Você está na página 1de 9

Article

pubs.acs.org/EF

Eight-Lump Kinetic Modeling of Vacuum Residue Catalytic Cracking


in an Independent Fluid Bed Reactor
Haohua Gao,†,‡ Gang Wang,*,† Chunming Xu,† and Jinsen Gao†

State Key Laboratory of Heavy Oil Processing, China University of Petroleum, Beijing 102249, China

National Institute of Clean-and-Low-Carbon Energy, Shenhua Group Corporation Limited, Beijing 102211, China

ABSTRACT: This study proposes an eight-lump kinetic model to describe the reaction behavior of the vacuum residue (VR)
catalytic cracking in a conceptual catalytic cracking process. The proposed lump model has 20 kinetic constants and one for
catalyst deactivation, which is specifically suitable for a VR catalytic cracking process. In this reaction system, the VR is divided
into three lumps on the basis of the composition of the structure group, so as to expand the application of the model. The
experimental data were obtained from a fixed fluidized-bed reactor and a pilot plant. The kinetic constants were estimated using a
special program compiled on the basis of Marquardt’s algorithm. The model shows high simulation accuracy, with the predicted
yields being in close agreement with the experimental results. In addition, model simulations were performed to determine the
effects of two parameters on product yields. By setting the unit characteristic factors, the model was applied to a pilot plant, and
the concentration profiles of the components along the reactor were described.

1. INTRODUCTION two reaction zones. In the first zone, a relatively light


The recent upsurge in the demand for high value-added hydrocarbon feedstock is contacted with the first catalyst
petroleum products, together with the worldwide increasing stream comprising the spent catalyst. In the second zone,
trend for heavier and inferior crude oil supply, has contributed another relatively heavy hydrocarbon feedstock is contacted
to the increased use of residue feedstock in the refineries.1−4 with the second catalyst stream comprising freshly regenerated
Fluid catalytic cracking (FCC) is one of the most important catalyst. Besides, Herbst et al.12 proposed a multiple riser
technologies in oil refining industries, in which the heavy catalytic cracking process, which involves the following
fractions of the crude oil are converted or cracked into a variety processes: (1) conversion of the first hydro-deficient heavy
of lighter products.5 Consequently, FCC is being widely used hydrocarbon feedstock in the first riser; (2) conversion of a
for converting the residues into more useful products,6 hydrogen-rich hydrocarbon feedstock in the lower region of the
especially in countries like China.7 second riser, and (3) feeding a second relatively hydro-deficient
The conventional residue fluid catalytic cracking (RFCC) heavy hydrocarbon feedstock into the upper region of the
process involves blending residues, such as atmospheric residue second riser. However, no further advancements of these
(AR) or vacuum residue (VR), deasphalted oil, aromatic processes have been reported.
extracts, and so on, into vacuum gas oil (VGO). It is well- Recently, our research group developed a conceptual
known that the distillation range and physicochemical proper- catalytic cracking process on the basis of the reaction
ties of VR dramatically differ from those of VGO. Compared characteristics.13 Figure 1 shows the schematic of the
with VGO, VR usually has a higher molecular weight and conceptual catalytic cracking process proposed by our research
boiling point and contains more sulfur and nitrogen group. Contrary to the state-of-the-art practice of the RFCC
heteroatom species that cause poisoning of the acid sites.7−9 process, the conceptual catalytic cracking process involves two
Most catalyst contaminant metals in crude oil, such as nickel, reactors. Vacuum gas oil (VGO) is cracked in the conventional
vanadium, sodium, and iron, are concentrated in VR, which riser reactor, while VR is converted in the other modified
contributes to more contaminant coke and irreversible reactor. The objective of the improvement is to weaken the
deactivation of catalysts. These inherent differences between competitive adsorption effect between VGO and VR and
VGO and VR contribute to their diverse reaction character- provide favorable reaction conditions. Thus far, a number of
istics. However, during the typical RFCC process, VGO and studies, including prestage feasibility analysis and process
VR are usually premixed and then cracked in one common parameters, has been conducted on this process.13 However,
reactor at the same reaction conditions. Given their diverse to the best of our knowledge, there are not many studies on the
reaction characteristics, cracking them in one common reactor reaction kinetics, which might actually facilitate the optimiza-
ultimately causes a competitive adsorption effect between them, tion of this process.
thereby further retarding the reaction.7,10 In order to reduce It is well-known that building a kinetic model is highly
this retardation effect and improve the product distribution by imperative to deeply describe the reaction characteristics after
considering the reaction characteristics of different feedstock
used in FCC, some novel FCC processes have been proposed Received: June 4, 2014
in the literature.11,12 For instance, Harandi et al.11 proposed a Revised: September 2, 2014
multizone catalytic cracking process, which generally comprises Published: September 3, 2014

© 2014 American Chemical Society 6554 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562
Energy & Fuels Article

of the catalytic cracking processes as well as for the lump kinetic


study. Unfortunately, previous studies in this regard have
mainly focused on the VGO and heavy oil coking deactivation.7
Only few deactivation models could be applied for pure VR
catalytic cracking. Therefore, it is highly imperative to establish
a coking deactivation model for the VR catalytic cracking.
To this end, this study presents a different eight-lump model
for describing the reaction behaviors of pure VR catalytic
cracking. Moreover, the relationship between microactivity
index and coke content is also investigated, in which the coking
deactivation model for VR catalytic cracking has been
established. The kinetic parameters have been estimated using
a custom program based on the Marquardt’s algorithm.
Furthermore, we have also investigated the effects of the two
parameters on the product distribution. By setting the unit
characteristic factor, the model has been applied to a pilot plant.

