Professional Documents
Culture Documents
A Refined Hydroprocessing Multicatalyst Deactivation and Reactor Performance ModelsPilot-Plant Accelerated Test Applications
Mohammed Juraidan, Mahmoud Al-Shamali, Hasan Qabazard, and Ezra K. T. Kam*
Petroleum Research and Studies Center, Kuwait Institute for Scientific Research, P.O. Box 24885, Safat 13109, Kuwait ReceiVed December 21, 2005. ReVised Manuscript ReceiVed March 23, 2006
A hydroprocessing multicatalyst deactivation and reactor performance model was originally developed to predict the life and performance of atmospheric residue hydroprocessing catalysts in a pilot-plant long-life test. The model has been subsequently applied to predict catalyst and reactor behavior in an accelerated life test using Boscan crude oil as the feedstock. Simulations from the original model did not compare well with accelerated test run data. Because Boscan crude oil has very high metal and asphaltene contents, mass balance terms that emphasized noncatalytic hydrothermal reactions in the model were needed, especially when catalysts were severely deactivated. Excellent comparisons were obtained after including hydrothermal reaction terms to refine the model. The model was then used to simulate the Boscan crude oil hydroprocessing in constantsulfur-mode operations, and it was found that it would be more beneficial to operate the atmospheric residue desulfurization unit at a lower initial temperature.
1. Introduction The hydroprocessing units in petroleum refineries are becoming increasingly important with continuing depletion in the quality of crude oils, stringent environmental regulations, increasing world demand for middle distillates, and especially rocketing crude oil prices. Atmospheric residue desulfurization (ARDS) processes are being used globally to retrieve more valuable products out of the bottoms of the barrels. Efforts are focused on enhancing unit performance in terms of operation parameters and catalyst improvement. The conventional way of evaluating the performance of a combination of several hydrotreating catalysts as a single catalyst system is to conduct a long-life test in the laboratory using bench or pilot-plant reactors under operational conditions close to those of commercial plants. This is a very expensive, time-consuming, and labor-intensive exercise. It costs at least $1 million per run and takes 12 months to complete. Repeated runs may be required to ensure quality control and quality assurance. An alternative way of solving this multiparameter problem is to test the catalysts in the laboratory using bench or pilotplant reactors under accelerated conditions.2 Accelerated tests are commonly employed in industry for comparing catalyst performance, screening catalysts for catalyst selection, or process optimization. They offer a more economical means to reach a set of conditions more severe than normal and can provide a similar indication of the deactivation pattern of the catalyst system within a shorter time scale.3-5 They can be undertaken
* To whom correspondence should be addressed. E-mail: ekam@ prsc.kisr.edu.kw. (1) Kam, E. K. T.; Al-Shamali, M.; Juraidan, M.; Qabazard, H. A hydroprocessing multicatalyst deactivation and reactor performance model - Pilot-plant life treat applications. Energy Fuels 2005, 19, 753-764. KISR 7488. (2) Yamamoto, Y.; Mizutani, Y.; Shibata, Y.; Kitou, Y.; Yamazaki, H. Cosmo resid hydroconversion catalyst: Catalyst combination technology. Stud. Surf. Sci. Catal. 1995, 100, 181-188.
either by higher operation severity or by employing high-metal residual feedstocks. Special consideration must be applied to the accelerated test because the results are catalyst-specific and there may be modifications in catalyst deactivation characteristics and mechanisms, which may lead to changes in the startof-run (SOR), middle-of-run (MOR), and end-of-run (EOR) conditions. One of the objectives for undertaking the accelerated test in our laboratories was to obtain the maximum metals uptake in the selected catalyst system. Boscan crude oil was selected as the feedstock in the study because it has a 10-times-higher metal content than those in the atmospheric residue from Kuwait export crudes (KEC-AR). Simultaneously, a mathematical model,1 which covers chemical reaction kinetics of various hydrocracking and hydrotreating reactions, deactivation reactions, and catalyst performance in ARDS reactors, was used to interpret the experimental results. Apart from the very high metal content, Boscan crude also contains substantially higher levels of asphaltenes compared to KEC-AR. Asphaltenes are considered a very complex mixture of high-boiling, aromatic, and molecular-weight compounds that may precipitate after partial catalytic processing. These precipitations can cause catalytic deactivation as coke deposits, especially at a reaction temperature over 668 K.6-7 The EOR conditions are expected to prevail sooner in hydroprocessing
(3) Birtill, J. J. But will it last until the shutdown? Deciphering catalyst decay. Catal. Today 2003, 81, 531-545. (4) Petrov, L. Problems and challenges about accelerated testing of the catalytic activity of catalysts. In Principles and Methods for Accelerated Catalyst Design and Testing; Derouane, E. G., Parmon, V., Lemos, F., Ribeiro, F. R., Eds.; Kluwer Academic Publisher: The Netherlands, 2002; pp 13-69. (5) Silvy, R. P. Future trends in the worldwide oil refining industry catalyst market. In Principles and Methods for Accelerated Catalyst Design and Testing; Derouane, E. G., Parmon, V., Lemos, F., Ribeiro, F. R., Eds.; Kluwer Academic Publishers: The Netherlands, 2002; pp 337-353. (6) Bartholomew, C. H. Catalyst deactivation in hydrotreating of residues. In Catalyst Hydroprocessing of Petroleum and Distillates; Obella, M. C., Shih, S. S., Eds.; Marcel Dekker: New York, 1994; pp 1-32.
