- Email: [email protected]
Effects of experimental operations on the Fischer-Tropsch product distribution
Catalysis Today xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
Catalysis Today journal homepage: www.elsevier.com/locate/cattod
Effects of experimental operations on the Fischer-Tropsch product distribution ⁎
⁎
Ruyi Yanga,c, Liping Zhoub, , Junhu Gaob, Xu Haoa,b, Baoshan Wua,b, , Yong Yanga,b, Yongwang Lia,b a b c
State Key Laboratory of Coal Conversion, Institute of Coal Chemistry, Chinese Academy of Sciences, Taiyuan 030001, PR China National Energy Center for Coal to Liquids, Synfuels China Co., Ltd., Huairou District, Beijing 101400, PR China University of Chinese Academy of Sciences, Beijing 100049, PR China
A R T I C L E I N F O
A B S T R A C T
Keywords: Fischer-Tropsch synthesis Simulated distillation Product distribution Flash loss Separation
The present study provides insight into the effect of experimental operations on the Fischer-Tropsch synthesis (FTS) product distribution by combining experimental and theoretical analyses of the process. Experimental errors originated from GC data processing and product collecting conditions are addressed here based on the experimental data obtained using commercial iron-based catalyst in a tubular fixed-bed reactor. A corrected area normalization method, which is a combination of the area normalization and simulated distillation, is proved to be an accurate and convenient method for FTS heavy oil chromatogram quantification. The flash loss from the hot trap and the cold trap is found to be negligible under the normal operating conditions. However, water may be condensed remarkably in the hot trap at low temperature and wrapped by wax. In addition, the selectivity of the high-carbon-number alcohols will be seriously underestimated at low temperature of the hot trap. The FTS product distributions under a wide range of reaction conditions are exhibited.
1. Introduction Fischer-Tropsch Synthesis which converts carbonaceous materials such as coal, renewable biomass, and natural gas (via gasification) into clean transport fuels or other valuable organics has received great attention in the past decades [1,2]. Generally speaking, FTS is a polymerization-like reaction and the product spectrum consists of a very broad range of hydrocarbons (mainly linear paraffin and linear olefin) and oxygenates (mainly alcohol), with carbon number from C1 to even C120+. The Anderson-Schulz-Flory equation is usually introduced to describe the relative ratios of these polymers of different length, with one constant chain growth probability, regarded as ASF distribution. However, a straight line of the hydrocarbon distribution only occurs when the process produce light hydrocarbons, e.g. carbon number below fourteen. Tremendous experimental results showed that the hydrocarbon selectivity did not obey the ideal ASF distribution under the industrial operating conditions [3–13]. The compositions decrease gradually with the increasing carbon number but show some irregularities at the same time, as exhibited in Fig. 1. The main trend is approximately fitted by a so-called double-α (α1 and α2) distribution [14–17]. Besides, some worse irregularities deviating from the double-α distribution occurred [4–9]. Case 1 and case 4 are some saddle-like
⁎
shapes that appeared around C10–C25 and heavier hydrocarbons, respectively, depending on the specific catalysts and the experimental conditions. Case 2 and Case 3 refer to some positive and negative deviations at high-carbon-number region. The cause of these deviations is a truly controversial issue. Lots of studies tried to explain it by the olefin re-adsorption and the subsequent secondary reactions, by the promoters’ effects on the FTS catalysts which may lead to different kinds of active sites, or by the existence of multiple reaction mechanisms in the chain growth process [16,18]. Some others even attributed it to the cracking of heavier hydrocarbons which may happen under the FTS reaction conditions because of the acid-supported catalysts [7]. Our recent studies showed that the experimental artifacts in FTS data collection also played a very important role in the patterns of the product distribution curve. Considering the importance of the FTS experimental data to the reaction mechanism assumption, the product kinetic modeling and even the catalyst development and selectivity control, the present article further clarifies the influence of the experimental artifacts on the FTS product distribution, e.g., the product analysis methods, the product condensing and collecting and flash loss, based on the experimental results and the theoretical calculation.
Corresponding authors at: National Energy Center for Coal to Liquids, Synfuels China Co., Ltd., Huairou District, Beijing 101400, PR China. E-mail addresses: [email protected] (L. Zhou), [email protected], [email protected] (B. Wu).
http://dx.doi.org/10.1016/j.cattod.2017.05.056 Received 30 September 2016; Received in revised form 22 March 2017; Accepted 15 May 2017 0920-5861/ © 2017 Elsevier B.V. All rights reserved.
Please cite this article as: Yang, R., Catalysis Today (2017), http://dx.doi.org/10.1016/j.cattod.2017.05.056
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 1. The commonly appeared irregularities in the FTS product distribution. The solid line: approximately estimating of FTS product distribution by a double-α model; the dashed lines: case 1 and case 4 are some saddle-like shapes that appeared around C10-C25 and heavier hydrocarbons, respectively. Case 2 and Case 3 are some positive and negative deviations at high-carbon-number region. (Reprinted from Industrial & Engineering Chemistry Research, 51 (2012) 11618–11628.) [28].
2. Experimental Fig. 2. Flow scheme of the fixed-bed experimental setup. 1. purification unit (de-sulphur, de-oxygen, de-carbonyl, de-water); 2. mass flow controller; 3. thermocouple; 4. hot trap; 5. cold trap; 6 & 7. back pressure regulator; 8.wet flow meter.
2.1. Catalyst and reduction The industrial iron-based catalysts were provided by Synfuels China Co., Ltd. Typically 1.0 g catalyst with particle size 150–180 μm was used in order to eliminate the internal diffusion. Prior to its use, the catalysts were reduced by synthesis gas at H2/CO = 2, T = 260 °C, P = 0.1 MPa, space velocity (s.v.) = 1000 mL g−1 h−1 for 48 h. The detailed characterization and properties of the catalysts could be seen elsewhere [19].
In the feeding section, the hydrogen (purity > 99.999%) and the carbon monoxide (purity > 99.999%) passed through a series of purifiers to further remove oxygen, sulfur, carbonyls and water and the flow rates were measured by the thermal mass flow controllers (Brooks 5850E). The tubular fixed bed reactor is a stainless steel tube (total length 80 cm and internal diameter 1 cm) with a thermal-well (external diameter 0.1 cm) in the center. The reaction heat is removed by a molten salt bath, which is composed of 53% KNO3, 40% NaNO2 and 7% NaNO3. The salt bath is stirred strongly by a churning motor to obtain a better heat transfer efficiency and keep acceptable temperature uniformity (see in Fig. 3). Moreover, considering the strongly exothermic property of the FTS reaction, a remarkable dilution of the catalyst bed was performed by inert Silicon Carbide, with the same particle size and mass ratio of 30.The real temperature profile along the catalytic bed is shown in Fig. 3. The hot spot appears in the first 2 cm of the catalytic bed. The maximum temperature gradient is around 1.5 °C, which is negligible in the FTS reaction. The gas-liquid mixture of the reactor effluent enters into a hot trap (typically 170 °C), then passes through a back pressure regulator and enters into a cold trap (typically 2 °C), as shown in Fig. 2. The noncondensable gases release via a wet flow meter (Ritter TG 1–5). Accordingly, the product collection consists of three parts, the heavy oil (the so-called wax, C6–C70 paraffins and olefins) condensed in the hot trap, the light oil (C3–C35 paraffins and olefins), the oxygenates (mainly C1–C20 alcohols) and the water condensed in the cold trap and the non-condensable gases, based on the boiling range of the reactor effluent, as shown in Fig. 4. The detailed GC configuration parameters and the flow scheme of the FTS product analysis are shown in Fig. 4. The non-condensable gases include H2, CO, CO2, C1–C12 paraffins and C2–C12 olefins. These are analyzed by an online Agilent 7890 B GC equipped with two TCD channels and one FID channel, in which H2 is detected on TCD1 with a HayeSep Q column (0.5m × 1/8in., mesh size 80/100), CO, CO2, CH4 are detected on TCD2 with a Agilent MolSieve 13X column (1.83m × 1/8in., mesh size 60/80) and the C1–C9 hydrocarbons are detected on FID with a DM-Plot Al2O3/Na2SO4 column
2.2. Experimental setup In view of the close relationship between the experimental setup and the accuracy of the experimental data, the reactor to be used in product distribution study should be considered carefully. Several types of reactors have been used in bench-scale FTS experimental data collection in published papers, e.g., stirred slurry autoclave, spinning basket reactor, Berty-type internal recycle reactor and tubular fixed-bed reactor. The first three types all could be regarded as a “point” operation pattern, the temperature and concentration inside the reactor are in uniform. This property is convenient for FTS kinetic modeling. However, none of them is perfect for the FTS product distribution study. The accumulation of high-carbon-number hydrocarbons in stirred slurry autoclave would seriously disguise the true FTS product selectivity [12,20–22]. The Berty-type recycle reactor is suitable for light hydrocarbon synthesis, but it is hard to be run under real FTS conditions because of the blocking of the agitating shaft by the produced heavy wax. Continuous spinning basket reactor has all the merits of gradientless reactor and has no accumulation effect. However, it is difficult to know the real temperature and the real space velocity of the catalytic bed [23]. The tubular fixed-bed reactor has several distinct advantages in product distribution study, such as easily experimental operating, less product accumulation inside the reactor and shorter time for environment replacement when changing reaction conditions. By suitable catalyst dilution and molten-salt-bath heating, the temperature gradient commonly met in an industrial-scale fixed-bed reactor could be negligible in a laboratory reactor [24]. The flow scheme of the experimental setup used in this study is shown in Fig. 2. It consists of four sections: the feeding, the tubular fixed-bed reactor, the product collection and the product analysis. 2
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 3. The temperature profile along the tubular reactor with and without reaction. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4.