2. EIGHT-LUMP KINETIC MODEL


2.1. Model Description. In this study, the VR lump was divided
into three sublumps, namely, alkyl carbon lump (CP), naphthenic
carbon lump (CN), and aromatic carbon lump (CA). The numerical
values of the three lumps (CP, CN, and CA) were determined using the
modified Brown-Ladner method based on the data of VR 1H NMR.
This well-known method has been commonly used for the structural
analysis of petroleum and its fractions.26−28 The detailed values are
listed in Table 1. The product was divided into five groups according
Figure 1. Schematic of the conceptual catalytic cracking process.
Table 1. Average Molecular Weight and Initial Values of
separately feeding in the different reactors. The catalytic Eight-Lumps
cracking process is a complicated reaction system involving a
vast number of molecules. Therefore, it is often extremely average molecular weight,
(g·mol−1) initial values
difficult to characterize and describe the kinetics at the
molecular level.7 Consequently, such complicated reaction lumps CQ-VR JN-VR CQ-VR JN-VR
systems are usually studied by lumping the large numbers of CP 662 859 0.66 0.65
chemical compounds into several pseudocomponents, accord- CN 662 859 0.15 0.06
ing to their boiling points and molecular characteristics.14 To CA 662 859 0.19 0.29
this end, several lump kinetic models, including 3-lump,15 4- HCO 371a 371
lump,16 5-lump,17 6-lump,18 7-lump,19 8-lump,20 10-lump,21 11- light oil 130 130
lump,22 13-lump,23 and 14-lump,24 have been proposed in the LPG 40 40
literature. The majority of these models focus on the VGO or dry gas 16 16
heavy oil, while only a few lump kinetic models focus on pure coke 400b 400
a
VR catalytic cracking. Recently, our group worked on the Molecular weight of HCO reported by Peixoto and Medeiros.29
b
development of a lumping kinetic model with a product Molecular weight of coke lump reported by Xu et al.19 and Wang et
distribution for VR catalytic cracking in a fluidized bed al.30
reactor.25 The model did not take into consideration the
properties of the feedstock. Therefore, the model could not be
expanded to describe the reaction behaviors of other VR to their carbon number and boiling point ranges, as follows HCO
feedstock. One possible solution to the above-mentioned (350−500 °C), light oil (C5−350 °C), LPG (C3−C4), dry gas (C1−
problem is to divide the feed into several lumps. As it is well- C2), and coke. An eight-lump model with 20 reactions was established,
known, irrespective of the complexity in the composition and as shown in Figure 2.
molecular structure of VR, the hydrocarbon molecules can be 2.2. Mathematical Models. The following assumptions were
considered an aggregation of alkyl groups, cycloparaffin groups, made to develop the mathematical model:14,15 (1) the feedstock
and aromatic groups. The proportion of these three groups in vaporizes instantaneously; (2) plug flow for gas and catalyst and the
radial dispersion in the reactor are negligible; (3) the reactor is either
the hydrocarbon molecule can be represented by the ratio of isothermal or adiabatic; (4) the catalyst deactivation is nonselective.
carbon atoms of the alkyl side chain, naphthenic rings, and Accordingly, the continuity equation for the reactor can be written as
aromatic rings on the total number of carbon atoms.22−24 in the following eq 1.
Given this viewpoint, it is simpler and reasonable to divide the
VR lump on the basis of their structural group. ⎛ ∂ρCi ⎞ ⎛ ∂C ⎞
⎜ ⎟ + G V ⎜ i ⎟ = − ri
Furthermore, coking is an inevitable and important process ⎝ ∂t ⎠x ⎝ ∂x ⎠t (1)
during VR catalytic cracking. Coking tends to reduce the The reaction rate (ri) is proportional to the molar concentration of
activity of the catalysts and results in the variation of product lump i (ρ/Ci) and the ratio of the catalyst mass density to the gas
distribution, given the fact that the deposited coke covers parts volume (ρb/ε), as shown in eq 2. Furthermore, the rate constant ki′ is
of the active sites on the catalyst surface. Therefore, the study not a constant, rather it decreases with the deactivation of the catalyst.
of coking deactivation is helpful to an insightful understanding Accordingly,