10.1021/ef0504265 CCC: $33.50 2006 American Chemical Society Published on Web 05/09/2006
kinetic data at the same conditions as those of the catalytic hydroprocessing, Boscan crude at 0.092 L/h and hydrogen gas at 62 L/h, both at a pressure of 120 bar, were passed through the reactors at temperatures of 603, 623, 643, 663, 673, and 683 K. Each temperature was maintained for 3 days. Product samples were collected every 24 h at each temperature to analyze sulfur (S), vanadium (V), nickel (Ni), conradson carbon (CCR), and asphaltene (Asph) contents. Figure 1 presents experimental results of S removal by hydrothermal reactions at different temperatures.18 It clearly shows the greatest amount of S removal by hydrothermal desulfurization was achieved at a high temperature of 683 K. Therefore, it is appropriate to consider noncatalytic hydrothermal reactions to reflect on their contributions to Boscan feedstock hydroprocessing. 2. Model Description
The original model1 is a simple one-dimensional homogeneous plug-flow model with the following assumptions: 1. The physical properties of the fluid and catalysts are constant. 2. It is a plug-flow reactor, and axial dispersion is not important. 3. The reactor is isothermal only in the radial direction in a commercial plant reactor. 4. There is no interphase heat and mass-transfer resistance between fluid and catalyst particles, but the intraparticle masstransfer resistance is considered. 5. The heat released by the hydrodemetalation (HDM) of vanadium and nickel (HDV and HDNi), hydrodenitrogenation (HDN), or the hydrocracking of asphaltenes (HDAsph) is negligible compared with that of hydrodesulfurization (HDS). 6. The main hydroprocessing reactions, such as HDS, HDV, HDNi, and HDAsph, and coking reactions can be represented by separable kinetics. 2.1. Hydrothermal Treating Consideration. To account for the depletion of materials because of hydrothermal reactions, a new term is added to each mass-balance equation. Hence, the external field equations for the mass and heat balance in any one reactor, under a pseudo-steady state, become
LuL
Boscan crude than in KEC-AR because of the higher metal and asphaltene contents in Boscan crude. Asphaltenes are the most complex components in petroleum oils and can deactivate catalysts progressively. Because of their importance in defining the refined product quality (in terms of sulfur, metals, nitrogen, and aromatic contents) and product distribution, numerous studies of their structural characteristics and changes, reaction pathways, and kinetics due to catalytic hydroprocessing have been conducted.8-11 Investigations into the thermal cracking of asphaltenes and hydrocarbons without hydrogen gas have been carried out by Wang and Anthony,12 Burkle-Vitzthum et al.,13 and Singh et al.14 Research on the hydrocracking and hydrotreating of petroleum asphaltenes with or without a catalyst was also undertaken by Sebor et al.15 and Le Page et al.16 Table 1 shows the properties of KEC-AR and Boscan crude.17 Metal and asphaltene contents and those of some lighter materials are higher in Boscan crude than those in KEC-AR. Thus, hydrothermal processing reactions are expected to be more important in crudes than in residues. Hence, blank experiments were conducted using the pilot-plant designated for the accelerated life test. Only inert materials, such as R-aluminum balls and carborundum particles without catalysts, were loaded into the reactors. To obtain the noncatalytic hydrothermal reaction
(7) Minicucci, X.; Zou, X. Y.; Shaw, J. M. The impact of liquid-liquidvapour phase behaviour on coke formation from model coke precursors. Fluid Phase Equilib. 2002, 194, 353-360. (8) Asaoka, S.; Nakata, S.; Shiroto, Y.; Takeuchi, C. Asphaltene cracking in catalytic hydrotreating of heavy oils. 2. Study of changes in asphaltene structure during catalytic hydroprocessing. Ind. Eng. Chem. Process. Des. DeV. 1983, 22, 242-248. (9) Lapinas, A. T.; Klein, M. T.; Gates, B. D.; Macris, A.; Lyons, J. E. Ind. Eng. Chem. Res. 1991, 30, 42-48. (10) Seki, H.; Kumata; F. Deactivation of hydrodesulfurisation catalysts for resides: Effect of hydrodemetallization operation conditions. Proceedings of the 8th International Symposium on Catalyst DeactiVation, Brugge, Belgium, October 10-13, 1999; pp 357-364. (11) Ancheyta, J.; Centeno, G.; Trejo, F.; Marroquin, G.; Garcia, J. A.; Tenorio, E.; Torres, A. Extraction and characterization of asphaltenes from different crude oils and solvents. Energy Fuels 2002, 16, 1121-1127. (12) Wang, J. S.; Anthony, E. J. A study of thermal-cracking behavior of asphaltenes. Chem. Eng. Sci. 2003, 58, 157-162. (13) Burkle-Vitzthum, V.; Michels, R.; Scacchi, G.; Marquaire, P. M. Mechanistic modeling of the thermal cracking of decylbenzene. Application to the prediction of its thermal stability at geological temperatures. Ind. Eng. Chem. Res. 2003, 42, 5791-5808. (14) Singh, V.; Kumar, M. M.; Saxena, A. K.; Kumar, S. Studies on thermal cracking behaviour of residue feedstocks in a batch reactor. Chem. Eng. Sci. 2004, 59, 4505-4515. (15) Sebor, G.; Weisser, O.; Hajek, M. Chem. Technol. 1981, 33, 362365. (16) Le Page, J. F.; Morel, F.; Trassards, A. M.; Bousquet, J. Am. Chem. Soc., DiV. Pet. Chem. 1987, 32, 470-476. (17) Marafi, A.; Maruyama, M.; Absi-Halabi, M. ARDS catalyst system testing methodology. Kuwait Institute for Scientific Research, Final Report, Vol. III, Kuwait, 2003.
