a steam
processing. cracker.
The profitability of the aromatics is driven by Apart from the reactions and yields, feed
the reformate yield and distribution of compo- selection criteria strongly depend on opex and
nents in the reformate. Higher yield of xylenes the relative prices of fuels compared with pet-
over benzene is preferred from the aromatics, rochemicals. The latest process simulation tools
which could be achieved by selecting the right are critical for estimating optimum steam or
feed for the naphtha reformer. molecule routing based on stream composition.
Selection criteria for different feeds in refinery Molecular management is not limited to optimi-
process units depends on the predicted octane sation of refinery naphtha only, as the optimised
barrels produced by feed components. routings are applicable for streams from the pet-
Figure 2 explains typical preferences which can rochemical complex as well.
be used to select feeds for different process units
and product blending: Challenges
• Lighter hydrocarbons like C2 components in Quality parameters measured and analysed by
off-gas and LPG are preferred feedstock for a the refinery engineers are usually based on bulk
steam cracking unit as these molecules provide properties. Distillation and density are consid-
higher yield of lighter olefins. ered the most important properties for refining
• First preference for routing C5/C6 molecules units. Though gas chromatography is performed
should be the steam cracker due to the signifi- for selected streams in naphtha processing units,
cant premium for light olefins over gasoline. C6 components routed to refinery products are sel-

2 PTQ Q3 2020 www.digitalrefining.com/article/1002475

Figure 3 Petro-SIM model of an integrated refinery-petrochemical complex

dom analysed. The molecules lost in jet from the tha feed and do not simulate component based
crude unit and the composition of hydrocracker assays and the fractionation efficiency of refinery
naphtha are occasionally analysed in the refin- optimisation.
ery. On the other hand, the focus for petrochem- Although LP tools allow some flexibility to
icals is usually on pure components rather than simulate refinery-petrochemical integration, uti-
bulk properties. Some of these components can lising simplified LP models can give erroneous
be lumped together for monitoring purposes but results, for instance feed quality parameters in
detail to the level of isomers is required for a few the form of bulk PNA, which do not capture the
units. For example, N+2A is used as a feed qual- effect of carbon numbers in the steam cracking
ity indicator for a gasoline reformer, however operating parameters. Also, a significant fraction
C7/C9 distribution is important for an aromatics of components are continuously recycled in aro-
reformer. matics units, which has an impact on unit opex
Simulation tools for refinery and petrochem- and the maximum feed that can be processed.
ical units are developed based on the informa- Very few refiners simulate these recycles in an
tion available – bulk properties for a refinery and LP based model.
composition for petrochemicals. Integrating these
tools to create a complex-wide process model Proven basis of the digital twin
has been a challenge for process engineers in the KBC has developed and improved Petro-SIM
past. models for refinery-petrochemical integration
Tools based on linear programming (LP) are a over many years, working on design configura-
solution for bigger optimisation problems such tion and optimisation engagement with clients.
as crude selection, however they are rarely con- Petro-SIM has a seamless transition from bulk
figured for molecular management. Even when properties to component based simulation and
they are used, LPs have their own limitations. vice versa. Information for components avail-
Organisations focused on refining operations use able in crude assays, as well as generated from
comparatively simplistic techniques to simulate the refinery reactor models, is carried through to
petrochemical units in LPs; petrochemical oper- petrochemical models using Petro-SIM’s infra-
ators typically start optimisation at the naph- structure. Reactor models are calibrated based

