The need to be able to assess the

Transcription

1 EFINING igorous hydrotreater simulation The authors describe an integrated approach to dealing with the complexities of producing ultra low sulphur diesel, involving analytical support and process research as well as computer simulation Michael C Hu KBC Advanced Technologies Zbigniew ing Jenny Briker Mure Te National Centre for Upgrading Technology The need to be able to assess the process alternatives for future production of low and ultra-low sulphur diesel (ULSD) has resulted in the development of a rigorous hydrotreater simulator, capable of predicting feedstock and process condition effects on sulphur removal below 10ppmw. How low the sulphur level will eventually go throughout the world is still a subject of debate, but the trend points to a near zero sulphur requirement. Although the technologies required to produce low and ultra-low sulphur gasoline and diesel are somewhat different, a significant increase in overall refinery hydrotreating capacity or, more specifically, the degree of hydrodesulphurisation is inevitable. The additional sulphur reduction will have to be achieved either by feed hydrotreatment for catalytic cracking conversion units or hydrofinishing of final fuel products. This article is concerned with the subject of ULSD. There is no common strategy for refiners to plan their future capital investment. They must select the most economical process option and plan their investment strategy based on available technical know-how, refinery complexity and local market economics. Usually, refineries will work closely with process licensors and catalyst vendors to accomplish this task. Nevertheless, it is extremely useful to have a process simulator capable of predicting hydrotreater performance under ultra-low sulphur mode of operation for various feedstocks and process conditions. The development of this simulator requires a good understanding of diesel feed and product characterisation, comprehensive and realistic process data, and rigorous kinetics modelling. ecently, Canada s National Centre for Upgrading Technology (NCUT) and KBC Advanced Technologies, based in California, have been cooperating in analytical research, pilot plant testing, and process modelling to provide an integrated tool for assessing alternative S DBT process options for future low and ULSD production. After a brief literature review, the analytical and hydroprocessing pilot plant facility at NCUT is presented here, along with selected results of pilot plant testing in terms of sulphur type analysis. The technical basis of KBC Profimatics hydrotreater model, HT-SIM is described in detail. Technical issues To meet the current 500ppmw diesel sulphur specification, refineries have successfully increased their hydrodesulphurisation (HDS) capacity by either installing more active HDS catalysts, raising reactor temperature (at a cost of shorter catalyst cycle life), lowering reactor space velocity (eg, dense bed loading, lower throughput, or adding a new reactor), or improving reactor internals for good catalyst utilisation. Adjustments in other process variables, such as increased hydrogen partial pressure and reduced hydrogen sulphide concentration, in the treat gas recycle loop have also become more important. To reduce diesel sulphur content from 500ppmw to 15ppmw or less, substantial further increases in operating severity and reactor volume will be unavoidable. For example, Lamourelle and co-workers [Lamourelle, McKnight and Nelson; Clean fuels: route to low-sulphur low Figure 1 HDS mechanism K 1 K 2 BHP K 3 K 4 S THDBT CHB aromatic diesel; NPA annual meeting, 2001] reported a required 200 per cent increase in catalyst volume to switch from 500ppmw to 10ppmw S. Selection of the right type of HDS catalyst, cobalt/molybdenum (Co/Mo) versus nickel/molybdenum (Ni/Mo), and the catalyst from within these two groups that is most suitable for the feedstock at hand, also become much more important for ultra-low sulphur production. Some authors have reported the so-called sulphur wall effect [Mayo et al; Elegant solutions for ultra-low sulphur diesel; NPA annual meeting, Nash; Meeting the challenge of low aromatics diesel; NPA annual meeting, 1989]. This effect limits the lowest sulphur level in the diesel stream that is achievable by raising hydrotreater reactor temperature. The lowest achievable sulphur level depends on the catalyst type and hydrogen partial pressure. Under realistic hydrotreater conditions, it could be as high as 20ppmw. The wall effect can be explained by the hydrodesulphurisation mechanism. It is well known that there are two distinct routes for sulphur removal by hydrotreating (Figure 1). The first is direct hydrogenolysis. Almost all the sulphur present in the forms of mercaptans, sulphides, disulphides and thiophenes, as well as a majority of benzothiophenes and unsubstituted 85 PTQ SPING ptq.com