j ) 1-N
(1)
LuL
j )1-N
(2)
LuL
j ) 1-N
(3)
LuL
j ) 1-N (4)
( LuLFLCpL +
where z is the reactor axial coordinat and j is the reactor index, from 1 to N. R and are the correlation coefficients obtained from results of blank experiments undertaken in our laboratories. They can be considered as apparent rate constants and reaction orders of the various hydroprocessing reactions specific to our pilot-plant unit. Table 2 shows R and values at 663 K as an example.
(18) Marafi, A.; Al-Marri, M.; Al-Mutairi, A. Hydroprocessing Boscan crude over inert particles. Kuwait Insitute for Scientific Research, Internal Report, Kuwait, 2005.
Juraidan et al.
For a processing unit consisting of N reactors cascaded in series, the boundary conditions for each reactor are considered as follows, at z ) 0: CS ) CS,j-1; CV ) CV,j-1; CNi ) CNi,j-1; CAsph ) CAsph,j-1 j ) 1-N (6) T ) Tj-1 j ) 1-N (7)
than the SOR and EOR. Finally, at the EOR, a large amount of metal deposits accumulate and the process temperature reaches 683 K or more, at which the coke formation becomes more severe. Consequently, a second rapid catalyst deactivation is observed. Empirical decay functions for the initial rapid deactivation for the HDS catalyst in the SOR are obtained from our pilot-plant experimental results, and they can be expressed as follows: k o,i ) ko,i + 0.5ko,ii i ) -exp i ) 1-N (8) i ) 1-N (9)
{[
cp,i(Target)
]}
n
where CS,j-1, CNi,j-1, CV,j-1, CAsph,j-1, and Tj-1 are the inlet reactant concentrations and temperatures of each reactor. 2.2. Deactivation Characteristics. The catalyst deactivation in a multiple-catalyst system for residue hydroprocessing occurs typically in three stages. Initially, under the SOR conditions, the catalysts are fresh and their activity is the highest. It was suggested that the first initial rapid catalyst activity decay at the SOR might be due to coke deposits accumulating quickly on the catalyst surface.19-21 However, workers believed that the initial rapid catalyst deactivation was due to metal sulfide deposition on the catalyst surface.22-23 Second, after the SOR, a relatively long period of slower deactivation from the gradual build up of metal deposits on the catalyst pore surface follows, while coke deposition is greatly reduced. The cokes nature also changes with the process time, as suggested by Bartholomew6 and Furimsky and Massoth.24 This deactivation period is usually found in the MOR, which lasts longer
(19) Gualda, G.; Kasztelan, S. Initial deactivation of residue hydrodemetallization catalysts. J. Catal. 1996, 161, 319-337. (20) Higashi, H.; Takashi, T.; Kai, T. The effect of start-up conditions on deactivation of hydrotreating catalyst for heavy residue with high asphaltenes content. Catal. SurV. Jpn. 2002, 5 (2), 111-119. (21) Ancheyta, J.; Betancourt, G.; Centeno, G.; Marroquin, G.; Alonso, F.; Garciafigueroat, E. Catalyst deactivation during hydroprocessing of Maya heavy Crude Oil. 1. Evaluation at constant operating conditions. Energy Fuels 2002, 16, 1438-1443. (22) Tamm, P. W.; Harnsberger, H. F.; Bridge, A. G. Effects of feed metals on catalyst aging in hydroprocess residuum. Ind. Eng. Chem. Process Des. DeV. 1981, 20, 262-273. (23) Baumgart, J.; Yang, Y. T.; Ernst, W. R.; Carruthers, J. D. Characteristics of laboratory-coked resid HDS catalysts. J. Catal. 1990, 126, 477-488. (24) Furimsky, E.; Massoth, F. E. Deactivation of hydroprocessing catalysts. Catal. Today 1999, 52, 381-495.
i ) 1-N i ) 1-N
(10) (11)
where k o,i and ko,i are apparent rate constants for species i in the initial rapid deactivation period and reported kinetics, respectively, i is the species index (1 for S, 2 for V, 3 for Ni, and 4 for Asph), i is a decay function for species i, cp and cf are concentrations of the product and feed, respectively, and n is the reaction order. When the decay function for sulfur becomes S e 0.01, the initial rapid deactivation period is terminated. The pore-filling model of Khang and Mosby,25 postulating catalyst deactivation by metal deposits, is adopted for the quasisteady-state deactivation in the MOR. The relative hydrotreating activity and relative metal loading are the two parameters to postulate the behavior of catalyst deactivation by two adjustable parameters of the Thiele modulii for fresh, o, and deactivated, z, catalysts. The activity equation is represented by a ) (Z/){(1/) tanh[2z (1 - Z)] + 1}/{(1/) + tanh[2z (1 - Z)]} (12) where a is the normalized activity, Z is the parameter for the unconverted metals, and and z are the two Thiele modulii to facilitate the catalyst deactivation model simulation. Moreover:
(25) Khang, S. J.; Mosby, J. F. Catalyst deactivation due to deposition of reaction product in macropores during hydroprocessing of petroleum residuals. Ind. Eng. Chem. Process Des. DeV. 1986, 25, 437-442.
