Processes 08 00032 PDF

processes
Article
Modeling and Simulation of Reaction and
Fractionation Systems for the Industrial Residue
Hydrotreating Process
Yalin Wang, Dandan Shang, Xiaofeng Yuan * , Yongfei Xue and Jiazhou Sun
School of Automation, Central South University, No. 932 of South Lushan Road, Changsha 410083, China;
ylwang@csu.edu.cn (Y.W.); sdandan@csu.edu.cn (D.S.); xueyongfei@csu.edu.cn (Y.X.);
174612233@csu.edu.cn (J.S.)
* Correspondence: yuanxf@csu.edu.cn

Received: 25 November 2019; Accepted: 25 December 2019; Published: 27 December 2019
Abstract: The residue hydrotreating process plays a significant role in the petroleum refining industry.
In this process, modeling and simulation have critical importance for process development, control,
and optimization. However, there is a lack of relevant reports of plant scale due to complexity
in characterizing feedstock and determining reaction mechanisms. In this paper, reaction and
fractionation models are constructed and simulated for a real-life industrial residue hydrotreating
process based on Aspen HYSYS/Refining. Considering the heavier and inferior residue, analytical
characterization is carried out for feedstock characterization based on laboratory analysis data.
Moreover, two reactor models with parallel structures are proposed to implement the intricate
reaction network, namely, a hydrocracker reactor and a plug flow reactor. The former simulates lighter
petroleum hydrotreating based on the built-in reaction network. The latter emulates the conversion
of a peculiar, heavier resin and asphaltene, using a six-lump model, which expands the scope of the
feedstock and improves the accuracy of the model. To obtain a realistic simulation of fractionation,
the database-based delumping method is adopted to model it with proper pseudo-components. The
simulation results, including temperature rise, hydrogen consumption, temperature distribution,
product yield, product properties, indicate that the model is capable of reflecting the realistic
process accurately.
Keywords: asphaltene conversion; database-based delumping; feedstock characterization; parallel

structure reactors; residue hydroconversion model
1. Introduction
The global energy crisis has become an urgent issue in the 21st century [1]. The worldwide growth
in energy consumption and carbon emissions is about 1.5% and 1.4% annually over the past 5 years [2].
However, as the main form of energy, the remaining oil is forecasted to be available for only about
50 years [2]. Furthermore, the oil refineries face various challenges: growing demand for high-quality
and low-boiling point products, deterioration of crude oil, ever-increasing difficulty of petroleum
processing, and stringent environmental requirements. Since vacuum residue accounts for about half
of the crude oil, more attention has been focused on vacuum residue hydroconversion processes, which
are capable of converting heavy, inferior petroleum into light, valuable products such as gasoline, jet
fuel, and diesel [3,4]. A good process model is important for monitoring the plant-level operation
status, optimizing operating conditions, verifying the dynamic adjustment scheme, and maximizing
the profit [5–7].
The residue hydrotreating process mainly consists of a reactor and a fractionation unit. The core
of the process model is the kinetic model, which describes the catalytic cracking reactions based on
Processes 2020, 8, 32; doi:10.3390/pr8010032 www.mdpi.com/journal/processes

Processes 2020, 8, 32 2 of 19
mechanism knowledge. For the kinetic model, the inputs are the feedstock properties and reaction
conditions. The model outputs are product yield, product properties, reactor temperature rise, and
hydrogen consumption at different operation conditions. In the past years, two main kinds of methods
have been studied to establish the kinetic model, which are based on molecular composition and
lumped composition, respectively.
Based on molecular composition for the kinetic model, structure-oriented lumping (SOL) proposed
by Quann et al. [8] is one of the most widely used techniques. In this technique, the feedstock mixture
was expressed as a matrix, whose rows indicated different atomic elements like C, H, S, N, O,
and columns represented different structural groups. This technique was then applied to model
the fluid catalytic cracking (FCC) of the naphtha hydrodesulfurization [9]. Additionally, a two-step
reconstruction algorithm was proposed by Hudebine et al. [10]. This algorithm could rebuild petroleum
at a molecular level from overall petroleum analyses. Moreover, Peng et al. [11] investigated another
approach called as the molecular type homologous series (MTHS), in which homologous series, along
with carbon number information were considered for kinetic modeling. This approach obtained
excellent predictions for a coal tar hydrogenation process [12]. However, these methods are with a
shortcoming of the computation expensiveness since extensive molecular compositions are generated
in the kinetic model. Additionally, another shortcoming is that routine chemical analysis and 13 C
nuclear magnetic resonance (NMR) are required. However, these analyses are mostly performed with
low frequency, which limits these molecular-level methods on applying in residue hydroconversion
process simulations.
The lumped method, proposed by Weekman et al. [13], was originally utilized for a three lumped
model of a catalytic cracking process. In this method, the mixture of hydrocarbons was characterized
by several lumped compositions classified by their reaction characteristics, which can alleviate the
difficulty of building the complicated kinetic model. Then, the kinetic model is further extended to
four [14], five [15], seven [16], and even thirty-seven lumped models [17]. However, those models only
reflect the productivity, whereas some detailed information, such as sulfur and nitrogen content, is not
included. Besides, once the lumps of kinetic model are determined, the cutting scopes of gasoline and
distillate oil are permanent. However, the cutting scopes generally vary from refinery to refinery with
different optimal economic benefits, which results in poor adaptability of the simulation models.
Hence, it is necessary to construct a reasonable model based on proper material compositions.
The compositions should contain theoretical, circumstantial information sufficient to reflect the process.
Meanwhile, proper number of compositions are required to easily make the calculation and give
good extrapolation results. Fortunately, another approach is utilized in the Aspen HYSYS/Refining
hydrocracker model to simulate the reaction process at the molecular level, in which 97 lumps are
selected to characterize petroleum. Also, these lumps take both the carbon number and molecular
structure information into consideration, which are in accordance with the results in the reported
literatures [18,19]. These lumps can be grouped into three classes of paraffin (P), naphthene (N), and
aromatics (A). They can be tested using mass spectrometry and adopted to describe light petroleum
(i.e., the PNA method). However, the heavy vacuum residue is usually characterized by the method
four components of saturate, aromatic, resin, and asphaltene (SARA). The PNA characterization method
is different from SARA, which may be improper for simulating the vacuum residue hydrogenation
process. In addition, high-boiling point resin and asphaltene are not represented by the lumps in the
hydrocracker [20,21], which results in the lower high-boiling points of simulated petroleum. Moreover,
the colloidal structure of resin, and asphaltene (RA) have a great impact on the overall kinetics [22].
To alleviate the aforementioned problems, two parallel structure reactors model is proposed in
this paper to simulate a real industrial residue hydrogenation process. The conversion process of the
heavier petroleum regarded as an asphaltene lump is modeled in a plug flow reactor (PFR), which
is in parallel with the original hydrocracker reactor (HCR) existing in Aspen HYSYS/Refining. In
the PFR, the asphaltene conversion reaction network is established and its lumps are necessary to be
characterized. To describe the asphaltene conversion, a six-lump reaction model is adopted. For the
Processes 2020, 8, x FOR PEER REVIEW 3 of 20
Processes 2020, 8, 32
PFR, the asphaltene conversion reaction network is established and its lumps are necessary 3toofbe 19
characterized. To describe the asphaltene conversion, a six-lump reaction model is adopted. For the
lump characterization,
lump characterization, thethe property
property of
of asphaltene
asphaltene is is calculated
calculated using
using the
the group
group contribution
contribution method,
method,
and the remaining lumps in the reaction model are represented utilizing the
and the remaining lumps in the reaction model are represented utilizing the substitute mixtures substitute mixtures of
of real
real components
components (SMRCs)(SMRCs)
methodmethod [23].the
[23]. And AndHCRthesimulates
HCR simulates lighter petroleum
lighter petroleum hydrogenation
hydrogenation based on
based on the built-in reaction network
the built-in reaction network and lumps. and lumps.
As the
As the outermost
outermost layer
layer of
of residue
residue hydrogenation
hydrogenation model,model, the process model
the process model seeks
seeks to
to integrate
integrate
the reactor
the reactor model
model with with the
the fractionation
fractionation model.
model. Due
Due to to the complexity of
the complexity residues, most
of residues, most previous
previous
works paid attention to only one aspect of the reactor model or the fractionator
works paid attention to only one aspect of the reactor model or the fractionator model for the model for the for the
for
plant wide residue hydrogenation process [24]. To make contribution to this
the plant wide residue hydrogenation process [24]. To make contribution to this aspect, a residue aspect, a residue
hydroconversion process
hydroconversion process model
model is
is comprehensively
comprehensively developed
developed by by exploiting
exploiting delumping,
delumping, which
which is is aa
committed step to concatenate the reactor with the fractionation
committed step to concatenate the reactor with the fractionation model. model.
The paper
The paper is is structured
structured asas follows:
follows: Section
Section 22 describes
describes the residue hydrogenation
the residue hydrogenation process
process in in
detail. Section 3 illustrates the kinetic model describing the hydrogenation process of
detail. Section 3 illustrates the kinetic model describing the hydrogenation process of built-in residuebuilt-in residue
oil simulated
oil simulated in an HCR
in an HCR andand asphaltene
asphaltene simulated
simulated in in aa PFR.
PFR. Section
Section 44 demonstrates
demonstrates the the framework
framework
for building
for building the
the whole
whole process
process model,
model, which
which includes
includes characterization
characterization of of the
the feedstock
feedstock mixture,
mixture,
establishment of a reactor model of two parallel structure reactors, and delumping of the of
establishment of a reactor model of two parallel structure reactors, and delumping the reactor
reactor model
effluent. Section 5 presents the simulation results to clarify the model effectiveness. Section 6 provides6
model effluent. Section 5 presents the simulation results to clarify the model effectiveness. Section
provides
the the conclusion.
conclusion.
2. Process
2. Process Description
Description
The residue
The residue hydrogenation
hydrogenationtechnology
technologyisisone oneofofthe the most
most effective
effective ways
ways to deeply
to deeply hydro
hydro treattreat
the
the heavy
heavy oil. oil.
TheThe process
process is implemented
is implemented in fixed
in fixed bedsbeds for removal
for removal of majority
of the the majority of impurities
of impurities fromfrom
oils,
oils, which
which can supply
can supply excellent
excellent raw materials
raw materials to thetosubsequent
the subsequent catalytic
catalytic crackingcracking
process.process.
In theIn the
fixed
fixed bed reactors, both hydrotreating and hydrocracking reactions take
bed reactors, both hydrotreating and hydrocracking reactions take place in appropriate temperature, place in appropriate
temperature,
pressure, pressure, ratio,
hydrogen–oil hydrogen–oil
and liquid ratio, and space
hourly liquidvelocity
hourly (LHSV)
space velocity
conditions.(LHSV) conditions.
Generally, four
Generally, four reactors are adopted and connected in series, and the main function
reactors are adopted and connected in series, and the main function of each reactor is to remove some of each reactor is
to remove some
specific impurities. specific impurities.
A schematic
A schematicdiagram
diagram of of
thethe residue
hydrogenation unit isunit is displayed
displayed in Figurein 1.Figure 1. materials
The raw The raw
materials are blended with vacuum heavy wax oil, coke gas oil, tar wax oil
are blended with vacuum heavy wax oil, coke gas oil, tar wax oil and catalytic circulating oil. Beforeand catalytic circulating
oil. Before
entering theentering
reactionthe reaction
system, system, the
the blended blended
oil is filteredoil in aisbuffer
filtered in aThen
tank. bufferthetank.
oil isThen thein
boosted oiltheis
boostedand
pump in the pump and
exchanged heatexchanged
with the heat with
reactor the reactor
effluent. effluent. the
Afterwards, Afterwards, the oil goes
oil goes through the through
heating
the heating
furnace before furnace before
entering entering
the first the Thus,
reactor. first reactor.
the firstThus,
reactor the first reactoris temperature
temperature mainly affected is mainly
by the
affected furnace.
heating by the heating furnace.
RG QHT QHT QHT H 2,makeup
Gas
compressor
Heating Condenser
furnace
off-gas
Vacuum heavy HT HT HT HT Naphtha