6555 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562


Energy & Fuels Article

dC 2 P MW
=− (k6 + k 7 + k 8 + k 9 + k10)C 2 Φ
dX S WHRT (11)
dC 3 P MW
=− (k11 + k12 + k13)C3Φ
dX S WHRT (12)
dC4 P MW
=− [ν14k1C1 + ν24k6C2 + ν34k11C3
dX S WHRT
− (k14 + k15 + k16 + k17)C4]Φ (13)
dC5 P MW
=− [ν15k5C1 + ν25k10C2 + ν35k13C3
dX S WHRT
− (k18 + k19 + k 20)C5]Φ (14)
dC6 P MW
=− (ν16k 2C1 + ν26k 7C2 + ν46k17C4 + ν56k18C5)Φ
dX S WHRT
(15)
Figure 2. Kinetic scheme of the eight-lump model.
dC 7 P MW
=− (ν17k4C1 + ν27k 9C2 + ν47k15C4 + ν57k 20C5)Φ
ρb dX S WHRT
ri = ki′(ρCi)
ε (2) (16)
Substituting eq 2 in eq 1, we can deduce the following eq 3. C8 = (1 − C1M1 − C2M 2 − C3M3 − C4M4 − C5M5 − C6M6)
⎛ ∂ρCi ⎞ ⎛ ∂C ⎞ ρ /M 7 (17)
⎜ ⎟ + G V ⎜ i ⎟ = − ki′(ρCi) b
⎝ ∂t ⎠x ⎝ ∂x ⎠t ε (3) where the stoichiometric coefficient, vij, is equal to MWi/MWj. The
In the case of the steady-state fluidized-bed reactors, the time partial concentration of the coke’s lump C8 can be calculated by using a mass
derivative is zero for gas-phase plug flow. Accordingly, replacing x with balance.
the dimensionless length X = x/L, eq 3 can be rewritten as follows: 2.3. Catalyst Deactivation Model. For the kinetic studies of
lumping models during catalytic cracking, the catalyst deactivation
G V dCi ρ function (Φ) is usually described by a function that depends on the
= − ki′(ρCi) b
L dX ε (4) time-on-stream (TOS, the residence time of catalyst) or catalyst coke
content (COC). Previous studies in this regard have predominantly
By definition, GV = ((SWHρbL)/ε), and hence, eq 4 can be written as adopted the catalyst deactivation functions that depend on TOS.
dCi 1 However, functions depending on COC are much more appropriate
=− k i′(ρCi) for VR catalytic cracking, given the fact that coking is one the most
dX S WH (5) important reactions during VR catalytic cracking.31,32 Therefore, in
Here, assuming that the oil gas in the reactor is an ideal gas, this paper, a deactivation function depending on COC was adopted
and could be described as7
P MW
ρ= f (CC) = Φ = (1 + βCC)−M (18)
RT (6)
n where β and M are the COC dependent parameters that can be
∑i = 0 Ci MWi 1 determined by coke deposition experiments.
MW = =
n
∑i = 0 Ci
n
∑i = 0 Ci (7) To this end, the coking experiments were carried out in a fixed
fluidized bed reactor, which is a batch system operated in the fluidized
where MW is the average molecular weight of the oil gas and MWi is mode. In each experiment, the commercial equilibrium catalyst LVR-
the average molecular weight of lump i. In this paper, although the VR 60R and Changqing (CQ) VR were used and the details have been
lump was divided into three lumps (CP, CN, and CA), we assume that described elsewhere.25 The reaction conditions are as follows: the
the molecular weights of them are equal to that of VR.22 The average reaction temperature is 500 °C, catalyst loading is 60 g, and the weight
molecular weights of the lumps in the eight-lump model are hourly space velocity (WHSV) is 25 h−1; the feeding time ranges from
summarized in Table 1. 1 to 40 s. Eight spent catalysts with different coke contents were
The actual rate constant k′i is equal to the product of the intrinsic collected. The coke content on catalysts was determined by the high
rate constant ki and the catalyst deactivation function (Φ), as is shown frequency infrared carbon sulfur analyzer HIR-944B, and their
in the following eq 8. microactivity was tested on the basis of the ASTM D5154-2003
method. Their detailed values are listed in Table 2. The COC
ki′ = ki Φ (8) dependent parameters in eq 18 were determined through the least-
Hence, eq 9 can be deduced from eqs 5−8, as follows: squares regression analysis of the experimental data, and the R-squared
was up to 0.99 as shown in Figure 3. Accordingly, the parameters were
dCi P MW determined to be β = 0.93 and M = −0.68.
=− kiCi Φ
dX S WHRT (9) Figure 3 reveals the variations in deactivation functions for different
feedstock catalytic crackings.7 As is seen, the microactivity tends to
According to the reaction network of the eight-lump, the differ for the same coke content of a catalyst. The observed variation
mathematical equations of the kinetic models can be written as follows the sequence: VGO > heavy oil > VR. Compared to VGO and
follows: heavy oil, VR contains large-sized and refractory polycyclic aromatic
dC1 P MW hydrocarbons. Thus, the reactants can access the inner pores of zeolite
=− (k1 + k 2 + k 3 + k4 + k5)C1Φ only after thermal cracking on the external of the catalyst substrate.
dX S WHRT (10) Thus, for the same coke content of the catalyst, there is a decrease in