Energy & Fuels, Vol. 20, No. 4, 2006 1357 (13) (14) (15) (MMAX)j ) (16) (Mi)j )
{[(C 10 W
4 j i)1
F L FL
V,f
) 0 /z
2
Z ) (1 - M/M0)1/2
{[(C
V,f
where k0 is the initial reaction rate constant in the active sites of fresh catalysts, Dz is the diffusivity in the deposited layer, rp is the initial pore radius, kz is the reaction rate constant in the deposited layer, M is the metal-on-catalyst, and M0 is the maximum metalon-catalyst. Separable kinetics for the main and deactivation reactions26 are used to describe the continuous coke deposition at the SOR, MOR, and EOR. The coking precursors are mainly asphaltenes.27-28 The aromaticity of asphaltenes in the residue oil is a clear index of the coking of HDS catalysts, which was confirmed by Seki and Kumata.10 The reduction of catalyst activity, s, depends on the deactivation kinetics and first order to the concentration of the coking precursors, that is, ds - ) [kf1 f1(Cfl,T) + kf2 f2(Cf2,T)]s dt 0ezeL (17)
[W (M ) ]
j i j
WAB Mi )
j)1
W
j)1 N j
(24)
j
[W (M
WAB MMAX )
j)1 N j)1
MAX)j]
(25)
j
where kf1 and kf2 are the apparent rate constants for parallel and series coking, respectively, Cf1 and Cf2 are the concentrations of the asphaltenes and products from the cracked asphaltenes species, and T is the reaction temperature. The term fi(Cfi,T) is defined as fi(Cfi,T) ) Cfi exp
[ ( )]
To Efi 1RTo T at t ) 0 at z ) 0 at z ) 0
where FL and FL are the volumetric flow and density of the feed oil, respectively, CV,f and CV,p are the concentrations of V in the feed and product, respectively, CNi,f and CNi,p are the concentrations of Ni contents in the feed and product, respectively, Wj is the weight of a type of catalyst in the catalyst system, ti is the time between two consecutive samples at the ith interval, and N is the number of catalyst types in the system.
i ) 1-2
(18)
where Efi is the activation energy for the parallel or series coking reaction, R is the gas constant, and To is the initial reaction temperature. The initial conditions for a parallel fouling mechanism are s)1 (19) (20) (21)
where t is the process time. Deactivation by coke deposition is not pronounced at low temperatures or coke precursor concentrations. However, it becomes more important when the temperature is sufficiently high and asphaltene conversion to desirable products is greatly reduced, as experienced at the EOR. This may contribute to the second period of rapid deactivation. 2.3. Metal-on-Catalyst. The metal-on-catalyst, Mi, is expressed as the weight of metal deposited on 100 g of the catalyst. It provides an additional way to verify the model by comparing Mi from pilotplant mass balance data to that of model simulations. Equations to determine Mi, at any process time, t, and the maximum metal-oncatalyst, MMAX, of any one type of catalyst in the catalyst system, and the weight-average-bed (WAB) values of Mi and MMAX of the whole catalyst system are
(26) Kam, E. K. T.; Hughes, R. The effect of catalyst fouling on the performance of adiabatic packed bed reactors - A theoretical study. Chem. Eng. J. 1979, 18, 93-102. (27) McKnight, C. A.; Nowlan, V. Metals accumulation and particle mixing in a commercial residue hydroprcessor with continuous catalyst addition. Proceedings of the 205th National Meeting of the American Chemical Society, Denver, CO, March 28-April 2, 1993; pp 1-3. (28) Gray, M. R.; Zhao, Y. X.; McKnight, C. M.; Komar, D. A.; Carruthers, J. D. Coking of hydroprocessing catalyst by residue fractions of bitumen. Energy Fuels 1999, 13, 1037-1045.
4. Model Applications and Performance Simulation The model reaction unit consists of two reactors and three catalysts. The first reactor was packed with a HDM catalyst only, while the second was packed with HDS and HDS/HDN catalysts. The catalyst types and their respective ratios, the initial process temperature, the liquid hour space velocity (LHSV), and the feedstock were identical to those used in our pilot-plant accelerated life test. Input parameters were feed properties, catalyst system configurations and their kinetic properties, reactor configurations, and operation conditions. The values of
Juraidan et al.
