Filippo Gorra
Praoil srl
Milan
Giorgio Scribano
Praoil srl
Gela, Italy
Preben Christensen, Karin Vibeke Andersen, Osvaldo Gaetano Corsaro
Haldor Topsoe A/S
Lyngby, Denmark
Prompted by decreasing catalyst activity and unit P run lengths, an Italian refiner made several modifications to its coker gas oil desulfurization unit equipment, catalyst, and operations.
Results of the project include improved catalyst activity at start-of-run, increased unit capacity at end-of-run, and improved plant economics.
BACKGROUND
Praoil srl was experiencing deteriorating cycle lengths in the desolforazione flussanti (hydrodesulfurization, or HDS) unit at its Gela, Italy, refinery.
Praoil asked Haldor Topsoe A/S and an engineering consultant to make recommendations for improving unit performance. Prior to submitting its recommendation, Haldor Topsoe performed a pilot-plant test to select the most suitable catalyst system for the unit.
The unit was commissioned in May 1988 with the proposed catalyst loading and reactor modifications. The continuous operation of the unit since that time has significantly reduced operating costs.
PROCESS DESCRIPTION
Despite the fact that the unit processes highly unsaturated feedstocks, which cause exotherms of 60-70 C. (108-126 F.), it was designed with a single reactor and a single catalyst bed. The liquid feed is preheated by heat exchange with various streams within the unit and is brought to the required temperature in a fired heater. The treat gas (combined makeup and recycle gas) is preheated using only heat exchange, then mixed with the hot liquid feed just upstream of the reactor.
The reactor effluent is cooled then separated in a high-pressure, high-temperature separator. The gas from this separator is further cooled and separated in a high-pressure, low-temperature separator.
The gas from the high-pressure, low-temperature separator is scrubbed in a monoethanolamine (MEA) wash, then recycled back to the reactor. The liquid from this separator is depressurized and separated in a low-pressure, low-temperature separator.
The liquid from the low-pressure, low-temperature separator is then mixed with the liquid from the high-pressure, high-temperature separator and fed to the product stripper column.
The reactor is operated at a total pressure of 65-70 kg/sq cm gauge (925-995 psig) and a liquid hourly space velocity (LHSV) of 0.9-1.2 hr-1. The treat-gas-to-oil ratio is typically 1,000 normal (N) cu m/cu in oil (5,900 scf/bbl oil), which equals an H2-to-oil ratio of 675 N cu in H2/CU M OR (4,000 scf H2/bbl oil).
The unit has been used primarily to treat heavy, coker gas oil (HCGO) and a blend of light coker gas oil (LCGO) and straight-run gas oil (SRGO) in "blocked out" operation (or, in other words, separately). The hydrotreated HCGO was used as feedstock for the fluid catalytic cracking unit (FCCU), and the hydrotreated LCGO/SRGO blend is used to produce diesel fuel.
Typical properties of the two feedstocks are shown in Table 1.
OPERATIONAL HISTORY
Fig. 1 shows changes in relative catalyst activity during periods of operation on HCGO feedstock for the five unit runs between 1981 and 1987.
During the first run using fresh NiMo catalyst, the typical feed rate was 45 cu m/hr (6,800 b/d). This equates to an LHSV of 0.53 hr-1. A feedstock other than HCGO was processed in 1981-82, which explains the portion of the graph missing during that time.
As Fig. 1 shows, the initial processing of HCGO resulted in significant catalyst deactivation. More activity than was lost, however, was recovered during periods when other feedstocks were processed. The first run lasted a little longer than 2 years.
During the second run using fresh CoMo catalyst, the typical feed rate was 65 cu m/hr (9,800 b/d). This is equivalent to an LHSV of 0.76 hr-1.
HCGO was processed about 30% of the time during the second run (1983-84). The increase in activity during the first part of this run may have indicated that the catalyst was not presulfided properly.
After the catalyst activity stabilized, the deactivation rate was calculated to be about 0.3 C./month (0.91 F./month). The second run lasted about I'/2 years.
For the third run (1985-86), the regenerated NiMo catalyst from the first run was reinstalled. The typical feed rate was similar to that during the second run, but HCGO was processed about 40% of the time.