6556 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562


Energy & Fuels Article

Table 2. Coke Contents on Catalyst vs Relative Microactivity operated in the fluidized mode. The application simulation
experiments were performed in a technical pilot scale riser with
feeding coke content, relative
number quantity, g time, s wt % microactivity a throughput of 2.0 kg/h and a catalyst holdup of 12 kg. Details
1 0.50 1 0.16 0.89
on the specifications of the equipment and the experimental
2 1.61 3 0.40 0.80
data have been published elsewhere.13,25
3 3.08 6 0.66 0.73
3.1. Estimation of Kinetic Parameters. In order to
4 5.13 10 1.01 0.63
determine the kinetic constants, a program was compiled in
5 9.15 18 1.33 0.58 Matlab. The kinetic constants of the eight-lump model at 460,
6 13.25 26 1.78 0.48 480, 500, and 520 °C were estimated using the experimental
7 20.34 32 2.03 0.48 data obtained from CQ-VR catalytic cracking in a fixed fluidized
8 20.26 40 2.44 0.46 bed reactor. The values of kinetic constants thus obtained are
listed in Table 3. In the case of the CP lump cracking, the rate
constant k5 was found to be much larger than the other rate
constants. More quantitatively, the following trend was
observed in the rate constants: k5 (CP to light oil) > k2 (CP
to LPG) > k4 (CP to dry gas) > k3 (CP to coke) > k1 (CP to
HCO). This indicates that the main product of CP cracking is
the light oil. On the other hand, in the case of the lump CN
cracking, the rate constant k10 was found to be larger than the
other rate constants. The variation trend of rate constant was:
k10 (CN to light oil) > k6 (CN to HCO) > k8 (CN to coke) > k7
(CN to LPG) > k9 (CN to dry gas). However, k10 is slightly
smaller than k5. This implies that CN is also mostly converted
into light oil during the catalytic cracking process. The aromatic
carbons do not undergo ring-opening reactions, which would
not have produced small molecular products, LPG, and dry gas.
The variation trend of rate constant is k13 (CA to light oil) > k12
Figure 3. Catalyst deactivation functions for different feed catalytic
(CA to coke) > k11 (CA to HCO). At various temperatures, the
cracking. reaction rate ratios of CA (k13/k11 and k13/k12) are less than that
of CP (k5/k1 and k5/k3) and CN (k10/k6 and k10/k8). This
the catalytic coke, which in turn has a significant effect on the indicates that the light oil selectivity of CA lump cracking is
corresponding microactivity. With an increase in the coke content of lower than the other lump cracking. In the case of coke and
the catalyst to 1.2 wt %, the relative microactivity decreases to 0.6. heavy oil production, the variation sequence of the rate
constant is k12 (CA to coke) > k8 (CN to coke) > k3 (CP to
3. RESULTS AND DISCUSSION coke) and k11 (CA to HCO) > k6 (CN to HCO) > k1 (CP to
The VR catalytic cracking of CQ and Jinan (JN) was conducted HCO), respectively, where k12 and k11 are much larger than the
in a fixed fluidized bed reactor. The reactor is a batch system others. This indicates that the aromatic carbons are the main

Table 3. Kinetic Rate Constants

reaction temperature (°C)


−3 −1 −1
reaction network kinetic rate constants ((kg·m ) ·h ) 460 480 500 520
CP → HCO k1 0.15 0.19 0.24 0.30
CP → LPG k2 9.04 11.19 13.69 16.60
CP → coke k3 0.45 0.51 0.58 0.66
CP → dry gas k4 0.76 0.94 1.15 1.39
CP → light oil k5 42.43 46.13 49.94 53.84
CN → HCO k6 4.98 6.05 7.29 8.70
CN → LPG k7 2.46 3.13 3.95 4.92
CN → coke k8 4.86 5.49 6.16 6.87
CN → dry gas k9 0.39 0.49 0.61 0.75
CN → light oil k10 41.34 45.11 48.99 52.99
CA → HCO k11 12.06 13.80 15.69 17.72
CA → coke k12 20.20 21.87 23.57 25.32
CA → light oil k13 20.94 26.02 31.96 38.85
HCO → light oil k14 1.76 2.62 2.96 3.33
HCO → dry gas k15 0.49 0.68 0.81 0.93
HCO → coke k16 0.23 0.27 0.38 0.43
HCO → LPG k17 1.44 2.07 2.27 2.67
light oil → LPG k18 0.03 0.04 0.05 0.06
light oil → coke k19 0.15 0.19 0.24 0.28
light oil → dry gas k20 0.03 0.04 0.05 0.06

6557 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562


Energy & Fuels Article

cause underlying the production of coke and heavy oil, which is


consistent with the result reported in the literature.7
Table 4 lists the frequency factors and apparent activation
energies calculated using the Arrhenius equation. As is seen,