(50m × 0.32 mm × 5 μm). In the cold trap, the light oil and the water layered and the oxygenates dispersed into the two phases. The oil phase components are analyzed by an offline Agilent 7890A GC with a HPPONA column (50m × 200 μm × 0.5 μm), while the water phase components are analyzed by an offline Agilent 7890A GC with an ABINOWAX column (50m × 200 μm × 0.5 μm). The wax sample is first dissolved in CS2 and then analyzed by an offline Agilent 7890A GC with a UA1-30M-0.5F column (32m × 530 μm × 0.5 μm).
collecting enough products for analysis. The total mass balances and the elemental balances (C, H and O) of the experimental data were around 97%–102%. 3. Results and discussion 3.1. Influence of GC data processing on the FTS product distribution FTS products are less complex than petroleum-based samples, but containing some heavier components (especially those from C70 to nearly C120 or more), e.g., the heavy oil condensed in the hot trap showed in Fig. 4. The characterization of these high boiling point hydrocarbons requires high temperature GC (oven up to 450 °C) in order to elute the components. However, the typical columns for detailed analysis of FTS products are limited to around 350 °C because of the limited thermal stabilities of capillary column and the stationary phase. This will lead to loss of peaks of heavier components in reality.
2.3. Experimental procedure To prevent temperature runaway, the reactor temperature was cooled down to 220 °C after the catalyst reduction, and then the operating parameters are adjusted gradually to meet the desired reaction conditions. After a steady-state transition period, normally 48 h needed for sure, the experimental data collection for product distribution study could be started. Generally 10 h are necessary for
Fig. 4. Flow scheme of separation and analysis procedure of the FTS products.
3
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 5. Influence of chromatographic data processing methods of wax on the product distributions. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/ CO = 2.4, Thottrap = 170 °C, Pcoldtrap = 3.0 MPa.
distillation (HTSD, based on the ASTM D7169 method). The recovered mass% value Fn (n denotes carbon number Cn) obtained from HTSD, multiplying by the area normalization data, would be necessary to eliminate the overestimation originated from residue fraction which does not distill at the maximum operating temperature. The Fn values of FTS heavy oil and light oil are shown in Fig. 6. The distillation yield of light oil at 520 °C(up to C34) is nearly 100%, while that of heavy oil is only around 41%. At 645 °C(up to C70), the Fn value of heavy oil is only 53.4%. The results indicate that the AN method is suitable for light oil chromatogram quantification within the scheme of separation and analysis procedure of FTS products showed in Fig. 4, but the Fn value must be checked and added in heavy oil chromatogram quantification.
Therefore, one should be very careful in GC data processing of FTS products because some traditional methods may introduce errors. The internal standard (IS) method is a general way in chromatogram quantification, but considering the complexity of FTS heavy oil, it is boring to find a special component that is similar in physico-chemical properties to the analytes but does not co-elute with them. An alternative way, standard addition method (SA, showed in Eq. (1)) is used here [25,26]. Tetradecane is served as a reference. The results in Fig. 5 showed that the paraffin, the olefin and the total hydrocarbon distributions are all smooth lines without any saddle-like shapes. The paraffin distribution exhibited remarkable double-α property compared to the olefin distribution, which is close to a straight line.
% component a =
f Aa m × a × s × ws × 100% As fs mt
% component a =
(1)
∑
Where As and Aa are peak areas of tetradecane in the standard solution and component a in the sample solution, respectively; fs and fa are the response factors of tetradecane and the component a, respectively; ms and mt are the mass of the standard solution and the sample solution; ws is the mass percentage of tetradecane in the standard solution. The SA (see also IS) method requires a precisely prepared standard solution and a weighed heavy oil sample. Moreover, the normally used CS2 solvent in sample preparation is highly volatile. These make it easier to introduce experimental errors in practical operations. Area normalization (AN), showed in Eq. (2), is a simple and convenient method to the quantification of chromatograms data. It does not require additional calibration standards, and small variations in sample preparation and/or detector operating parameters have approximately the same relative effect on each component. Nevertheless, this method will lead to remarkable over-estimation of the analytes from around C12 in the heavy oil (showed in Fig. 5) because some heavier components could not be recorded by the detector.
% component a =
fa Aa × 100% ∑ fi Ai
fa Aa n
× Fn × 100%
fi Ai
i =1
(3)
Where faAa is the corrected peak area of component a, ∑fiAi is the sum of the corrected peak areas in the chromatogram, and Fn is the recovered mass% to Cn obtained from the simulated distillation. It can be seen from Fig. 5 that the product distribution curves calculated by CAN method are quite consistent with those calculated by SA method, but with relatively simple operation and less opportunities for experimental error introducing. Therefore, CAN method could be a recommended method for accurate and trustable FTS heavy oil chromatogram quantification. 3.2. Influence of experimental errors on the FTS product distribution As discussed in Fig. 4, the FTS product collection and analysis refer to a number of steps, in which the product outflow from the catalytic bed, the product condensation in the two traps, the separation of oil phase and water phase and all these steps may introduce random errors. Hence, it is necessary to check the effect of the random experimental errors on the repeatability and accuracy of FTS product distribution. Four experimental data within 50 h are collected and investigated at a constant reaction condition [27]. The repeatability of hydrocarbon, alcohol and ratio of olefin to paraffin (O/P) distribution curves of four measurements has been compared in Fig. 7. The relative standard deviation(RSD) of the components, which reflects how widely spread the measured values are on either side of the mean, is shown in Table 1. The comparison of
(2)
Where faAa is the corrected peak area of component a;∑fiAi is the sum of the corrected peak areas in the chromatogram. In this paper, a corrected area normalization method (CAN, showed in Eq. (3)), which is a combination of the area normalization and simulated distillation, is developed for FTS heavy oil chromatogram quantification. An Agilent 7890A GC with a DB-1 column (5m × 530 μm × 0.17 μm) is used for high temperature simulated 4
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 6. The HTSD results of the FTS light oil and heavy oil.
Fig. 7. Repeatability of hydrocarbon and alcohol distributions and the ratio of olefin to paraffin. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4.
5
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
trap is critical to the accuracy of the alcohol distribution. As shown in Fig. 9, the selectivity of C1–C6 alcohols is repeatable at different temperatures, but that of the heavy alcohols at 170 °C is significantly higher than that at 110 °C. This could be explained by the fact that the lower temperature is set, the more alcohols will be condensed and dissolved in the wax (as mentioned above, this part of alcohols are hard to be identified in wax analysis). The distribution trends at 170 °C and 190 °C are close to each other, which indicate that at least this temperature level should be maintained to guarantee a real alcohol distribution. These accurate experimental results are very important to reaction mechanism and kinetic studies. It should be noted that the quantity of alcohol dissolved in the wax is almost two orders of magnitude less than that of hydrocarbon products, as shown in Table 3. Therefore, the error introduced into hydrocarbon distribution by alcohol condensation is negligible. This is also proved by the results in Fig. 8, in which the hot trap was operated at different temperatures, indicating different levels of alcohol condensation. During the product collection from the cold trap, the liquid light hydrocarbons may be vaporized and lost because of the suddenly environment changes from high pressure and low temperature to low pressure and high temperature, e.g., 0.1 MPa and 25 °C. The effect of this flash loss on the FTS product distribution is investigated. By adjusting the back pressure regulator behind the cold trap (Fig. 2), the operating pressure of the cold trap is experimentally changed from 3.0 MPa to 0.3 MPa, and the results are shown in Fig. 10. The quantities of the oil plus tail gas are almost the same at such a broad range of pressures (Fig. 10a), which means that the flash loss in oil collection is negligible. Thus, it is feasible to see that the total product distributions at different operating pressures of the cold trap are overlapped perfectly.
