[29b] - A User's Guide to the Chemistry, Kinetics and Basic Reactor Engineering of Hydroprocessing

Woodrow K. Shiflett, Advanced Refining Technologies L.P. Houston, Texas 2002 Advanced Refining Technologies L.P. Prepared for Presentation at the AIChE Spring Meeting March 2002, Session 29, Tutorial: Desulfurization Unpublished AIChE Shall Not Be Responsible For Statements or Opinions Contained in Papers or Printed in its Publications

Introduction In the last decade, desulfurization of petroleum feedstocks has become critical to ensuring that refiners worldwide are able to conform to the ever-tightening transportation fuels regulations put in place to enable better air quality and a cleaner environment in the future. While the science of catalysis and reactor engineering technology has made profound progress in the same period, the actual practice of operating refinery hydrodesulfurization units must by necessity meld catalytic science, practical reactor engineering and a long history and lore of empirical practice. The practicing process engineer, through the rapid growth and application of data historian system, is deluged with a myriad of operating and analytical data that all too often is of dubious quality and suffers from the high degree of variability in feedstock quality often driven by refining economics. The questions to be addressed include: What does it all mean? How is it all to be best addressed and interpreted? How can all the data be compiled in order to enable optimal catalyst system choices in the future? What are the elements to consider when unit operation is failing to meet objectives? Hydrodesulfurization, HDS for short, is accompanied to varying degrees by concurrent reactions that also result in the removal of nitrogen compounds, saturation of unsaturates, removal of contaminant metals and some molecular weight reduction. Therefore a tutorial of desulfurization must address elements of hydroprocessing in general. The intent here is two-fold: to provide a useful set of tools and information for the practicing engineer (or those aspiring to be) and to inspire those working at the cutting edges of catalysis and reactor engineering to continue to strive to bridge the gap between elegant theory and sometimes ugly practice.

Scope This tutorial is comprised of seven basic sections: Introduction to hydroprocessing General chemistry Catalysts Process overview Kinetics and reactor engineering Some rules of thumb Summary

Owing to the scope limitations established in advance, no attempt will be made to review the latest developments in hydoprocessing catalysis or reactor engineering. The reader is encouraged to seek out reviews and specific articles of interest in the literature. Hydrocracking, the specific area of hydroprocessing that focuses on the significant reduction of molecular weight in order to convert appreciable quantities of lower-value, heavier hydrocarbons into more highervalue, lighter hydrocarbons is also outside of the scope of this tutorial. Introduction to Hydroprocessing The complete area of hydroprocessing includes the broad categories of hydrotreating and hydrocracking. These categories are by no means exclusive and significant hydrotreating is required to enable hydrocracking and occurs during it. Similarly, severe hydrotreating can be accompanied by significant levels of hydrocracking, although such phenomena are often dubbed mild hydrocracking. Unfortunately, across the industry, the semantics can be loose and illdefined. Our focus here is predominantly hydrotreating. Briefly, hydrotreating can be defined as the removal of petroleum stock contaminants or otherwise unwanted components at elevated temperatures and pressures in the presence of a hydrogen-rich gas phase and a heterogeneous catalyst system. In some cases, desirable compounds may be converted as well. In general, the classes of materials removed or activities occurring encompass the following: Sulfur and nitrogen Olefin saturation Metals (vanadium, nickel, iron, silicon, etc.) Aromatics and polynuclear aromatic (PNA) compound saturation High molecular weight compounds (coke precursors, carbon residue)

Most units involve down-flow of both liquid and gas phases over solid catalysts. Most units are described as being in trickle phase operation, although this is by no means universal, nor should it be assumed the case without validation. Pulse flow of the liquid can also be encountered. Another exception is naphtha hydrotreating which will typically be conducted 100% in the gas phase over solid catalysts. It is instructive to look at the typical coking and cracking conversion refinery and note how hydrotreating is integral to the processing of many refinery streams. Figure 1 is an example.

Figure 1. How Hydrotreating Fits into the Typical Coking/Cracking Refinery

General Chemistry Hydrotreating reactions are carried out in a matrix consisting of many, many hydrocarbons. Table 1 summarizes the basic types of compounds encountered. A brief discussion of the roles that these compounds play in refining gives perspective to the hydrotreating reactions. Paraffins in gasoline-boiling ranges, they are a low octane component. In diesel-boiling ranges, they provide good cetane but also can contribute to unsatisfactory cloud and pour point properties. Iso-paraffins provide good octane in gasoline range, and reasonable cetane in the diesel range. Olefins (mono) are high octane components in the gasoline range. They are found predominantly in streams coming from coking operations and the FCC. Olefins (di) are undesirable owing to their tendency to foul process equipment, as well as catalysts, from condensation/polymerizations reactions. Their highest concentrations arise in coker stocks, but they are also found in FCC products.