Table 4. Frequency Factors and Apparent Activation


Energies
frequency factor activation energy
reaction network ((kg·m−3)−1·h−1) (kJ·mol−1)
CP → HCO A1 1436.55 E1 56.05
CP → LPG A2 29436.77 E2 49.39
CP → coke A3 64.07 E3 30.24
CP → dry gas A4 2643.87 E4 49.80
CP → light oil A5 1012.32 E5 19.36
CN → HCO A6 8349.86 E6 45.36
CN → LPG A7 25591.10 E7 56.45 Figure 4. Comparison between the experimental yields (points) and
CN → coke A8 492.75 E8 28.22 the predicted yields (line) for CQ-VR in the fixed bed at 480 °C.
CN → dry gas A9 2121.76 E9 52.41
CN → light oil A10 1118.79 E10 20.16
CA → HCO A11 2018.28 E11 31.25 Besides, the secondary reaction rates are also relatively smaller.
CA → coke A12 407.48 E12 18.35 These results indicate that VGO has good crack-ability and
CA → light oil A13 78433.00 E13 50.20 favors secondary reaction due to the high content of CN and CP
HCO → light oil A14 5900.55 E14 49.06 carbons. In contrast to the residual oil catalytic cracking, the
HCO → dry gas A15 2348.19 E15 51.46 rate constants of VR catalytic cracking are closer to each other.
HCO → coke A16 1644.02 E16 54.31 However, the ratio of primary and secondary reactions for
HCO → LPG A17 3210.88 E17 46.65 generating light oil is higher. This indicates that the light oil
light oil → LPG A18 200.52 E18 53.69 lump favors the reaction path for VR catalytic cracking (VR →
light oil → coke A19 487.80 E19 49.13 light oil).
light oil → dry gas A20 364.45 E20 57.04 Table 6 lists the apparent activation energies of VGO,
residual oil, and VR catalytic cracking. The activation energy for
generating light oil (32.6−38.5 kJ·mol−1) during VGO catalytic
most activation energies are in the range of 40−60 kJ·mol−1, cracking is lower than that of residual oil and VR catalytic
close to the values reported in the literature14 but lower than cracking. Moreover, the activation energy for generating coke
those reported for thermal cracking (210−290 kJ·mol−1).7 This during VGO catalytic cracking (81.5 kJ/mol) is far less than
result confirms the practical feasibility of dividing the VR into that of VR catalytic cracking (18.4 kJ/mol). These results
CP, CN, and CA lumps, consistent with the catalytic cracking substantiate the fact that the VGO and VR have diverse
reaction mechanism. The activation energy E5 (CP to light oil) reaction characteristics and hence require different reaction
is closer to E10 (CN to light oil) but far less than E13 (CA to light operation strategies. Therefore, a better product distribution
oil). This shows that the CP and CN can readily be converted can be realized via separate feeding.13,25
into light oil, while the conversion of CA is rather difficult. 3.2. Prediction of Product Yields. One of the important
Nevertheless, the activation energies of generating coke and applications of the kinetic model is to predict the product
heavy oil are just the opposite. More quantitatively, E12 (CA to distribution. Some fundamental information can be obtained
coke) < E8 (CN to coke) ≈ E3 (CP to coke) and E11 (CA to from the model for the optimization of the reaction process.
HCO) < E6 (CN to HCO) < E1 (CP to HCO), indicating that The forthcoming section presents some simulation results and
the aromatic carbons are more inclined to generate coke and related discussions.
heavy oil. Moreover, the apparent activation energies for Figure 5 shows the predicted relationship between product
cracking CN and CP into gas (E2, E4, E7, E9) are larger than that yield and the reaction temperature at a WHSV of 20 h−1 and
for generating light oil (E5, E10). This shows that, with an CTO of 6 and LVR-60R as the catalyst for JN-VR. The details
increase in the reaction temperature, the reaction rate for of the feedstock have been described elsewhere.13,25 It was
producing gas increases more quickly than in the other clearly observed that the simulation values were close to the
reactions. Therefore, a relatively low temperature is favorable experimental ones. This indicated that the eight-lump kinetic
for the process producing liquid products. model can fit the experimental data well. With an increase in
Figure 4 shows the comparison of the experimental yields the reaction temperature, the light oil yield tends to decrease,
(points) and those predicted by the model (line) for the while the yields of dry gas and LPG increase continuously. This
catalytic cracking of CQ-VR at 480 °C. As is seen, the predicted trend is in agreement with serial kinetics. The temperature, as a
yields are almost closer to the experimental ones. This indicates main process parameter, significantly influences the product
that the eight-lump kinetic model adopted in this study fits well distribution caused by the thermodynamics of the catalytic
with the experimental data and that the predicted results are cracking reactions. Catalytic cracking being an endothermic
reliable. reaction is obviously favored by a higher reaction temperature.
Table 5 lists the kinetic rate constants of VGO,22 residual However, the product distribution of a parallel-series reaction is
oil,19 and VR catalytic cracking. Compared with the VGO closely associated with the reaction depth. The excessive
catalytic cracking, the primary reaction rates of VR catalytic conversion caused by high reaction temperature will lead to the
cracking are lower by more than an order of magnitude. generation of undesirable products, such as dry gas and coke.
6558 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562
Table 5. Kinetic Rate Constants for Differernt Lumped Kinetic Models
VGO 11-lump22 residual 7-lump19 VR 8-lump
a 3 −1 −1 a 3 −1 −1
items reaction network rate constants, (kg/m ) ·h reaction network rate constants, (kg/m ) ·h reaction network rate constants, (kg/m3)−1·h−1
primary reaction HFO → LFO 56.0−155.0 RFO → HFO 14.9 (CP, CN, CA) → HCO 0.24−15.69
Energy & Fuels

HFO → gasoline 77.8−144.1 RFO → LFO/gasolie 5.8−11.7 (CP, CN, CA) → light oil 31.96−49.94
HFO → gas + coke 12.7−41.8 RFO → gas 0.4−3.6 (CP, CN, CA) → gas 0.61−13.69
RFO → coke 11.6 (CP, CN, CA) → coke 0.58−23.57
secondary reaction LFO → gasoline 28.5−86.5 HFO → LFO/gasolie 0.9−5.8 HCO → light oil 2.96
LFO → coke 0.5−14.0 HFO → gas 0.01−0.1 HCO → gas 0.81−2.27
gasoline → coke 2.9 HFO → coke 0.3 HCO → coke 0.38
LFO → gas + coke <0.1 light oil → gas + coke 0.05−0.24
a
RFO: residual oil; HFO: heavy fuel oil; LFO: light fuel oil; gas: LPG and dry gas.