wax oil Pump
Heat
Coke gas oil exchanger Separator Diesel
Pump stripper
Tar wax oil
Diesel
Catalytic Separator Separation
circulating oil Buffer Desulfurization system
.
Separator
.
tank tower
.
Hydrofining
heavy oil
Separator
Steam
Pump
Figure 1. Process flow diagram of the residue hydrogenation unit. (HT: hydrotreater; RG: recycled
Figure 1. Process flow diagram of the residue hydrogenation unit. (HT: hydrotreater; RG: recycled
hydrogen gas).
hydrogen gas).
The temperatures of the rest ones are controlled by hydrogen quenching to attain the required
operation conditions. In the hydrotreater (HT), sulfur, nitrogen, and oxygen compounds are mostly
The
Processes temperatures
2020, 8, 32 of the rest ones are controlled by hydrogen quenching to attain the required 4 of 19
operation conditions. In the hydrotreater (HT), sulfur, nitrogen, and oxygen compounds are mostly
converted to hydrogen sulfide, ammonia, and water, while polycyclic aromatic hydrocarbons and
converted to hydrogen sulfide, ammonia, and water, while polycyclic aromatic hydrocarbons and
some unsaturated hydrocarbons are hydro saturated.
some unsaturated hydrocarbons are hydro saturated.
The outflow from the last HT is then transferred to a high and low pressure separation system
The outflow from the last HT is then transferred to a high and low pressure separation system
(HLPS) in which three-phase separation of the gas, oil and water takes place. The HLPS mainly
(HLPS) in which three-phase separation of the gas, oil and water takes place. The HLPS mainly consists
consists of four separators, as illustrated in Figure 1. The gas of the separators then enters a
of four separators, as illustrated in Figure 1. The gas of the separators then enters a desulfurization
desulfurization tower to desorb hydrogen sulfide. After this operation, the hydrogen purity of
tower to desorb hydrogen sulfide. After this operation, the hydrogen purity of recycled gas is improved.
recycled gas is improved. Meanwhile, the off-gas discharges a part of the H 2 . The hydrogen is then
Meanwhile, the off-gas discharges a part of the H2 . The hydrogen is then compressed and blended
compressed and blended with fresh hydrogen ( H 2, m akeup ).The blended gas recycles back to the four
with fresh hydrogen (H2,makeup ).The blended gas recycles back to the four reactors, in which a part
reactors,
of in which
the recycled gasa(RG)
part isofmixed
the recycled gas oil,
with feed (RG)and
is mixed with
the other feed
gas oil, and the
functioned other
as gas gas functioned
quenching (QHT )
quenches the beds( Q
as gas quenching ) quenches
toHTensure the and
security bedsprotect
to ensure
thesecurity
catalyst.and
Andprotect the catalyst.
the bottom streamAnd theHLPS
of the bottom is
stream oftothe
delivered theHLPS is delivered
downstream: to the downstream:
the fractionation the fractionation
process, which process, which
includes a fractionator, includes
a diesel stripper, a
afractionator,
condenser, anda diesel
a pump.stripper,
The topa condenser, and
product of the a pump. The
fractionator top product
is naphtha, of the
the middle onefractionator
is diesel, and is
naphtha,
the bottom the middle
one is tailone
oil. is diesel, and the bottom one is tail oil.
3. Kinetic Model
3. Kinetic Model of
of the
the Residue
Residue Hydrotreating
Hydrotreating Unit
Unit
In
In this
this paper,
paper, the
the reaction
reaction process
process isis modeled
modeled with
with aa two-parallel
two-parallel reactors
reactors model:
model: an
an HCR
HCR and
and aa
user-defined
user-defined PFR.
PFR. For
For the
the HCR,
HCR, the
the lighter
lighter petroleum
petroleum hydrogenation
hydrogenation process
process is
is simulated
simulated using
using the
the
built-in
built-in residue
residue oil
oil hydrogenation
hydrogenation reaction
reaction network
network with
with 97
97 lumps.
lumps. For
For the PFR, the
the PFR, the heavy
heavy asphaltene
asphaltene
hydrogenation
hydrogenationprocess
processis modeled
is modeledwithwith
the six-lump kinetickinetic
the six-lump model. model.
Figure 2 Figure
macroscopically illustrates
2 macroscopically
the reaction
illustrates mechanism
the utilized in utilized
reaction mechanism this paper.
in this paper.
H2S C1
HCR model Gas
NH3 C2
C4 C3
Built-in residue oil Naphtha

Paraffins
Naphthenes
Residue oil Aromatics
Sulfur Components
Asphaltene Nitrogen Diesel
Components
Asphaltene
PFR model Tail oil

Solid residue
2. The reaction mechanism of the two-reactors model for residue

Figure 2. residue reactions
reactions process.
process. (HCR:
reactor; PFR:
hydrocracker reactor; PFR: plug
plug flow
flow reactor).
reactor).
3.1. Kinetic Model

3.1. Kinetic Model of
of the
the Built-In
Built-In Residue
Residue Oil
Oil Hydrotreating
Process
In
In the
the HCR
HCR model,
model,the
thereaction
reactionnetwork
networkisisextremely
extremelycomplicated
complicated with
with177
177reactions. According
reactions. According to
reference [24],[24],
to reference the reactions can be
the reactions can grouped into six
be grouped types:
into (1) paraffin
six types: hydrocracking;
(1) paraffin (2) ring-opening;
hydrocracking; (2) ring-
(3) dealkylation of aromatics, naphthenes, nitrogen lumps, and sulfur lumps;
opening; (3) dealkylation of aromatics, naphthenes, nitrogen lumps, and sulfur lumps; (4)(4) saturation of saturation
aromatics,
non-basic nitrogen
of aromatics, lumps,
non-basic and hindered
nitrogen lumps, sulfur lumps; (5)
and hindered hydrodesulfurization
sulfur (HDS) of unhindered
lumps; (5) hydrodesulfurization (HDS)
sulfur lumps; and (6) hydrodenitrogenation (HDN) of nitrogen lumps. Each reaction
of unhindered sulfur lumps; and (6) hydrodenitrogenation (HDN) of nitrogen lumps. Each reaction rate equation
is based
rate on the
equation is classical
based ontheory of Langmuir–Hinshelwood–Hougen–Watson
the classical theory of Langmuir–Hinshelwood–Hougen–Watson (LHHW) [25], which
(LHHW)
demonstrates that the heterogeneous
[25], which demonstrates that the catalytic reactions
heterogeneous undergoreactions
catalytic three basic stages: three
undergo adsorption,
basic surface
stages:
reaction, and desorption. Figure 3 illustrates the part of the reaction network employed
adsorption, surface reaction, and desorption. Figure 3 illustrates the part of the reaction network in the HCR
model of Aspen HYSYS for the reaction of nitrogen lumps.
employed in the HCR model of Aspen HYSYS for the reaction of nitrogen lumps.
Processes 2020, 8, 32 5 of 19
Figure 3. TheThe
Figure reaction network
reaction of nitrogen
network lumpslumps
of nitrogen adopted in theinHCR
adopted the model
HCR [24].
model(HDN:
[24].
hydrodenitrogenation).
(HDN: Reproduced
hydrodenitrogenation). with permission
Reproduced from copyright
with permission American
from copyright Chemical
American Society,
Chemical
[2011]. [2011].
Society,
According to the LHHW theory,