2 EFINING dibenzothiophenes, is removed by this route. The relative reactivity of the various sulphur compounds differs significantly and within each compound group. It has also been reported that the reactivity of sulphur compounds also decreases monotonically with true boiling point [Sau, Narasimhan and Verma; Studies in surface science and catalysis, 1997]. The Co/Mo HDS catalysts are most effective in removing sulphur via this route, even under modest pressures. There is a small, positive effect for higher hydrogen partial pressure and a negative effect of hydrogen sulphide in the recycle gas. The second route, particularly effective with Ni/Mo catalysts, requires partial hydrogenation of aromatic rings in the dibenzothiophene structure prior to the removal of sulphur by hydrogenolysis. This route is much slower than the direct hydrogenolysis route. It is strongly influenced by hydrogen partial pressure and subjected to thermodynamic equilibrium limitation. The sulphur wall effect is believed to be the result of the equilibrium limitation at high reactor temperature. Due to steric hindrance, some of the substituted dibenzothiophenes, particularly those with alkyl-substituents at the 4 and 6 positions, must be desulphurised through the hydrogenation route. One of the most unreactive sulphur compounds in the diesel range is the 4,6- dimethyl-dibenzothiophene. When the diesel sulphur level is lowered to below 100ppmw, almost all of the remaining sulphur belongs to the dibenzothiophene class. The removal of remaining sulphur to the 10 to 15ppmw level is almost entirely carried out via the hydrogenation route. Therefore, Ni/Mo catalysts with better hydrogenation activity are preferred for ULSD production. In order to accurately model the ultra-low sulphur HDS process, the kinetics of hydrogenation and hydrogenolysis of sulphur must be formulated in the programme. PP/1 capabilities at a glance Maximum temperature 450 C 842 F Maximum pressure 180bar 2610psi Feed rate g/h lb/h Maximum gas rate 300NL/h 11.2scf/h ecycle rate 5-200gh/h lb/hr Table 1 Compressor Feed pump H 2 (H 2 S, NH 3 ) eactor 1 Interstage sampler eactor 2 High pressure phase separator Figure 2 PP-1 in series flow hydrocracker configuration Analytical methodology For several years, NCUT has been involved in the development of analytical methods to characterise distillates in terms of their hydrocarbon type and elemental composition. A new methodology has been developed that determines detailed hydrocarbon type and elemental composition of individual fractions of a total liquid product (TLP) without distillation, by providing these compositions as by boiling-point distributions. The sample is analysed in a few steps. The first step is conventional PIONA. A 0.1 L injection results in the distribution by carbon number (and hence boiling point), boiling point of paraffins, iso-paraffins, olefins, naphthenes, and aromatics found in the portion of the TLP that boils up to 200 C. The heavier material is separated by the PIONA prefractionator and discarded. The second step involves solid phase extraction (SPE) analysis of another 20mg sample of TLP. This method is based on the original SPE method for class-type separation and was modified at NCUT to facilitate olefins separation. Four known-volume fractions, namely saturates, aromatics, olefins and polars, are generated. The SPE fractions are further analysed by gas chromatograph with a flame ionisation detector (GC- FID) and gas chromatograph with a mass selective detector (GC-MS). The GC-FID analysis handles simultaneously two tasks. First, the quantification of each SPE fraction using external calibration, and second, determination of the simulated distillation (SimDis) curve for each of the SPE fractions. SimDis is required for calculation of by-boiling-point hydrocarbon-type distribution of the GC-MS results. The GC-MS- and GC-FID-related calculations are performed between 200 C and final boiling-point (FBP) of the sample. Quantitation of the olefin and polar Gas product to meter and analyser N 2 Distillation column ecycle pump Stabiliser column Liquid product receiver SPE fractions require only a single GC- FID run each. The results of the individual steps are fed to software developed at NCUT that reconciles and summarises final results in the form of a characterisation table [Briker et al; Distribution of aromatics, HC ratio and olefins in cracked liquid products from MAT new approach for analyzing small size samples; 40th International Petroleum Conference, Bratislava, Sept, 2001]. This table contains mass fractions of individual hydrocarbon types in userselected boiling-point fractions of the TLP. The method may significantly reduce the total time required to complete analysis, but its main advantage is that it requires a very small sample. It is particularly useful to look at product quality of individual fractions of TLP when samples are inherently available in very small quantities (eg, FCC-MAT microactivity test or micro-hydroprocessing reactor). Following the treatment detailed in the previous paragraphs, the TLP sample is subjected to elemental analysis using GC-AED that provides sulphur, carbon, and hydrogen distributions by boiling-point. The sulphur emission at 181nm is used to collect the sulphur signal. At the same time, the carbon signal is collected at 179nm, which allows distribution by boiling-point of the sulphur-to-carbon ratio. The carbon and hydrogen emissions of 496nm and 486nm respectively are used to collect the signals for carbon and hydrogen and calculate the hydrogen-to-carbon ratio distributed by boiling-point. The nitrogen content and distribution by boiling-point is determined by using a nitrogen specific detector (NCD). For all the elements, the distribution by boiling-point is calculated by using the calibration parameters obtained from analysing the mixture of 87 PTQ SPING 2002