Table 4. Accelerated Test Run Experimental Conditions17 process parameter temperature pressure LHSV H2 to oil ratio process time units K bar h-1 mL/mL h KEC-AR 648.0 120.0 0.28 570.0 336.0, 336.0 Boscan crude 663.0 120.0 0.28 570.0 1272.0
Table 5. Estimated Apparent Rate Constants at 663 K for the Boscan Crude29 parameter kSo wt % h-1 kVo h-1 kNio h-1 kAsph h-1 HDM catalyst 0.4299 0.0622 0.1237 0.0430 HDS catalyst 1.500 0.008 0.040 0.340 HDS/HDN catalyst 1.500 0.008 0.040 0.340
the parameters for the initial KEC-AR hydroprocessing were similar to those used in the previous life test.1 After switching the feed to Boscan crude, values of the adjustable parameters of the contacting efficiency and Thiele modulii had to be determined through the optimization procedure from the first 14 days of the accelerated run data with Boscan feedstock. Moreover, because no dedicated kinetics experiments were conducted in the accelerated life test run to establish the apparent rate constants of HDS, HDV, HDNi, and HDAsph reactions using Boscan crude, the kinetic values were also determined through parameter estimation procedures. Table 5 shows the estimated values of the apparent rate constants of HDS, HDV, HDNi, and HDAsph reactions on the three catalysts at 663 K. 4.1. Model Verification with Long Pilot-Plant Life Test. Before applying the refined model to predict reactor and catalyst system behaviors in the accelerated test, it was judged useful to assess simulations from both the original and refined models by comparing their data with the previous long pilot-plant life test data on KEC-AR.1 The values of MMAX for the HDM, HDS, and HDS/HDN catalysts in the system are 60, 39, and 39 g/100 g of catalyst, respectively. The changes of S content in the product from the last reactor with the process time are presented in Figure 2. Both simulated results followed the pilot-plant data (symbols) closely in the SOR, until the EOR at 9696 h (404 days). Nevertheless, as the catalyst gradually deactivated, a small divergence between simulations of the refined (solid line) and original (dotted line) models emerged. The refined model, which included hydrothermal reactions, simulated lower S content values. 4.2. Model Verification and Prediction. Simulations were conducted using the first 336 h (14 days) of accelerated test data on the Boscan crude feedstock to perform parameter estimation and ensure good matches between measured and simulated values. The main performance indicators, such as HDS, HDV, HDNi, and HDAsph variation with process time from the two fixed bed reactors, were used for model verification. After satisfactory comparison between the pilot-plant data and simulations, the model was used to predict the complete accelerated test run, starting from the 336 h (14 days) of KECAR stabilization, through 1272 h (53 days) of Boscan crude accelerated deactivation, to another 336 h (14 days) of KEC-
AR activity confirmation. The model was then applied for a parametric study of catalyst combinations. Figure 3 presents the product S content varying with process time in the first reactor, in which only the HDM catalyst was loaded. Three curves were plotted; pilot-plant data is represented by squares, simulated results from the original model without hydrothermal reactions by a dotted line, and simulated results from the refined model with hydrothermal reactions by a solid line. The y axis represents the S contents in weight percent, and the x axis indicates the process time in hours. The pilot-plant data clearly showed an initial enhancement in hydrodesulfurization activity of the HDM catalyst on KECAR feed. Simulated results of such catalyst activity enhancement from both models are also depicted. Once the feed was switched to Boscan crude, very rapid catalyst deactivation was observed. Most catalyst activity diminished within the first 744 h (31 days) after switching the feedstock, and no more increase in S content in the product was observed thereafter. Simulated results from the refined model (solid line) matched those of the pilot-plant very well. However, simulated results from the original model (dotted line), which did not account for the S removal through noncatalytic hydrothermal reactions, showed a higher concentration, especially when the catalyst activity had been severely lost. After returning the feedstock to KEC-AR, there was a drop in S content in the product oil for both plant data and simulations because of the lesser S content in KEC-AR feedstock. However, in the last 336 h (14 days) of operation of the test, all three of the graphs did not match. The difference between the two simulations is due to the effects of hydrothermal desulfurization. The discrepancy between the pilot-plant data and those of the simulations may be due to the deposited vanadium and nickel materials, which could catalyze hydrodesulfurization in the pilotplant experiment. This can be an area in which the model can be improved further. S removal due to hydrothermal desulfurization reactions from Boscan crude and KEC-AR can be estimated by taking the difference between the two sets of simulated S values; that is, Sp is the difference between the solid (with hydrothermal HDS) and dotted (no hydrothermal HDS) lines. When KEC-AR was hydroprocessed at a lower reaction temperature (648 K) with a very active HDM catalyst up to 336 h (14 days), Sp was averaged at 0.01 wt %. After switching the feed to Boscan crude, hydrodesulfurization was conducted at a higher reaction temperature (663 K) throughout the next 1272 h (53 days). The activity of the HDM catalyst continuously declined, and the maximum Sp reached 0.61 wt % at 1088th hour, which can be considered the tmax as used in eq 23. Sp remained at this level until the feedstock was returned to KEC-AR. Afterward, because the KEC-AR was hydroprocessed at a lower reaction temperature over the nearly inactive HDM catalyst, the average Sp became 0.29 wt %. This observation clearly indicates that hydrothermal desulfurization reactions over the HDM catalyst have lesser effects on residue feedstock than on crude oil, either when the catalyst has high HDS activity or at low reaction temperatures. Similar observations were also found for the hydrothermal reactions on the changes in V and Ni contents with the process time in the first reactor, as shown in Figures 4 and 5, respectively. The values of simulated metal contents in the product oil from both models were slightly higher when the feedstock was switched from KEC-AR to Boscan crude. After a further 312 h (13 days) of operation, however, the simulations from the refined model (solid line) matched the pilot-plant data far better than those of the original model (dotted line), until the feedstock returned to KEC-AR.