theory, reversible
reversible and irreversible
irreversible reactions can be expressed by
Equations
Equations (1)
(1) and
and (2).
(2).
m
((ADS
ADS iiCCi iADS /eq−−ADS
ADSHH2 ((PPH 2H) )m/ eq ADSj Cj C
j )j )
Rr R=r =RRwhole
whole w (1)
2 2
w ADS (1)
ADS
ADSHH2((PP m m
ADS CiiADS
ADS iiC H )
H 2 )2
Rir R=ir =RRwhole
whole w
2 (2)
(2)
ADS
ADS
where Rr is the reversible reaction rate and Rir is the irreversible reaction rate; Rwhole is the whole
where Rr is the reversible reaction rate and Rir is the irreversible reaction rate; Rwhole is the whole
reaction rate; w represents the inherent rate constant ascertained by basic study; ADSi and ADS j
reactionthe
indicate rate; w represents
adsorption the inherent
coefficients of lumpsrate constant
i and ascertained
j, respectively; lumpbyi and
basic study; H
hydrogen ADSareand ADS j
reactants;
2 i
indicate
lump j isthe
theadsorption
resultant; C coefficients of the
i and C j are lumps i and j , of
concentration respectively;
lumps i and lump i and hydrogen
j, respectively; H2
PH2 is the
partial pressure
are reactants; lump of j is the resultant; Ci and j are the concentration of lumps i and j ,
hydrogen; m is the reaction order of
C hydrogen; eq is the equilibrium constant of the
reaction; and ADS indicates the whole LHHW adsorption coefficient, which expresses the competitive
respectively; PH is the partial pressure of hydrogen; m is the reaction order of hydrogen; eq is
2
adsorption with diverse restrainers.
the equilibrium constant of the reaction; and A D S indicates the whole LHHW adsorption coefficient,
which expresses
3.2. Kinetic Modelthe competitive
of the Asphalteneadsorption
HydrogenationwithProcess
diverse restrainers.
The hydrogenation
3.2. Kinetic of resin and
Model of the Asphaltene asphaltene with
Hydrogenation the high boiling point, and density also involves
Process
involute and heterogeneous catalytic reactions. Hence, by virtue of the complexity of high boiling
pointThe hydrogenation
petroleum of resin
and its main and asphaltene
products, a simplewith
lumpthestrategy
high boiling point, and
is employed density
[26]. also involves
For simplification,
involute
resin and and heterogeneous
asphaltene are togethercatalytic reactions.
regarded as the Hence, by virtue
asphaltene lump. of In the complexity of high
the hydrogenation, threeboiling
phase
point petroleum and its main products, a simple lump strategy is employed
products are decomposed from asphaltene. The liquid mixtures are grouped into three lumps: the [26]. For simplification,
resinoil
light and asphaltene
(initial boilingare together
point (IBP) regarded
< 350 ◦ C) as the asphaltene
lump, the middle lump. In the hydrogenation,
oil (350−540 ◦ C) lump, and the three phase
heavy oil
products◦ are decomposed from asphaltene. The liquid mixtures are grouped
(>540 C) lump. Gaseous gas and solid coke were taken as the gas lump and coke lump, respectively. into three lumps: the
light oil
Also, it is(initial
notableboiling
that thepoint (IBP)
active < 350 °C)reaction
hydrogen lump, the middle
is not oil (350−540
considered in this°C) lump,
paper andthe
since thetetralin
heavy
oil (>540 °C) lump. Gaseous gas and solid coke were taken as the gas lump and coke lump,
respectively. Also, it is notable that the active hydrogen reaction is not considered in this paper since
Processes 2020, 8, 32 6 of 19
(its
the donor),
tetralin as
(itsshown inas
donor), [26], doesin
shown not exist
[26], in the
does notreal
existrefinery production
in the real refineryprocess. Figure
production 4 illustrates
process. Figure
the reaction network of the asphaltene conversion process. The components substituting
4 illustrates the reaction network of the asphaltene conversion process. The components substitutingfor lumps of
the network
for lumps ofare
theplaced
networkin are
the placed
right and described
in the in detail
right and in Section
described 4.1.in Section 4.1.
in detail
Figure 4.
Figure 4. Kinetic scheme for the six-lump model of asphaltene hydrogenation.
hydrogenation.
Since
Since the
the constants
constants ofof LHHW
LHHW in in Equations
Equations (1)
(1) and
and (2)
(2) ascertained
ascertained by by basic
basic research
research are
are scarcely
scarcely
found, general kinetic reactions are applied to describe the reaction network of petroleum
found, general kinetic reactions are applied to describe the reaction network of petroleum based based on the
on
Arrhenius equation.
the Arrhenius And researchers
equation. have different
And researchers opinionsopinions
have different about theabout
orderthe
of the reactions
order of the[27,28]. In
reactions
this paper, it is presumed that the whole reaction pathway consists of first-order reactions.
[27,28]. In this paper, it is presumed that the whole reaction pathway consists of first-order reactions. In addition,
due to the low
In addition, duegastoyield of the
the low gasentire
yieldreaction system,
of the entire the volume
reaction system,ofthe
thevolume
reactionof system is assumed
the reaction to
system
be immutable.
is assumed to be immutable.
The rate RRcan
The reaction rate
reaction also
can bebe
also expressed
expressedin in
thethe
form of of
form thethe
power
power law model:
law model:
R= − kCim,m,i
i m, j
C m,j (3)
R = −kCi jC j (3)
where k denotes the reaction rate constant; Ci and C j are the molar concentrations of lump i
where k denotes the reaction rate constant; Ci and C j are the molar concentrations of lump i and j,
and j , respectively; m, i and m, j denote the mth reaction orders of lump i and j , respectively.
respectively; m, i and m, j denote the mth reaction orders of lump i and j, respectively. In this paper, all
In this paper, all constants for m, i are set to 1.
constants for m, i are set to 1.
In fact,
In fact,the
thedata
dataobtained
obtainedfrom
fromthethe factory
factory is based
is based on mass
on mass flowflow
rate.rate. It is difficult
It is difficult to acquire
to acquire mole
mole fractions of the lumps. Instead, the mass concentration is adopted to calculate
fractions of the lumps. Instead, the mass concentration is adopted to calculate the kinetic parameters. the kinetic
parameters. Accordingly,
Accordingly, the rate
the rate equation can equation can as:
be rewritten be rewritten as:
Rl = -k l yim,i y m,
j
j
(4)
m,i m,j
Rl = −kl yi y j (4)
where yi and yj indicate the mass concentrations of lump i and j , respectively; kl is the rate
where
constantyi and y j indicate
of the the mass
lth reaction for lump i and j ,of
concentrations lumpbased
Thus, i and on
j, respectively;
the reactionknetwork
l is the rate
of constant of
asphaltene
the lth reaction
conversion, for lump i and j, Thus, based on the reaction network of asphaltene
the mathematical equations of the kinetic model can be rewritten in the following forms: conversion, the
mathematical equations of the kinetic model can be rewritten in the following forms:
dy1
= −(k1 y1 + k2 y1 + k3 y1 + k4 y1 ) (5)
dy1 dt
= −(k1 y1 + k2 y1 + k3 y1 + k4 y1 ) (5)
dt
dy2
dy2 = k1 y1 − (k5 y2 + k6 y2 + k7 y2 ) (6)
dt k1 y1 − (k5 y2 + k6 y2 + k7 y2 )
= (6)
dt
dy3 dy3
= k=2ky21y1++kk55 y22 -−
k8 ky38 y3 (7)
(7)
dt dt
dy4
= k6 y2 + k8 y3 (8)
dtdy4 = k y + k y (8)
6 2 8 3
dt
dy5
Processes 2020, 8, 32 = k3 y1 + k7 y2 7 of
(9)19
dt
dy5 dy6
= k3=y1k4+y1k7 y2 (9)
(10)
dt dt
dy6
= k4 y1 (10)
dt
4. Workflow of Developing an Integrated Residue Hydrogenation Process Model
4. Workflow of Developing
The workflow an Integrated
for developing Residue
an integrated Hydrogenation
residue Process
hydrogenation Model
model is shown in Figure 5
withThe
the workflow
Aspen HYSYS/Refining
for developing an integrated residue hydrogenation model is the
simulation software. A similar workflow for hydrocracking
shown in Figure 5
process was proposed in [24]. However, it is different for the heavy petroleum refining
with the Aspen HYSYS/Refining simulation software. A similar workflow for the hydrocrackingprocess.
process was proposed in [24]. However, it is different for the heavy petroleum refining process.
Figure
Figure 5.
5. Workflow of building
Workflow of building an
an integrated
integrated residue
hydrogenation process
process model.
model.
First
First ofofall,
all,residue
residue tends
tendsto become
to become inferior and heavier
inferior since the
and heavier crude
since theoilcrude
gets heavier.
oil getsHowever,
heavier.
the feed library
However, the feed for residue
library foris predetermined in the simulation
residue is predetermined in the software.
simulationFurthermore, the asphaltene
software. Furthermore, the
with higherwith
asphaltene boiling
higherpoint is notpoint
boiling considered in the feedinlibrary.
is not considered the feedInlibrary.
addition, owing toowing
In addition, the absence
to the
of asphaltene,
absence its product
of asphaltene, lumps are
its product lacking
lumps in the feed
are lacking library.
in the Thus, itThus,
feed library. is required to make
it is required to some
make
adjustments
some adjustments for thefor feed
theinfeed
the HCR
in thelibrary and characterize
HCR library the added
and characterize six lumps.
the added Moreover,
six lumps. due to
Moreover,
the
duedifferent feed properties
to the different and reaction
feed properties andconditions, it is necessary
reaction conditions, it is to determine
necessary to the kinetic parameters
determine the kinetic
of the reactor
parameters of model.
the reactor In the
model.HCR Inmodel,
the HCR themodel,
reaction theactivity
reactionvariables are to beare
activity variables estimated for the
to be estimated
built-in residueresidue
for the built-in oil hydrogenation.
oil hydrogenation. Correspondingly,
Correspondingly, in theinPFR
the model,
PFR model,the kinetic
the kineticparameters
parameters are
determined for the additional asphaltene hydrogenation. Finally, the reactor
are determined for the additional asphaltene hydrogenation. Finally, the reactor model is established model is established with
kinetic lumps
with kinetic grouped
lumps by reaction
grouped characteristics,
by reaction characteristics, whereas the fractionation
whereas the fractionation modelmodelis built based
is built on
based
the pseudo-components
on the pseudo-components according
according to to
fractionation
fractionation characteristic
characteristic(i.e.,
(i.e.,the
theboiling
boilingpoint).
point). Therefore,
the adopted kinetic lumps of reactor model may be be inappropriate
inappropriate for aa fractionator
fractionator modeling
modeling of
residue
hydrogenationprocess. process. Consequently,
Consequently, it isitprerequisite to transform
is prerequisite kinetic kinetic
to transform lumps into lumpssuitable
into
pseudo-components through a process called delumping. The
suitable pseudo-components through a process called delumping. The detailed workflow of the detailed workflow of the residue
hydrogenation
residue hydrogenationmodel can be described
model as follows:
can be described as follows:
(1) Obtain data from factories and select the required data for modeling. In this work, the set data
(e.g., the feed property, flow) is the input of the model and the target data (e.g., the product yield,
property) are the object value to be attained.
product yield, property) are the object value to be attained.
(2) Characterize the feedstock based on laboratory testing data.
(3) Develop the reactor model according to the real process data.
(4) Delump the effluent from the reactor to build the fractionator.
(5) Test the effectiveness of the model by comparing the model results with actual plant data.
Processes 2020, 8, 32 8 of 19
4.1. Feedstock Mixture Characterization