Table 1. Typical Hydrocarbons in Hydrotreating Applications

Class Paraffins Example
CH3-CH2-(CH2)x-CH3 R |

Iso-Paraffins

CH3-CH-(CH2)x-CH3

Olefins (mono)

CH3-CH=CH-(CH2)x-CH3

Olefins (di)

CH2=CH-(CH2)x -CH=CH2

Naphthenes provide poor octane in gasoline but marginal cetane in diesel stocks. Aromatics (mono) provide excellent octane in gasoline and are found in high concentrations from FCC units. Aromatics (di) those found in the gasoline range provide good octane while they are a poor contributor to cetane in the diesel range. Aromatics (poly) promote deactivation of the catalyst, referred to as coking of the catalyst surface, as well as fouling of the catalyst bed.

This roughly describes the hydrocarbon matrix in which the desulfurization reactions take place. Concurrent reactions unavoidably take place with a number of these compounds as hydrotreating proceeds: olefins are saturated, some aromatic compounds are saturated, modest amounts of isomerization are possible. Petroleum-derived streams contain a myriad of sulfur compounds. A short summary of the principle sulfur compounds encountered in refinery streams is given in Table 2. The relative amounts of each in any given refinery stream is dependent upon the crude sources involved and the nature of up-stream conversion units. For example, benzothiophenes and dibenzothiophenes are somewhat more common in coker-derived stocks and quite common in FCC products.

Table 2. Principle Sulfur Compounds in Refinery Streams

The hydrodesulfurization of these various sulfur compounds proceeds with differing degrees of reactivity. As Figure 2 illustrates, mercaptans submit to easy sulfur removal, while thiophenic sulfur compounds are more difficult. In either case, hydrogen is consumed and H2S is produced as a by-product. The H2S is a poison for most hydrotreating reactions. HDS reactions are exothermic as are concurrent reactions such as hydrodenitrogenation (HDN), saturation of olefins and aromatics.

Figure 2. Basic HDS Reactions

Harder 4H2

As environmental regulations drive fuel sulfur levels to dramatically lower levels, the nature of specific sulfur compounds in feedstocks becomes more and more important in understanding how hydrotreating units -- either existing, those slated for a revamp or new units in design -- are to be able to reliably meet product sulfur requirements. This is particularly true in the area of ultra-low sulfur diesel (ULSD) production where unit sulfur levels will in the future need to be ca. 10 ppm or less in the U.S. and even lower in the EU later this decade. There are also implications for catalyst selection which will touched upon in the next section. Figure 3 shows this importance. Classes of compounds that were ultimately required to be removed for 500 ppm sulfur diesel by current regulations unsubstituted dibenzothiophenes (DBTs) or lowsubstituted DBTs, proceeded most effectively by C-S bond scission without saturation of an adjacent aromatic ring. Such reactions are generally most easily catalyzed by cobaltmolybdenum (CoMo) based catalysts. As lower sulfur levels become required in diesel-boiling feedstocks, the less-reactive, more difficult sulfur species need to be removed. These are multi-substituted DBTs in which alkyl groups tend to make access to the sulfur atom difficult with respect to the catalyst surface. 4,6-Dimethyldibenzothiophene (DMDBT) is one such compound that has been studied somewhat extensively recently. Removal of sulfur from these types of compounds dictates that saturation of an adjacent aromatic ring be required first, to essentially make the molecule more flexible and enable blocking alkyl groups to move and permit access to the C-S bonds for catalytic activation. Such saturation activity is much more easily facilitated by nickel-molybdenum (NiMo) based catalysts. Figure 4 provides a graphical perspective on the 4,6-DBT molecule.

Figure 3. HDS of Dibenzothiophenes

Reaction Pathways:
500 ppm Sulfur Diesel <50 ppm Sulfur Diesel
CoMo
S
-H S 2 +H 2

NiMo
Very Difficult
+H 2

S
-H S 2

CoMo
+H 2

NiMo
Hard
+H 2

Extremely Difficult

S
-H S 2

S
-H S 2

Figure 4. 4,6-Dimethyldibenzothiophene

Methyl groups hinder the sulfur atom from accessing the active catalyst site.