Figure 2. Comparison of simulated S and pilot-plant data from the previous KEC-AR long-life test.1
Figure 3. Comparison of simulated results of S and pilot-plant data in the first reactor.
Figure 4. Comparisons of simulated results of V and pilot-plant data in the first reactor.
Figure 6 shows comparisons of the asphaltene content in the product of the first reactor between pilot-plant data and model simulations. The simulated results with hydrothermal HDAsph reactions from the refined model (solid line) compared well with the pilot-plant data and were much better than those of the simulation from the original model (dotted line). However, because of the breakdown of the automatic asphaltene analyzer in our laboratory, no more analytical values were obtained after 1200 h.
The changes in product S content with the process time from the second reactor are presented in Figure 7. The characteristics of the graphs are different than those of the first reactor. Because there was no leveling-off of the S content, the catalysts in the second reactor may not be completely deactivated. Simulations from the original model (dotted line) and refined model (solid line) compared very well initially with the pilot-plant data. After 1100 h of operation, however, the simulation from the original model, which did not include the hydrothermal reactions,
Juraidan et al.
Figure 5. Comparisons of simulated results of Ni and pilot-plant data in the first reactor.
Figure 6. Comparisons of simulated results of asphaltenes and pilot-plant data in the first reactor.
Figure 7. Comparisons of simulated results of S and pilot-plant data in the last reactor.
diverged from pilot-plant values, while the simulated results from the refined model matched the pilot-plant results for the entire run, even after the feedstock changed from Boscan crude back to KEC-AR. One interesting observation was found in the comparison between the pilot-plant data and the simulation results from the original model. The difference in S content in the final product from the simulation and pilot-plant results was drastically increased after 1152 h (48 days) of operation when the HDM
catalyst in the previous reactor had lost most of its activity. Metal spilled over to the second reactor caused the HDS and HDS/HDN catalysts in the second reactor to assume the duty of the HDM catalyst of capturing the metal deposits, and they deactivated very rapidly. Moreover, because the original model did not account for the sulfur removal through hydrothermal reactions, a greater S content would be expected in the product oils. This observation did not apply to simulated results from the refined model.
Figure 8. Comparisons of simulated results of V and pilot-plant data in the last reactor.
Figure 9. Comparisons of simulated results of Ni and pilot-plant data in the last reactor.
Figure 10. Comparisons of simulated results of Asph and pilot-plant data in the last reactor.
Comparisons of V, Ni, and Asph contents in the final product between simulation results and pilot-plant data are presented in Figures 8, 9, and 10, respectively. In all three figures, simulated results from the original model (dotted line) and refined model (solid line) matched the pilot-plant values at the SOR when KEC-AR was used. After switching the feedstock to Boscan crude, only simulation results from the refined model followed the pilot-plant data consistently. Very rapid deactivation was also found in the hydroprocessing of all three
components after the HDM catalyst in the first reactor lost most of its activity. 4.3. Maximum Metal-on-Catalyst, MMAX. There are different ways to define the maximum metal-on-catalyst, MMAX, also known as the maximum capacity of metal uptake. Yamamoto and co-workers2 suggested that the MMAX is at the breakpoint (intersection) between the two gradients of the loss of catalyst activity in the MOR and EOR. Our evaluation is to adopt the MMAX when the catalyst system is close to inactivity.
Juraidan et al.
Figure 11. Comparison of metal-on-catalyst between pilot-plant data and models simulation.
Table 6. Comparison of WAB MMAX Values Obtained from Different Routes conditions 1 2 3 4 Boscan crude accelerated life test (tmax ) 1088 h) calculated from oil mass balance simulated from refined model determined from catalyst analysis30 70% of mass balance value simulated from original model 1st reactor 82.10 78.62 57.40 59.28 WAB MMAX (g/100 g) 2nd reactor 18.78 16.64 13.13 13.63 catalyst system 36.65 37.02 25.66 27.84
From our previous study,1 the optimum values of MMAX in the selected systems of HDM, HDS, and HDS/HDN catalysts were 60, 39, and 39 g/100 g of catalyst, respectively. A comparison of the metal-on-catalyst, Mi, from the pilotplant data and those of simulations from both the original and refined models is shown in Figure 11. The changes in the Mi with the process time of the HDM catalyst in the first reactor, the combined HDS and HDS/HDN catalysts in the second reactor, and the WAB values of catalysts in both reactors are presented. The Mi values from the pilot-plant data are represented in solid symbols; the simulated results from the original model are plotted in dotted lines with hollow symbols, and simulations from the refined model are plotted in solid lines. The vertical line indicates the maximum process time (tMAX) of 1088 h (45.3 days) at which S, V, Ni, and asphaltene contents in the product oil are leveling-off at the end of the HDM catalysts life. The simulated Mi values of the HDM catalyst from the original model (dotted line and ) symbol) followed the pilotplant data (( symbol) very closely until 816 h (34 days), when the HDM catalyst became catalytically inactive. The simulated results from the refined model (solid line) matched the pilotplant data for the entire process time but at slightly lower values than the pilot-plant data. Hence, the simulations from the refined model give a better comparison than original model. The Mi values of the combined HDS and HDS/HDN catalysts between the pilot-plant data (b symbol) and simulations from the original model (dotted line with O symbol) compared very well at the SOR and EOR, but those of the simulations were lower in the MOR. The comparison between the pilot-plant data (b symbol) and simulations from the refined model (solid line) matched very well at the SOR and MOR, but it was higher than the pilot-plant data at the EOR. For the catalyst system, comparisons between the pilot-plant data (2 symbol) and simulation results from the original model (dotted line with 4 symbol) matched well at the SOR. When the HDM catalyst started to lose most of its activity, however,