www.digitalrefining.com/article/1002475 PTQ Q3 2020 3

 on the performance of the operating unit, which refinery-petrochemical complex. According to
allows process engineers to track information at the design of the integrated complex, C5 from a
a molecular level that is rarely available at the cracked gasoline hydrotreater was to be routed
operating unit. as a feed to the steam cracker. The C5 material
The main function and value of the com- was highly olefinic and therefore would give
plex-wide process simulation model is under- poor cracking yields unless it was hydrogenated
standing the interaction between different to saturate the olefins. Based on the relatively
process units and product yield. This can be small quantity of C5, a dedicated hydrogenation
quickly and easily done, as the equilibrium and system was not economically justified. The next
kinetic based models for all the refinery, steam best option was to blend C5 into the gasoline pool
cracker, and aromatics units are in a single Petro- as it had relatively high RON and would boost
SIM simulation flowsheet (see Figure 3). gasoline production owing to its lower specific
This combination holistically and seamlessly gravity. However, routing this stream to the
captures the effect of feed molecules. The inte- gasoline pool without any other change would
grated process model carries detailed compo- have reduced petrochemical yield and would
sition and non-linear kinetics and can be used therefore have given lower overall margins. The
to augment the LP model to boost profitability other potential constraint was the high RVP of C5
through effective molecular management. LP which would have caused issues with the RVP
models are still required for bigger optimisation and VLI of gasoline.
problems, however an integrated process model KBC reviewed all available streams that can
is needed to routinely validate the optimised be used to replace C5 as a cracker feed. Raffinate
solution estimated by an LP. The optimised solu- from the aromatics extraction unit was partially
tion analysed through the integrated process returned to the cracker as feed, with the balance
model should be used to validate various con- sent to the gasoline pool to meet volume demand
straints and to estimate operating targets. for gasoline. The raffinate was rich in paraffins
Various process streams are exchanged with C6-C8 components, an ideal feedstock for
between refinery and petrochemical units in steam cracking. KBC proposed to replace the C5
the complex-wide model. Apart from feeds to feed to the steam cracker with raffinate, routing
cracker and aromatics, many other streams are the C5s to the gasoline pool.
also mapped, for instance off-gases and propyl- A complex-wide model was used to capture
ene from a refinery, integration of the hydrogen the impact on the cracker yield for both feeds
network, and so on. and changes in gasoline RVP/VLI. Due to dif-
KBC has deep experience of using integrated ferences in the product specifications, summer
models for refinery-petrochemical complexes and winter operations were analysed and it was
for identifying and implementing opportunities found that all of the C5 could be absorbed in the
for profit improvement for the parties involved: gasoline pool. There was an overall increase in
re-routing of streams, changing unit severities, petrochemical yield due to better cracking from
and modifying the degree of fractionation. In joint raffinate while maintaining gasoline target vol-
venture projects, a single integrated model is val- ume production. This resulted in an economic
uable for transfer pricing of intermediate streams benefit of almost 7c/bbl of crude. An overview of
between respective parties, based on stream qual- this case study is shown in Figure 4.
ity. After all, it is the streams’ molecular nature
which ultimately determines its value in use. The A digital twin for unit monitoring
use of the integrated model extends to the cost Extending the availability of real-time unit oper-
side too, with integrated modelling of utilities. ations data to refinery and petrochemical pro-
The integrated model can be used to estimate cess engineers can transform their situational
transfer pricing, as well as optimise the over- awareness of unit performance. With a greater
all complex for the mutual benefit of all parties understanding of the factors affecting unit per-
involved. formance, in real-time, deeper unit optimisation
can be achieved.
Case study Utilising the complex-wide Petro-SIM model,
A KBC client was planning to build a grassroots with detailed embedded reactor representa-

4 PTQ Q3 2020 www.digitalrefining.com/article/1002475

tions calibrated using operat-
ing data, allows the integrated Ethylene

model to represent true unit Aromatics BTX

Hydrotreated
operations over a wide range.
Cracker
complex Extraction
Pygas Fractionation
To constitute an integrated pro- To cracker

cess digital twin for evaluating

Raffinate to cracker

unit operations in real-time, Raffinate to blending

Cracked Surplus for
the complex-wide process sim- gasoline C5s blending

ulation model can be put into Refinery

Gasoline Change after implementation
operation with real-time oper- Gasoline blend
sales
Increased Reduced
ations data from sites’ DCS, components
Gasoline No change Discontinued
historian (OSIsoft, PI or other), blending

and laboratory information sys-

tems (LIMS) to deliver real-time, Figure 4 Case study for refinery and cracker complex integration
high-fidelity virtual representa-
tions of hydrocarbon molecule transforma- Case study
tion and associated operating conditions. This KBC worked with a refinery which was designed
involves activating the in-built capabilities of the to produce light and middle distillates for the
process simulation model without middleware domestic market. A new aromatics complex was
being required. commissioned near to the refinery which was
To track unit performance, well-defined KPIs operated and managed by a different organisa-
are used for ongoing scrutiny by unit operations tion. Most of the feed for the aromatics unit was
engineers. For each KPI identified, actionable rec- imported naphtha.
ommendations and procedures for correcting the Although a part of the naphtha stream from
poor or inefficient performance are developed. the refinery was transferred to the aromatics
One of the key KPI measures is the comparison plant, both the refinery and the aromatics plant
of reconciled data, with results from process sim- maintained their own production accounting and
ulation and the LP model used to validate the reported financial performance using a naphtha
performance of these tools. transfer price. The refinery was using naphtha as
Petro-SIM’s data reconciliation utility assesses low flash blendstock for middle distillates with
the health of raw data available through the the surplus sold to the aromatics plant. As the
historian and LIMS. This includes raw mass transfer price of naphtha to the aromatics plant
balancing of sulphur and nitrogen for naphtha was lower, blending in middle distillate was
processing and petrochemical units, as well as profitable for the refinery.
carbon and hydrogen. KBC compared the quality of imported and
The digital twin can estimate parameters which refinery naphtha and found that the quality of
require complex calculations such as cyclone refinery naphtha was inferior to imported naph-
velocity in a RFCC, chemical hydrogen consump- tha in terms of carbon number. The paraffin
tion in each reactor, and remaining catalyst life content of refinery naphtha was higher, and it
in hydroprocessing units. Margin improvement was rich in C7 whereas imported naphtha had
indicators (MII) configured in the digital twin higher C8/C9 content. With this comparison it
indicate the loss of margin or gap that needs to was agreed that the refinery naphtha should be
be closed to improve the profitability of the com- priced at a discount to imported naphtha.
plex. The impact of MIIs on margin is estimated The discount for refinery vs imported naphtha
using complex- wide digital twin calculations was validated using the integrated process sim-
and is configured in the digital twin of a process ulation model. In a sensitivity case, a fraction of
unit. Monitoring of MIIs ensures that the process the imported naphtha was replaced with refin-
engineers and operators continue to optimise the ery naphtha. Both the refinery and the aromatics
unit. Deviations from target and loss associated plant were considered as one holistic optimisa-
with this can then be reported on a continuous tion for this sensitivity, meaning that a transfer
basis and acted on, thereby developing a culture price was not required to estimate the economic
of profitability in the organisation. impact of this change for the overall complex.