As in the second run, an increase in catalyst activity was seen during the first part of the run, which again could indicate insufficient presulfiding. A cause has not been found for the sudden decrease in activity at the beginning of 1985.
After stabilization of the catalyst activity, the deactivation rate was calculated to be approximately 0.6 C./month (1.1 F. /month). The third run lasted about 2 years.
In the fourth run, the regenerated CoMo catalyst from the second run was reinstalled. The typical feed rate was increased further to 90 cu m/hr (13,500 b/d). This feed rate is equal to an LHSV of 1.05 hr-1, and the unit was run on HCGO about 50% of the time.
The catalyst activity was very low from start-of-run and the deactivation rate was as high as 3 C./month (5.4 F./month). The run therefore lasted only, about 6 months. A new NiMo catalyst was then installed in the fifth run. The feed rate typically was maintained at 90 cu m/hr (13,500 b/d), but the amount of HCGO was reduced slightly, to include only about 46% of the run days.
The catalyst deactivated very quickly from the beginning of the run, reaching the level of the previous charge after only about 2 months. It was suspected that this unsatisfactory performance was caused mainly by bad flow distribution in the catalyst bed, and the catalyst was therefore dumped and reloaded. This, however, did not improve performance.
Three samples of the spent catalyst from the fifth run were analyzed to determine the cause of the severe deactivation. The three samples were found to be very similar in terms of carbon, vanadium, iron, and silicon. All samples contained about 0.2 wt % each of vanadium and iron and 0.4-0.7 wt % silicon.
These contaminant levels normally would not have significantly, influenced catalyst activity. The amount of coke present, however, was 13-15 wt % for all samples. Laboratory studies of hydrotreating catalyst coke deactivation have shown such high levels of coke to reduce catalyst activity by 60-80% .1 2
Based on the previous two runs, Haldor Topsoe and the engineering consultant were asked to provide recommendations to improve the unit's performance. Haldor Topsoe carried out bench-scale testing to investigate the performance and deactivation rate expected for this unit.
CATALYST TESTING
The catalyst tested in the laboratory unit was Topsoe's standard NiMo catalysts TK-551. Catalyst properties are listed in Table 2.
The properties of the HCGO on which the catalyst was tested are given in Table 3. As can be seen, these properties are similar to the typical feedstock properties given in Table 1.
The test was performed in a 50-ml isothermal bench-scale reactor with once-through hydrogen in down-flow mode. The catalyst was diluted by 25 vol % with inert glass beads (0.2-0.3 mm OD) to improve liquid distribution in the reactor.
The test was carried out in three parts:
- Operation for about 100 hr at conditions close to those used in the desulfurization unit (Condition A)
- Operation for about 350 hr at high space velocity and high temperature to accelerate deactivation (Condition B)
- Operation for 200 lay at initial conditions to measure the extent of the deactivation (Condition A).
The two sets of conditions and a summary of the test results are given in Table 4. The partial pressure of hydrogen applied during the test corresponds approximately to that used in the desulfurization unit.
The test temperature is equal to the estimated start-of-run weighted-average temperature for operation on HCGO. (The estimated start-of-run reactor inlet temperature is 306 C., or 583 F., and the estimated start-of-run reactor outlet temperature is 380 C., or 716 F.) To compensate for the test reactor's lower gas linear velocities - which can lead to liquid maldistribution - the H2-to-oil ratio was 50% higher than in the desulfurization unit.
Topsoe's experience with a wide range of atmospheric' vacuum, and cracked gas oils indicates that the kinetics of desulfurization are best described as combined second-order/first-order kinetics. Second-order kinetics are assumed to apply for the removal of 84% of the sulfur, and first-order kinetics for higher conversions.
The rate constants calculated using these kinetics are given in the column in Table 4 headed K,il.
For hydrodenitrogenation (HDN), first-order kinetics were assumed to apply.3
Table 4 also includes normalized rate constants in the column headed
These rate constants are calculated at constant temperature and pressure (375 C., or 707 F., and 50 atm, or 735 psi) using the Arrhenius expression. The calculation assumes activation energies of 30 kcal/mole for HDS and 24 kcal/mole for HDN, and a pressure dependency of 0.4.
To find the deactivation rate, the initial data points of the test were omitted and a regression analysis was performed on both the HDS and HDN reactions. The conclusion was that much better operation would be possible with the new catalyst.