3 EFINING paraffin standards (C 5 to C 44 ) at the same conditions as for the analysed sample. In the next step, sulphur compounds found in the sample are identified by retention-time matching, using the NCUT database. For high-sulphur feedstocks, another SPE method has been developed to separate saturate and aromatic sulphur species using silver nitrate impregnated silica gel. This method, together with 60m-long chromatographic columns used in the GC-AED and GC-MS instruments, facilitates good peak separation and identification. The experimental results used in this article are limited to this method. Total S, ppmv 10,000 1, Figure 3 Total sulphur vs WABT Pilot plant Some pilot plant units used by NCUT to conduct contract hydrotreating and hydrocracking pilot plant testing programmes consist of two reactors with an integrated fractionator and recycling of unconverted products. These units have consistently generated precise data on catalyst stability, yield distribution, gas make, and hydrogen consumption under commercially realistic conditions. The pilot plant facility has been extensively used in catalyst evaluation programmes for commercial hydrotreating and hydrocracking reactors, for process optimisation, and to support process model development. The ULSD study partially presented here was carried out using NCUT s PP-1 unit. Table 1 summarises PP-1 capabilities WABT, C HT SIM Pilot plant data and Figure 2 provides a schematic diagram of this unit. The feedstock used in this test was a commercially sampled FCC light cycle oil (LCO). The LCO has been considered as the most refractory 88 PTQ SPING 2002

4 EFINING Feed analysis and pilot plant test conditions Feed stream FCC LCO Density, g/ml Carbon, wt% Hydrogen, wt% 9.78 Total S, wt% Total N, wt% Distillation, D86, C IBP vol% vol% vol% 304 FBP 377 Test Conditions Pressure, bar 70 WABT, C Space velocity, L/L/hr 1.5 vol% H 2 S in make-up H Table 2 feed component in a commercial distillate hydrotreater and practically always requires hydrotreating before being used as a diesel-blending component. The pilot plant reactor was packed with 150ml of a highly active commercial Ni/Mo hydrotreating catalyst. Table 2 summarises the feed analysis and pilot plant test conditions. esults The total product sulphur content was plotted against reactor temperature in Figure 3. Each of the data points representing the experimental data is, in fact, two overlapping squares since two mass balance runs were conducted at each temperature. This illustrates the high quality of the data generated in this study. The difficulty of producing 10ppmw sulphur product can be clearly seen from the plot. To better understand the kinetics of sulphur removal, the sulphur analytical procedure outlined earlier was used to identify and quantify each individual sulphur component in both the feed and products. In Figures 4, 5 and 6, only the sulphur chromatograms for feed and products from experiments at 375 C and 385 C, respectively, are presented. At about 375 C, the thiophene and benzothiophene spices have been removed completely. Consequently, the parts of the product chromatograms could be removed up to retention time of 90 minutes. Above 385 C reactor temperatures, the only remaining compounds were 4,6-dimethyl-dibenothiophene and related compounds with more and longer side chains. The unsubstituted dibenzothiophenes and substituted dibenzothiophenes at positions other than 4 and 6 disappeared completely. Close inspection of the amount of conversion for each individual peak revealed that we could group the sulphur compounds into three types: the easy, difficult and very difficult groups. The easy sulphur group included sulphides, mercaptans, thiophenes and benzothiophenes. The difficult sulphur included dibenzothiophenes with no substituents at the 4 and 6 positions. The dibenzothiophene isomers substituted at the 4 and 6 positions were classified as the very difficult sulphur. Similar classification was also proposed by Tippett and co-workers [Tippett, Knudsen and Cooper; Ultra low sulfur diesel: catalyst and process options; NPA annual meeting, 1999] Kinetics modelling There are two extreme approaches in modelling the HDS reaction. Traditionally, a simple n-th order kinetics (1.5 to 2) was applied to model the removal of total sulphur. This simple approach was successfully used in reactor sizing for down to the 500ppmw sulphur level. For sulphur levels less than that, the n- th order kinetics model becomes too optimistic and is inadequate. The other extreme is based on a structural approach where HDS reactions are modelled based on each