Table 1 Relative standard deviations of product distributions and O/P in experimental repeatability. Carbon number
Total hydrocarbon (%)
Paraffin (%)
Olefin (%)
O/P (%)
Alcohol (%)
1 2 3 4 5–9 10–15 16–20 21–34 35+
7.5 7.1 7.4 7.6 7.4 5.0 10.3 10.8 11.5
7.5 6.8 7.7 8.1 8.9 7.9 12.6 11.8 11.7
– 7.3 7.3 7.3 6.7 3.0 8.4 11.3 –
– 0.9 0.6 0.8 2.4 5.4 6.8 9.9 –
4.9 6.2 7.0 8.0 7.5 6.5 6.4 – –
the curves shows that the multi-times experiments provide a good repeatability of the detailed product distributions. However, the repeatability decreases gradually with the increasing carbon number, e.g., RSD values of paraffin from 7.5% to around 12%, as shown in Table 1. The experimental O/P ratios are normally trustable (RSD below 5%) except for these at high-carbon-number region where some iso-olefin peaks could not be separated accurately from paraffin peaks in GC analysis.
3.3. Influence of product collecting conditions on the product distribution During the multistep flash separation processes in FTS product condensation and collection, the operating temperature and pressure of the two traps (shown in Fig. 2) are two key parameters. They determine the partition of the products with different boiling point in the gas and liquid phase, and also the extent of flash loss of the products during the product collecting process. The effects of these collecting conditions on the FTS product distribution are investigated here. Table 2 shows the sensitivity of the quantities of water and liquid hydrocarbons partitioned in the two traps to temperature changes, respectively. At 110 °C, most of the produced water is condensed in the hot trap. This water could easily be wrapped by wax during product collection. Therefore, the quantity of wax may be overestimated and the product distribution would be lack fidelity [28]. This phenomenon could be improved by increasing the temperature of the hot trap, e.g., the water in the hot trap is only 0.14 g/h at 190 °C and this will even be negligible at 200 °C. Although the temperature variation changed the partition of hydrocarbons in the wax phase and the oil phase, the total amount of the liquid hydrocarbon is almost a constant and this partition have no effect on the accuracy of GC analysis. Hence, the hydrocarbon distribution is repeatable at different temperatures of the hot trap, as shown in Fig. 8, provided that one has complete wax and wrapped water separation after the wax collection from the hot trap. Few people paid attention to the oxygenate (mainly alcohols) distribution in FTS. Actually, we find that the temperature of the hot
3.4. Simulation of product separation and collection process For an in-depth and comprehensive insight into the effect of collecting process conditions on the final products distribution, the product separating and collecting process is modeled by Aspen Plus software. The schematic simulation diagram is shown in Fig. 11. The PRODUCTS stream denotes the total effluent from the reactor, which includes CO, H2, CO2, H2O, C1–C60 paraffins, C2–C35 olefins and C1–C19 n-alcohols. It goes through a hot trap (HOTTRAP) and a cold trap (COLDTRAP) successively as the real experimental condition. A three phase flash module in Aspen Plus is adopted to describe the two traps [29]. The WATER1 stream simulates the accumulated water in the hot trap. The WATER2 stream simulates the water phase products condensed in the cold trap. FLASHOIL and FLASHWAX are added to simulate the operating conditions for flash loss calculation during oil and wax collection. The TAILGAS, OIL-L and WAX-L streams simulate the tail gas, oil and wax products obtained in real situation, respectively. OIL-G and WAX-G streams denote the quantities of lost hydrocarbons from the hot trap and the cold trap because of flash loss. RKSOAVE equation of state in Aspen Plus is chosen for the flash equilibrium calculation. The physical property data of the light hydrocarbons are obtained directly from the PURE 11 database in Aspen Plus, while these required of the heavy hydrocarbons are estimated by extrapolation through the ABC methods introduced by Marano and Holder [30]. In the following simulation, the experimental data obtained under the reaction condition mentioned above (T = 260 °C, P = 3.3 MPa, GHSV = 28000 mL g−1 h−1, H2/CO = 2.4) are used as the compositions of the PRODUCTS stream. The operating parameters, e.g., temperature and pressure, of the two traps (HOTTRAP and COLDTRAP) refer to the experimental conditions mentioned above. The temperature and pressure of the FLASHOIL and FLASHWAX set as 25 °C and 0.1 MPa, respectively.
Table 2 Influence of temperature of hot trap on the repartition of reactor effluent. Thottrap(°C)
Water (g/h)
Heavy oil(g/ h)
Light oil(g/ h)
Tailgas(g/ h)
6.37 6.25
Cold trapa
Hot trapb
110
0.56
1.41
1.21
0.20
170 190
1.73 1.86
0.14 0.00
1.06 0.88
0.44 0.51
6.40
a
Water condensed in the cold trap. Water condensed in the hot trap. Reaction T = 270 °C,P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, H2/CO = 4. b
3.4.1. Simulation of the product separation To clarify the trends of the product repartition in a vapor and in a
conditions:
6
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 8. Influence of temperature of hot trap on hydrocarbon distributions.Reaction conditions: T = 270 °C, P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, H2/CO = 4.
liquid phase during condensation, simulation is performed where vapor-liquid equilibrium is calculated at different temperatures and pressures of the two traps for the same PRODUCT stream. With increasing temperature from 100 °C to 200 °C, the molar flow rate of alcohol outputting from the hot trap decreases remarkably, which means alcohols are heavily condensed in the hot trap at low temperature and are omitted, especially these from C3 to C16 (Fig. 12). Due to the co-elution of alcohols and iso-hydrocarbons during the chromatography analysis of wax it is not easy to quantify the content of alcohol condensed in wax. Therefore, to avoid underestimating the alcohol contents the temperature of hot trap should be kept at a high level. This is in agreement with the experimental results shown in Fig. 9. The water content, defined as the molar ratio between the condensed water in the hot trap and the total water produced, is shown in Fig. 13. It clearly exhibits the relationships among temperature, pressure and water content in the hot trap, so that provides useful information for operators to estimate the possible water content at different operating conditions. It should be noted that the exact
Table 3 Quantity of alcohol dissolved in wax predicted by Aspen Plus modeling. Carbon number
Hydrocarbon (mol/ h)
Total alcohol (mol/h)
Alcohol dissolved in wax (mol/h)a
1 2 3 4 5 6–10 11–18
6.79E-03 1.75E-03 1.97E-03 1.36E-03 9.12E-04 2.48E-03 1.24E-03
7.51E-04 7.18E-04 1.64E-04 1.15E-04 9.01E-05 1.91E-04 4.09E-05
5.01E-06 9.05E-06 3.81E-06 5.15E-06 7.46E-06 6.09E-05 3.67E-05
a Reaction conditions: T = 270 °C,P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, CO = 4, Thottrap = 110 °C, Tcoldtrap = 3 °C, Phottrap = 2.7 MPa, Pcoldtrap = 2.7 MPa.
H2/
temperature value of the hot trap could not be given to guarantee no water condensation because it also depends on the trap size, CO conversion and CO2 selectivity in reality. The vapor-liquid equilibrium of the mixture in the cold trap is shown in Fig. 14. The liquid phase is the condensed oil products, while
Fig. 9. Influence of temperature of hot trap on alcohol distributions. Reaction conditions: T = 270 °C, P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, H2/CO = 4.
7
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 10. Influence of pressure of cold trap on (a) light oil plus tail gas and (b) total hydrocarbon distributions. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4.
Fig. 15. With or without considering the flash loss, the flow rates of the components are almost the same (Fig. 15a). Fig. 15b shows the relative flash loss errors of the flow rates of each component, which indicates that the flash loss mainly happens in C2 − C8 region.
the gas phase is the outputting tail gas. It showed that most of C12+ components are condensed in the liquid phase even at low pressure of the cold trap, while most of C4- are in the gas phase even at high pressure of the trap. This intrinsic property makes the accuracy of the product distribution in this carbon number range.
error in heavy oil collection of Cn % Cn component in WAX-G stream = Cn component in PRODUCTS stream
3.4.2. Simulation of the flash loss in product collection The flash losses during wax and oil product collection are estimated by simulating the sudden pressure drop and/or temperature raise. The influence of flash loss on the flow rates of hydrocarbons is shown in
Fig. 11. Flow scheme to simulate the real product collecting system in FTS.