Test results are shown in Fig. 2, together with a regression line of all data subsequently obtained for TK-551 in the industrial unit. The test predicted actual operation very well. The HDS deactivation rate calculated from test data is a little higher than for the industrial data. The test shows that, by careful planning of bench-scale tests, it is possible to predict actual industrial operation reliably.
UNIT MODIFICATIONS
Based on the recommendations, a chimney tray was installed to improve liquid distribution in the catalyst bed. (Previously only a baffle tray had been placed on the reactor inlet to distribute the liquid.1) Furthermore, Topsoe suggested that a top layer of special ring-shaped catalyst be installed to further improve liquid distribution.
As mentioned, there had been indications that previous charges had not been presulfided properly, thus resulting in inferior performance from start-of-run. The special nature of the feedstocks available at the Gela refinery imposed additional high risk of coke formation on the fresh catalyst because of the olefinic and polynuclear-aromatic (PNA) components. Olefins are present, even in the straight-run gas oils, because of the injection of coker gas oils into the oil wells to dilute the heavy Gela crude oil.
A special start-up procedure was therefore recommended by Topsoe to improve presulfiding without excessive coke formation on the fresh catalyst. A short summary of the procedure - based on many years of research at Topsoe's laboratories-follows.4
- A diesel-range hydrotreated gas oil is used for presulfiding, thus eliminating the risk of coke formation caused by olefins and PNAs in the untreated gas oils.
- The hydrotreated gas oil is spiked with dimethyldisulfide (DMDS) to produce a total feed sulfur content of at least 1 wt %.
- The recycle-gas MEA wash is bypassed to secure a high hydrogen-sulfide partial pressure in the reactor during presulfiding.
- A final catalyst sulfiding is carried out at a reactor outlet temperature of 350 C. (662 F.) to ensure optimal presulfiding.
- The use of hydrotreated gas oil feedstock is continued for 24 hr after completion of presulfiding before introducing coker gas oil into the unit.
The conclusion drawn from Topsoe's research is that the catalyst is not fully sulfided if presulfiding is stopped at too low a reactor temperature. Topsoe therefore recommended that the final sulfiding be carried out at a higher reactor temperature (350 C.).
Because the unit was designed for a reactor exotherm of 60-70 C. (108-126 F.), the maximum reactor temperature obtainable with the DMDS-spiked hydrotreated gas oil was about 350 C. (662 F.).
CATALYST LOADING
The new charge of TK-551 NiMo catalyst was installed in the unit in April 1988. The total catalyst quantity was about 58 metric tons (128,000 lb). As mentioned, a range of different particle sizes and shapes of catalyst were installed in layers, to improve liquid distribution. AU catalysts were sock loaded in the configuration given in Table 5.
A 6-in. deep layer of 1/8-in. cylinders of TK-551 was installed in the bottom of the reactor bed to prevent the 1/16-in. cylinders in the next layer from migrating down into the support layer of 1/4-in. inert ceramic bags.
The TK-10 used as a top layer is a special "hold-down" material that loads with a high void fraction. This material replaces the ceramic balls normally used as hold-down material in hydrotreaters. Below the TK-10 layer, the active, ring-shaped catalyst was installed. The placement of these rings was designed to improve liquid distribution in the catalyst bed and provide storage capacity for deposits that could otherwise cause pressure-drop problems.
Compared to, for instance, ceramic balls, the TK-10 and 3/16-in. TK-551 rings tend to produce better radial spread of liquid in the catalyst bed. The top two layers will therefore even out a possible non-ideal liquid distribution in the top of the catalyst bed and ensure better utilization of the bulk catalyst.'
CATALYST HISTORY
The unit was brought on stream in May 1988. To date, the average HCGO feed rate during the run on TK-551 is 83 cu m/hr (12,500 b/d) equivalent to an LHSV of 0.97 hr - 1. HCGO has been treated about 39% of the run days.
Beginning year-end 1990, HCGO is no longer being processed in the hydrotreater because of the commissioning of a new gas oil finer, which now treats the HCGO. Since that time, the hydrotreater has been used solely for treatment of the LCGO/SRGO blend. The relative catalyst activity as a function of time is shown in Fig. 3 for HCGO and in Fig. 4 for the blend.