individual molecule type. This article takes an intermediate approach and models HDS based on three types of sulphur as classified above. Sugimoto and co-workers proposed a kinetics model based on four types of sulphur [Sugimoto, Tsuchiya and Sagara; Deep hydrodesulferization of light gas oil; NPA annual meeting, 1992]. However, as shown below, the three types of sulphur used in our work can simulate the kinetics of total sulphur removal to ultra-low levels with sufficient accuracy. For easy and difficult sulphur, the removal is mainly realised via the direct hydrogenolysis route, which can be described by the following formula: r S = k 0 e E/T [S] [H 2 P.P.] n (1) where, r S = rate of conversion for easy or difficult sulphur k 0 = pre-exponential factor E = activation energy [S] = concentration of easy or difficult sulphur n = n-th order H 2 partial pressure effect The k 0 is hindered by the presence of H 2 S. The pressure effect is usually minor for this route. A typical n value of 0.5 to 0.6 is applied for commercial HDS unit. The modelling of the removal of the very difficult sulphur is more complex since there are two parallel routes (direct and hydrogenation) to be considered. Moreover, the hydrogenation route is a reversible reaction. Based on the reaction network given in Figure 1 and using 4,6-DMDBT as an example, the HDS kinetics for the very difficult sulphur can be described as follows. Direct route The kinetics expression is identical to Equation 1. r S1 = k 1 [4,6-DMDBT] [H 2 P.P.] n1 (2) where, r S1 = rate of conversion of the very difficult sulphur via direct route k 1 = k 0, 1 e E 1/ T [4,6-DMDBT] = concentration of 4,6- dimethyldibenzothiophene n1 = n-th order H 2 partial pressure effect for direct route Hydrogenation route r S2 = k 2 [4,6-DMDBT] [H 2 P.P.] n2 k3 [4,6-DMTHDBT] (3) r 3 = k 2 [4,6-DMDBT] [H 2 P.P.] n2 k3 [4,6-DMTHDBT] k4 [4,6-DMTHDBT] [H 2 P.P.] n3 (4) where, r S2 = rate of conversion of the very difficult sulphur via hydrogenation route r 3 = rate of formation of the partially saturated 4,6-DMDBT n2 = n-th order H 2 partial pressure effect for hydrogenation n1 = n-th order H 2 partial pressure effect for sulphur extraction Vanrysselberghe and Froment reported that the amount of partially saturated 4,6-DMDBT was negligible. Therefore, we assumed the net rate of formation of the partially saturated 4,6- DMDBT, r 3 is approximately zero. We also assumed that n 3 is much smaller than n 2 and can be ignored. By substituting [4,6-DMTHDBT] into equation 3 and rearranging the formula, the final rate equation derived for the hydrogenation route is: rs2 = k 2 [4,6-DMDBT] [H 2 P.P.] n2 (k 3 /k 4 +1) (5) where k 2, k 3 and k 4 have their specific activation energies E 2, E 3 and E 4, respectively. The total rate of removal of the very difficult sulphur is the sum of the rates detailed in Equations 2 and 5. The pilot plant data for the removal of the very difficult sulphur were used to test this kinetic model. Equation 5 fits the data quite well. To fit the sulphur data at the ppmw range, a value of 60kcal/mole was required for E2-E3. The relatively high activation energy for dehydrogenation reaction promotes the reverse reaction at high temperature and prevents further reduction in the sulphur content. This thermodynamic limitation has long been known for hydrodenitrification 89 PTQ SPING 2002

5 EFINING reactions, but it became a limiting factor for sulphur removal only recently as a result of ULSD specifications. As discussed in the previous section, the lowest achievable sulphur level depends on feed quality, catalyst type, and process conditions, particularly AED signal counts hydrogen partial pressure. Moreover, in a commercial hydrotreater, when the catalyst is approaching the end of run weighted average bed temperature (WABT), the bottom portion of the reactor could have already been in the equilibrium controlled temperature range. LCO feed chromatogram (GC AED) DBT=dibenzothiophene, NTH=naphthothiophene Thiophenes and benzothiophens 4MDBT C1 DBTs & C1 NTHs Figure 4 LCO feed sulphur chromatogram AED signal columns MDBT DBT EDBT, 46 DMDBT C2 DBTs & C2 NTHs C3 DBTs & C3 NTHs C4 DBTs & C4 NTHs etention time, min 375 C Effluent 4,6 DMDBT C3 DBTs C3 naphthothiophenes Other C2 DBTs C2 naphthothiophenes C4 DBTs C4 naphthothiophenes etention time, min Figure 5 Product sulphur chromatogram (WABT = 375 C) AED signal columns MDBTs 385 C Effluent 4, 6 DMDBTs C 3 DBTs C 3 naphthothiophenes Other C 2 DBTs C 2 naphthothiophenes etention time, min Figure 6 Product sulphur chromatogram (WABT = 385 C) C 4 DBTs C 4 naphthothiophenes Therefore, it is critical to consider this factor when sizing the distillate hydrotreater and estimating reactor cycle life. Steady state simulator In order to accurately evaluate alternative process options for ULSD production or optimise an existing unit, many other distillate product qualities (eg, aromatic content, cetane number) and process parameters (yields, hydrogen consumption) must be considered in the study. Moreover, when catalyst deactivates, all those variables shift along with higher reactor WABT. Therefore, the reaction