8
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 12. Simulated quantity of alcohol condensed in wax with temperature changes of the hot trap. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/ CO = 2.4, Pcoldtrap = 3.0 MPa, Tcoldtrap = 3 °C.
error in light oil collection of Cn % =
total error of Cn % =
Cn component in OIL-G stream Cn component in PRODUCTS stream
reaction temperature from 260 to 290 °C and reaction pressure from 2.0 to 4.0 Mpa, and accordingly, CO conversion changes from 7.5% to 96.1%. The commonly appeared irregularities in Fig. 1 never happened even under such a wide range of reaction conditions. Fig. 16a and b presents some so-called two-alpha distributions when producing more heavy hydrocarbons, while Fig. 16c and d present gradually changed smooth curves when producing more light hydrocarbons.
Cn component in (OIL-G stream + WAX-G stream) Cn component in PRODUCTS stream
The maximum error appears in C4, around 5% at current situation. Moreover, the results show that the flash loss mainly comes from the cold trap, and the contribution originated from the hot trap is close to zero. This is feasible because the wax mainly consists of heavier hydrocarbons with high boiling point.
4. Conclusions The effects of the operating conditions and experimental errors on the FTS product distribution are investigated. The traditional two-step (a hot trap and a cold trap) product condensation scheme is analyzed based on a wide range of experimental operating conditions and theoretical simulation. The flash loss from the hot trap and the cold trap is experimentally checked and the influence on the product distribution is negligible. The
3.4.3. Product distribution at various reaction conditions By carefully controlling the operating parameters of the two traps mentioned above, together with the CAN method for GC data processing, experimental product distribution of FTS at various conditions are exhibited, as shown in Fig. 16. The reaction conditions include inlet H2/ CO from 1.6 to 5.0, s.v. from 15000 mL h−1 g−1 to 80000 mL h−1 g−1,
Fig. 13. Simulated results of relationship among operating temperature and pressure of hot trap and water condensation in hot trap. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4, Pcoldtrap = 3.0 MPa, Tcoldtrap = 3 °C.
9
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 14. Simulated molar flow rates of tail gas and light oil at different pressures of cold trap.Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/ CO = 2.4, Thottrap = 170 °C, Tcoldtrap = 3 °C.
Fig. 15. (a) Simulated total product molar flow rate with and without considering flash loss; (b) Estimating the relative flash loss errors in product collecting from cold trap and hot trap.Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4, Thottrap = 170 °C, Tcoldtrap = 3 °C, Pcoldtrap = 3 MPa.
10
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 16. Experimental FTS product distributions at various conditions: (a) T = 280 °C, P = 3.0 MPa, s.v. = 20000 mL h−1 g−1; (b) P = 4.0 MPa, s.v. = 32000 mL h−1 g−1, H2/ CO = 4.0; (c) T = 280 °C, P = 3.0 MPa,H2/CO = 1.6; (d) d1：T = 260 °C, P = 3.9 MPa,H2/CO = 1.6, s.v.=32000 mL h−1 g−1, XCO = 26.4%; d2: T = 290 °C, P = 2.0 MPa,H2/ CO = 1.6, s.v. = 16000 mL h−1 g−1, XCO = 62.6%; d3: T = 290 °C, P = 4.0 MPa,H2/CO = 2.5, s.v. = 15000 mL h−1 g−1,XCO = 96.1%.
References
trends of the relationship between the operating parameters and the quantity of the flash loss is also calculated by Aspen Plus modeling, which are in agreement with the experimental results. The temperature of the hot trap should maintain a high level (e.g., 190 °C) to avoid water condensation in the trap and avoid underestimating the selectivity of the high-carbon-number alcohols. The GC data processing methods are discussed and a more convenient and accurate method, which named corrected area normalization, is developed to quantify the heavy oil products. The FTS product distribution at various reaction conditions are exhibited. These results are very important in mechanism study, kinetic modeling and parameter estimation.
[1] P.M. Maitlis, A. de Klerk, Greener Fischer-Tropsch Processes for Fuels and Feedstocks, John Wiley & Sons, 2013. [2] J. Xu, Y. Yang, Y.-W. Li, Curr. Opin. Chem. Eng. 2 (2013) 354–362. [3] Y.Y. Ji, H.W. Xiang, J.L. Yang, Y.Y. Xu, Y.W. Li, B. Zhong, Appl. Catal. A: Gen. 214 (2001) 77–86. [4] Jun Yang, Ying Liu, Jie Chang, Yi-Ning Wang, Liang Bai, Yuan-Yuan Xu, HongWei Xiang, Yong-Wang Li, Bing Zhong, Ind. Eng. Chem. Res. 42 (2003) 5066–5090. [5] B.-T. Teng, J. Chang, C.-H. Zhang, D.-B. Cao, J. Yang, Y. Liu, X.-H. Guo, H.W. Xiang, Y.-W. Li, Appl. Catal. A: Gen. 301 (2006) 39–50. [6] J. Chang, L. Bai, B. Teng, R. Zhang, J. Yang, Y. Xu, H. Xiang, Y. Li, Chem. Eng. Sci. 62 (2007) 4983–4991. [7] D.B. Bukur, L. Nowicki, X.S. Lang, Energy Fuels 9 (1995) 620–629. [8] D.B. Bukur, X.S. Lang, Ind. Eng. Chem. Res. 38 (1999) 3270–3275. [9] D.B. Bukur, L. Nowicki, X. Lang, Chem. Eng. Sci. 49 (1994) 4615–4625. [10] E. Iglesia, S.C. Reyes, R.J. Madon, J. Catal. 129 (1991) 238–256. [11] R.J. Madon, W.F. Taylor, J. Catal. 69 (1981) 32–43. [12] T.J. Donnelly, C.N. Satterfield, Appl. Catal. 56 (1989) 231–251. [13] I.C. Yates, C.N. Satterfield, Energy Fuels 6 (1992) 308–314. [14] R.J. Madon, E. Iglesia, S.C. Reyes, Non-flory product distributions in FischerTropsch synthesis catalyzed by ruthenium, cobalt, and iron, selectivity in catalysis, Am. Chem. Soc. (1993) 383–396. [15] T.J. Donnelly, I.C. Yates, C.N. Satterfield, Energy Fuels 2 (1988) 734–739. [16] B. Schliebs, J. Gaube, Ber. Bunsenges. Phys. Chem. 89 (1985) 68–73. [17] J. Gaube, H.F. Klein, J. Mol. Catal. A: Chem. 283 (2008) 60–68. [18] N.O. Egiebor, W.C. Cooper, Appl. Catal. 17 (1985) 47–56. [19] M. Ding, Y. Yang, B. Wu, T. Wang, H. Xiang, Y. Li, Fuel Process. Technol. 92 (2011) 2353–2359. [20] B. Shi, B.H. Davis, Appl. Catal. A: Gen. 277 (2004) 61–69. [21] Y. Liu, S. Zheng, B. Shi, J. Li, J. Mol. Catal. A: Chem. 276 (2007) 110–115. [22] C.M. Masuku, W.D. Shafer, W. Ma, M.K. Gnanamani, G. Jacobs, D. Hildebrandt,
Acknowledgments This work was financially supported by the Major State Basic Research Development Program of China (973 Program) (No. 2012CB215305). The authors are grateful to Synfuels China technology Co., Ltd for the financial and academic supports.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.cattod.2017.05.056. 11
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
D. Glasser, B.H. Davis, J. Catal. 287 (2012) 93–101. [23] C. Perego, S. Peratello, Catal. Today 52 (1999) 133–145. [24] E. Lox, F. Coenen, R. Vermeulen, G.F. Froment, Ind. Eng. Chem. Res. 27 (1988) 576–580. [25] S. Król, B. Zabiegała, J. Namieśnik, J. Chromatogr. A 1249 (2012) 201–214. [26] B.E.H. Saxberg, B.R. Kowalski, Anal. Chem. 51 (1979) 1031–1038. [27] B. Sarup, B.W. Wojciechowski, Can. J. Chem. Eng. 62 (1984) 249–256.
[28] J. Gao, B. Wu, L. Zhou, Y. Yang, X. Hao, J. Xu, Y. Xu, Y. Li, Ind. Eng. Chem. Res. 51 (2012) 11618–11628. [29] A. Plus, Aspen Plus User Guide, Aspen Technology Limited, Cambridge, Massachusetts, United States, 2003. [30] J.J. Marano, G.D. Holder, Fluid Phase Equilib. 138 (1997) 1–21.