In Fig. 3, the activity is compared to that of the previous catalyst charge. The relative catalyst activities of both HCGO and the LCGO/SRGO blend indicate that presulfiding has improved, compared to previous runs. The current run has exhibited high activity from start-of-run and no significant signs of increasing activity during the early part of the run.
When operating on HCGO (Fig. 3), the relative catalyst activity was reduced from about 80 to about 60 between June 1988 and August 1990 (a 26-month period). This activity reduction equals about 7 C. (13 F.) and therefore indicates a deactivation rate of about 0.3 C./month (0.5 F./month).
When processing the LCGO/SRGO blend, the relative catalyst activity decreased from about 95 to about 25 between June 1988 and August 1992 (a 50-month period). This reduction is equivalent to about 33 C. (59 F.), which equates to a deactivation rate of about 0.7 C./month (1.3 F./month).
The deactivation rate probably was lower during the treatment of HCGO, as compared to operation on the blend, because the unit was run in blocked-out operation on the two feedstocks. Some of the catalyst activity lost during HCGO operation was recovered during the periods when the blend was used. When analyzing only the relative Catalyst activity when running HCGO, the apparent deactivation rate will therefore appear to be less than if only HCGO had been treated.
ECONOMICS
The implementation of these modifications has improved the unit operation, especially when compared to the previous two runs. As compared to the runs during the first half of the 1980s, the major improvements are in terms of a higher feed rate and a higher percentage of HCGO processed. The smooth operation of the unit, together with a high catalyst activity right from the beginning of the run, has improved plant economics significantly. Furthermore, the installation of the ring-shaped catalyst reduced the pressure drop toward the end of the run, resulting in increased plant capacity. Table 6 shows a comparison of operating costs, including estimates of the value of lost production, for these three runs:
- The run on regenerated NiMo catalyst from year-end 1984 to year-end 1986. This was a fairly good run-somewhere between the worst and best runs (Fig. 1).
- The run on fresh NiMo catalyst from mid-1987 to mid-1988. This run was the worst run experienced in the unit (Fig. 2).
- The run on TK-331 from mid-1988 to present. This run is the best run experienced on the unit to date (see Figs. 2 and 3). The comparison in Table 6 is based on several economic indicators:
- Total production
- Number of turnarounds during run
- Value of lost production
- Cost of unit turnaround (catalyst handling, DMDS, etc.)
- Cost of catalyst
- Benefit of Increased capacity.
The average value of lost production was estimated based on a downtime of 10 days/turnaround. The production loss is calculated to be 2,000 tons/day at $39/ton.
The average cost of a unit turnaround is based on an average catalyst-handling cost of $42,500 and a DMDS cost of $18,-;00. The average catalyst cost is based on a unit price of $5/kg of catalyst ($2.27/lb). The total increased capacity at end-of-run has been estimated to be worth $400,000. The values for Run C are given as greater than or less than because the run is not yet completed.
REFERENCES
- Wivel, Per, Zeuthen, Per, and Jacobsen, Andreas C., "Initial Coking and Deactivation of Hydrotreating Catalysts by Real Feed," Studies in Surface Science and Catalysis, Vol. 68, pp. 2-;7-64.
- Zeuthen, Per, Progress Report No. 2, under EEC Contract No. EN3C-0026-DK(B), Coke Deactivation of Catalysts for Hydroprocessing of Heavy Petroleum Feedstocks, pp. 34-40.
- Christensen, Helle, and Cooper, Barry H., "The influence of catalyst and feedstock properties in FCC pretreatment," AlChE Spring National Meeting, Paper No. 44b, 1990.
- Zeuthen, Per, Blom, Peder, Muegge, Brian, and Massoth, F. E., "Temperature-programmed Sulfidation and Oxidation of NiMo/Alumina Catalysts and Reaction with Ammonia," Applied Catalysis, Vol. 68, 1991, pp. 117-30.
- Jameson, G. J., "A Model for Liquid Distribution in Packed Columns and Trickle-bed Reactors," Transactions, Institution of Chemical Engineers, Vol. 44, 1966, pp. 198-206.
Copyright 1993 Oil & Gas Journal. All Rights Reserved.