Figure 5 shows the distributions of sulfur compounds in several commonly-encountered diesel feedstocks. The C1 and C2 nomenclature indicates the number of carbons in alkyl constituents on the core molecular structure (BT or DBT). The SR diesel is a straight-run stream that has come directly from a crude column without any other processing. The LCO is light cycle oil, which is the diesel-boiling range product from an FCC unit. The HT diesel is a diesel stream that has been subjected to some prior hydrotreating and is a draw stream from the fractionator of a heavy oil hydrotreater. A few comments are in order to show the complexity that is possible. SR Diesel A and SR Diesel B differ from one another in that the former has a higher distillation boiling range and endpoint. Thus, more difficult sulfur species are present in higher levels. The LCO stream has a very high level of the most difficult to process sulfur compounds. This is a consequence of the FCC conversion process itself. While the HT diesel

has relatively lower levels of sulfur than the other streams, virtually all of the remaining sulfur compounds are the most difficult to treat category. Figure 5. Sulfur Species Distributions in Diesel Stocks

4,500 4,000 3,500 Concentration, ppm 3,000 2,500 2,000 1,500 1,000 500 0

Sulfur Speciation of Diesel Stocks

Gasoline Range Benzothiophene C1-BT's C2-BT's C3-BT's C4+-BT's Dibenzothiophene
LCO SR Diesel A SR Diesel B HT Diesel

C1-DBT's C2-DBT's 4,6-DMDBT C3+-DBT's

Catalysts Most catalysts used for hydrotreating are based on active components of either nickel + molybdenum (NiMo) or cobalt + molybdenum (CoMo) oxides dispersed on a high surface area, highly porous alumina, or in some cases, silica-alumina substrate. The substrate is also referred to as the carrier or catalyst support. -alumina is a common support and final catalyst surface areas are typically in the range of 200 300 m2/g. (For perspective, note then that a handful of catalyst has the approximate surface area of a U.S. football field.) Nickel-cobaltmolybdenum oxide formulations are also encountered, and tungsten oxide is a component in certain specialty catalysts used in lube applications or for aromatics saturation. For exceptionally low sulfur and nitrogen feedstocks, noble metals such as platinum and palladium are employed for selective saturation of unsaturated compounds. In external appearance, hydrotreating catalysts will be approximately sized in standard diameters ranging from as low as 1 mm (1/25 inch) to more common sizes such as 1.3 mm (1/20 inch) and up to 3.2 mm (1/8 inch) with typically length/diameter ratios falling in the 1:1 to 2:1 range. Catalyst size selection is often dictated by concern (or lack of) for excessive pressure drop. Owing to the need for catalyst pellets to be robust and able to withstand up to 10 15 m (30 45

feet) of catalyst bed weight above them, most pellets are solid geometric shapes having either three or four rounded lobes or are cylindrical in shape. Hollow cylinders and other open geometries do find some application in the tops of catalyst beds where high void fraction is beneficial in slowing pressure drop build up from contaminant solids and deposition of carbonaceous materials. The levels of the catalyst metals components typically lie in the following ranges: Molybdenum 5 20 % wt. metals basis Cobalt or nickel 1 3 % wt. metals basis Tungsten somewhat higher than molybdenum

Other additives or promoters will be found in catalysts. Phosphorus is one common promoter encountered. These are typically proprietary to the catalyst suppliers. Under process conditions, hydrotreating catalysts are active in the sulfided state of the active metals components. The sulfiding of the metals oxides must be done under controlled conditions in the reactor, preferably using specific sulfur containing compounds as agents such that heat release can be controlled. This is denoted as in situ sulfiding. Some common examples of these agents, although by no means inclusive, are dimethyl sulfide (DMS) and dimethyldisulfude (DMDS). A number of polysulfide compounds are also in common use. Some service providers will treat hydrotreating catalysts with sulfur compounds that remain adsorbed or otherwise bound to the catalyst in order to simplify the sulfiding process and potentially speed up the unit start up. This is referred to ex situ sulfiding, although complete conversion to the metals sulfides does not typically occur until the recommended start-up process is followed in the reactor vessel. Much catalysis work has been done over the last two decades to understand exactly how hydrotreating catalysts function. A central focus has been just what exactly constitutes an active site for the reaction of interest. It is believed that an active site is associated with cobalt or nickel atoms that are decorating the edges of flat MoS2 slabs on the carrier surface. The goal, then, in using hydrotreating catalysts, is to maximize the number of active sites on a high surface area component in a reactor volume. This is by no means simple, and a high degree of both science, development effort and art are employed. There is no unambiguous means to count active sites, much less specify how to optimally distribute them across a substrate surface in manufacturing processes at the present. Furthermore, maximum dispersion of the active oxides does not insure the final sulfides will be likewise dispersed. It is essential that catalyst users work closely with their catalyst technology suppliers to match the correct catalyst to the application. While properties such as metals loading, surface area and pore volume are important for good performance, it is ultimately the suppliers databases of pilot plant and commercial performance and their modeling correlations upon which they are based that is the measure of the final catalyst performance. In Figure 6, an attempt is made to visualize the high level of molecular traffic and activity that is occurring on a microscopic scale on a catalyst in a hydrotreating application. Active sites are dispersed in the pores of the carrier, where virtually all of the surface area is found. Typically, such pores that are effective will be in the range of 6 15 nm. Pores much smaller are prone to blockage by coke, the catch-all term for carbonaceous deposition products arising from

undesirable polymerization or condensation/dehydrogenation reactions involving such compound as polynuclear aromatics and diolefins. This pore blockage is particularly the case also when heavier hydrocarbons containing nickel and vanadium metals are treated. Smaller pores also can invite more severe diffusion limitations for larger molecular species. On the other hand, significantly increasing pore size results in loss of surface area, and ultimately catalyst activity is sacrificed. One compromise in designing hydrotreating catalysts is that of balance between surface area and pore size. Nonetheless, the deposition of coke remains a prevailing and unavoidable means of catalyst deactivation in most applications.