12
Contents lists available at ScienceDirect
Catalysis Today journal homepage: www.elsevier.com/locate/cattod
Effects of experimental operations on the Fischer-Tropsch product distribution ⁎
⁎
Ruyi Yanga,c, Liping Zhoub, , Junhu Gaob, Xu Haoa,b, Baoshan Wua,b, , Yong Yanga,b, Yongwang Lia,b a b c
State Key Laboratory of Coal Conversion, Institute of Coal Chemistry, Chinese Academy of Sciences, Taiyuan 030001, PR China National Energy Center for Coal to Liquids, Synfuels China Co., Ltd., Huairou District, Beijing 101400, PR China University of Chinese Academy of Sciences, Beijing 100049, PR China
A R T I C L E I N F O
A B S T R A C T
Keywords: Fischer-Tropsch synthesis Simulated distillation Product distribution Flash loss Separation
The present study provides insight into the effect of experimental operations on the Fischer-Tropsch synthesis (FTS) product distribution by combining experimental and theoretical analyses of the process. Experimental errors originated from GC data processing and product collecting conditions are addressed here based on the experimental data obtained using commercial iron-based catalyst in a tubular fixed-bed reactor. A corrected area normalization method, which is a combination of the area normalization and simulated distillation, is proved to be an accurate and convenient method for FTS heavy oil chromatogram quantification. The flash loss from the hot trap and the cold trap is found to be negligible under the normal operating conditions. However, water may be condensed remarkably in the hot trap at low temperature and wrapped by wax. In addition, the selectivity of the high-carbon-number alcohols will be seriously underestimated at low temperature of the hot trap. The FTS product distributions under a wide range of reaction conditions are exhibited.
1. Introduction Fischer-Tropsch Synthesis which converts carbonaceous materials such as coal, renewable biomass, and natural gas (via gasification) into clean transport fuels or other valuable organics has received great attention in the past decades [1,2]. Generally speaking, FTS is a polymerization-like reaction and the product spectrum consists of a very broad range of hydrocarbons (mainly linear paraffin and linear olefin) and oxygenates (mainly alcohol), with carbon number from C1 to even C120+. The Anderson-Schulz-Flory equation is usually introduced to describe the relative ratios of these polymers of different length, with one constant chain growth probability, regarded as ASF distribution. However, a straight line of the hydrocarbon distribution only occurs when the process produce light hydrocarbons, e.g. carbon number below fourteen. Tremendous experimental results showed that the hydrocarbon selectivity did not obey the ideal ASF distribution under the industrial operating conditions [3–13]. The compositions decrease gradually with the increasing carbon number but show some irregularities at the same time, as exhibited in Fig. 1. The main trend is approximately fitted by a so-called double-α (α1 and α2) distribution [14–17]. Besides, some worse irregularities deviating from the double-α distribution occurred [4–9]. Case 1 and case 4 are some saddle-like
⁎
shapes that appeared around C10–C25 and heavier hydrocarbons, respectively, depending on the specific catalysts and the experimental conditions. Case 2 and Case 3 refer to some positive and negative deviations at high-carbon-number region. The cause of these deviations is a truly controversial issue. Lots of studies tried to explain it by the olefin re-adsorption and the subsequent secondary reactions, by the promoters’ effects on the FTS catalysts which may lead to different kinds of active sites, or by the existence of multiple reaction mechanisms in the chain growth process [16,18]. Some others even attributed it to the cracking of heavier hydrocarbons which may happen under the FTS reaction conditions because of the acid-supported catalysts [7]. Our recent studies showed that the experimental artifacts in FTS data collection also played a very important role in the patterns of the product distribution curve. Considering the importance of the FTS experimental data to the reaction mechanism assumption, the product kinetic modeling and even the catalyst development and selectivity control, the present article further clarifies the influence of the experimental artifacts on the FTS product distribution, e.g., the product analysis methods, the product condensing and collecting and flash loss, based on the experimental results and the theoretical calculation.
Corresponding authors at: National Energy Center for Coal to Liquids, Synfuels China Co., Ltd., Huairou District, Beijing 101400, PR China. E-mail addresses: [email protected] (L. Zhou), [email protected], [email protected] (B. Wu).
http://dx.doi.org/10.1016/j.cattod.2017.05.056 Received 30 September 2016; Received in revised form 22 March 2017; Accepted 15 May 2017 0920-5861/ © 2017 Elsevier B.V. All rights reserved.
Please cite this article as: Yang, R., Catalysis Today (2017), http://dx.doi.org/10.1016/j.cattod.2017.05.056
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 1. The commonly appeared irregularities in the FTS product distribution. The solid line: approximately estimating of FTS product distribution by a double-α model; the dashed lines: case 1 and case 4 are some saddle-like shapes that appeared around C10-C25 and heavier hydrocarbons, respectively. Case 2 and Case 3 are some positive and negative deviations at high-carbon-number region. (Reprinted from Industrial & Engineering Chemistry Research, 51 (2012) 11618–11628.) [28].
2. Experimental Fig. 2. Flow scheme of the fixed-bed experimental setup. 1. purification unit (de-sulphur, de-oxygen, de-carbonyl, de-water); 2. mass flow controller; 3. thermocouple; 4. hot trap; 5. cold trap; 6 & 7. back pressure regulator; 8.wet flow meter.
2.1. Catalyst and reduction The industrial iron-based catalysts were provided by Synfuels China Co., Ltd. Typically 1.0 g catalyst with particle size 150–180 μm was used in order to eliminate the internal diffusion. Prior to its use, the catalysts were reduced by synthesis gas at H2/CO = 2, T = 260 °C, P = 0.1 MPa, space velocity (s.v.) = 1000 mL g−1 h−1 for 48 h. The detailed characterization and properties of the catalysts could be seen elsewhere [19].
In the feeding section, the hydrogen (purity > 99.999%) and the carbon monoxide (purity > 99.999%) passed through a series of purifiers to further remove oxygen, sulfur, carbonyls and water and the flow rates were measured by the thermal mass flow controllers (Brooks 5850E). The tubular fixed bed reactor is a stainless steel tube (total length 80 cm and internal diameter 1 cm) with a thermal-well (external diameter 0.1 cm) in the center. The reaction heat is removed by a molten salt bath, which is composed of 53% KNO3, 40% NaNO2 and 7% NaNO3. The salt bath is stirred strongly by a churning motor to obtain a better heat transfer efficiency and keep acceptable temperature uniformity (see in Fig. 3). Moreover, considering the strongly exothermic property of the FTS reaction, a remarkable dilution of the catalyst bed was performed by inert Silicon Carbide, with the same particle size and mass ratio of 30.The real temperature profile along the catalytic bed is shown in Fig. 3. The hot spot appears in the first 2 cm of the catalytic bed. The maximum temperature gradient is around 1.5 °C, which is negligible in the FTS reaction. The gas-liquid mixture of the reactor effluent enters into a hot trap (typically 170 °C), then passes through a back pressure regulator and enters into a cold trap (typically 2 °C), as shown in Fig. 2. The noncondensable gases release via a wet flow meter (Ritter TG 1–5). Accordingly, the product collection consists of three parts, the heavy oil (the so-called wax, C6–C70 paraffins and olefins) condensed in the hot trap, the light oil (C3–C35 paraffins and olefins), the oxygenates (mainly C1–C20 alcohols) and the water condensed in the cold trap and the non-condensable gases, based on the boiling range of the reactor effluent, as shown in Fig. 4. The detailed GC configuration parameters and the flow scheme of the FTS product analysis are shown in Fig. 4. The non-condensable gases include H2, CO, CO2, C1–C12 paraffins and C2–C12 olefins. These are analyzed by an online Agilent 7890 B GC equipped with two TCD channels and one FID channel, in which H2 is detected on TCD1 with a HayeSep Q column (0.5m × 1/8in., mesh size 80/100), CO, CO2, CH4 are detected on TCD2 with a Agilent MolSieve 13X column (1.83m × 1/8in., mesh size 60/80) and the C1–C9 hydrocarbons are detected on FID with a DM-Plot Al2O3/Na2SO4 column
2.2. Experimental setup In view of the close relationship between the experimental setup and the accuracy of the experimental data, the reactor to be used in product distribution study should be considered carefully. Several types of reactors have been used in bench-scale FTS experimental data collection in published papers, e.g., stirred slurry autoclave, spinning basket reactor, Berty-type internal recycle reactor and tubular fixed-bed reactor. The first three types all could be regarded as a “point” operation pattern, the temperature and concentration inside the reactor are in uniform. This property is convenient for FTS kinetic modeling. However, none of them is perfect for the FTS product distribution study. The accumulation of high-carbon-number hydrocarbons in stirred slurry autoclave would seriously disguise the true FTS product selectivity [12,20–22]. The Berty-type recycle reactor is suitable for light hydrocarbon synthesis, but it is hard to be run under real FTS conditions because of the blocking of the agitating shaft by the produced heavy wax. Continuous spinning basket reactor has all the merits of gradientless reactor and has no accumulation effect. However, it is difficult to know the real temperature and the real space velocity of the catalytic bed [23]. The tubular fixed-bed reactor has several distinct advantages in product distribution study, such as easily experimental operating, less product accumulation inside the reactor and shorter time for environment replacement when changing reaction conditions. By suitable catalyst dilution and molten-salt-bath heating, the temperature gradient commonly met in an industrial-scale fixed-bed reactor could be negligible in a laboratory reactor [24]. The flow scheme of the experimental setup used in this study is shown in Fig. 2. It consists of four sections: the feeding, the tubular fixed-bed reactor, the product collection and the product analysis. 2
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 3. The temperature profile along the tubular reactor with and without reaction. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4.