kinetics other than hydrodesulfurization and catalyst deactivation must be considered simultaneously. These reaction kinetics include cracking, hydrodenitrification, aromatic saturation, naphthenic ring opening, etc. A good feed property characterisation procedure is also required to account for feedstock effects on various types of reactions. The need for this additional detail leads to the conclusion that a comprehensive kinetics-based process model is needed to perform proper ULSD evaluation. KBC Profimatics has developed and commercialised a rigorous and kineticsbased hydrotreater simulator, HT-SIM, for commercial unit monitoring, process optimisation, and LP studies. The technical basis of this model has been described in detail elsewhere. The HT- SIM model rigorously simulates the reactor section of a commercial hydrotreater including the catalyst beds, high pressure separator, makeup gas system, recycle/quench gas system, and vent gas (purge) system. Fresh feedstocks are broken down into carbon number pseudo-components. Each carbon number pseudocomponent is assigned different types of carbon in terms of aromatic, naphthenic and paraffinic carbon fractions, as well as sulphur, nitrogen and olefins, using bulk feed properties and given distribution rules. In the reaction section of the model, each pseudo-component is included in a set of reaction kinetics formulas for cracking, hydrodesulfurization, hydrodenitrification, aromatic saturation, naphthene ring opening, and olefins saturation. Then, numerical integration is carried along the catalyst beds. Finally, the liquid product stream is split into desirable product fractions for final product quality determination. The model is also designed to simulate catalyst deactivation phenomena. For example, it will track changes in the intrinsic reaction rate constants based on coke deactivation rate and catalyst age. Effects of changing operating 90 PTQ SPING 2002

6 EFINING conditions and feed properties on coke deactivation rate are also included in the simulation. Finally, the HT-SIM hydrotreater model is embedded in a Microsoft Excel-based interface for easy data entry and results presentation. Once calibrated with commercial test run data or unit design data, the HT-SIM model can be used to study feed effect and process condition effects (eg, WABT, reactor throughput, catalyst volume, pressure effect, make up H 2 purity). Moreover, the kinetic rate constants can be easily adjusted to reflect different type of catalysts (Co/Mo versus Ni/Mo). In terms of HDS, a Co/Mo catalyst will be more effective in the direct route than a Ni/Mo catalyst. On the other hand, a Ni/Mo catalyst will be much more active in the hydrogenation route. Some catalyst vendors have proposed a concept of stacked-bed catalyst configuration for ULSD application. This configuration could be easily simulated by the HT- SIM model, since the model provides the user with the option of up to three different catalysts in series. Based on the work performed by NCUT, KBC was able to enhance the capability of the HT-SIM simulator for ULSD study. The three types of sulphur and the parallel direct and hydrogenation HDS routes have been implemented in the simulator. To demonstrate the predictability of HT-SIM, the model was first tuned against the pilot plant data at the lowest WABT (320 C) and then it was used to predict sulphur content (extrapolate) at higher temperatures (model predict mode). A good match is evident in Figure 3, where the model-predicted product sulphur content was compared with pilot plant data as a function of WABT. The model generates additional useful process information for example, chemical hydrogen consumption as a function of WABT. Initially, hydrogen consumption increases with WABT. However, after reaching the sulphur level of about 10 to 15ppmw, the hydrogen consumption starts levelling off and goes through a maximum. This coincides with the aromatic content in the product as a function of temperature. The thermodynamic equilibrium limitation for aromatic saturation has been attributed to the deterioration of distillate product quality (eg, diesel cetane number, smoke point and Saybolt colour) associated with the high WABT operation, particularly at EO. Of course, one of the very important applications for a hydrotreater simulator is to evaluate various refinery feed streams for potential processing in a diesel hydrotreater. In principle, a refinery could analyse the feed sample for sulphur types, quantify the refractory sulphur-containing feed components, and determine how difficult it would be to remove total sulphur from that feed. However, work is going on to develop a sulphur-type correlation based on bulk feed properties such as type of feed, density, and boiling-point distribution. A preliminary correlation has already been built into the model, and KBC will continue working with NCUT to further refine this correlation. Particularly, it would be beneficial to expand this correlation to heavier feed streams, such as VGO cuts, for VGO hydrotreater application. You can air your views on this subject on our eaders Forum: 91 PTQ SPING 2002