Figure 6. A Hydrotreating Catalyst in Action

H2

6 15 nm

The current scope does not permit time to discuss hydrotreating catalysts in greater detail. The reader is encouraged to seek further information in the non-proprietary literature or to seek out application information from the hydrotreating catalyst technology suppliers. Process Overview While other parts of this tutorial session cover process aspects in greater detail, a brief overview is in order to provide consistency in terms for subsequent information and to set the stage for later rules of thumb discussion. While it is difficult to select a typical hydrotreating unit for

discussion, Figure 7 embodies most features encountered. All hydrotreating units will have one or more reactor vessels that themselves may include one or more beds of catalyst. Feed pumps and gas compressors are required to elevate the reacting components to the appropriate pressure. After feed/effluent exchanger trains, reactant liquid and gases are heated to reaction temperatures by either two-phase furnaces or separate single-phase furnaces. Depending upon the heat released in the process at hand, quench of cold recycle gas may be provided either between reactors in a multi-reactor plant or between beds in a multi-bed vessel, or both. Some units, such as naphtha HDS units may not have a recycle compressor and use only single-pass gas flow. The make-up compressor is required to supply hydrogen-rich gas to accommodate the unit hydrogen consumption from the various hydrotreating reactions as well as to help maintain adequate hydrogen partial pressures in the unit needed to keep catalyst deactivation by coking within design or processing constraints. Owing to the significant amounts of hydrogen sulfide and ammonia produced by hydrotreating reactions, water wash of the reactor effluent is frequently employed to lessen downstream corrosion potential. Further cooling prior to gas-liquid separation may be facilitated by an air cooler. One or more separators are used for phase separation. Most major units will have a product fractionator for directing products to storage as appropriate or to downstream processing units. A gas scrubber on the recycle gas loop is an option for high severity units in which the inhibiting influence of H2S on the catalyst system cannot be economically tolerated. As ULSD and other applications for ultra-low sulfur fuels become more common, likewise gas scrubbers will become more of a requirement as well.

Figure 7. General Flow Diagram of a Typical Hydrotreating Plant

HP SEPARATOR

Typical operating conditions for hydrotreating units are summarized in Table 3a (U.S. units) and Table 3b (metric units) for both prior experience as well as more recent demands for such units in the ULS fuels area. The liquid hourly space velocity, LHSV, is defined as the volume of oil flow at standard conditions per volume of active catalyst per hour. LHSV is therefore inversely proportional to the catalyst required. Two observations are clear: lighter feedstocks require lower pressure and less amounts of catalyst than heavier feedstocks, and trends driven by fuels regulations for ULS will result in larger catalyst volumes being employed and the utilization of higher pressure when revamps or new unit construction permit that option. In general, temperature ranges are limited by the prevailing metallurgy of the reactor vessels. The maximum temperature is therefore often the end-of-run (EOR) temperature at a bed outlet. In other cases, feedstock quality can limit temperature such as in the case of diesel color and color stability, or furnace duty can be a limitation. In addition, at temperatures in the range of 400 425 oC (750 800 oF) at typical unit limitations on hydrogen partial pressure, catalyst deactivation phenonema tend to become unacceptably rapid.

Table 3a. Typical Hydroprocessing Conditions (U.S. units)

RECENT HISTORY Service Naphtha Jet/Diesel FCC Feed Pretreat CURRENT TRENDS Service Naphtha Jet/Diesel FCC Feed Pretreat Pressure (psig) 200 - 500 500 - 1,200 750 - 2,000 LHSV (hr-1) Temperatures (F) 2-8 550 - 700 2-4 600 - 750 1.0 - 3.0 700 - 800

Pressure (psig) 200 - 750 500 - 1,500 1,000 - 3,000

LHSV (hr-1) Temperatures (F) 2-6 550 - 700 0.5 - 3 600 - 750 0.5 - 2.0 700 - 800

Table 3b. Typical Hydroprocessing Conditions (metric units)

RECENT HISTORY Service Naphtha Jet/Diesel FCC Feed Pretreat
-1 Pressure (kPa gauge) LHSV (hr ) Temperatures (C) 1,380 - 3,450 2-8 290 - 370 3,450 - 8,270 2-4 315 - 400 5,170 - 13,800 1.0 - 3.0 370 - 425