(50m × 0.32 mm × 5 μm). In the cold trap, the light oil and the water layered and the oxygenates dispersed into the two phases. The oil phase components are analyzed by an offline Agilent 7890A GC with a HPPONA column (50m × 200 μm × 0.5 μm), while the water phase components are analyzed by an offline Agilent 7890A GC with an ABINOWAX column (50m × 200 μm × 0.5 μm). The wax sample is first dissolved in CS2 and then analyzed by an offline Agilent 7890A GC with a UA1-30M-0.5F column (32m × 530 μm × 0.5 μm).
collecting enough products for analysis. The total mass balances and the elemental balances (C, H and O) of the experimental data were around 97%–102%. 3. Results and discussion 3.1. Influence of GC data processing on the FTS product distribution FTS products are less complex than petroleum-based samples, but containing some heavier components (especially those from C70 to nearly C120 or more), e.g., the heavy oil condensed in the hot trap showed in Fig. 4. The characterization of these high boiling point hydrocarbons requires high temperature GC (oven up to 450 °C) in order to elute the components. However, the typical columns for detailed analysis of FTS products are limited to around 350 °C because of the limited thermal stabilities of capillary column and the stationary phase. This will lead to loss of peaks of heavier components in reality.
2.3. Experimental procedure To prevent temperature runaway, the reactor temperature was cooled down to 220 °C after the catalyst reduction, and then the operating parameters are adjusted gradually to meet the desired reaction conditions. After a steady-state transition period, normally 48 h needed for sure, the experimental data collection for product distribution study could be started. Generally 10 h are necessary for
Fig. 4. Flow scheme of separation and analysis procedure of the FTS products.
3
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 5. Influence of chromatographic data processing methods of wax on the product distributions. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/ CO = 2.4, Thottrap = 170 °C, Pcoldtrap = 3.0 MPa.
distillation (HTSD, based on the ASTM D7169 method). The recovered mass% value Fn (n denotes carbon number Cn) obtained from HTSD, multiplying by the area normalization data, would be necessary to eliminate the overestimation originated from residue fraction which does not distill at the maximum operating temperature. The Fn values of FTS heavy oil and light oil are shown in Fig. 6. The distillation yield of light oil at 520 °C(up to C34) is nearly 100%, while that of heavy oil is only around 41%. At 645 °C(up to C70), the Fn value of heavy oil is only 53.4%. The results indicate that the AN method is suitable for light oil chromatogram quantification within the scheme of separation and analysis procedure of FTS products showed in Fig. 4, but the Fn value must be checked and added in heavy oil chromatogram quantification.
Therefore, one should be very careful in GC data processing of FTS products because some traditional methods may introduce errors. The internal standard (IS) method is a general way in chromatogram quantification, but considering the complexity of FTS heavy oil, it is boring to find a special component that is similar in physico-chemical properties to the analytes but does not co-elute with them. An alternative way, standard addition method (SA, showed in Eq. (1)) is used here [25,26]. Tetradecane is served as a reference. The results in Fig. 5 showed that the paraffin, the olefin and the total hydrocarbon distributions are all smooth lines without any saddle-like shapes. The paraffin distribution exhibited remarkable double-α property compared to the olefin distribution, which is close to a straight line.
% component a =
f Aa m × a × s × ws × 100% As fs mt
% component a =
(1)
∑
Where As and Aa are peak areas of tetradecane in the standard solution and component a in the sample solution, respectively; fs and fa are the response factors of tetradecane and the component a, respectively; ms and mt are the mass of the standard solution and the sample solution; ws is the mass percentage of tetradecane in the standard solution. The SA (see also IS) method requires a precisely prepared standard solution and a weighed heavy oil sample. Moreover, the normally used CS2 solvent in sample preparation is highly volatile. These make it easier to introduce experimental errors in practical operations. Area normalization (AN), showed in Eq. (2), is a simple and convenient method to the quantification of chromatograms data. It does not require additional calibration standards, and small variations in sample preparation and/or detector operating parameters have approximately the same relative effect on each component. Nevertheless, this method will lead to remarkable over-estimation of the analytes from around C12 in the heavy oil (showed in Fig. 5) because some heavier components could not be recorded by the detector.
% component a =
fa Aa × 100% ∑ fi Ai
fa Aa n
× Fn × 100%
fi Ai
i =1
(3)
Where faAa is the corrected peak area of component a, ∑fiAi is the sum of the corrected peak areas in the chromatogram, and Fn is the recovered mass% to Cn obtained from the simulated distillation. It can be seen from Fig. 5 that the product distribution curves calculated by CAN method are quite consistent with those calculated by SA method, but with relatively simple operation and less opportunities for experimental error introducing. Therefore, CAN method could be a recommended method for accurate and trustable FTS heavy oil chromatogram quantification. 3.2. Influence of experimental errors on the FTS product distribution As discussed in Fig. 4, the FTS product collection and analysis refer to a number of steps, in which the product outflow from the catalytic bed, the product condensation in the two traps, the separation of oil phase and water phase and all these steps may introduce random errors. Hence, it is necessary to check the effect of the random experimental errors on the repeatability and accuracy of FTS product distribution. Four experimental data within 50 h are collected and investigated at a constant reaction condition [27]. The repeatability of hydrocarbon, alcohol and ratio of olefin to paraffin (O/P) distribution curves of four measurements has been compared in Fig. 7. The relative standard deviation(RSD) of the components, which reflects how widely spread the measured values are on either side of the mean, is shown in Table 1. The comparison of
(2)
Where faAa is the corrected peak area of component a;∑fiAi is the sum of the corrected peak areas in the chromatogram. In this paper, a corrected area normalization method (CAN, showed in Eq. (3)), which is a combination of the area normalization and simulated distillation, is developed for FTS heavy oil chromatogram quantification. An Agilent 7890A GC with a DB-1 column (5m × 530 μm × 0.17 μm) is used for high temperature simulated 4
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 6. The HTSD results of the FTS light oil and heavy oil.
Fig. 7. Repeatability of hydrocarbon and alcohol distributions and the ratio of olefin to paraffin. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4.
5
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
trap is critical to the accuracy of the alcohol distribution. As shown in Fig. 9, the selectivity of C1–C6 alcohols is repeatable at different temperatures, but that of the heavy alcohols at 170 °C is significantly higher than that at 110 °C. This could be explained by the fact that the lower temperature is set, the more alcohols will be condensed and dissolved in the wax (as mentioned above, this part of alcohols are hard to be identified in wax analysis). The distribution trends at 170 °C and 190 °C are close to each other, which indicate that at least this temperature level should be maintained to guarantee a real alcohol distribution. These accurate experimental results are very important to reaction mechanism and kinetic studies. It should be noted that the quantity of alcohol dissolved in the wax is almost two orders of magnitude less than that of hydrocarbon products, as shown in Table 3. Therefore, the error introduced into hydrocarbon distribution by alcohol condensation is negligible. This is also proved by the results in Fig. 8, in which the hot trap was operated at different temperatures, indicating different levels of alcohol condensation. During the product collection from the cold trap, the liquid light hydrocarbons may be vaporized and lost because of the suddenly environment changes from high pressure and low temperature to low pressure and high temperature, e.g., 0.1 MPa and 25 °C. The effect of this flash loss on the FTS product distribution is investigated. By adjusting the back pressure regulator behind the cold trap (Fig. 2), the operating pressure of the cold trap is experimentally changed from 3.0 MPa to 0.3 MPa, and the results are shown in Fig. 10. The quantities of the oil plus tail gas are almost the same at such a broad range of pressures (Fig. 10a), which means that the flash loss in oil collection is negligible. Thus, it is feasible to see that the total product distributions at different operating pressures of the cold trap are overlapped perfectly.