6559
Table 6. Apparent Activation Energies of Typical FCC Lumped Kinetic Model
VGO 11-lump22 residual 7-lump19 VR 8-lump
a −1 a −1
items reaction network activation energy (kJ·mol ) reaction network activation energy (kJ·mol ) reaction network activation energy (kJ·mol−1)
primary reaction HFO → LFO 38.5 RFO → HFO 50.7 (CP, CN, CA) → HCO 50.0
HFO → gasoline 32.6 RFO → LFO/gasolie 50.7 (CP, CN, CA) → light oil 45.6
HFO → gas + coke 81.5 RFO → gas 16.2 (CP, CN, CA) → gas 15.4−18.4
RFO → coke 16.2 (CP, CN, CA) → coke 18.7
secondary reaction LFO → gasoline 32.6 HFO → LFO/gasolie 50.7 HCO → light oil 49.1
LFO → coke 46.0 HFO → gas 59.8 HCO → gas 46.7−51.5
gasoline → coke 124.1 HFO → coke 59.8 HCO → coke 54.3
LFO → gas + coke 59.8−78.5 light oil → gas + coke 49.1−57.0
a
RFO: residual oil; HFO: heavy fuel oil; LFO: light fuel oil; gas: LPG and dry gas.
Article

dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562


Energy & Fuels Article

the reaction time is increased further, the partial pressure of


light oil increases correspondingly, resulting in higher
secondary reactions and leading to more final products. As a
result, a better production distribution for JN-VR can be
obtained under a relatively low WHSV of 15 h−1.
3.3. Application in a Pilot Plant. Although the proposed
eight-lump model takes into account several key factors, it is
still not capable of describing the entire catalytic cracking
process comprehensively. In addition to the reaction
conditions, feedstock properties, and catalysts considered in
this model, other important factors affecting the product
distributions include the reactor types, structures and sizes,
feedstock atomization, mixing between catalysts and oil, slide
and radial distributions of the catalyst and bubbles in the riser,
and the separation velocities between the catalysts and oil at the
exit, and so on. All the above-mentioned parameters can affect
the yield and quality of the product. However, it is rather
Figure 5. Predicted (lines) and experimental (points) yields as a difficult to establish a model that takes into consideration all the
function of reaction temperature at WHSV of 20 h−1 and CTO of 6
above-mentioned factors. A more common and feasible
and LVR-60R as the catalyst for JN-VR.
approach is to set the characteristic factors to calibrate the
Therefore, relatively low reaction temperature is more favorable model’s deviation. Accordingly, in this study, five characteristic
for realizing a high total yield of light oil products. factors of the pilot plant are being set up on the basis of the
Figure 6 shows the predicted relationship between product product-oriented principle for the eight-lump model. All the
yield and WHSV at a reaction temperature of 480 °C and CTO unit factors were regressed by the modified Levenberg−
Marquardt algorithm with two sets of plant data. The calculated
unit characteristic factor’s values are listed in Table 7. As is
seen, the values range from 0.8 to 1.7, which are in the
reasonable range.19,33,34
Figure 7 shows the comparison of the experimental yields
(points) and the predicted yields (line) for CQ-VR in the pilot

Figure 6. Predicted (lines) and experimental (pionts) yields as a


function of WHSV at reaction temperature of 480 °C and CTO of 6
and LVR-60R as the catalyst for JN-VR.

of 6 and LVR-60R as the catalyst for JN-VR. With a decrease in Figure 7. Comparison between the experimental yields (points) and
WHSV, there is an increase in the yield of light oil. This trend the predicted yields (line) for CQ-VR in the pilot plant.
reaches the maximum at 15 h−1 and then declines thereafter.
The yields of dry gas plus LPG and coke increase monotoni-
cally. The increase is rather significant for WHSV less than 10 plant at various operation conditions. It can be seen that the
h−1. These results can be explained on the basis of the fact that, deviations between the predicted values and experimental
upon prolonging the reaction time, the cracking of some large- values are rather small. Most absolute deviations are lesser than
sized hydrocarbons is promoted, thereby improving the 1.5%, with the maximum deviation for the light oil lump being
conversion and increasing the yield of light oil. However, as 4.34%. It can be concluded that the predicted and experimental

Table 7. Characteristic Factors of the Pilot Plant

characteristic factors (CF) reaction networks rate constants values


CF(1) CP → HCO, CN → HCO, CA→ HCO k1, K6, K11 1.20
CF(2) CP → light oil, CN → light oil, CA → light oil, HCO → light oil K5, K10, K13, k7 0.82
CF(3) CP → LPG, CN → LPG, HCO → LPG, light oil → LPG K2, K7, K17, K18 1.32
CF(4) CP → dry gas, CN → dry gas, HCO → dry gas, light oil → dry gas K4, K9, K15, K20 1.49
CF(5) CP → coke, CN → coke, CA → coke, HCO → coke, light oil → coke K3, K8, K12, k16, K19 1.61

6560 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562


Energy & Fuels Article

results are rather close, which proves that the predicted kinetic deviations between the simulated and experimental values,
parameters and estimated characteristic factors of the pilot proving that the eight-lump kinetic model fits well with the
plant are reliable. experimental data.
The model also describes the concentration profiles of the In addition, five characteristic factors of the pilot plant were
components along the reactor. The typical profiles are shown in set up according to the product-oriented principle. The values
Figure 8. It can be seen that more than 80 wt % feedstock has thus obtained were found to be within the reasonable range.
The model was applied to a pilot plant unit, and the
component concentration profiles along the reactor were
analyzed and presented.