deviation occurred. In contrast, the refined model simulation results (solid line) and the pilot-plant data matched well for the entire run. This was due to the higher Mi values of the HDS and HDS/HDN catalysts in the second reactor compensating the lower Mi values of the HDM catalyst in the first reactor. The maximum metal-on-catalyst, MMAX, of the HDM catalyst at 82.1 g/100 g of catalyst was calculated from the mass balance in the oil samples. Although a large quantity of metals in the Boscan crude had been removed by the HDM catalyst, the amount of metal leakage from the guard bed, according to the metal analyses from pilot-plant samples, was still higher than the feed metal contents of V and Ni in the KEC-AR feedstock. Hence, the life of the subsequent HDS and HDS/HDN catalysts, which did not have larger pores as in the HDM catalyst to accommodate metal deposits, was shortened considerably. A lower MMAX value for catalysts in the second reactor was found at 18.78 g/100 g of catalyst. The WAB MMAX of the catalyst system reached 36.65 g/100 g of catalyst. Table 6 shows a summary of the WAB MMAX values of the first and second reactors and the catalyst system. They were estimated from different routes, the mass balance between the oil feed and products, a catalyst sample analysis,30 and from simulations of the refined model and original models. These values were determined up to a tMAX of 1088 h (45.3 days) of process time. Values of WAB MMAX determined from analyzing catalyst samples were generally 30% less than the mass balance values. It is also interesting to note that the routes to determine the WAB MMAX values can be roughly categorized into two groups. The first group covered the routes in which the WAB MMAX results were obtained from the mass balance of oil feed and product and simulation from the refined model. Values from this group included both catalytic hydroprocessing and hydro(29) Kam, E. K. T.; Al-Shamali, M.; Juraidan, M. Studies of ARDS catalyst systems and reactor performance - model development and applications. Kuwait Institute for Scientific Research, Final Report, Vol. IV, Kuwait, 2003. (30) Marafi, A. Private correspondence, 2005.
Figure 12. Simulated WABT and product S content from hydroprocessing Boscan crude at 648 and 663 K in constant-sulfur-mode operations.
thermal noncatalytic reactions. The other group estimated the WAB MMAX values from catalyst sample analysis,30 which was very compatible with the simulation of the original model, and covered only catalytic hydroprocessing reactions. Hence, logically, they should provide a more appropriate representation to estimate the metal-on-catalyst. Nevertheless, the MMAX value of the HDM catalyst depends on the properties of the feedstock, process conditions, and catalyst deactivation characteristics. It affects the MMAX values of subsequent hydroprocessing catalysts of the system and those of the entire catalyst system. Subsequently, it can influence the active life span of the catalyst system. Further research must be conducted to obtain a better understanding in defining and determining the MMAX. 4.4. Constant-Sulfur-Mode Simulation. In commercial operations, a WABT against process time curve is more useful to process engineers in the ARDS unit operation because the refinery ARDS units are always operated in the constant-sulfurmode. The input values of process and operation parameters required in the following model simulation were the same as those obtained from the optimization procedures in the constanttemperature study. Figure 12 shows the changes of WABT and product S content with the process time at two reaction temperatures of 663 and 648 K of hydroprocessing Boscan crude oil. At 663 K, the WABT (curve A) approached the metallurgical temperature limit of the ARDS unit (685 K) in 1032 h (43 days), but the product S content had not breached the S target limit of 0.7 wt % until reaching 1068 h (44.5days). The continuation of sulfur removal for the extended period was due to hydrothermal noncatalytic reactions. This observation did not appear in the simulation from the original model.1 When the reaction temperature at the SOR was reduced to 648 K, the fuel requirement of the two ARDS feed heaters was greatly reduced because it took 792 h (33 days) to raise the reactor temperature to 663 K to compensate for the loss of catalyst activity. The product S content (curve D) could match the S target until 1128 h (47 days) of process time was reached. Both the process time and extended period from operating at the SOR temperature of 648 K were longer than those at 663 K. Moreover, when comparing the two S content curves, there was a considerable amount of quality giveaway of the product S content operating at 663 K (curve C) with 648 K (curve D) in the SOR when the catalyst was at its prime conditions. Hence, it will be more beneficial to operate the ARDS unit in hydroprocessing Boscan crude oil at a lower SOR
temperature. Nevertheless, because the objectives of an accelerated catalyst deactivation test were to reduce the time and cost to obtain a complete set of catalyst deactivation data, a higher SOR temperature might be better to serve these purposes. 5. Conclusions A model to study the hydrotreating reactor performance and catalyst deactivation was refined by including the hydrothermal noncatalytic hydroprocessing reactions. The model is based on a simple plug-flow model in which the heat and mass conservation, and its associated process parameters covering the apparent rate constants of HDS, HDV, HDNi, and HDAsph reactions, process conditions, catalyst system configurations, feed properties, and maximum values of metal-on-catalyst were considered. The model was verified with pilot-plant accelerated life test run data using Boscan crude as the feedstock at an elevated reaction temperature of 663 K. A catalyst system, containing three different types of catalystssa HDM, HDS, and HDS/HDN catalystswas packed into two reactors and operated in a constant-temperature mode. The simulations from the refined model that included the hydrothermal reactions achieved a marked improvement over those of the original model.1 The hydrothermal noncatalytic reactions were found to be important when catalyst activity had been substantially reduced or at higher reaction temperatures. The hydrothermal noncatalytic reactions became more pronounced when both conditions were met. From the model simulation, it was also realized that the maximum metal capacity of the HDM catalyst is highly influenced by the feedstock properties and processing conditions. Because it is one of the important parameters to determine the residue hydroprocessing catalyst performance and life, especially in a multicatalyst system, more research must be conducted to gain a better understanding to determine this quantity more conveniently and reliably. The simulated results from operating under a constant-sulfur mode indicate an extended period of substantial sulfur removal, which is due to hydrothermal noncatalytic reactions. Moreover, the results suggest that it will be more beneficial to operate the ARDS unit using the current catalyst system to hydroprocess Boscan crude oil at a lower SOR temperature of 648 K, in terms of catalyst life, energy requirements, and quality giveaway.