(2) Feed
Characterize
quality the
hasfeedstock based on
a significant laboratory
influence on testing
residuedata.
hydrogenation process. Thus, it is
(3) Developto
fundamental the reactor model
characterize the according
feedstock to the real
mixture process data.
according to its bulk properties. In this work, the
(4) Delump
“residue” the effluent
fingerprint typefrom the reactor
is selected in thetobasic
buildfeed
the fractionator.
library of HCR for the residue hydrogenation
(5) TestAnd
process. the effectiveness
the feed dataofsection
the model by comparing
inputs the model results
are the experimental with actual
bulk properties of plant
feed, data.
such as its
density, boiling point, refractive index, sulfur content, and nitrogen content. The corresponding
4.1. Feedstock
outputs Mixture
are the Characterization
simulated bulk properties and composition contents. Generally, the experimental
bulk Feed
properties diverge from the simulated.
quality has a significant influence on Figure
residue6a hydrogenation
show the residue properties
process. Thus,ofit aisdataset in the
fundamental
HCR.
to It can be the
characterize found that themixture
feedstock simulated initial boiling
according pointproperties.
to its bulk (IBP) and final boiling
In this work,point (FBP) are
the “residue”
lower than their experimental values. Meanwhile, the simulated total light oil percentage
fingerprint type is selected in the basic feed library of HCR for the residue hydrogenation process. is slightly
higher in Table 1, which may have impact on the simulated reactor performance. It
And the feed data section inputs are the experimental bulk properties of feed, such as its density, is because that
the conversion
boiling of actual index,
point, refractive weightsulfur
content, and nitrogenis only 15–20%,
content. Thewhich is definedoutputs
corresponding as the sum
are
of the weight proportions of gas and light oil after reaction. Therefore, two problems
the simulated bulk properties and composition contents. Generally, the experimental bulk properties demand to be
solved. from the simulated. Figure 6a show the residue properties of a dataset in the HCR. It can
diverge
be found that the simulated initial boiling point (IBP) and final boiling point (FBP) are lower than
their Table 1. Simulated
experimental residue
values. feed composition
Meanwhile, contents with
the simulated totalbasic
lightand
oilimproved
percentage feedislibrary in HCR.
slightly higher in
Table 1, which may have impact on theNaphthasimulatedDistillate
reactor performance. It is because
Gas Oil Total Light Oil that the conversion
of actual weight residue
Basichydrogenation
library 0%is only 15–20%,
1.72% which
12.04%is defined as the sum of the weight
13.76%
proportions of gas Improved
and light library 0%
oil after reaction. 2.79% two
Therefore, 8.72% 11.01% to be solved.
problems demand
Figure
Figure 6. Comparisonsofofthe
6. Comparisons theboiling
boilingpoints
points forfor
thethe residue
residue feed
feed in HCR
in HCR with
with different
different values:
values: (a)
(a) experimental
experimental value
value and
and simulated
simulated valuewith
value withbasic
basiclibrary;
library;(b)
(b)experimental
experimentalvalue,
value, simulated
simulated values
values
with
with basic
basic and
and improved
improved library
library (LV%:
(LV%: the
the liquid
liquid volume
volume fraction).
fraction).
Table 1. Simulated residue feed composition contents with basic and improved feed library in HCR.
The first problem is that the lumped compositions whose normal boiling point is lower the real
IBP value should be removed. Naphtha Distillate
Therefore, their proportion shouldGas Oilto zero, which
be set Total Light Oil on
is based
the separation
Basic library efficiency factor
0% proposed by 1.72%García et al. [29]. 12.04%
In this way, the proportions
13.76% of
Improvedlumps
remaining libraryinevitably increase.
0% To handle 2.79% 8.72% to adjust the11.01%
this problem, it is necessary proportions,
which has been presented in the previous work [30]. Figure 6b and Table 1 show the results of the
experimental value, simulated
The first problem value
is that the withcompositions
lumped basic and improved library boiling
whose normal of the residue
point isfeed.
lowerSince the
the real
IBP value should be removed. Therefore, their proportion should be set to zero, which is based on the
separation efficiency factor proposed by García et al. [29]. In this way, the proportions of remaining
lumps inevitably increase. To handle this problem, it is necessary to adjust the proportions, which has
been presented in the previous work [30]. Figure 6b and Table 1 show the results of the experimental
value, simulated value with basic and improved library of the residue feed. Since the distillation range
of naphtha is lower than that of residue, the contribution of naphtha is 0% in Table 1.
The other problem is that the simulated FBP of the feed is lower than its experimental value.
Therefore, it is indispensable to add the composition representing the higher boiling point petroleum
Processes 2020, 8, 32 9 of 19
lump, and other lumps in the reaction network, which is displayed in Figure 4. To represent the added
petroleum, the composition of Tahe residue heavy resin is adopted from [31], since its properties are
closer to the real petroleum cut. Its properties of molecular weight and density are also given in
reference [31]. Although the software has a powerful ability to calculate physical property, at least
three properties are required to characterize the lumps more precisely. Fortunately, the normal boiling
point of petroleum cut can be estimated using the group contribution method [32]. In this method, the
composition (considered as a molecule here) is composed of 77 structural groups. Each group has
its own contribution and can be referred to in reference [33]. The molecule property is calculated by
summing the products over each group numbers and its corresponding contribution. Meanwhile,
the method of first-order group contributions is applied to predict the normal boiling point, which is
shown in the following equation [34]:
Tb X
exp( )= (ni ∆Tb,i ) + CIT,b (11)
Tb,0
i
where ∆Tb,i represents the group contribution of group i; ni represents the number of group i in the
molecule; Tb,0 represents a regulatable parameter (Tb,0 = 307.63K), and CIT,b represents a boiling point
corrector, which can be computed as follows:
CIT,b = 0.2285NR2 + 0.4687NR (12)
where NR represents the number of rings in a molecule.

According to Equations (11) and (12), the normal boiling point of asphaltene is estimated, and
other properties are calculated by the software. The results are shown in Table 2.
Table 2. Lumps and their properties utilized in plug flow reactor (PFR).
Normal Critical Critical Critical

Density MW
Lump Boiling Temperatures Pressures Volumes
(kg·m−3 ) (g·mol−1 )
Point (◦ C) (◦ C) (kpa) (m3 ·kg−1 ·mol−1 )
ASP 889 1052 1057 989.6 378.7 2.78
C47 563 825 661 643 375 3.1217
n-C30 449.719 811.898 422.780 589.85 868 1.72353
n-C14 253.508 762.913 198.380 420.85 1620.18 0.82999
C2 −88.600 355.683 30.070 32.27801 4883.85 0.148
C / 1642.06 / / / /
ASP, asphaltene; C47: C47paraffins; n-C30: n-Triacontane; n-C14: 1,2-Tetradecanediol; C2: Ethane; C, Coke.
Then, the remaining five lumps need to be characterized in the reaction network. Here, the
method of substitute mixtures of real components (SMRCs) is used for lump characterization. This
method enhances comprehension of the mechanism of the reaction since the structure of a substituted
component is included. In this method, each lump is substituted by one component whose normal
boiling point approximates the average distillation range of the lump. Since multiple components
may be available for a specified boiling point, it is important to select the proper components. The
main principles are as follows: (1) The components exist in the system and can be detected by analysis;
(2) The low-boiling-point components, the low-carbon components and their isomers can be tested
by analytical techniques. (3) For the high-carbon components, it is preferable to choose paraffin and
aromatics. All the selected lumps and their properties are shown in Table 2.
4.2. Reactor Model Construction with Two Parallel Structures