■ AUTHOR INFORMATION
Corresponding Author
*Tel.: 8610-8973-3085. Fax: 8610-6972-4721. E-mail:
wanggang@cup.edu.cn.
Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
The authors acknowledge the financial support provided by the
National Natural Science Foundation of China (21176252), the
State Key Program of National Natural Science Foundation of
China (21336011), the Program for New Century Excellent
Talents in University of China (NCET-13-1029), and Key
Technologies Research and Development Program of China
(2012BAE05B02).

■ NOMENCLATURE
a = void volume fraction of fluidized bed
ρ = gas density, g·cm−3
ρb = catalyst bed density, g·cm−3
vij = stoichometric coefficient for the reaction between lump i
and lump j
k = rate constant of reaction, (kg·m−3)−1·h−1
Figure 8. Component concentration profiles along the reactor for JN- L = effective reactor length, cm
VR and CQ-VR. MW = average molecular weight of the lump, g·mol−1
P = reaction pressure, Pa
R = gas constant, 8.314 J·mol−1·K−1
been cracked in the 30% mark of the riser from the bottom, SWH = weight hourly space velocity, h−1
yielding a good product distribution. As we look at the reactor t = residence time of oil gas, s
upward, it can be realized that the conversion increases, while T = reaction temperature, K
the light oil percentage is not increased. This implies that the x = reactor length at x cross-section, cm
secondary reaction is enhanced. Therefore, the modified model Ci = concentration of the lump i, mol/ggas
can be used to simulate the pilot reactors as well as to gain Cc = coke content on catalyst, %
insights on the performance of RCCUS, which will allow the X = nondimensional length


suppression of the secondary reaction in order to obtain more
light oil. REFERENCES
4. CONCLUSIONS (1) Vafi, K.; McCaffrey, W. C.; Gray, M. R. Minimization of coke in
thermal cracking of athabasca vacuum residue in a high-temperature
A different eight-lump kinetic model for understanding the VR short-residence time continuous flow aerosol reactor. Energy Fuels
catalytic cracking was presented in this work, which contains 2012, 26, 6292−6299.
CP, CN, CA, light oil, HCO, LPG, dry gas, and coke as lumps. (2) Long, J.; Shen, B. X.; Ling, H.; Zhao, J. Nonconventional vacuum
Besides, the model has 20 kinetic constants and a new one for residue upgrading blended with coal tar by solvent deasphalting and
catalyst deactivation which is specifically suitable for VR fluid catalytic cracking. Ind. Eng. Chem. Res. 2012, 51, 3058−3068.
catalytic cracking. (3) Zhang, Y. M.; Yu, D. P.; Li, W. L.; Wang, Y.; Gao, S. Q.; Xu, G.
The rate constants at 460, 480, 500, and 520 °C and the W. Fundamentals of petroleum residue cracking gasification for
corresponding apparent activation energies have been calcu- coproduction of oil and syngas. Ind. Eng. Chem. Res. 2012, 51, 15032−
15040.
lated. The crack-ability of CP and CN are found to be close, with (4) Fathi, M. M.; Pereira-Almao, P. Kinetic modeling of Arab light
their main product being light oil. This is significantly different vacuum residue upgrading by quaprocessing at high space velocities.
from CA, which acts as the main source of coke and heavy oil. Ind. Eng. Chem. Res. 2013, 52, 612−623.
The reliability of the lumping model proposed in this study was (5) Wang, G.; Yang, G. F.; Xu, C. M.; Gao, J. S. A novel conceptional
tested by comparing the simulated results with the correspond- process for residue catalytic cracking and gasoline reformation dual-
ing experimental values. The comparison indicates small reactions mutual control. Appl. Catal., A 2008, 341, 98−105.