Acknowledgment. The authors acknowledge the financial support of the Kuwait Institute for Scientific Research (KISR); the
1364 Energy & Fuels, Vol. 20, No. 4, 2006 Petroleum Energy Center (Japan); Japan Cooperation Center, Petroleum (JCCP); and the Ministry of Economics, Trade, and Industry (Japan). The authors also acknowledge the support of Kuwait National Petroleum Company in supplying the required petroleum feedstocks for the pilot-plant study on the performance of selected catalysts, as well as for their useful discussion of the results. Finally, special appreciation is due to the Petroleum Research and Studies Center pilot-plant staff for supplying the experimental data. This publication bears a KISR reference number of KISR 8071.
Juraidan et al. MMAX ) maximum metal-on-catalyst (g/100 g cat.) n ) correlation coefficient or reaction order (-) N ) total number of reactors (-) N ) total number of catalyst types (-) rp ) initial pore radius (m) r( ) ) rate of reaction function (-) R ) gas constant (kcal kmol-1 K-1) s ) catalyst activity (-) Sf, Sp ) sulfur content in feed and in product (wt ppm) t ) on-stream time (h) T ) process temperature (K) To ) initial reaction temperature (K) uG ) linear gas flow rate (m h-1) uL ) linear liquid flow rate (m h-1) VL ) volumetric flow rate of liquid (m3 h-1) w ) weight of deposited metal (kg) W ) total weight of the catalyst system (kg) WAB ) weight-average bed (-) z ) reactor axial coordinate (m) Z ) parameter for unconverted metals (-) ZRj ) dimensionless axial coordinate of the jth reactor Subscripts i ) species index (1 for sulfur, 2 for vanadium, 3 for nickel, and 4 for asphaltenes) j ) reactor index, between 1 and N Greek Symbols R ) deactivation constant R ) correlation coefficient for a hydrothermal processing reaction constant ) correlation coefficient for a hydrothermal processing reaction constant -H ) heat of reaction (kcal mol-1) ti ) time at ith interval (h) 3 -1 G ) gas holdup in catalyst bed (m kg ) 3 kg-1) ) liquid holdup in catalyst bed (m L ) relative catalyst activity (-) ) normalized o as defined in eq 14 (-) o ) Thiele modulus based on the bulk conditions (-) z ) Thiele modulus based on conditions in deactivated pores (-) ) deactivation function FG ) gas density (kg m-3) FL ) density of oils (kg m-3) i ) a decay function for species i ) cumulative feed-to-catalyst ratio (m3 kg-1)
EF0504265
Nomenclature
a ) normalized activity (-) A ) deactivation correlation coefficients (-) cp, cf ) concentrations of product and feed, respectively (wt % or wt ppm) C ) the actual amount of carbon on the catalyst C1, C2 ) concentrations of asphaltenes and products from the cracked asphaltenes species Cmax ) the maximum carbon deposition of a complete monolayer CF ) the concentration of the foulant CL ) concentration in the oil feed (wt % or wt ppm) CAsph ) asphaltenes concentration (wt %) CS ) sulfur concentration (wt %) CNi,CV ) nickel and vanadium concentrations, respectively (wt ppm) Cf,i, Cp,i ) feed and product concentrations, respectively (wt % or wt ppm) Cf1, Cf2 ) parallel and series fouling precursor concentration, respectively (wt % or wt ppm) Co ) the amount of coke on the catalyst (wt %) CpG ) heat capacity of gas (kcal kmol-1) CpL ) heat capacity of liquid (kcal kmol-1) Dz ) diffusivity in the deposited layer (m2 h-1) Efi ) activation energy for parallel or series coking reaction (kcal kmol-1) fi( ) ) function of ( ) FL, FG,j ) mass flow rate of gas and liquid, respectively (kg m-2 h-1) k0 ) initial reaction rate constant in the active sites of fresh catalysts kz ) reaction rate constant in the deposited layer k o,i, ko,i ) apparent rate constants for species i in the initial rapid deactivation period and reported kinetics, respectively kf1, kf2 ) apparent rate constants for parallel and series coking, respectively K ) adsorption constant LHSV ) liquid hour space velocity (h-1) Mi ) metal-on-catalyst (g/100 g cat.)