After feedstock mixture characterization, the next step is to develop reactor model. First, the
process data (e.g., catalyst loading, feed rate, feedstock analysis, reactor inlet temperature, and reactor
pressure) is adopted to synchronously identify the parameters of the rate equations and reactor design
pressure)
Processes is8,adopted to synchronously identify the parameters of the rate equations and reactor
pressure)2020,
is 32
adopted to synchronously identify the parameters of the rate equations and reactor 10 of 19
design equations. Then, the deviation of model prediction from plant data is minimized by adjusting
design equations. Then, the deviation of model prediction from plant data is minimized by adjusting
the reaction activity variable in the HCR model and the user-defined PFR. For the reactor model,
the reaction activity
equations. variable inofthe HCRprediction
model and the user-defined PFR. For thebyreactor model,
modifying Then, the deviation
the activity model
factor is essential, since the from
feed plant data reactor
property, is minimized adjusting
configuration, the
catalyst
modifying
reaction the variable
activity activity infactor
the is essential,
HCR model since
and the the feed property,
user-defined PFR. Forreactor
the configuration,
reactor model, catalyst
modifying
activity, and operating conditions differ greatly in various refineries. The procedure of adjusting the
activity,
the activityandfactor
operating conditions
is essential, differ
since the greatly
feed in various
property, refineries.
reactor The procedure
configuration, of adjustingand
catalyst the
activity factor to lessen the disparity is called “calibration”. For the user-defined PFR,activity,
the kinetic
activity
operating factor to
conditions lessen the disparity
differ greatly is
in variouscalled “calibration”. For the user-defined PFR, the kinetic
parameters are identified similarly. The tworefineries. The procedure
parallel structure of adjusting
reactor model is the activity factor
demonstrated in
parameters
to lessen the are identified
disparity is similarly.
called The two For
“calibration”. parallel
the structure reactor
user-defined PFR, model
the is demonstrated
kinetic parameters arein
Figure 7.
Figure 7.
identified similarly. The two parallel structure reactor model is demonstrated in Figure 7.
Figure 7. Two
Two parallel structure reactor model.
Figure 7. Two parallel structure reactor model.
For
For the
the HCR
HCR model,
model, the
the calibration
calibration procedure
procedure is is already
already described
described in in [24]. Here, there
[24]. Here, there are
are some
some
For the HCR model, the calibration procedure is already described in [24]. Here, there are some
additional steps in the calibration procedure for the residue hydrogenation process.
additional steps in the calibration procedure for the residue hydrogenation process. In the tuning In the tuning step,
additional steps in the calibration procedure for the residue hydrogenation process. In the tuning
it is discovered
step, that the
it is discovered reaction
that intensity
the reaction of the catalyst
intensity bed competes
of the catalyst againstagainst
bed competes each othereachasother
shown as
step, it is discovered that the reaction intensity of the catalyst bed competes against each other as
in Figure 8. The reaction intensity is reflected by the temperature rise since the reactions
shown in Figure 8. The reaction intensity is reflected by the temperature rise since the reactions are are overall
shown in Figure 8. The reaction intensity is reflected by the temperature rise since the reactions are
exothermal. That is, That
overall exothermal. whenis,other
when parameters are unaltered,
other parameters the global
are unaltered, theactivity factor offactor
global activity the first catalyst
of the first
overall exothermal. That is, when other parameters are unaltered, the global activity factor of the first
bed will increase,
catalyst bed will which might
increase, result
which in a decline
might result in of temperature
a decline of rise for the remaining
temperature rise for catalyst bed, as
the remaining
catalyst bed will increase, which might result in a decline of temperature rise for the remaining
demonstrated
catalyst bed, as in Figure 8. Furthermore,
demonstrated in Figure it is
8. recommended
Furthermore, thatit is the simulated temperature
recommended rises are
that the simulated
catalyst bed, as demonstrated in Figure 8. Furthermore, it is recommended that the simulated
overall lower than the real temperature rise. It may be because the exothermal hydrodemetallization
temperature rises are overall lower than the real temperature rise. It may be because the exothermal
temperature rises are overall lower than the real temperature rise. It may be because the exothermal
reactions are excludedreactions
hydrodemetallization in the HCR aremodel,
excluded which occur
in the HCR inmodel,
practical production.
which occur in practical production.
hydrodemetallization reactions are excluded in the HCR model, which occur in practical production.
8. Four
Fourreactor
reactortemperature
temperaturerises
riseswith
withglobal
globalactivity
activityof of
different reactors:
different (a) reactor
reactors: 1; (b)
Figure
Figure 8. Four reactor temperature rises with global activity of different reactors: (a) (a) reactor
reactor 1;
1; (b)
reactor
(b) 3. 3.
reactor
reactor 3.
Since a PFR substitutes for four reactors of the real process, the arithmetic averages of operating
parameters are adopted, such as temperature, pressure, volume, and so on. For the identification of
kinetic parameters, reference [26] adopted genetic algorithm method for minimizing the deviation
of the prediction of seven-lump kinetic model from the experimental data. In the established kinetic
Processes 2020, 8, 32 11 of 19
model, the activation energies were from 106.07 to 237.5 kJ. mol −1 . However, due to the deviations
of the feed, operation condition between the reference and the actual process, adjustments of kinetic
parameters are required to match the real process. In addition, it is impossible to directly determine the
kinetic parameters for the conversion of asphaltene because the petroleum is part of the inlet residue
oil in a real industrial process. To handle this problem, the consistent rule and the whole asphaltene
conversion are employed. And considering the existence of a catalyst in the real process, only the
activation energy is changed. The consistent rule is that the adjusted distribution of reaction activation
energy should be similar to that published in [26]. The kinetic parameters applied in this work are
shown in Table 3. It can be observed that the adjusted value is almost 0.66–0.76 times its original value.
Table 3. Apparent activation energies with their frequency factors of the six-lump model in [26] and
this paper. Reproduced with permission from copyright American Chemical Society, [2017].
From [26] This Paper

Reaction Frequency Factor (h−1 ) Activation Energy (kJ·mol−1 ) Activation Energy (kJ·mol−1 )
Asp → HO A1 6.70 × 108 E1 106.07 E1 82.2
Asp → MO A2 7.78 × 108 E2 109.06 E2 83.05
Asp → G A3 8.13 × 109 E3 175.56 E3 126
Asp → C A4 5.06 × 109 E4 169.17 E4 125
HO → MO A5 9.92 × 108 E5 130.78 E5 86.5
HO → MO A6 9.92 × 109 E6 155.43 E6 112
HO → G A7 9.91 × 108 E7 237.50 E7 156.75
MO → LO A8 2.24 × 1011 E8 153.63 E8 110
Asp, asphaltene; HO, C47; MO, N-C30; LO, N-C14; G, C2; C, coke
4.3. Delumping of the Reactor Model Effluent and Fractionator Model Development
Delumping the outflow of the reactor model works as a fundamental step combining the reactor
model to the fractionator model. Firstly, lumps in the reactor model are more concerned with the
carbon number and structure, whereas the lumps in the fractionator model pay more attention to the
accuracy of thermodynamic behavior. Secondly, the lumps used in the kinetic model belong to two
different reactor models and are not integrated in this paper.
Numerous researchers have contributed to the development of this aspect. Haynes et al. [35]
applied the Gauss–Legendre quadrature to calculate the vapor–liquid equilibrium (VLE) of the
petroleum whose properties are expressed by critical properties and acentric factors. Yang et al. [36]
determined appropriate fractionation lumps and their physical properties by interpolation, cutting,
and normalization of the distillation curve of reactor effluent for the residue hydrogenation process.
Chang et al. [24] utilized the Gauss–Legendre quadrature to delump the overflow of the reactor model
into 20 pseudo-components for the medium-pressure hydrocracking (MP HPR) and high-pressure
hydrocracking (HP HPR) units.
However, a common disadvantage in these methods is that some of the related correlations are
controversial. Additionally, the residue feed covers a wider range of distillation and contains more
complex compositions, making the delumping tougher. Moreover, for the high and heavy fractions,
property correlation may generate questionable results because the formulas are usually summarized
by low boiling-point fractions. In this paper, a data-based method is applied to delump the reactor
model outflow and develop the fractionator model. The component list of the method is named
“Assay Components Celsius to 850 C” in ASPEN/HYSYS software. The software has collected and
checked the data from the world’s most respected sources [37]. Thus, its data source has high credibility.
Due to page limitations, Table 4 shows only part of the pseudo-components from 340–850+C * and
its property of true boiling point (TBP), ideal liquid density(ILD), molecular weight(MW). More detailed
information about the complete components from 36–850+C * can be found in the Supplementary
Materials Table S1. For the distillation range from 36 to 460 ◦ C, each temperature interval with 10 ◦ C is
Processes 2020, 8, 32 12 of 19
represented as a lump; for the distillation boiling point range from 460 to 600 ◦ C, approximately an
interval of 20 ◦ C was regarded as a lump; for the range from 600 to 850 ◦ C, each interval with 25 ◦ C is
regarded as a lump; the range over 850 ◦ C is considered as a lump. The stream cutter is adopted to
connect the two models belonging to different physical property packages. The purpose of the stream
cutter is to recalculate stream enthalpy. In the calculation, component A (TBP) in physical property
pack 1 is mapped to pseudo-component A* in physical property pack 2 based on the boiling point.
Table 4. Part of boiling point-based pseudo-components and their properties.
Pseudo- ILD MW Pseudo- ILD MW

TBP (◦ C) TBP (◦ C)
Components (kg·m−3 ) (g·mol−1 ) Components (kg·m−3 ) (g·mol−1 )
340–350 C * 350 877 299 520–540 C * 540 958 538
350–360 C * 360 882 309 540–560 C * 560 967 565
360–370 C * 370 887 320 560–580 C * 580 975 593
370–380 C * 380 892 330 580–600 C * 600 984 622
380–390 C * 390 894 341 600–625 C * 625 990 668
390–400 C * 400 898 353 625–650 C * 650 997 718
400–410 C * 410 903 367 650–675 C * 675 1006 770
410–420 C * 420 908 381 675–700 C * 700 1015 828
420–430 C * 430 912 394 700–725 C * 725 1019 911
430–440 C * 440 917 408 725–750 C * 750 1023 1002
440–450 C * 450 921 421 750–775 C * 775 1030 1093
450–460 C * 460 926 434 775–800 C * 800 1036 1190
460–480 C * 480 932 461 800–825 C * 825 1043 1292
480–500 C * 500 941 487 825–850 C * 850 1049 1398
500–520 C * 520 950 512 850+C * 770 1062 998
5. Simulation Results
In this section, five operation points are used to demonstrate the efficacy of the established model.
Three parts are displayed for the simulation: comparison of a single HCR and the proposed two-reactor
model, comparison of the default HCRSRK method in Aspen HYSYS/Refining and the proposed
delumping method, and the prediction of the process model for the whole residue hydrotreating
conversion. The HCRSRK is a built-in fluid package in the HCR model (with a special HCR.cml
component list).
5.1. Comparison of a Single HCR Model with the Proposed Two Reactor Model of HCR and PFR
Five datasets are utilized to verify the effectiveness of the model with two parallel structure
reactors. It is easily seen from Figure 9 that compared with a single HCR, the proposed model can
provide more precise prediction at high boiling points for tail oil and slightly better results for the
lighter product. Table 5 presents the prediction average relative deviation of boiling points for tail oil
with the HCR and the proposed model in detail. The proposed reactor model extends the simulated
distillation range of the products through the added simulation of asphaltene conversion process.
Hence, the distillation curve of the tail oil can be predicted better by the proposed reactor model.
However, the predicted initial boiling point of naphtha is lower than the real value in Figure 9. In
the simulated naphtha, the additional gas that does not exist in the actual stream contributes to the
strange result.
Table 5. Comparison of the prediction average relative deviations of boiling point for tail oil with the
HCR model and HCR + PFR model.
Dataset 1 Dataset 2 Dataset 3 Dataset 4 Dataset 5