CURRENT TRENDS -1 Pressure (kPa gauge) LHSV (hr ) Temperatures (C) Service Naphtha 1,380 - 5,170 2-6 290 - 370 Jet/Diesel 3,450 - 10,300 0.5 - 3 315 - 400 FCC Feed Pretreat 6,900 - 20,700 0.5 - 2.0 370 - 425

Kinetics and Reactor Engineering To the hydrotreating unit process engineer, the initial question that arises in this section is What can I use basic kinetics and reactor engineering concepts to do for me? There are at least three areas that can be addressed in this respect: Use as a tool in predicting required process changes needed to maintain product specifications A means to decrease operational signal/noise ratio and assess an objective catalyst aging rate and thus enable benchmarking with prior unit cycles A means to trouble-shoot unanticipated performance and/or unexpected behavior

The beginning of this analysis is to apply the most reasonable, yet simple, model that describes the process. Downflow, trickle-phase units are anticipated to behave somewhat near to an ideal plug-flow reactor (PFR). The simplifying assumptions appropriate are as follows: A differential radial slice of the reactor is uniform across its diameter in composition, flow rate and temperature. Back-mixing of chemical species is negligible, which in plain language means no dead zones, eddy pockets or non-uniform flow anywhere. Radial uniformity is flatly assumed not to be violated despite the three-phase system present (solid catalyst, liquid hydrocarbons and hydrogen-rich gas)

A differential species balance, on species A, follows directly.

rAdV = v0dC A

Differential balance on species

rA is rate of reaction, v 0 is volumetric flow C A is concentration, V is catalyst volume

Equation 1.

Integration across the reactor volume from feed concentration CAf to product concentration CAp yields the basic equation into which the kinetics are fed:

dV dC A = v0 rA

C A p dC dV A = CA f (r ) v0 A

or

C A p dC 1 A = CA f (r ) LHSV A

Equation 2. The general kinetics expression can be simply stated:

rA = k * function(C A )
Equation 3.

The rate constant k will by necessity depend upon temperature, feedstock, hydrogen partial pressure, hydrogen sulfide partial pressure and likely other operating conditions. Owing to the heterogeneous catalytic reactions occurring, it might be assumed that LangmuirHinshelwood/Hougen-Watson (LHHW) kinetics are applied to the respective hydrotreating reactions. In the industrial refining applications, this has not been the case. Instead, first-order or nth-order kinetics have been widely used. There are several practical and historic reasons for this. They hinge upon the quality of process data, analytical speed and cost. Historically, the frequency and quality of operating and analytical data from hydrotreating units did not support the need to fit the various parameters inherent to LHHW kinetics. Detailed analyses of different sulfur species that are reasonably quick and inexpensive have not been readily available until relatively recently, so it was recognized early on that lumped nth-order kinetics was one means to simply accommodate the various sulfur specie reactivities. In this respect, the apparent order n can be expected to exhibit variations with feedstock as well as with degree of conversion. The first-order and nth-order definitions of the rate constants follow:

k = LHSV * ln

C Af C Ap
Equation 4.

k=

LHSV 1 1 n 1 n ( n 1) C Ap 1 C A f
Equation 5.

Temperature dependence is correlated by a standard, apparent Arrhenius relationship,

k = A* e

EA

RT

Equation 6. where EA is the apparent activation energy and T is absolute temperature. The ramifications of the kinetic assumptions on irreversible first or nth-order kinetics can appear somewhat daunting in light of the prior description of the chemistry involved: All chemical reactant species are assumed to have the same average reactivity, reaction order and activation energy. There are no equilibrium limitations; 100 % conversion is presumably possible. No account is made for sequential reactions or reaction intermediates. The last sulfur compound removed would appear to be no more difficult than the first.

The integration in equation 2 is carried out over an increasing temperature profile, since hydrotreating reactions are exothermic. The definition of an appropriate average temperature for the reactor catalyst system is needed to readily track operational data as well as to harmonize with isothermal data developed in pilot plants and assess relative performance of new catalysts. One of the common averages defined and in widespread use is the Weighted Average Bed Temperature or WABT defined below:

2 2 WABT = Tin + (Tout Tin ) = Tin + T 3 3

Equation 7. This may be also called the WAT, and equivalent isothermal temperature or EIT is not uncommon as well. Some usage is found of the arithmetic average bed temperature which is often called the catalyst average temperature or CAT. When average temperature is discussed here, it will assumed to be WABT. Unit WABTs are the catalyst volume weighted averages of the individual beds. In the last two or three decades, a significant amount of commercial operating data as well as pilot plant data has been treated by these simple nth-order kinetics. As interpolative tools, such handling of the data has been largely productive. Table 4 summarizes the collective, publiclyavailable history of the observed nth-order kinetics for the several boiling ranges of feedstocks processed for HDS, HDN and hydrodemetallization reactions (HDM) for nickel and vanadium removal. Table 4. Kinetic Parameters for Plug Flow
FEEDS Hydrodesulfurization (HDS) Reaction Activation Energy Order n (kcal/mole) 1.0 - 1.5 23 -27 1.4 - 1.7 1.6 - 2.0 1.4 - 2.0 25 - 27 27 - 35 30 - 35 Hydrodenitrogenation (HDN) Reaction Activation Energy Order n (kcal/mole) 1 15 1 1 1 15 27 - 33 30 - 33 Hydrodemetallization (HDM) Reaction Activation Energy Order n (kcal/mole) 1.5 - 1.7 1.0 - 1.5 14 - 20 29 - 31

Naphthas Jet - Diesel Heavy Gas Oils Residues

Corrections to the rate constants for operational changes are empirically done for the most part. The influence of hydrogen partial pressure is handled as follows:

k 2 pH 2 = k1 pH1

0.5 x 1.0 for HDS 1.0 x 2.0 for HDN

Equation 8.

The affect of catalyst treat gas (recycle + make-up) rate is handling analogously. Inhibition by hydrogen sulfide has been correlated as follows, where ko is the rate constant without H2S present:

k=

k0 pH 2 S 1+ K * p total
Equation 9.

K is approximately 20 for HDS reactions. Figure 8 shows the influence of H2S on required WABT for a fixed level of HDS. Since unit operating windows for temperatures can be often on the order of 55 C degrees (100 F degrees), it can be noted that the effect can be appreciable in terms of lost activity or cycle length.

Figure 8. H2S Inhibition on HDS Activity

H2S Inhibition on HDS Activity

60.00 50.00 Activity Loss, F
o

40.00 30.00 20.00 10.00 2% 0.00 0 20,000 40,000 60,000 80,000 100,000 120,000 H2S Level, ppm volume 4% 6% 8% 10%

A few observations on the rate constants defined in this approach are appropriate. First of all, rate constants are highly dependent upon the feedstock:

k = function( crude, % cracked stocks, boiling point, etc.)

Equation 10.

Such information is often difficult to access in the public domain. In addition, rate constants by definition are constrained to contain all of the information on catalyst activity and deactivation. As measured, they represent an averaging across the reactor system with respect to operating conditions such as hydrogen partial pressure, H2S levels and multiple beds and interbed quench. It is also of some interest to look at the influences of the two principle kinetic parameters that are fit for correlating hydrotreating reactions: apparent reaction order n and apparent activation energy EA. Several observations can be made: Lower order n means less temperature increase needed to achieve a lower product specification with the same feed concentration. Lower apparent activation energy EA forces a higher temperature increase to achieve a lower product specification, with the same feed concentration. Lower apparent activation energy EA also means higher temperature increase needed for a higher LHSV (through-put) in order to keep same product specification as before.

It becomes apparent that the reaction order n and apparent activation energy EA ought be most realistically viewed as through-put and temperature response parameters respectively. No deeper interpretation is merited. As a predictive tool, the approaches discussed can be used to project future operating requirements. Typically, the key control variable for hydrotreating units is the reactor inlet temperature which is directly related to the furnace outlet temperature. Therefore, adjustments in hydrotreater operation are usually done in the context of the WABT required for the new condition. The procedure for estimating a change proceeds as follows: Carefully select appropriate existing operating conditions (typical or average) and calculate the rate constant from these feed/product concentrations, LHSV, hydrogen partial pressure, etc. values Calculate the rate constant based on the needed future conditions and correct Told to the new temperature, Tnew:

Tnew =

1 R knew 1 ln Told E A kold

Equation 11.

The normalization of operating data is undertaken in a parallel manner to how the process change was predicted. In the normalization case, our desire is to predict what the WABT would have been at some fixed reference conditions. Clearly, this process is most easily done in a spreadsheet format. The process is undertaken as follows: Carefully select appropriate reference conditions (typical or average) and calculate the reference rate constant from these feed/product concentrations, LHSV, hydrogen partial pressure, etc. values

Calculate the rate constant based on the actual prevailing conditions and correct Tact back to the normalized temperature:

Tnorm =

1 1 R knorm ln Tact E A kact

Equation 12.