www.digitalrefining.com/article/1002475 PTQ Q3 2020 5

 accurate understanding of its
Import naphtha Aromatics Paraxylene future potential.
complex
Benzene KBC’s digital twin is built
Surplus
naphtha on our rigorous site-wide first
Refinery Naphtha for
Change after
implementation
principles simulator, Petro-
naphtha blending
Increased
SIM. The plant model inside
Refinery Middle
distillates Reduced
the simulator synchronises
Refinery middle
sales
No change with the plant model inside
distillates
Middle distillate Discontinued the enterprise process histo-
blending rian to ensure their separate
representations of the plant
Figure 5 Case study for refinery and aromatics integration remain aligned. Data gath-
ered from the plant in real-
This sensitivity case showed benefits of >$5 mil- time is validated and statistically reconciled to
lion/y without any investment. ensure that all physical and chemical laws are
The model indicated that processing the differ- respected; and electronic noise and dynamic
ent quality of refinery naphtha reduced p-xylene effects are eliminated through filtering, thereby
production and increased benzene yield. With the ensuring that data quality issues are identified
higher price of p-xylene compared to benzene, and mitigated.
there seemed to be an economic penalty. The mid- Petro-SIM enables identification of opportuni-
dle distillate production was also reduced. Even ties for site-wide improvements in performance,
with all these changes, the sensitivity case indi- as well as benefit sustainment through real-time
cated improvement in overall profitability. This monitoring of the entire complex. The poten-
was due to the significant premium paid by the tial for improvement of an integrated refinery-
aromatics plant for imported naphtha. petrochemical unit is significant and requires
The proposed opportunity was implemented tools that are able to analyse accurately for all of
by the refinery and the aromatics plant. The inte- the key optimisation factors. Additional bene-
grated process simulation model for the complex fits can be achieved through effective molecular
calculated the breakeven price of refinery naph- management for a refinery and petrochemical
tha. A discount to the breakeven price at which complex aided with a detailed carbon number
the aromatics plant would purchase naphtha breakdown from crude assays through to blend-
from the refinery was agreed between the two ing and petrochemical units.
organisations. With a revised transfer price, there Jitendra Chellani is Senior Consultant with KBC (A Yokogawa
was improved profitability of both organisations. Company) in Abu Dhabi. He has 15 years of experience in
With the integrated process model able to mon- delivering consulting solutions for refining and petrochemical
itor the streams for quality changes, an update industry. He is SME for identifying, evaluating, and implementing
to the operating strategies of both plants was opportunities for margin improvement.
able to be implemented. An overview of this case Sachin Srivastava is Petrochemical Consultant with KBC (A
study is shown in Figure 5. Yokogawa Company) in Mumbai. He has 11 years of experience
in process optimisation of olefins complexes with a focus on
refinery petrochemical integration. He holds a bachelor’s degree
Conclusion in chemical engineering from Indian Institute of Technology,
A digital twin aims to be an accurate representa- Kanpur, India.
tion of an asset over its full range of operation
and its full lifecycle. It is ideally created during
the initial study to evaluate the feasibility of the LINKS
asset. It is used and further developed during
the design, construction, and commissioning More articles from: KBC (A Yokogawa Company)
of the asset. It facilitates the optimum design of More articles from the following categories:
the asset and the training of the staff who will Energy Efficiency and Energy Management
operate the asset. It works in the present, mirror- Instrumentation, Automation and Process Control
ing the actual plant in simulated mode, but with Process Modelling and Simulation
full knowledge of its historical performance and

6 PTQ Q3 2020 www.digitalrefining.com/article/1002475