HCR 12.97% 14.20% 13.50% 14.85% 14.54%
HCR + PFR 5.92% 6.22% 6.06% 7.00% 6.05%
Processes 2020, 8, 32 13 of 19
Figure 9.
Figure 9. Predictions of the ASTM D86 distillation curves of liquid products with the HCR model and
HCR ++ PFR
PFR model: (a)
(a) Dataset
Dataset 1;
1; (b)
(b) Dataset
Dataset 3.
3.
5.2. Comparison of the HCRSRK

Table 5. Comparison Method and
of the prediction the Proposed
average relative Delumping
deviations ofMethod
boiling point for tail oil with the
HCR model tests
Sensitivity and HCRand +predictions
PFR model. of properties of the products are performed to demonstrate the
rationality and accuracy of results with1theDataset
Dataset HCRSRK 2 and the 3proposed
Dataset Datasetdelumping
4 Datasetmethod.5 Sensitivity
tests are essential to HCR ensure that12.97%
delumped14.20% lumps are13.50%capable of14.85% producing rational
14.54% results [38]. A
rational result is that HCRthe relationships
+ PFR 5.92% of the side draw
6.22% rate change
6.06% to the corresponding
7.00% 6.05% distillation
curve and temperature are consecutive rather than stepwise. Figures 10 and 11 show the sensitivity
5.2. results
test Comparisonof theofside
the HCRSRK Method and
products naphtha andthe Proposed
diesel for theDelumping
HCRSRKMethod method (a default method in HCR
model) and the data-based
Sensitivity delumpingofmethod.
tests and predictions properties Apparently, the HCRSRK
of the products method to
are performed generates stepwise
demonstrate the
relationships at most distillation points with the change of draw
rationality and accuracy of results with the HCRSRK and the proposed delumping method. rates, especially in naphtha, as seen in
Figure 10a. tests
Sensitivity The data-based
are essentialdelumping
to ensure thatmethod produces
delumped lumpscontinuous
are capable relationships
of producing at most distillation
rational results
points, except the acceptable ASTM D86 10% of naphtha and ASTM D86
[38]. A rational result is that the relationships of the side draw rate change to the corresponding 70% of diesel. Actually, the
ASTM D86 10% of naphtha is easily affected by the contents of its light
distillation curve and temperature are consecutive rather than stepwise. Figures 10 and 11 show the components. Meanwhile,
the actual value
sensitivity of ASTM
test results D86side
of the 10%products
of naphtha is veryand
naphtha small.
dieselThus,
for the
the relative
HCRSRK deviation
method may seem
(a default
large even for a small absolute deviation. As can be seen from Figure
method in HCR model) and the data-based delumping method. Apparently, the HCRSRK method 10d, the large relative deviation
for ASTM stepwise
generates D86 70%relationships
appears afteratthe change
most of sidepoints
distillation draw is large
with thethan 6%. of
change Therefore,
draw rates, theespecially
stepwise
relationship is acceptable in this case. Figure 11 illustrates that the
in naphtha, as seen in Figure 10a. The data-based delumping method produces continuous relationship between side draw rate
and temperature is smooth with the two methods. In this way, the proposed
relationships at most distillation points, except the acceptable ASTM D86 10% of naphtha and ASTM delumping method is
able
D86 to70% predict rational
of diesel. result.the
Actually, TheASTMresultsD86
also10%show of that it is necessary
naphtha to delump
is easily affected by the
the reactor
contents model.
of its
light components. Meanwhile, the actual value of ASTM D86 10% of naphtha is very small. Thus, the
The experiments of Figure 12 are performed with the feed of fractionator from a HCR since the
added ashphaltene lump is out of the range of the HCRSRK. Figure 12
relative deviation may seem large even for a small absolute deviation. As can be seen from Figure shows the simulation accuracy
comparison
10d, the large of relative
productsdeviation
properties,forincluding
ASTM D86 the70%distillation
appearscurve andchange
after the standard of gravity
side drawwith is these
large
two methods. Figure 12a,d show that the data-based method has better prediction
than 6%. Therefore, the stepwise relationship is acceptable in this case. Figure 11 illustrates that the for the distillation
range of naphtha,
relationship between diesel
sidefuel
drawandratetheand
standard densityisofsmooth
temperature the products.
with theFor two the distillation
methods. curve
In this of
way,
tail oil, the two methods have almost the same prediction. The detailed
the proposed delumping method is able to predict rational result. The results also show that it is average relative deviations
(ARD)
necessary of distillation
to delump the with the HCRSRK
reactor model. and data-based method are 5.59% and 2.56% for naphtha,
respectively; they are 6.48% and 3.61% for diesel fuel, respectively; and they are 16.41%, 13.19% for
tail oil, respectively. The deficiency of the higher boiling point RA accounts for large ARD of tail oil.
Meanwhile, the average relative deviations (ARD) of density of three products are 6.4% and 2.8% with
the HCRSRK and data-based method. The lumps with the data-based method, which possesses more
number and better distillation properties account for the higher accuracy.
Processes 2020, 8, 32 14 of 19
Processes
Processes 2020,
2020, 8,
8, xx FOR
FOR PEER
PEER REVIEW
REVIEW 14
14 of
of 20
20
Figure
Figure 10.
10. Sensitivity
Sensitivity results
results of
of the
the side
side draw
draw rate
rate and
and ASTM
ASTM D86D86 curve
curve change
change of
of different
different products
products
and methods: (a) naphtha with the HCRSRK method; (b) naphtha with the data-based
data-based
and methods: (a) naphtha with the HCRSRK method; (b) naphtha with the data-based delumping delumping
delumping
method;
method; (c)
(c) diesel
diesel with
with the
the HCRSRK method;
HCRSRK method; (d)
method; (d) diesel
(d) diesel with
diesel with the
with the data-based
the data-based delumping
data-based delumping method.
delumping method.
method.
Figure 11. Sensitivity results of the side draw rate and temperature change of different products and
Figure 11. Sensitivity results of the side draw rate and temperature change of different products and
methods:
methods: (a) naphtha with
with the HCRSRK
HCRSRK method; (b) (b) naphtha with
with thedata-based
data-based delumping method;
method;
methods: (a) naphtha
naphtha with the
the HCRSRK method;
method; (b) naphtha
naphtha with the
the data-based delumping
delumping method;
(c)
(c) diesel with the HCRSRK method; (d) diesel with the data-based delumping method.
(c) diesel
diesel with
with the
the HCRSRK
HCRSRK method;
method; (d)
(d) diesel
diesel with
with the
the data-based
data-based delumping
delumping method.
method.
The
The experiments
experiments of
of Figure
Figure 12
12 are
are performed
performed with
with the
the feed
feed of
of fractionator
fractionator from
from aa HCR
HCR since
since the
the
added ashphaltene lump is out of the range of the HCRSRK. Figure 12 shows the simulation accuracy
added ashphaltene lump is out of the range of the HCRSRK. Figure 12 shows the simulation accuracy
(ARD) of distillation with the HCRSRK and data-based method are 5.59% and 2.56% for naphtha,
respectively; they are 6.48% and 3.61% for diesel fuel, respectively; and they are 16.41%, 13.19% for
tail oil, respectively. The deficiency of the higher boiling point RA accounts for large ARD of tail oil.
Meanwhile, the average relative deviations (ARD) of density of three products are 6.4% and 2.8%
with the2020,
Processes HCRSRK
8, 32 and data-based method. The lumps with the data-based method, which possesses15 of 19
more number and better distillation properties account for the higher accuracy.
Figure 12. Predictions