An example of a data normalization exercise is shown in Figure 9. It summarizes the first six weeks of an HDS operation and establishes that the new catalyst system is more active as indicated by the lower WABTs required at consistent conditions. Based on the data shown, it is questionable as to the extent S/N in the operating data has been improved. Figure 9. Data Normalization Example

Normalization Basis: 0.8 %wt. Feed sulfur to 300 ppm product sulfur, 1.0 LHSV at 470 psia pH2

Prior Catalyst

WABT

CK 400

10

15

20

25 Days on Stream

30

35

40

45

50

Norm. T

Actual Avg. WABT

Prior Cycle

Linear (Norm. T)

Linear (Prior Cycle)

As a trouble-shooting tool, observed kinetics have a role. Two response factors can be indicative of a problem. If the apparent activation energy is abnormally low, as in significantly less than 15 kcal/mole (ca. 4 kJ/mole), then there are several potential causes to be investigated: Severe diffusion limitations Reaction is not irreversible but equilibrium limited Non-catalytic problems such as feed/effluent exchanger leaks

If an observed reaction order significantly greater than 2 is encountered, this is one consequence of potential catalyst bed mal-distribution of the reactants and associated catalyst by-passing. It should be emphasized that such kinetic tests and observations should be carried out in the context of a full appreciation of the noise in the operating data, the time required for the unit to attain a new steady state after a change, the error limits on the analytical techniques and the application of statistical techniques to objectively assess the data.

Rules of Thumb As was stated in the introduction, the general area of hydrotreating, as well as HDS specifically, has a long history of technical lore and empirical practice. We strive to summarize the more important of these in this final section. These observations will be grouped as follows: General operation Limiting catalyst deactivation Estimating hydrogen consumption Areas to troubleshoot for performance problems

In general for hydrotreaters, the normal or well-behaved unit will typically operate with a start-of-run pressure drop in the range of 11 23 kPa/m (0.5 1.0 psi/ft) of bed depth, and an apparent mass flux in the range of 3.4 5.4 kg/m2-s (2,500 4,000 lb/ft2-hr). Operations outside of these ranges could be subject to less than optimal performance and utilization of the full catalyst activity. Good gas-liquid distribution hardware is becoming a must above catalyst beds, and top bed size, void fraction and activity grading is a requirement. The latter area is best addressed for a specific unit in collaboration with catalyst technology suppliers. Two basic means of loading catalyst are employed: sock loading and dense loading. The former uses a cloth sock to dump catalyst into a reactor vessel in a controlled manner. The latter uses rotating mechanical equipment that can be selected from a variety of sources to orient the catalyst pellets as they are projected into the reactor. The dense loading technique results in more catalyst being loaded per unit reactor volume owing to the more geometric uniformity of the pellets orientation in the catalyst bed and is preferred in almost all cases. It is one means to limit potential flow mal-distribution that could result from a poor loading of catalyst. The deposition of various metal and other inorganic contaminants (e.g. Si, Fe, Na, Ni, V) can be easily estimated as follows: Removal of 1 ppm of contaminant at 1 LHSV for a period of one year deposits on average across all of the catalyst about 1 % of the catalyst weight. The

calculation is easily scaled; 2 ppm at 5 LHSV yields ca. 10% wt. in a year. It is also typical that more reactive contaminants will be found in amounts 3 5 times the average number calculated in the top of catalyst beds. This can lead to bed plugging with resultant excessive pressure drop and possible poor flow distribution. Excessive catalyst deactivation often results from inadequate hydrogen partial pressure to hydrogenate coke precursor compounds either in the oil or absorbed on the catalyst surface. It is therefore highly recommended that recycle hydrogen gas rates be 3 4 times the estimated hydrogen consumption. Gas purity must be taken into account in setting the recycle gas rates when hydrogen purity is low in the system. A second contributor to excessive catalyst deactivation is exposing the catalyst to temperatures in excess of what is required to achieve the operating objective. When available, make use of quench even early in a cycle to minimize outlet catalyst bed temperatures. Also, when operating at mild conditions or processing an easier feed, avoid over-treating by adjusting the processing temperatures lower. The degree of temperature reduction can be estimated using the kinetic relations provided earlier. Estimating hydrogen consumption from process gas flows is often not a very accurate approach. Meter calibrations frequently drift significantly, particularly when gas compositions change appreciably, and meter maintenance may not always have the attention it deserves. Approximate hydrogen consumptions can be obtained from analytical data on feed and product liquids: Each 1 wt.% sulfur removed results in about 18 20 Nm3/m3 (110 - 120 SCFB) feed of hydrogen consumption. Each 1,000 ppmw nitrogen removed yields 5.9 6.1 Nm3/m3 (35 - 36 SCFB) of hydrogen consumption. Hydrogen consumption for aromatics removal is about 5.0 8.4 Nm3/m3 (30 - 50 SCFB) per 1 wt.% aromatic carbons removed. (Use half these values if aromatics are reported on volume % basis.) Olefin removal requires about 8 SCFB per unit bromine number reduction.