Table 1 Relative standard deviations of product distributions and O/P in experimental repeatability. Carbon number
Total hydrocarbon (%)
Paraffin (%)
Olefin (%)
O/P (%)
Alcohol (%)
1 2 3 4 5–9 10–15 16–20 21–34 35+
7.5 7.1 7.4 7.6 7.4 5.0 10.3 10.8 11.5
7.5 6.8 7.7 8.1 8.9 7.9 12.6 11.8 11.7
– 7.3 7.3 7.3 6.7 3.0 8.4 11.3 –
– 0.9 0.6 0.8 2.4 5.4 6.8 9.9 –
4.9 6.2 7.0 8.0 7.5 6.5 6.4 – –
the curves shows that the multi-times experiments provide a good repeatability of the detailed product distributions. However, the repeatability decreases gradually with the increasing carbon number, e.g., RSD values of paraffin from 7.5% to around 12%, as shown in Table 1. The experimental O/P ratios are normally trustable (RSD below 5%) except for these at high-carbon-number region where some iso-olefin peaks could not be separated accurately from paraffin peaks in GC analysis.
3.3. Influence of product collecting conditions on the product distribution During the multistep flash separation processes in FTS product condensation and collection, the operating temperature and pressure of the two traps (shown in Fig. 2) are two key parameters. They determine the partition of the products with different boiling point in the gas and liquid phase, and also the extent of flash loss of the products during the product collecting process. The effects of these collecting conditions on the FTS product distribution are investigated here. Table 2 shows the sensitivity of the quantities of water and liquid hydrocarbons partitioned in the two traps to temperature changes, respectively. At 110 °C, most of the produced water is condensed in the hot trap. This water could easily be wrapped by wax during product collection. Therefore, the quantity of wax may be overestimated and the product distribution would be lack fidelity [28]. This phenomenon could be improved by increasing the temperature of the hot trap, e.g., the water in the hot trap is only 0.14 g/h at 190 °C and this will even be negligible at 200 °C. Although the temperature variation changed the partition of hydrocarbons in the wax phase and the oil phase, the total amount of the liquid hydrocarbon is almost a constant and this partition have no effect on the accuracy of GC analysis. Hence, the hydrocarbon distribution is repeatable at different temperatures of the hot trap, as shown in Fig. 8, provided that one has complete wax and wrapped water separation after the wax collection from the hot trap. Few people paid attention to the oxygenate (mainly alcohols) distribution in FTS. Actually, we find that the temperature of the hot
3.4. Simulation of product separation and collection process For an in-depth and comprehensive insight into the effect of collecting process conditions on the final products distribution, the product separating and collecting process is modeled by Aspen Plus software. The schematic simulation diagram is shown in Fig. 11. The PRODUCTS stream denotes the total effluent from the reactor, which includes CO, H2, CO2, H2O, C1–C60 paraffins, C2–C35 olefins and C1–C19 n-alcohols. It goes through a hot trap (HOTTRAP) and a cold trap (COLDTRAP) successively as the real experimental condition. A three phase flash module in Aspen Plus is adopted to describe the two traps [29]. The WATER1 stream simulates the accumulated water in the hot trap. The WATER2 stream simulates the water phase products condensed in the cold trap. FLASHOIL and FLASHWAX are added to simulate the operating conditions for flash loss calculation during oil and wax collection. The TAILGAS, OIL-L and WAX-L streams simulate the tail gas, oil and wax products obtained in real situation, respectively. OIL-G and WAX-G streams denote the quantities of lost hydrocarbons from the hot trap and the cold trap because of flash loss. RKSOAVE equation of state in Aspen Plus is chosen for the flash equilibrium calculation. The physical property data of the light hydrocarbons are obtained directly from the PURE 11 database in Aspen Plus, while these required of the heavy hydrocarbons are estimated by extrapolation through the ABC methods introduced by Marano and Holder [30]. In the following simulation, the experimental data obtained under the reaction condition mentioned above (T = 260 °C, P = 3.3 MPa, GHSV = 28000 mL g−1 h−1, H2/CO = 2.4) are used as the compositions of the PRODUCTS stream. The operating parameters, e.g., temperature and pressure, of the two traps (HOTTRAP and COLDTRAP) refer to the experimental conditions mentioned above. The temperature and pressure of the FLASHOIL and FLASHWAX set as 25 °C and 0.1 MPa, respectively.
Table 2 Influence of temperature of hot trap on the repartition of reactor effluent. Thottrap(°C)
Water (g/h)
Heavy oil(g/ h)
Light oil(g/ h)
Tailgas(g/ h)
6.37 6.25
Cold trapa
Hot trapb
110
0.56
1.41
1.21
0.20
170 190
1.73 1.86
0.14 0.00
1.06 0.88
0.44 0.51
6.40
a
Water condensed in the cold trap. Water condensed in the hot trap. Reaction T = 270 °C,P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, H2/CO = 4. b
3.4.1. Simulation of the product separation To clarify the trends of the product repartition in a vapor and in a
conditions:
6
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 8. Influence of temperature of hot trap on hydrocarbon distributions.Reaction conditions: T = 270 °C, P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, H2/CO = 4.
liquid phase during condensation, simulation is performed where vapor-liquid equilibrium is calculated at different temperatures and pressures of the two traps for the same PRODUCT stream. With increasing temperature from 100 °C to 200 °C, the molar flow rate of alcohol outputting from the hot trap decreases remarkably, which means alcohols are heavily condensed in the hot trap at low temperature and are omitted, especially these from C3 to C16 (Fig. 12). Due to the co-elution of alcohols and iso-hydrocarbons during the chromatography analysis of wax it is not easy to quantify the content of alcohol condensed in wax. Therefore, to avoid underestimating the alcohol contents the temperature of hot trap should be kept at a high level. This is in agreement with the experimental results shown in Fig. 9. The water content, defined as the molar ratio between the condensed water in the hot trap and the total water produced, is shown in Fig. 13. It clearly exhibits the relationships among temperature, pressure and water content in the hot trap, so that provides useful information for operators to estimate the possible water content at different operating conditions. It should be noted that the exact
Table 3 Quantity of alcohol dissolved in wax predicted by Aspen Plus modeling. Carbon number
Hydrocarbon (mol/ h)
Total alcohol (mol/h)
Alcohol dissolved in wax (mol/h)a
1 2 3 4 5 6–10 11–18
6.79E-03 1.75E-03 1.97E-03 1.36E-03 9.12E-04 2.48E-03 1.24E-03
7.51E-04 7.18E-04 1.64E-04 1.15E-04 9.01E-05 1.91E-04 4.09E-05
5.01E-06 9.05E-06 3.81E-06 5.15E-06 7.46E-06 6.09E-05 3.67E-05
a Reaction conditions: T = 270 °C,P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, CO = 4, Thottrap = 110 °C, Tcoldtrap = 3 °C, Phottrap = 2.7 MPa, Pcoldtrap = 2.7 MPa.
H2/
temperature value of the hot trap could not be given to guarantee no water condensation because it also depends on the trap size, CO conversion and CO2 selectivity in reality. The vapor-liquid equilibrium of the mixture in the cold trap is shown in Fig. 14. The liquid phase is the condensed oil products, while
Fig. 9. Influence of temperature of hot trap on alcohol distributions. Reaction conditions: T = 270 °C, P = 2.7 MPa, s.v. = 32000 mL h−1 g−1, H2/CO = 4.
7
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 10. Influence of pressure of cold trap on (a) light oil plus tail gas and (b) total hydrocarbon distributions. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4.
Fig. 15. With or without considering the flash loss, the flow rates of the components are almost the same (Fig. 15a). Fig. 15b shows the relative flash loss errors of the flow rates of each component, which indicates that the flash loss mainly happens in C2 − C8 region.
the gas phase is the outputting tail gas. It showed that most of C12+ components are condensed in the liquid phase even at low pressure of the cold trap, while most of C4- are in the gas phase even at high pressure of the trap. This intrinsic property makes the accuracy of the product distribution in this carbon number range.
error in heavy oil collection of Cn % Cn component in WAX-G stream = Cn component in PRODUCTS stream
3.4.2. Simulation of the flash loss in product collection The flash losses during wax and oil product collection are estimated by simulating the sudden pressure drop and/or temperature raise. The influence of flash loss on the flow rates of hydrocarbons is shown in
Fig. 11. Flow scheme to simulate the real product collecting system in FTS.