Figure 12. Predictions of
of the
the properties
properties of different products
of different products with
with the
the two
two methods:
methods: (a)
(a) ASTM
ASTM D86D86
distillation curve of naphtha; (b) ASTM D86 distillation curve of diesel fuel; (c) ASTM D86 distillation
distillation curve of naphtha; (b) ASTM D86 distillation curve of diesel fuel; (c) ASTM D86 distillation
curve of tail oil; (d) density of three products.
products.
5.3. The
5.3. The Predictions
Predictions of
of the
the Whole
Whole Residue
Residue Hydrotreating
Process Model
Model
The industrial
The industrial residue
residuehydrotreating
hydrotreatingprocess
processisismodeled
modeledwithwitha reactor
a reactorand
andfractionator
fractionator model
model in
thisthis
in paper.
paper. TheThe
model
modelpredictions of four
predictions HT reactor
of four temperature
HT reactor rises are
temperature displayed
rises in Figure
are displayed 13. In the
in Figure 13.
reactor model, the catalyst reactor temperature rise is computed with the
In the reactor model, the catalyst reactor temperature rise is computed with the given inletgiven inlet temperature. The
average absolute ◦ for
temperature. Thedeviations (AAD) of
average absolute each reactor
deviations (AAD)temperature rise is 1.22,
of each reactor 0.95, 2.10,rise
temperature andis1.80
1.22, C0.95,
the residue
2.10, and 1.80 hydrorefining reactor.
°C for the residue Thus, it is concluded
hydrorefining that the
reactor. Thus, it ismodel generates
concluded that good predictions
the model generatesfor
the temperature rise of the HT reactor, which plays an important role in evaluating
good predictions for the temperature rise of the HT reactor, which plays an important role in product conversion.
The temperature
evaluating product rises of reactor
conversion. R3temperature
The and R4 are evidently higher
rises of reactor R3than R2 are
and R4 andevidently
R1. Thishigher
is because
than
different
R2 and R1. catalysts
This isare loadeddifferent
because in the reactor. Reactor
catalysts R3 andinR4
are loaded themainly removes
reactor. Reactor sulfide
R3 andandR4nitrogen
mainly
while
removes R2 sulfide
and R1and mainly remove
nitrogen metals.
while R2 and Figure 14 shows
R1 mainly the prediction
remove metals. Figureof H14 shows
2,makeup flow
therate for the
prediction
residue
of hydrorefining
H 2,makeup flow rate process and the ARD
for the residue is 2.27%. Obviously,
hydrorefining process andthe excellent
the ARD ishydrogen consumption
2.27%. Obviously, the
prediction is achieved by the developed model.
excellent hydrogen consumption prediction is achieved by the developed model.
To assess the performance of the simulated fractionator, temperature profile, product yields,
distillation curves, and density of liquid products are important indicators for model evaluation.
Temperature distribution is meaningful for evaluating energy consumption, which is also important
to energy optimization and product cutting for engineers. Figure 15 shows the prediction of the
temperature distribution of the column. The Pearson coefficients of the temperature distribution are
0.977, 0.985, 0.987, 0.983, 0.985 for dataset 1 to 5, respectively. Apparently, the model demonstrates
good capability in predicting column temperature distribution.
Three products are produced in the residue hydrogenation process, namely, naphtha, diesel fuel,
and tail oil. Figure 16 illustrates the model predictions on product yields: naphtha, diesel fuel, and
tail oil. And Table 6 shows the results of five datasets in detail. It is observed that the model can
accomplish accurate prediction on product yields with consideration of the relative deviation of each
product, which can reveal more detailed information.
Processes 2020, 8, 32 16 of 19
Figure 13. Predictions of temperature rise of the HT reactor for the residue hydrorefining process.
Figure
Figure 13.
13. Predictions
Figure Predictions of
13. Predictions temperature
of of temperaturerise
temperature riseof
rise of the
of HT
the HT reactor
HTreactor for
reactorfor the
forthe residue
residue
the hydrorefining
hydrorefining
residue process.
process.
hydrorefining process.
Figure 14. Predictions of H 2,makeup flow rate for the residue hydrorefining process. STD m3. h−1 is the
volume flow per hour of fluid at 15.556 °C and 101.325 kpa.
To assess the performance of the simulated fractionator, temperature profile, product yields,
Temperature distribution is meaningful for evaluating energy consumption, which is also important
to energy optimization and product cutting for engineers. Figure 15 shows the prediction of the
temperature distribution of the column. The Pearson coefficients of the temperature distribution are
14. Predictions of H flowrate
of H 2,makeupflow rate for
for the residue
residuehydrorefining process. STDSTD
m3. h
m3ish−1
−1 theis the
Figure
0.977, 0.985, 0.987, 0.983, 0.985 for dataset
2,makeup 1 to 5,the hydrorefining
respectively. Apparently, process.
the model demonstrates
◦C
goodvolume
volume
Figure flow flow
capability
14. per
perin
hour
Predictionshour
of of fluid
of fluid
predicting at at
H column 15.556
15.556
flow °Cand
and 101.325
for101.325
temperature
rate kpa.
kpa. hydrorefining process. STD m3. h−1 is the
thedistribution.
residue
2,makeup
volumeTo flow
assessper
thehour of fluid at of
performance 15.556 °C and 101.325
the simulated kpa.
fractionator, temperature profile, product yields,
To assess the
Temperature performance
distribution of the simulated
is meaningful fractionator,
for evaluating temperature
energy consumption, profile,
which is alsoproduct
importantyields,
distillation
to energycurves, and density
optimization of liquid
and product products
cutting are important
for engineers. Figure 15indicators
shows thefor model evaluation.
prediction of the
temperature
Temperature distribution
distribution is of the column.for
meaningful The Pearson coefficients
evaluating of the temperature
energy consumption, which distribution are
is also important
0.977, 0.985,
to energy 0.987, 0.983,
optimization and0.985 for dataset
product cutting1 tofor
5, respectively. Apparently,
engineers. Figure the model
15 shows the demonstrates
prediction of the
good capability
temperature in predicting
distribution of the column
column.temperature
The Pearson distribution.
coefficients of the temperature distribution are
0.977, 0.985, 0.987, 0.983, 0.985 for dataset 1 to 5, respectively. Apparently, the model demonstrates
good capability in predicting column temperature distribution.
Figure
Figure 15.15. Predictionofofthe
Prediction the temperature
temperature profile
profileofofthe fractionator.
the fractionator.
Table 6. Prediction relative deviations of the product yield.
Naphtha Diesel Fuel Tail Oil

Dataset 1 −0.32% 4.79% −2.11%
Dataset 2 −0.25% −0.91% −1.99%
Dataset 3 4.08% 1.52% 0.27%
Dataset 4 −0.01%of the temperature profile
Figure 15. Prediction −2.52%
of the fractionator. 0.06%
Dataset 5 −2.62% −0.07% −0.57%
Figure 15. Prediction of the temperature profile of the fractionator.

Naphtha Diesel fuel
Table 6. Prediction relative deviations of the product yield.
Tail Oil
Dataset 1 −0.32% 4.79% −2.11%
Naphtha Diesel fuel Tail Oil
Dataset 2 −0.25% −0.91% −1.99%
Dataset 1 −0.32% 4.79% −2.11%
Dataset 3 4.08% 1.52% 0.27%
Dataset 2 −0.25% −0.91% −1.99%
Dataset 4
Processes 2020, 8, 32 −0.01% −2.52% 0.06% 17 of 19
Dataset 3 4.08% 1.52% 0.27%
Dataset 5 −2.62% −0.07% −0.57%
Dataset 4 −0.01% −2.52% 0.06%
Dataset 5 −2.62% −0.07% −0.57%
Figure16.
Figure 16.Predictions
Predictionsof
ofthe
theproduct
product yield:
yield: (a)
(a) naphtha;
naphtha;(b)
(b)diesel
dieselfuel;
fuel;(c)
(c)tail
tailoil.
oil.
Figure 16. Predictions of the product yield: (a) naphtha; (b) diesel fuel; (c) tail oil.
Figure
Figure1717illustrates
illustratesmodel
modelpredictions
predictions on on the
the distillation curvesof
distillation curves ofmajor
majorproducts:
products:naphtha,
naphtha,
Figure 17 illustrates model predictions on the distillation curves of major products: naphtha,
diesel
diesel fuel, and tail oil (in Dataset 1 and 2). Figure 18 shows the density predictions of three
diesel fuel, and tail oil (in Dataset 1 and 2). Figure 18 shows the density predictions of three majormajor
fuel, and tail oil (in Dataset 1 and 2). Figure shows the density predictions of threemajor
products.
products. The
The pinpoint
pinpoint predictions
predictions indicate
indicate the proposed
proposed method
method can
can well
wellmodel
model
products. The pinpoint predictions indicate the proposed method can well model the distillation the
thedistillation
distillation
curves
curves and
and
curves density
density
and predictions
predictions
density ofof
predictions major
ofmajor products.
majorproducts.
products.
Figure
Figure
Processes17. 17. Predictions
Predictions
2020, 8, x of ASTMD86
of REVIEW
FOR PEER ASTM D86distillation
distillation curves
curvesofofliquid
liquidproducts:
products:(a) (a)
Dataset 1; (b)1;Dataset
Dataset 2.18 of 20
(b) Dataset 2.
Figure 17. Predictions of ASTM D86 distillation curves of liquid products: (a) Dataset 1; (b) Dataset 2.
Figure
18. Predictions of of
thethe densityofofproducts:
density products: (a)
(a)naphtha;
naphtha;(b)(b)
diesel fuel;
diesel (c) tail
fuel; oil. oil.
(c) tail
6. Conclusions
6. Conclusions
In paper,
In this this paper, a comprehensiveresidue
a comprehensive residuehydrotreating
hydrotreating process
processmodelmodel is established based
is established on on
based
plant data with consideration of the product yields, qualities, and process operations.
plant data with consideration of the product yields, qualities, and process operations. A novel two A novel two
parallel structure is developed for reactor model. In this model, resin and asphaltene with high
parallel structure is developed for reactor model. In this model, resin and asphaltene with high boiling
boiling point and density are added as a lump, which makes it more consistent with the actual
point and density are added as a lump, which makes it more consistent with the actual feedstock
feedstock and production conditions. The proposed reactor model can predict the boiling point of
and production
product moreconditions. The proposed
accurate. Especially, reactor
the average model
relative can predict
deviations theare
of tail oil boiling point
only about at of product
almost
more6.25%.
accurate. Especially, the average relative deviations of tail oil are only about at
The accuracy is improved by 55% compared with the original HCR model. For the fractionator, almost 6.25%.
a database-based delumping method is proposed for transforming kinetic lumps of HCR into data-
based pseudo-components. The method not only provides a consecutive response to the changes of
column specification, but also has a slightly higher prediction accuracy. Five datasets are used to
verify the effectiveness of the established integrated model. The developed process model can predict
temperature rise of the four catalyst beds with the AAD smaller than 2.10 °C, the H 2,makeup flow rate
with the ARD of 2.27%, the temperature distribution of the column with Pearson coefficients over
Processes 2020, 8, 32 18 of 19
The accuracy is improved by 55% compared with the original HCR model. For the fractionator, a
database-based delumping method is proposed for transforming kinetic lumps of HCR into data-based
pseudo-components. The method not only provides a consecutive response to the changes of column
specification, but also has a slightly higher prediction accuracy. Five datasets are used to verify
the effectiveness of the established integrated model. The developed process model can predict
temperature rise of the four catalyst beds with the AAD smaller than 2.10 ◦ C, the H2,makeup flow rate
with the ARD of 2.27%, the temperature distribution of the column with Pearson coefficients over
0.977, the product yield with the ARD smaller than 5%.
Supplementary Materials: The following are available online at http://www.mdpi.com/2227-9717/8/1/32/s1,