It should be noted that lack of analytical data on aromatics content or olefin content (bromine number) can lead to serious errors in estimating hydrogen consumption. As a cross check to the above estimates, a unit increase in feed API gravity results in roughly 17 Nm3/m3 (100 SCFB) hydrogen consumption, as does a unit increase in cetane for diesel stocks. In Figure 10, we revisit Figure 7 to briefly touch upon areas of focus for troubleshooting unit problems or potential problems. When unit performance is inexplicably and significantly worse than projected performance or prior performance, there are several areas to explore: Feed/effluent exchanger leaks can profoundly lower apparent HDS performance. Changes in feedstock, such as higher quantities of cracked stocks or higher distillation endpoints, can have a noted effect on hydrotreating performance. Noteworthy lowering of make-up gas or recycle gas purities can lower HDS and HDN performance as well as to dramatically accelerate catalyst deactivation. Mal-distribution in catalyst beds can result in appreciably lower apparent catalyst activity owing to quantities of catalyst by-passed. Evidence for mal-distribution can be in the form of

non-uniform temperature profiles and unusual kinetics as described earlier. Convincing evidence is typically not found until catalyst unloading or by using radioactive scanning techniques provided by appropriate service companies. Catalysts are often the first suspected problem when unit performance is substandard. While catalyst quality assurance needs to be verified in such cases, most catalyst problems tend to ultimately be related to other causes less apparent at first analysis.

Two other areas of concern are related to the increasingly higher and higher levels of ammonium bisulfide salts and other salts that are a by-product of the much more severe hydrotreating operations needed to make lower sulfur fuels. Corrosion in air coolers arising from salt deposition can ultimately lead to cooler failure and the safety risk of a potential fire. Avoidance of this problem is addressed by making certain that water wash operations are well-maintained and performing as designed. Salt deposition in compressors can also cause problems, near-term in compressor performance, and longer-term with respect to corrosion/erosion degradation of compressor components. Figure 10. Typical Hydroprocessing Problems
PURITY? P

SALT DEPOSITION SALT

CORROSION CORR

MALDISTRIBUTIO MALDISTRIBUTION

HP
DIFFERENT? DIF

F/E F/E EXCHANGER LEAKS

PURITY? PUR

CATALYST? CATAL

Summary The hydrotreating area, and hydrodesulfurization in particular, is a complex application of catalytic science and art, basic kinetic and reactor design approaches and unit operations knowhow. Those new to the area will find it necessary to seek out resident experts, whether they are process engineers or senior operators, to understand specific operations. A high degree of understanding comes only through a healthy attitude of asking questions, and then, asking more

questions. While historic approaches to interpreting hydrotreater operation, performance and analysis have been adequate, there is significant room for improvement and better understanding. The greater demands for ultra-low sulfur transportation fuels coupled with the swings in refining economics dictate that hydrotreating units be run well and reliably. Conventional wisdom is not always wise in the face of new challenges: enhanced optimization of hydrotreating processes will require the adroit application of new tools, and perhaps as well, some old ones that have been neglected. Acknowledgements Grateful appreciation is given to Ron Meeker and Bill Johns of ChevronTexaco for granting permission to use selected graphics from their presentation Hydroprocessing Basics 2001 given at the 2001 NPRA Clean Fuels Meeting in August in Houston, Texas. I am also indebted to the numerous experts and knowledgeable people who have answered my countless questions concerning hydroprocessing over the last two decades. Selected References for Further Reading Bartholomew, C. H., Catalytic Processing of Petroleum and Distillates, M. C. Oballa and S. S. Shih (Eds.). Marcel Dekker, 1994. Chianelli, R. R., Daage, M. and Ledoux, M. J., Advances in Catalysis 40, 177, 1994. Froment, G. F. and Bischoff, K. B., Chemical Reactor Analysis and Design, Wiley, 1990. Gary, J. H. and Handwerk, G. E., Petroleum Refining, Marcel Dekker, 2001. Gates, B. C., Katzer, J. R. and Schuit, G. C. A., Chemistry of Catalytic Processes, McGraw-Hill, 1979. Girgis, M. J. and Gates, B. C., Industrial Engineering & Chemistry Research 30, 2021, 1991. Hill, C. G., An Introduction to Chemical Engineering Kinetics and Reactor Design, Wiley, 1977. Isoda, T., Ma, X. and Michoda, I., Preprints, Division of Petroleum Chemistry, ACS, 39, 584, 1994. Ma, X., Sakanishi, K. and Michido, I., Industrial Engineering & Chemistry Research 33, 218, 1994. Meyers, R. A., Handbook of Petroleum Refining Processes, McGraw-Hill, 1996. Levenspiel, O., Chemical Reaction Engineering, Wiley, 1972. Shah, Y. T., Gas-Liquid-Solid Reactor Design, McGraw-Hill, 1979. Topse, H., Clausen, B. S. and Massoth, F. E., Hydrotreating Catalysis: Science and Technology, Springer-Verlag, 1996. Whitehurst, D. D., Isoda, T. and Michoda, I., Advances in Catalysi 42, 345, 1998.