8
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 12. Simulated quantity of alcohol condensed in wax with temperature changes of the hot trap. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/ CO = 2.4, Pcoldtrap = 3.0 MPa, Tcoldtrap = 3 °C.
error in light oil collection of Cn % =
total error of Cn % =
Cn component in OIL-G stream Cn component in PRODUCTS stream
reaction temperature from 260 to 290 °C and reaction pressure from 2.0 to 4.0 Mpa, and accordingly, CO conversion changes from 7.5% to 96.1%. The commonly appeared irregularities in Fig. 1 never happened even under such a wide range of reaction conditions. Fig. 16a and b presents some so-called two-alpha distributions when producing more heavy hydrocarbons, while Fig. 16c and d present gradually changed smooth curves when producing more light hydrocarbons.
Cn component in (OIL-G stream + WAX-G stream) Cn component in PRODUCTS stream
The maximum error appears in C4, around 5% at current situation. Moreover, the results show that the flash loss mainly comes from the cold trap, and the contribution originated from the hot trap is close to zero. This is feasible because the wax mainly consists of heavier hydrocarbons with high boiling point.
4. Conclusions The effects of the operating conditions and experimental errors on the FTS product distribution are investigated. The traditional two-step (a hot trap and a cold trap) product condensation scheme is analyzed based on a wide range of experimental operating conditions and theoretical simulation. The flash loss from the hot trap and the cold trap is experimentally checked and the influence on the product distribution is negligible. The
3.4.3. Product distribution at various reaction conditions By carefully controlling the operating parameters of the two traps mentioned above, together with the CAN method for GC data processing, experimental product distribution of FTS at various conditions are exhibited, as shown in Fig. 16. The reaction conditions include inlet H2/ CO from 1.6 to 5.0, s.v. from 15000 mL h−1 g−1 to 80000 mL h−1 g−1,
Fig. 13. Simulated results of relationship among operating temperature and pressure of hot trap and water condensation in hot trap. Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4, Pcoldtrap = 3.0 MPa, Tcoldtrap = 3 °C.
9
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 14. Simulated molar flow rates of tail gas and light oil at different pressures of cold trap.Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/ CO = 2.4, Thottrap = 170 °C, Tcoldtrap = 3 °C.
Fig. 15. (a) Simulated total product molar flow rate with and without considering flash loss; (b) Estimating the relative flash loss errors in product collecting from cold trap and hot trap.Reaction conditions: T = 260 °C, P = 3.3 MPa, s.v. = 28000 mL h−1 g−1, H2/CO = 2.4, Thottrap = 170 °C, Tcoldtrap = 3 °C, Pcoldtrap = 3 MPa.
10
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
Fig. 16. Experimental FTS product distributions at various conditions: (a) T = 280 °C, P = 3.0 MPa, s.v. = 20000 mL h−1 g−1; (b) P = 4.0 MPa, s.v. = 32000 mL h−1 g−1, H2/ CO = 4.0; (c) T = 280 °C, P = 3.0 MPa,H2/CO = 1.6; (d) d1：T = 260 °C, P = 3.9 MPa,H2/CO = 1.6, s.v.=32000 mL h−1 g−1, XCO = 26.4%; d2: T = 290 °C, P = 2.0 MPa,H2/ CO = 1.6, s.v. = 16000 mL h−1 g−1, XCO = 62.6%; d3: T = 290 °C, P = 4.0 MPa,H2/CO = 2.5, s.v. = 15000 mL h−1 g−1,XCO = 96.1%.
References
trends of the relationship between the operating parameters and the quantity of the flash loss is also calculated by Aspen Plus modeling, which are in agreement with the experimental results. The temperature of the hot trap should maintain a high level (e.g., 190 °C) to avoid water condensation in the trap and avoid underestimating the selectivity of the high-carbon-number alcohols. The GC data processing methods are discussed and a more convenient and accurate method, which named corrected area normalization, is developed to quantify the heavy oil products. The FTS product distribution at various reaction conditions are exhibited. These results are very important in mechanism study, kinetic modeling and parameter estimation.
[1] P.M. Maitlis, A. de Klerk, Greener Fischer-Tropsch Processes for Fuels and Feedstocks, John Wiley & Sons, 2013. [2] J. Xu, Y. Yang, Y.-W. Li, Curr. Opin. Chem. Eng. 2 (2013) 354–362. [3] Y.Y. Ji, H.W. Xiang, J.L. Yang, Y.Y. Xu, Y.W. Li, B. Zhong, Appl. Catal. A: Gen. 214 (2001) 77–86. [4] Jun Yang, Ying Liu, Jie Chang, Yi-Ning Wang, Liang Bai, Yuan-Yuan Xu, HongWei Xiang, Yong-Wang Li, Bing Zhong, Ind. Eng. Chem. Res. 42 (2003) 5066–5090. [5] B.-T. Teng, J. Chang, C.-H. Zhang, D.-B. Cao, J. Yang, Y. Liu, X.-H. Guo, H.W. Xiang, Y.-W. Li, Appl. Catal. A: Gen. 301 (2006) 39–50. [6] J. Chang, L. Bai, B. Teng, R. Zhang, J. Yang, Y. Xu, H. Xiang, Y. Li, Chem. Eng. Sci. 62 (2007) 4983–4991. [7] D.B. Bukur, L. Nowicki, X.S. Lang, Energy Fuels 9 (1995) 620–629. [8] D.B. Bukur, X.S. Lang, Ind. Eng. Chem. Res. 38 (1999) 3270–3275. [9] D.B. Bukur, L. Nowicki, X. Lang, Chem. Eng. Sci. 49 (1994) 4615–4625. [10] E. Iglesia, S.C. Reyes, R.J. Madon, J. Catal. 129 (1991) 238–256. [11] R.J. Madon, W.F. Taylor, J. Catal. 69 (1981) 32–43. [12] T.J. Donnelly, C.N. Satterfield, Appl. Catal. 56 (1989) 231–251. [13] I.C. Yates, C.N. Satterfield, Energy Fuels 6 (1992) 308–314. [14] R.J. Madon, E. Iglesia, S.C. Reyes, Non-flory product distributions in FischerTropsch synthesis catalyzed by ruthenium, cobalt, and iron, selectivity in catalysis, Am. Chem. Soc. (1993) 383–396. [15] T.J. Donnelly, I.C. Yates, C.N. Satterfield, Energy Fuels 2 (1988) 734–739. [16] B. Schliebs, J. Gaube, Ber. Bunsenges. Phys. Chem. 89 (1985) 68–73. [17] J. Gaube, H.F. Klein, J. Mol. Catal. A: Chem. 283 (2008) 60–68. [18] N.O. Egiebor, W.C. Cooper, Appl. Catal. 17 (1985) 47–56. [19] M. Ding, Y. Yang, B. Wu, T. Wang, H. Xiang, Y. Li, Fuel Process. Technol. 92 (2011) 2353–2359. [20] B. Shi, B.H. Davis, Appl. Catal. A: Gen. 277 (2004) 61–69. [21] Y. Liu, S. Zheng, B. Shi, J. Li, J. Mol. Catal. A: Chem. 276 (2007) 110–115. [22] C.M. Masuku, W.D. Shafer, W. Ma, M.K. Gnanamani, G. Jacobs, D. Hildebrandt,
Acknowledgments This work was financially supported by the Major State Basic Research Development Program of China (973 Program) (No. 2012CB215305). The authors are grateful to Synfuels China technology Co., Ltd for the financial and academic supports.
Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.cattod.2017.05.056. 11
Catalysis Today xxx (xxxx) xxx–xxx
R. Yang et al.
D. Glasser, B.H. Davis, J. Catal. 287 (2012) 93–101. [23] C. Perego, S. Peratello, Catal. Today 52 (1999) 133–145. [24] E. Lox, F. Coenen, R. Vermeulen, G.F. Froment, Ind. Eng. Chem. Res. 27 (1988) 576–580. [25] S. Król, B. Zabiegała, J. Namieśnik, J. Chromatogr. A 1249 (2012) 201–214. [26] B.E.H. Saxberg, B.R. Kowalski, Anal. Chem. 51 (1979) 1031–1038. [27] B. Sarup, B.W. Wojciechowski, Can. J. Chem. Eng. 62 (1984) 249–256.
[28] J. Gao, B. Wu, L. Zhou, Y. Yang, X. Hao, J. Xu, Y. Xu, Y. Li, Ind. Eng. Chem. Res. 51 (2012) 11618–11628. [29] A. Plus, Aspen Plus User Guide, Aspen Technology Limited, Cambridge, Massachusetts, United States, 2003. [30] J.J. Marano, G.D. Holder, Fluid Phase Equilib. 138 (1997) 1–21.
12