Table S1: BP-based pseudo-components and their properties.
Author Contributions: D.S. performed all the simulations and parameter estimation procedures, guided by Y.W.
and Y.X.; X.Y. and J.S. conceived the experimental research. X.Y. wrote the paper with feedback; D.S. and J.S.
revised the different intermediate versions until the final document. All authors have read and agreed to the
published version of the manuscript.
Funding: This work is supported by the National Key Research and Development Program of China
(Grant. 2018AAA0101603), the National Natural Science Foundation of China (Grant 61590921 and Grant
U1911401), and the Hunan Postgraduate Research Innovation Project (2019zzts148).
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Sheu, E.Y. Petroleum asphaltene properties, characterization, and issues. Energy Fuels 2002, 16, 74–82.
[CrossRef]
2. Bob, D. BP Statistical Review of World Energy; BP: London, UK, 2019.
3. Huc, A.-Y. Heavy Crude Oils: From Geology to Upgrading: An Overview; Editions Technip: Paris, France, 2010.
4. Pascal, R.; Hervé, T. Catalysis by Transition Metal Sulphides: From Molecular Theory to Industrial Application, 1st
ed.; Editions Technip: Paris, France, 2013.
5. Zobel-Roos, S.; Mouellef, M.; Ditz, R.; Strube, J. Distinct and Quantitative Validation Method for Predictive
Process Modelling in Preparative Chromatography of Synthetic and Bio-Based Feed Mixtures Following a
Quality-by-Design (QbD) Approach. Processes 2019, 7, 580. [CrossRef]
6. Arias, A.; Mores, P.; Scenna, N.; Caballero, J.; Mussati, S.; Mussati, M. Optimal Design of a Two-Stage
Membrane System for Hydrogen Separation in Refining Processes. Processes 2018, 6, 208. [CrossRef]
7. Li, K.; Yan, H.; He, G.; Zhu, C.; Liu, K.; Liu, Y. Seasonal Operation Strategy Optimization for Integrated
Energy Systems with Considering System Cooling Loads Independently. Processes 2018, 6, 202. [CrossRef]
8. Quann, R.J.; Jaffe, S.B. Structure-oriented lumping: Describing the chemistry of complex hydrocarbon
mixtures. Ind. Eng. Chem. Res. 1992, 31, 2483–2497. [CrossRef]
9. Ghosh, P.; Andrews, A.T.; Quann, R.J.; Halbert, T.R. Detailed kinetic model for the hydro-desulfurization of
FCC Naphtha. Energy Fuels 2009, 23, 5743–5759. [CrossRef]
10. Hudebine, D.; Vera, C.; Wahl, F.; Verstraete, J.J. Molecular representation of hydrocarbon mixtures from
overall petroleum analyses. In Proceedings of the 1th AIChE Spring Meeting, New Orleans, LA, USA, 10–14
March 2002; pp. 10–14.
11. Peng, B. Molecular Modelling of Petroleum Processes. Ph.D. Thesis, University of Manchester, Manchester,
UK, 1999.
12. Dai, F.; Gong, M.; Li, C.; Li, Z.; Zhang, S. New kinetic model of coal tar hydrogenation process via carbon
number component approach. Appl. Energy 2015, 137, 265–272. [CrossRef]
13. Weekman, V.W., Jr. Model of catalytic cracking conversion in fixed, moving, and fluid-bed reactors. Ind. Eng.
Chem. Proc. Des. Dev. 1968, 7, 90–95. [CrossRef]
14. Lee, L.S.; Chen, Y.W.; Huang, T.N.; Pan, W.Y. Four-lump kinetic model for fluid catalytic cracking process.
Can. J. Chem. Eng. 1989, 67, 615–619. [CrossRef]
15. Bollas, G.M.; Lappas, A.A.; Iatridis, D.K.; Vasalos, I.A. Five-lump kinetic model with selective catalyst
deactivation for the prediction of the product selectivity in the fluid catalytic cracking process. Catal. Today
2007, 127, 31–43. [CrossRef]
Processes 2020, 8, 32 19 of 19
16. Xu, O.-G.; Su, H.-Y.; Mu, S.-J.; Chu, J. 7-lump kinetic model for residual oil catalytic cracking. J. Zhejiang
Univ. Sci. A 2006, 7, 1932–1941. [CrossRef]
17. Verstraete, J.J.; Le Lannic, K.; Guibard, I. Modeling fixed-bed residue hydrotreating processes. Chem. Eng.
Sci. 2007, 62, 5402–5408. [CrossRef]
18. Korre, S.C.; Klein, M.T.; Quann, R.J. Hydrocracking of polynuclear aromatic hydrocarbons. Development of
rate laws through inhibition studies. Ind. Eng. Chem. Res. 1997, 36, 2041–2050. [CrossRef]
19. Denayer, J.F.; Baron, G.V.; Souverijns, W.; Martens, J.A.; Jacobs, P.A. Hydrocracking of n-Alkane Mixtures on
Pt/H− Y Zeolite: Chain Length Dependence of the Adsorption and the Kinetic Constants. Ind. Eng. Chem.
Res. 1997, 36, 3242–3247. [CrossRef]
20. Wiehe, I.A. Process Chemistry of Petroleum Macromolecules; CRC Press: Boca Raton, FL, USA, 2008.
21. Wiehe, I.A. The pendant-core building block model of petroleum residua. Energy Fuels 1994, 8, 536–544.
[CrossRef]
22. Ancheyta, J.; Rana, M.S.; Furimsky, E. Hydroprocessing of heavy petroleum feeds: Tutorial. Catal. Today
2005, 109, 3–15. [CrossRef]
23. Ba, A.; Eckert, E.; Vaněk, T. Procedures for the Selection of Real Components to Characterize. Chem. Pap.
2003, 57, 53–62.
24. Chang, A.-F.; Liu, Y.A. Predictive modeling of large-scale integrated refinery reaction and fractionation
systems from plant data. Part 1: Hydrocracking processes. Energy Fuels 2011, 25, 5264–5297. [CrossRef]
25. Balakos, M.W.; Chuang, S.S.C. Dynamic and LHHW kinetic analysis of heterogeneous catalytic
hydroformylation. J. Catal. 1995, 151, 266–278. [CrossRef]
26. Sheng, Q.; Wang, G.; Zhang, Q.; Gao, C.; Ren, A.; Duan, M.; Gao, J. Seven-Lump Kinetic Model for
Non-catalytic Hydrogenation of Asphaltene. Energy Fuels 2017, 31, 5037–5045. [CrossRef]
27. Li, N.; Yan, B.; Xiao, X. Kinetic and reaction pathway of upgrading asphaltene in supercritical water.
Chem. Eng. Sci. 2015, 134, 230–237. [CrossRef]
28. Trejo, F.; Rana, M.S.; Ancheyta, J. Thermogravimetric determination of coke from asphaltenes, resins and
sediments and coking kinetics of heavy crude asphaltenes. Catal. Today 2010, 150, 272–278. [CrossRef]
29. García, C.L.; Hudebine, D.; Schweitzer, J.M.; Verstraete, J.J.; Ferré, D. In-depth modeling of gas oil
hydrotreating: From feedstock reconstruction to reactor stability analysis. Catal. Today 2010, 150, 279–299.
[CrossRef]
30. Wang, Y.; Shang, D.; Yuan, X. A correction method for the proportion of key components in basic HYSYS
library based on an improved squirrel search algorithm. In Proceedings of the 12th Asian Control Conference
(ASCC), Kitakyusyu International Conference Center, Fukuoka, Japan, 9–12 June 2019; IEEE: Piscataway, NJ,
USA, 2019; pp. 236–241.
31. Ren, W.; Chen, H.; Yang, C. Application of Molecular Simulation in Density Characterization and Structure
Validation of Heavy Oil Fractions. CIESC J. 2009, 60, 1883–1888.
32. Hudebine, D. Reconstruction Moléculaire de Coupes Pétrolières. Ph.D. Thesis, École normale supérieure
(sciences), Lyon, France, 2003.
33. De Oliveira, L.P.; Vazquez, A.T.; Verstraete, J.J.; Kolb, M. Molecular reconstruction of petroleum fractions:
Application to vacuum residues from different origins. Energy Fuels 2013, 27, 3622–3641. [CrossRef]
34. Constantinou, L.; Gani, R. New group contribution method for estimating properties of pure compounds.
AIChE J. 1994, 40, 1697–1710. [CrossRef]
35. Haynes, H.W., Jr.; Matthews, M.A. Continuous-mixture vapor-liquid equilibria computations based on true
boiling point distillations. Ind. Eng. Chem. Res. 1991, 30, 1911–1915. [CrossRef]
36. Yang, X.; Yang, X.; Li, R. Lumped parameter conversion for flowsheeting of hydrocracking. Chem. Ind. Eng.
Prog. China 2011, 30, 1915–1918.
37. AspenTech. Aspen HYSYS Simulation Basis; AspenTech: Cambridge, MA, USA, 2006.
38. Kaes, G. A Practical Guide to Steady State Modeling of Petroleum Processes (Using Commercial Simulators); Athens
Printing Company: Athens, GA, USA, 2000.
© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).

Processes 08 00032 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Processes 08 00032 PDF

Uploaded by

Copyright:

Available Formats

processes

Keywords: asphaltene conversion; database-based delumping; feedstock characterization; parallel

Processes 2020, 8, 32; doi:10.3390/pr8010032 www.mdpi.com/journal/processes

RG QHT QHT QHT H 2,makeup

Vacuum heavy HT HT HT HT Naphtha

Built-in residue oil Naphtha

PFR model Tail oil

2. The reaction mechanism of the two-reactors model for residue

3.1. Kinetic Model

According to the LHHW theory,

4.1. Feedstock Mixture Characterization

CIT,b = 0.2285NR2 + 0.4687NR (12)

where NR represents the number of rings in a molecule.

Normal Critical Critical Critical

4.2. Reactor Model Construction with Two Parallel Structures

From [26] This Paper

Table 4. Part of boiling point-based pseudo-components and their properties.

Pseudo- ILD MW Pseudo- ILD MW

Dataset 1 Dataset 2 Dataset 3 Dataset 4 Dataset 5

5.2. Comparison of the HCRSRK

Figure 12. Predictions

Processes 2020, 8, x FOR PEER REVIEW 16 of 20

Table 6. Prediction relative deviations of the product yield.

Naphtha Diesel Fuel Tail Oil

Figure 15. Prediction of the temperature profile of the fractionator.

Supplementary Materials: The following are available online at http://www.mdpi.com/2227-9717/8/1/32/s1,

You might also like