6561 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562


Energy & Fuels Article

(6) Dupain, X.; Makkee, M.; Moulijn, J. A. Optimal conditions in (28) Zhang, A. G.; Gao, J. S.; Wang, G.; Xu, C. M.; Lan, X. Y.; Ning,
fluid catalytic cracking: A mechanistic approach. Appl. Catal., A: Gen. G. Q.; Liang, Y. M. Reaction performance and chemical structure
2006, 297, 198−219. changes of oil sand bitumen during the fluid thermal process. Energy
(7) Chen, J. W.; Cao, H. C. Catalytic Cracking Technology and Fuels 2011, 25, 3615−3623.
Engineering; China Petrochemical Press: Beijing, China, 2005. (29) Peixoto, F. C.; Medeiros, J. L. Reaction in multi-indexed
(8) Devard, A.; de la Puente, G.; Passamonti, F.; Sedran, U. continuous mixtures: Catalytic cracking of petroleum fractions. AIChE
Processing of resid−VGO mixtures in FCC: Laboratory approach. J. 2001, 47, 935−947.
Appl. Catal., A 2009, 353, 223−227. (30) Wang, H. L.; Wang, G.; Shen, B. J.; Xu, C. M.; Gao, J. S.
(9) Arandes, J. M.; Torre, I.; Azkoiti, M. J.; Erena, J.; Bilbao, J. Effect Upgrading residue by carbon rejection in a fluid-bed reactor and its
of atmospheric residue incorporation in the fluidized catalytic cracking multiple lump kinetic model. Ind. Eng. Chem. Res. 2011, 42, 6012−
(FCC) feed on product stream yields and composition. Energy Fuels 6019.
2008, 22, 2149−2156. (31) Landeghem, F. V.; Nevicato, D.; Pitault, I.; Forissier, M.; Turlier,
(10) Fan, J. Study of residue fluid catalytic cracking by multiple-position P.; Derouin, C.; Bernard, J. R. Fluid catalytic cracking: Modelling of an
feeding. Academic master’s dissertation, China University of Petroleum, industrial Riser. Appl. Catal., A: Gen. 1996, 138, 381−405.
Beijing, 1997. (32) Meng, X. H.; Xu, C. M.; Gao, J. S.; Li, L. Catalytic pyrolysis of
(11) Harandi, M. N.; Owen, H. Multiple zone catalytic cracking of heavy oils: 8-Lump kinetic model. Appl. Catal., A: Gen. 2006, 301, 32−
hydrocarbons. U.S. Patent 5,154,818, 1992. 38.
(12) Herbst, J. A.; Owen, H.; Schipper, P. H. Multiple riser fluid (33) Jiang, H. B.; Ouyang, F. S.; Weng, H. X. Lumped model for
catalytic cracking process utilizing hydrogen and carbon-hydrogen heavy oil catalytic cracking reaction: Simulation calculation of
contributing fragments. U.S. Patent 4,717,466, 1988. commercial units. J. Chem. Ind. Eng. (China) 2001, 52, 606−609.
(13) Gao, H. H.; Wang, G.; Wang, H.; Chen, J. L.; Xu, C. M.; Gao, J. (34) Mu, S. J.; Su, H. Y.; Li, W.; Chu, J. Reactor model for industrial
S. A conceptual catalytic cracking process to treat vacuum residue and residual fluid catalytic cracking based on six-lump kinetic model. J.
vacuum gas oil in different reactors. Energy Fuels 2012, 26, 1870−1879. Chem. Eng. Chin. Univ. 2005, 19, 630−635.
(14) Nace, D. M.; Voltz, S. E.; Weekman, V. W. Application of a
kinetic model for catalytic cracking effects of charge stocks. Ind. Eng.
Chem. Process Des. Dev. 1971, 10, 530−538.
(15) Weekman, V. W.; Nace, D. M. Kinetics of catalytic cracking
selectivity in fixed, moving and fluid bed reactors. AIChE J. 1970, 16,
397−404.
(16) 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.
(17) Corella, J.; Frances, E. Analysis of the riser reactor of a fluid
cracking unit. ACS Symp. Ser., Fluid Catal. Cracking II 1991, 452, 165−
182.
(18) Takatsuka, T.; Sato, S.; Morimoto, Y.; Hashimoto, H. A reaction
model for fluidized-bed catalytic cracking of residual oil. Int. Chem.
Eng. 1987, 27, 107−116.
(19) 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.
(20) Hagelberg, P.; Eilos, I.; Hiltunen, J.; Lipiainen, K.; Niemi, V. M.;
Aittamaa, J.; Krause, A. O. I. Kinetics of catalytic cracking with short
contact times. Appl. Catal., A: Gen. 2002, 223, 73−84.
(21) Jacob, S. M.; Gross, B.; Voltz, S. E.; Weekman, V. W. A lumping
and reaction scheme for catalytic cracking. AIChE J. 1976, 22, 701−
713.
(22) Wang, S. S.; Weng, H. X.; Mao, J. J.; Liu, F. Y. Investigation of
the lump kinetic model for catalytic IV: Determination of kintic
parameters in heavy gas oil network. Acta Pet. Sin. (Petrol. Process. Sect.)
1988, 4, 18−25.
(23) Deng, X. L.; Sha, Y. X.; Wang, L. Y. Study on a kinetic model of
resid catalytic cracking. Petrol. Process. Petrochem. (China) 1994, 26,
35−39.
(24) Sun, T. D.; Zhong, X. X.; Chen, Y.; Yu, B. T. Study and
application of a lumping FCC kinetics model for vacuum residue.
Petrol. Process. Petrochem. (China) 2001, 32, 41−44.
(25) Gao, H. H.; Wang, G.; Li, R.; Xu, C. M.; Gao, J. S. Study on the
catalytic cracking of heavy oil by proper cut for higher conversion and
desirable products. Energy Fuels 2012, 26, 1880−1891.
(26) Yoshida, R.; Miyazawa, M.; Ishiguro, H.; Itoh, S.; Haraguchi, K.;
Nagaishi, H.; Narita, H.; Yoshida, T.; Maekawa, Y.; Mitarai, Y.
Structure changes in Cold Lake oil-sand bitumen and catalytic
activities during catalytic hydrotreatment. Fuel Process. Technol. 1997,
51, 195−203.
(27) Zhao, S. Q.; Kotlyar, L. S.; Woods, J. R.; Sparks, B. D.; Gao, J.
S.; Kung, J.; Chung, K. H. A benchmark assessment of residues:
Comparison of Athabasca bitumen with conventional and heavy
crudes. Fuel 2002, 81, 737−746.

6562 dx.doi.org/10.1021/ef501260n | Energy Fuels 2014, 28, 6554−6562

Você também pode gostar